1 //===- ELF.cpp - ELF object file implementation ---------------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9
10 #include "llvm/Object/ELF.h"
11 #include "llvm/BinaryFormat/ELF.h"
12 #include "llvm/Support/LEB128.h"
13
14 using namespace llvm;
15 using namespace object;
16
17 #define STRINGIFY_ENUM_CASE(ns, name) \
18 case ns::name: \
19 return #name;
20
21 #define ELF_RELOC(name, value) STRINGIFY_ENUM_CASE(ELF, name)
22
getELFRelocationTypeName(uint32_t Machine,uint32_t Type)23 StringRef llvm::object::getELFRelocationTypeName(uint32_t Machine,
24 uint32_t Type) {
25 switch (Machine) {
26 case ELF::EM_X86_64:
27 switch (Type) {
28 #include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
29 default:
30 break;
31 }
32 break;
33 case ELF::EM_386:
34 case ELF::EM_IAMCU:
35 switch (Type) {
36 #include "llvm/BinaryFormat/ELFRelocs/i386.def"
37 default:
38 break;
39 }
40 break;
41 case ELF::EM_MIPS:
42 switch (Type) {
43 #include "llvm/BinaryFormat/ELFRelocs/Mips.def"
44 default:
45 break;
46 }
47 break;
48 case ELF::EM_AARCH64:
49 switch (Type) {
50 #include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
51 default:
52 break;
53 }
54 break;
55 case ELF::EM_ARM:
56 switch (Type) {
57 #include "llvm/BinaryFormat/ELFRelocs/ARM.def"
58 default:
59 break;
60 }
61 break;
62 case ELF::EM_ARC_COMPACT:
63 case ELF::EM_ARC_COMPACT2:
64 switch (Type) {
65 #include "llvm/BinaryFormat/ELFRelocs/ARC.def"
66 default:
67 break;
68 }
69 break;
70 case ELF::EM_AVR:
71 switch (Type) {
72 #include "llvm/BinaryFormat/ELFRelocs/AVR.def"
73 default:
74 break;
75 }
76 break;
77 case ELF::EM_HEXAGON:
78 switch (Type) {
79 #include "llvm/BinaryFormat/ELFRelocs/Hexagon.def"
80 default:
81 break;
82 }
83 break;
84 case ELF::EM_LANAI:
85 switch (Type) {
86 #include "llvm/BinaryFormat/ELFRelocs/Lanai.def"
87 default:
88 break;
89 }
90 break;
91 case ELF::EM_PPC:
92 switch (Type) {
93 #include "llvm/BinaryFormat/ELFRelocs/PowerPC.def"
94 default:
95 break;
96 }
97 break;
98 case ELF::EM_PPC64:
99 switch (Type) {
100 #include "llvm/BinaryFormat/ELFRelocs/PowerPC64.def"
101 default:
102 break;
103 }
104 break;
105 case ELF::EM_RISCV:
106 switch (Type) {
107 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def"
108 default:
109 break;
110 }
111 break;
112 case ELF::EM_S390:
113 switch (Type) {
114 #include "llvm/BinaryFormat/ELFRelocs/SystemZ.def"
115 default:
116 break;
117 }
118 break;
119 case ELF::EM_SPARC:
120 case ELF::EM_SPARC32PLUS:
121 case ELF::EM_SPARCV9:
122 switch (Type) {
123 #include "llvm/BinaryFormat/ELFRelocs/Sparc.def"
124 default:
125 break;
126 }
127 break;
128 case ELF::EM_AMDGPU:
129 switch (Type) {
130 #include "llvm/BinaryFormat/ELFRelocs/AMDGPU.def"
131 default:
132 break;
133 }
134 break;
135 case ELF::EM_BPF:
136 switch (Type) {
137 #include "llvm/BinaryFormat/ELFRelocs/BPF.def"
138 default:
139 break;
140 }
141 break;
142 case ELF::EM_MSP430:
143 switch (Type) {
144 #include "llvm/BinaryFormat/ELFRelocs/MSP430.def"
145 default:
146 break;
147 }
148 break;
149 default:
150 break;
151 }
152 return "Unknown";
153 }
154
155 #undef ELF_RELOC
156
getELFRelativeRelocationType(uint32_t Machine)157 uint32_t llvm::object::getELFRelativeRelocationType(uint32_t Machine) {
158 switch (Machine) {
159 case ELF::EM_X86_64:
160 return ELF::R_X86_64_RELATIVE;
161 case ELF::EM_386:
162 case ELF::EM_IAMCU:
163 return ELF::R_386_RELATIVE;
164 case ELF::EM_MIPS:
165 break;
166 case ELF::EM_AARCH64:
167 return ELF::R_AARCH64_RELATIVE;
168 case ELF::EM_ARM:
169 return ELF::R_ARM_RELATIVE;
170 case ELF::EM_ARC_COMPACT:
171 case ELF::EM_ARC_COMPACT2:
172 return ELF::R_ARC_RELATIVE;
173 case ELF::EM_AVR:
