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/ADT/StringExtras.h"
11 #include "llvm/BinaryFormat/ELF.h"
12 #include "llvm/Support/DataExtractor.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 
23 StringRef llvm::object::getELFRelocationTypeName(uint32_t Machine,
24                                                  uint32_t Type) {
25   switch (Machine) {
26   case ELF::EM_68K:
27     switch (Type) {
28 #include "llvm/BinaryFormat/ELFRelocs/M68k.def"
29     default:
30       break;
31     }
32     break;
33   case ELF::EM_X86_64:
34     switch (Type) {
35 #include "llvm/BinaryFormat/ELFRelocs/x86_64.def"
36     default:
37       break;
38     }
39     break;
40   case ELF::EM_386:
41   case ELF::EM_IAMCU:
42     switch (Type) {
43 #include "llvm/BinaryFormat/ELFRelocs/i386.def"
44     default:
45       break;
46     }
47     break;
48   case ELF::EM_MIPS:
49     switch (Type) {
50 #include "llvm/BinaryFormat/ELFRelocs/Mips.def"
51     default:
52       break;
53     }
54     break;
55   case ELF::EM_AARCH64:
56     switch (Type) {
57 #include "llvm/BinaryFormat/ELFRelocs/AArch64.def"
58     default:
59       break;
60     }
61     break;
62   case ELF::EM_ARM:
63     switch (Type) {
64 #include "llvm/BinaryFormat/ELFRelocs/ARM.def"
65     default:
66       break;
67     }
68     break;
69   case ELF::EM_ARC_COMPACT:
70   case ELF::EM_ARC_COMPACT2:
71     switch (Type) {
72 #include "llvm/BinaryFormat/ELFRelocs/ARC.def"
73     default:
74       break;
75     }
76     break;
77   case ELF::EM_AVR:
78     switch (Type) {
79 #include "llvm/BinaryFormat/ELFRelocs/AVR.def"
80     default:
81       break;
82     }
83     break;
84   case ELF::EM_HEXAGON:
85     switch (Type) {
86 #include "llvm/BinaryFormat/ELFRelocs/Hexagon.def"
87     default:
88       break;
89     }
90     break;
91   case ELF::EM_LANAI:
92     switch (Type) {
93 #include "llvm/BinaryFormat/ELFRelocs/Lanai.def"
94     default:
95       break;
96     }
97     break;
98   case ELF::EM_PPC:
99     switch (Type) {
100 #include "llvm/BinaryFormat/ELFRelocs/PowerPC.def"
101     default:
102       break;
103     }
104     break;
105   case ELF::EM_PPC64:
106     switch (Type) {
107 #include "llvm/BinaryFormat/ELFRelocs/PowerPC64.def"
108     default:
109       break;
110     }
111     break;
112   case ELF::EM_RISCV:
113     switch (Type) {
114 #include "llvm/BinaryFormat/ELFRelocs/RISCV.def"
115     default:
116       break;
117     }
118     break;
119   case ELF::EM_S390:
120     switch (Type) {
121 #include "llvm/BinaryFormat/ELFRelocs/SystemZ.def"
122     default:
123       break;
124     }
125     break;
126   case ELF::EM_SPARC:
127   case ELF::EM_SPARC32PLUS:
128   case ELF::EM_SPARCV9:
129     switch (Type) {
130 #include "llvm/BinaryFormat/ELFRelocs/Sparc.def"
131     default:
132       break;
133     }
134     break;
135   case ELF::EM_AMDGPU:
136     switch (Type) {
137 #include "llvm/BinaryFormat/ELFRelocs/AMDGPU.def"
138     default:
139       break;
140     }
141     break;
142   case ELF::EM_BPF:
143     switch (Type) {
144 #include "llvm/BinaryFormat/ELFRelocs/BPF.def"
145     default:
146       break;
147     }
148     break;
149   case ELF::EM_MSP430:
150     switch (Type) {
151 #include "llvm/BinaryFormat/ELFRelocs/MSP430.def"
152     default:
153       break;
154     }
155     break;
156   case ELF::EM_VE:
157     switch (Type) {
158 #include "llvm/BinaryFormat/ELFRelocs/VE.def"
159     default:
160       break;
161     }
162     break;
163   case ELF::EM_CSKY:
164     switch (Type) {
165 #include "llvm/BinaryFormat/ELFRelocs/CSKY.def"
166     default:
167       break;
168     }
169     break;
170   case ELF::EM_LOONGARCH:
171     switch (Type) {
172 #include "llvm/BinaryFormat/ELFRelocs/LoongArch.def"
173     default:
174       break;
175     }
176     break;
177   case ELF::EM_XTENSA:
178     switch (Type) {
179 #include "llvm/BinaryFormat/ELFRelocs/Xtensa.def"
180     default:
181       break;
182     }
183     break;
184   default:
185     break;
186   }
187   return "Unknown";
188 }
189 
190 #undef ELF_RELOC
191 
192 uint32_t llvm::object::getELFRelativeRelocationType(uint32_t Machine) {
193   switch (Machine) {
194   case ELF::EM_X86_64:
195     return ELF::R_X86_64_RELATIVE;
196   case ELF::EM_386:
197   case ELF::EM_IAMCU:
198     return ELF::R_386_RELATIVE;
199   case ELF::EM_MIPS:
200     break;
201   case ELF::EM_AARCH64:
202     return ELF::R_AARCH64_RELATIVE;
203   case ELF::EM_ARM:
204     return ELF::R_ARM_RELATIVE;
205   case ELF::EM_ARC_COMPACT:
206   case ELF::EM_ARC_COMPACT2:
207     return ELF::R_ARC_RELATIVE;
208   case ELF::EM_AVR:
209     break;
210   case ELF::EM_HEXAGON:
211     return ELF::R_HEX_RELATIVE;
212   case ELF::EM_LANAI:
213     break;
214   case ELF::EM_PPC:
215     break;
216   case ELF::EM_PPC64:
217     return ELF::R_PPC64_RELATIVE;
218   case ELF::EM_RISCV:
219     return ELF::R_RISCV_RELATIVE;
220   case ELF::EM_S390:
221     return ELF::R_390_RELATIVE;
222   case ELF::EM_SPARC:
223   case ELF::EM_SPARC32PLUS:
224   case ELF::EM_SPARCV9:
225     return ELF::R_SPARC_RELATIVE;
226   case ELF::EM_CSKY:
227     return ELF::R_CKCORE_RELATIVE;
228   case ELF::EM_VE:
229     return ELF::R_VE_RELATIVE;
230   case ELF::EM_AMDGPU:
231     break;
232   case ELF::EM_BPF:
233     break;
234   case ELF::EM_LOONGARCH:
235     return ELF::R_LARCH_RELATIVE;
236   default:
237     break;
238   }
239   return 0;
240 }
241 
242 StringRef llvm::object::getELFSectionTypeName(uint32_t Machine, unsigned Type) {
243   switch (Machine) {
244   case ELF::EM_ARM:
245     switch (Type) {
246       STRINGIFY_ENUM_CASE(ELF, SHT_ARM_EXIDX);
247       STRINGIFY_ENUM_CASE(ELF, SHT_ARM_PREEMPTMAP);
248       STRINGIFY_ENUM_CASE(ELF, SHT_ARM_ATTRIBUTES);
249       STRINGIFY_ENUM_CASE(ELF, SHT_ARM_DEBUGOVERLAY);
250       STRINGIFY_ENUM_CASE(ELF, SHT_ARM_OVERLAYSECTION);
251     }
252     break;
253   case ELF::EM_HEXAGON:
254     switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_HEX_ORDERED); }
255     break;
256   case ELF::EM_X86_64:
257     switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_X86_64_UNWIND); }
258     break;
259   case ELF::EM_MIPS:
260   case ELF::EM_MIPS_RS3_LE:
261     switch (Type) {
262       STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_REGINFO);
263       STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_OPTIONS);
264       STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_DWARF);
265       STRINGIFY_ENUM_CASE(ELF, SHT_MIPS_ABIFLAGS);
266     }
267     break;
268   case ELF::EM_MSP430:
269     switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_MSP430_ATTRIBUTES); }
270     break;
271   case ELF::EM_RISCV:
272     switch (Type) { STRINGIFY_ENUM_CASE(ELF, SHT_RISCV_ATTRIBUTES); }
273     break;
274   case ELF::EM_AARCH64:
275     switch (Type) {
276       STRINGIFY_ENUM_CASE(ELF, SHT_AARCH64_MEMTAG_GLOBALS_DYNAMIC);
277       STRINGIFY_ENUM_CASE(ELF, SHT_AARCH64_MEMTAG_GLOBALS_STATIC);
278     }
279   default:
280     break;
281   }
282 
283   switch (Type) {
284     STRINGIFY_ENUM_CASE(ELF, SHT_NULL);
285     STRINGIFY_ENUM_CASE(ELF, SHT_PROGBITS);
286     STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB);
287     STRINGIFY_ENUM_CASE(ELF, SHT_STRTAB);
288     STRINGIFY_ENUM_CASE(ELF, SHT_RELA);
289     STRINGIFY_ENUM_CASE(ELF, SHT_HASH);
290     STRINGIFY_ENUM_CASE(ELF, SHT_DYNAMIC);
291     STRINGIFY_ENUM_CASE(ELF, SHT_NOTE);
292     STRINGIFY_ENUM_CASE(ELF, SHT_NOBITS);
293     STRINGIFY_ENUM_CASE(ELF, SHT_REL);
294     STRINGIFY_ENUM_CASE(ELF, SHT_SHLIB);
295     STRINGIFY_ENUM_CASE(ELF, SHT_DYNSYM);
296     STRINGIFY_ENUM_CASE(ELF, SHT_INIT_ARRAY);
297     STRINGIFY_ENUM_CASE(ELF, SHT_FINI_ARRAY);
298     STRINGIFY_ENUM_CASE(ELF, SHT_PREINIT_ARRAY);
299     STRINGIFY_ENUM_CASE(ELF, SHT_GROUP);
300     STRINGIFY_ENUM_CASE(ELF, SHT_SYMTAB_SHNDX);
301     STRINGIFY_ENUM_CASE(ELF, SHT_RELR);
302     STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_REL);
303     STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELA);
304     STRINGIFY_ENUM_CASE(ELF, SHT_ANDROID_RELR);
305     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ODRTAB);
306     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_LINKER_OPTIONS);
307     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_CALL_GRAPH_PROFILE);
308     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_ADDRSIG);
309     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_DEPENDENT_LIBRARIES);
310     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_SYMPART);
311     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_PART_EHDR);
312     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_PART_PHDR);
313     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_BB_ADDR_MAP_V0);
314     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_BB_ADDR_MAP);
315     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_OFFLOADING);
316     STRINGIFY_ENUM_CASE(ELF, SHT_LLVM_LTO);
317     STRINGIFY_ENUM_CASE(ELF, SHT_GNU_ATTRIBUTES);
318     STRINGIFY_ENUM_CASE(ELF, SHT_GNU_HASH);
319     STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verdef);
320     STRINGIFY_ENUM_CASE(ELF, SHT_GNU_verneed);
321     STRINGIFY_ENUM_CASE(ELF, SHT_GNU_versym);
322   default:
323     return "Unknown";
324   }
325 }
326 
327 template <class ELFT>
328 std::vector<typename ELFT::Rel>
329 ELFFile<ELFT>::decode_relrs(Elf_Relr_Range relrs) const {
330   // This function decodes the contents of an SHT_RELR packed relocation
331   // section.
332   //
333   // Proposal for adding SHT_RELR sections to generic-abi is here:
334   //   https://groups.google.com/forum/#!topic/generic-abi/bX460iggiKg
335   //
336   // The encoded sequence of Elf64_Relr entries in a SHT_RELR section looks
337   // like [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
338   //
339   // i.e. start with an address, followed by any number of bitmaps. The address
340   // entry encodes 1 relocation. The subsequent bitmap entries encode up to 63
341   // relocations each, at subsequent offsets following the last address entry.
342   //
343   // The bitmap entries must have 1 in the least significant bit. The assumption
344   // here is that an address cannot have 1 in lsb. Odd addresses are not
345   // supported.
346   //
347   // Excluding the least significant bit in the bitmap, each non-zero bit in
348   // the bitmap represents a relocation to be applied to a corresponding machine
349   // word that follows the base address word. The second least significant bit
350   // represents the machine word immediately following the initial address, and
351   // each bit that follows represents the next word, in linear order. As such,
352   // a single bitmap can encode up to 31 relocations in a 32-bit object, and
353   // 63 relocations in a 64-bit object.
354   //
355   // This encoding has a couple of interesting properties:
356   // 1. Looking at any entry, it is clear whether it's an address or a bitmap:
357   //    even means address, odd means bitmap.