174 break;
175 case ELF::EM_HEXAGON:
176 return ELF::R_HEX_RELATIVE;
177 case ELF::EM_LANAI:
178 break;
179 case ELF::EM_PPC:
180 break;
181 case ELF::EM_PPC64:
182 return ELF::R_PPC64_RELATIVE;
183 case ELF::EM_RISCV:
184 return ELF::R_RISCV_RELATIVE;
185 case ELF::EM_S390:
186 return ELF::R_390_RELATIVE;
187 case ELF::EM_SPARC:
188 case ELF::EM_SPARC32PLUS:
189 case ELF::EM_SPARCV9:
190 return ELF::R_SPARC_RELATIVE;
191 case ELF::EM_AMDGPU:
192 break;
193 case ELF::EM_BPF:
194 break;
195 default:
196 break;
197 }
198 return 0;
199 }
200
getELFSectionTypeName(uint32_t Machine,unsigned Type)201 StringRef llvm::object::getELFSectionTypeName(uint32_t Machine, unsigned Type) {
202 switch (Machine) {
203 case ELF::EM_ARM:
204 switch (Type) {
205 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_EXIDX);
206 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_PREEMPTMAP);
207 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_ATTRIBUTES);
208 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_DEBUGOVERLAY);
209 STRINGIFY_ENUM_CASE(ELF, SHT_ARM_OVERLAYSECTION);
210 }
211 break;
212 case ELF::EM_HEXAGON:
213 switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_HEX_ORDERED); }
214 break;
215 case ELF::EM_X86_64:
216 switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_X86_64_UNWIND); }
217 break;
218 case ELF::EM_MIPS:
219 case ELF::EM_MIPS_RS3_LE:
220 switch (Type) {
221 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_REGINFO);
222 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_OPTIONS);
223 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_ABIFLAGS);
224 STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_DWARF);
225 }
226 break;
227 default:
228 break;
229 }
230
231 switch (Type) {
232 STRINGIFY_ENUM_CASE(ELF, SHT_NULL);
233 STRINGIFY_ENUM_CASE(ELF, SHT_PROGBITS);
234 STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB);
235 STRINGIFY_ENUM_CASE(ELF, SHT_STRTAB);
236 STRINGIFY_ENUM_CASE(ELF, SHT_RELA);
237 STRINGIFY_ENUM_CASE(ELF, SHT_HASH);
238 STRINGIFY_ENUM_CASE(ELF, SHT_DYNAMIC);
239 STRINGIFY_ENUM_CASE(ELF, SHT_NOTE);
240 STRINGIFY_ENUM_CASE(ELF, SHT_NOBITS);
241 STRINGIFY_ENUM_CASE(ELF, SHT_REL);
242 STRINGIFY_ENUM_CASE(ELF, SHT_SHLIB);
243 STRINGIFY_ENUM_CASE(ELF, SHT_DYNSYM);
244 STRINGIFY_ENUM_CASE(ELF, SHT_INIT_ARRAY);
245 STRINGIFY_ENUM_CASE(ELF, SHT_FINI_ARRAY);
246 STRINGIFY_ENUM_CASE(ELF, SHT_PREINIT_ARRAY);
247 STRINGIFY_ENUM_CASE(ELF, SHT_GROUP);
248 STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB_SHNDX);
249 STRINGIFY_ENUM_CASE(ELF, SHT_RELR);
250 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_REL);
251 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELA);
252 STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELR);
253 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ODRTAB);
254 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_LINKER_OPTIONS);
255 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_CALL_GRAPH_PROFILE);
256 STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ADDRSIG);
257 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_ATTRIBUTES);
258 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_HASH);
259 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verdef);
260 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verneed);
261 STRINGIFY_ENUM_CASE(ELF, SHT_GNU_versym);
262 default:
263 return "Unknown";
264 }
265 }
266
267 template <class ELFT>
268 Expected<std::vector<typename ELFT::Rela>>
decode_relrs(Elf_Relr_Range relrs) const269 ELFFile<ELFT>::decode_relrs(Elf_Relr_Range relrs) const {
270 // This function decodes the contents of an SHT_RELR packed relocation
271 // section.
272 //
273 // Proposal for adding SHT_RELR sections to generic-abi is here:
274 // https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
275 //
276 // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks
277 // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
278 //
279 // i.e. start with an address, followed by any number of bitmaps. The address
280 // entry encodes 1 relocation. The subsequent bitmap entries encode up to 63
281 // relocations each, at subsequent offsets following the last address entry.
282 //
283 // The bitmap entries must have 1 in the least significant bit. The assumption
284 // here is that an address cannot have 1 in lsb. Odd addresses are not
285 // supported.
286 //
287 // Excluding the least significant bit in the bitmap, each non-zero bit in
288 // the bitmap represents a relocation to be applied to a corresponding machine
289 // word that follows the base address word. The second least significant bit
290 // represents the machine word immediately following the initial address, and
291 // each bit that follows represents the next word, in linear order. As such,
292 // a single bitmap can encode up to 31 relocations in a 32-bit object, and
293 // 63 relocations in a 64-bit object.
294 //
295 // This encoding has a couple of interesting properties:
296 // 1. Looking at any entry, it is clear whether it's an address or a bitmap:
297 // even means address, odd means bitmap.
298 // 2. Just a simple list of addresses is a valid encoding.
299
300 Elf_Rela Rela;
301 Rela.r_info = 0;
302 Rela.r_addend = 0;
303 Rela.setType(getRelativeRelocationType(), false);
304 std::vector<Elf_Rela> Relocs;
305
306 // Word type: uint32_t for Elf32, and uint64_t for Elf64.
307 typedef typename ELFT::uint Word;
308
309 // Word size in number of bytes.
310 const size_t WordSize = sizeof(Word);
311
312 // Number of bits used for the relocation offsets bitmap.
313 // These many relative relocations can be encoded in a single entry.
314 const size_t NBits = 8*WordSize - 1;
315
316 Word Base = 0;
317 for (const Elf_Relr &R : relrs) {
318 Word Entry = R;
319 if ((Entry&1) == 0) {
320 // Even entry: encodes the offset for next relocation.
321 Rela.r_offset = Entry;
322 Relocs.push_back(Rela);
323 // Set base offset for subsequent bitmap entries.
324 Base = Entry + WordSize;
325 continue;
326 }
327
328 // Odd entry: encodes bitmap for relocations starting at base.
329 Word Offset = Base;
330 while (Entry != 0) {
331 Entry >>= 1;
332 if ((Entry&1) != 0) {
333 Rela.r_offset = Offset;
334 Relocs.push_back(Rela);
335 }
336 Offset += WordSize;
337 }
338
339 // Advance base offset by NBits words.
340 Base += NBits * WordSize;
341 }
342
343 return Relocs;
344 }
345
346 template <class ELFT>
347 Expected<std::vector<typename ELFT::Rela>>
android_relas(const Elf_Shdr * Sec) const348 ELFFile<ELFT>::android_relas(const Elf_Shdr *Sec) const {
349 // This function reads relocations in Android's packed relocation format,
350 // which is based on SLEB128 and delta encoding.
351 Expected<ArrayRef<uint8_t>> ContentsOrErr = getSectionContents(Sec);
352 if (!ContentsOrErr)
353 return ContentsOrErr.takeError();
354 const uint8_t *Cur = ContentsOrErr->begin();
355 const uint8_t *End = ContentsOrErr->end();
356 if (ContentsOrErr->size() < 4 || Cur[0] != 'A' || Cur[1] != 'P' ||
357 Cur[2] != 'S' || Cur[3] != '2')
358 return createError("invalid packed relocation header");
359 Cur += 4;
360
361 const char *ErrStr = nullptr;
362 auto ReadSLEB = [&]() -> int64_t {
363 if (ErrStr)
364 return 0;
365 unsigned Len;
366 int64_t Result = decodeSLEB128(Cur, &Len, End, &ErrStr);
367 Cur += Len;
368 return Result;
369 };
370
371 uint64_t NumRelocs = ReadSLEB();
372 uint64_t Offset = ReadSLEB();
373 uint64_t Addend = 0;
374
375 if (ErrStr)
376 return createError(ErrStr);
377
378 std::vector<Elf_Rela> Relocs;
379 Relocs.reserve(NumRelocs);
380 while (NumRelocs) {
381 uint64_t NumRelocsInGroup = ReadSLEB();
382 if (NumRelocsInGroup > NumRelocs)
383 return createError("relocation group unexpectedly large");
384 NumRelocs -= NumRelocsInGroup;
385
386 uint64_t GroupFlags = ReadSLEB();
387 bool GroupedByInfo = GroupFlags & ELF::RELOCATION_GROUPED_BY_INFO_FLAG;
388 bool GroupedByOffsetDelta = GroupFlags & ELF::RELOCATION_GROUPED_BY_OFFSET_DELTA_FLAG;
389 bool GroupedByAddend = GroupFlags & ELF::RELOCATION_GROUPED_BY_ADDEND_FLAG;
390 bool GroupHasAddend = GroupFlags & ELF::RELOCATION_GROUP_HAS_ADDEND_FLAG;
391
392 uint64_t GroupOffsetDelta;
393 if (GroupedByOffsetDelta)
394 GroupOffsetDelta = ReadSLEB();
395
396 uint64_t GroupRInfo;
397 if (GroupedByInfo)
398 GroupRInfo = ReadSLEB();
399
400 if (GroupedByAddend && GroupHasAddend)
401 Addend += ReadSLEB();
402
403 if (!GroupHasAddend)
404 Addend = 0;
405
406 for (uint64_t I = 0; I != NumRelocsInGroup; ++I) {
407 Elf_Rela R;
408 Offset += GroupedByOffsetDelta ? GroupOffsetDelta : ReadSLEB();
409 R.r_offset = Offset;
410 R.r_info = GroupedByInfo ? GroupRInfo : ReadSLEB();
411 if (GroupHasAddend && !GroupedByAddend)
412 Addend += ReadSLEB();
413 R.r_addend = Addend;
414 Relocs.push_back(R);
415
416 if (ErrStr)
417 return createError(ErrStr);
418 }
419
420 if (ErrStr)
421 return createError(ErrStr);
422 }
423
424 return Relocs;
425 }
426
427 template <class ELFT>
getDynamicTagAsString(unsigned Arch,uint64_t Type) const428 const char *ELFFile<ELFT>::getDynamicTagAsString(unsigned Arch,
429 uint64_t Type) const {
430 #define DYNAMIC_STRINGIFY_ENUM(tag, value) \
431 case value: \
432 return #tag;
433
434 #define DYNAMIC_TAG(n, v)
435 switch (Arch) {
436 case ELF::EM_HEXAGON:
437 switch (Type) {
438 #define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
439 #include "llvm/BinaryFormat/DynamicTags.def"
440 #undef HEXAGON_DYNAMIC_TAG
441 }
442
443 case ELF::EM_MIPS:
444 switch (Type) {
445 #define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
446 #include "llvm/BinaryFormat/DynamicTags.def"
447 #undef MIPS_DYNAMIC_TAG
448 }
449
450 case ELF::EM_PPC64:
451 switch (Type) {
452 #define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
453 #include "llvm/BinaryFormat/DynamicTags.def"
454 #undef PPC64_DYNAMIC_TAG
455 }
456 }
457 #undef DYNAMIC_TAG
458 switch (Type) {
459 // Now handle all dynamic tags except the architecture specific ones
460 #define MIPS_DYNAMIC_TAG(name, value)
461 #define HEXAGON_DYNAMIC_TAG(name, value)
462 #define PPC64_DYNAMIC_TAG(name, value)
463 // Also ignore marker tags such as DT_HIOS (maps to DT_VERNEEDNUM), etc.
464 #define DYNAMIC_TAG_MARKER(name, value)
465 #define DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
466 #include "llvm/BinaryFormat/DynamicTags.def"
467 #undef DYNAMIC_TAG
468 #undef MIPS_DYNAMIC_TAG
469 #undef HEXAGON_DYNAMIC_TAG
470 #undef PPC64_DYNAMIC_TAG
471 #undef DYNAMIC_TAG_MARKER
472 #undef DYNAMIC_STRINGIFY_ENUM
473 default:
474 return "unknown";
475 }
476 }
477
478 template <class ELFT>
getDynamicTagAsString(uint64_t Type) const479 const char *ELFFile<ELFT>::getDynamicTagAsString(uint64_t Type) const {
480 return getDynamicTagAsString(getHeader()->e_machine, Type);
481 }
482
483 template <class ELFT>
dynamicEntries() const484 Expected<typename ELFT::DynRange> ELFFile<ELFT>::dynamicEntries() const {
485 ArrayRef<Elf_Dyn> Dyn;
486 size_t DynSecSize = 0;
487
488 auto ProgramHeadersOrError = program_headers();
489 if (!ProgramHeadersOrError)
490 return ProgramHeadersOrError.takeError();
491
492 for (const Elf_Phdr &Phdr : *ProgramHeadersOrError) {
493 if (Phdr.p_type == ELF::PT_DYNAMIC) {
494 Dyn = makeArrayRef(
495 reinterpret_cast<const Elf_Dyn *>(base() + Phdr.p_offset),
496 Phdr.p_filesz / sizeof(Elf_Dyn));
497 DynSecSize = Phdr.p_filesz;
498 break;
499 }
500 }
501
502 // If we can't find the dynamic section in the program headers, we just fall
503 // back on the sections.
504 if (Dyn.empty()) {
505 auto SectionsOrError = sections();
506 if (!SectionsOrError)
507 return SectionsOrError.takeError();
508
509 for (const Elf_Shdr &Sec : *SectionsOrError) {
510 if (Sec.sh_type == ELF::SHT_DYNAMIC) {
511 Expected<ArrayRef<Elf_Dyn>> DynOrError =
512 getSectionContentsAsArray<Elf_Dyn>(&Sec);
513 if (!DynOrError)
514 return DynOrError.takeError();
515 Dyn = *DynOrError;
516 DynSecSize = Sec.sh_size;
517 break;
518 }
519 }
520
521 if (!Dyn.data())
522 return ArrayRef<Elf_Dyn>();
523 }
524
525 if (Dyn.empty())
526 return createError("invalid empty dynamic section");
527
528 if (DynSecSize % sizeof(Elf_Dyn) != 0)
529 return createError("malformed dynamic section");
530
531 if (Dyn.back().d_tag != ELF::DT_NULL)
532 return createError("dynamic sections must be DT_NULL terminated");
533
534 return Dyn;
535 }
536
537 template <class ELFT>
toMappedAddr(uint64_t VAddr) const538 Expected<const uint8_t *> ELFFile<ELFT>::toMappedAddr(uint64_t VAddr) const {
539 auto ProgramHeadersOrError = program_headers();
540 if (!ProgramHeadersOrError)
541 return ProgramHeadersOrError.takeError();
542
543 llvm::SmallVector<Elf_Phdr *, 4> LoadSegments;
544
545 for (const Elf_Phdr &Phdr : *ProgramHeadersOrError)
546 if (Phdr.p_type == ELF::PT_LOAD)
547 LoadSegments.push_back(const_cast<Elf_Phdr *>(&Phdr));
548
549 const Elf_Phdr *const *I =
550 std::upper_bound(LoadSegments.begin(), LoadSegments.end(), VAddr,
551 [](uint64_t VAddr, const Elf_Phdr_Impl<ELFT> *Phdr) {
552 return VAddr < Phdr->p_vaddr;
553 });
554
555 if (I == LoadSegments.begin())
556 return createError("Virtual address is not in any segment");
557 --I;
558 const Elf_Phdr &Phdr = **I;
559 uint64_t Delta = VAddr - Phdr.p_vaddr;
560 if (Delta >= Phdr.p_filesz)
561 return createError("Virtual address is not in any segment");
562 return base() + Phdr.p_offset + Delta;
563 }
564
565 template class llvm::object::ELFFile<ELF32LE>;
566 template class llvm::object::ELFFile<ELF32BE>;
567 template class llvm::object::ELFFile<ELF64LE>;
568 template class llvm::object::ELFFile<ELF64BE>;
569