358   // 2. Just a simple list of addresses is a valid encoding.
359 
360   Elf_Rel Rel;
361   Rel.r_info = 0;
362   Rel.setType(getRelativeRelocationType(), false);
363   std::vector<Elf_Rel> Relocs;
364 
365   // Word type: uint32_t for Elf32, and uint64_t for Elf64.
366   using Addr = typename ELFT::uint;
367 
368   Addr Base = 0;
369   for (Elf_Relr R : relrs) {
370     typename ELFT::uint Entry = R;
371     if ((Entry & 1) == 0) {
372       // Even entry: encodes the offset for next relocation.
373       Rel.r_offset = Entry;
374       Relocs.push_back(Rel);
375       // Set base offset for subsequent bitmap entries.
376       Base = Entry + sizeof(Addr);
377     } else {
378       // Odd entry: encodes bitmap for relocations starting at base.
379       for (Addr Offset = Base; (Entry >>= 1) != 0; Offset += sizeof(Addr))
380         if ((Entry & 1) != 0) {
381           Rel.r_offset = Offset;
382           Relocs.push_back(Rel);
383         }
384       Base += (CHAR_BIT * sizeof(Entry) - 1) * sizeof(Addr);
385     }
386   }
387 
388   return Relocs;
389 }
390 
391 template <class ELFT>
392 Expected<std::vector<typename ELFT::Rela>>
393 ELFFile<ELFT>::android_relas(const Elf_Shdr &Sec) const {
394   // This function reads relocations in Android's packed relocation format,
395   // which is based on SLEB128 and delta encoding.
396   Expected<ArrayRef<uint8_t>> ContentsOrErr = getSectionContents(Sec);
397   if (!ContentsOrErr)
398     return ContentsOrErr.takeError();
399   ArrayRef<uint8_t> Content = *ContentsOrErr;
400   if (Content.size() < 4 || Content[0] != 'A' || Content[1] != 'P' ||
401       Content[2] != 'S' || Content[3] != '2')
402     return createError("invalid packed relocation header");
403   DataExtractor Data(Content, isLE(), ELFT::Is64Bits ? 8 : 4);
404   DataExtractor::Cursor Cur(/*Offset=*/4);
405 
406   uint64_t NumRelocs = Data.getSLEB128(Cur);
407   uint64_t Offset = Data.getSLEB128(Cur);
408   uint64_t Addend = 0;
409 
410   if (!Cur)
411     return std::move(Cur.takeError());
412 
413   std::vector<Elf_Rela> Relocs;
414   Relocs.reserve(NumRelocs);
415   while (NumRelocs) {
416     uint64_t NumRelocsInGroup = Data.getSLEB128(Cur);
417     if (!Cur)
418       return std::move(Cur.takeError());
419     if (NumRelocsInGroup > NumRelocs)
420       return createError("relocation group unexpectedly large");
421     NumRelocs -= NumRelocsInGroup;
422 
423     uint64_t GroupFlags = Data.getSLEB128(Cur);
424     bool GroupedByInfo = GroupFlags & ELF::RELOCATION_GROUPED_BY_INFO_FLAG;
425     bool GroupedByOffsetDelta = GroupFlags & ELF::RELOCATION_GROUPED_BY_OFFSET_DELTA_FLAG;
426     bool GroupedByAddend = GroupFlags & ELF::RELOCATION_GROUPED_BY_ADDEND_FLAG;
427     bool GroupHasAddend = GroupFlags & ELF::RELOCATION_GROUP_HAS_ADDEND_FLAG;
428 
429     uint64_t GroupOffsetDelta;
430     if (GroupedByOffsetDelta)
431       GroupOffsetDelta = Data.getSLEB128(Cur);
432 
433     uint64_t GroupRInfo;
434     if (GroupedByInfo)
435       GroupRInfo = Data.getSLEB128(Cur);
436 
437     if (GroupedByAddend && GroupHasAddend)
438       Addend += Data.getSLEB128(Cur);
439 
440     if (!GroupHasAddend)
441       Addend = 0;
442 
443     for (uint64_t I = 0; Cur && I != NumRelocsInGroup; ++I) {
444       Elf_Rela R;
445       Offset += GroupedByOffsetDelta ? GroupOffsetDelta : Data.getSLEB128(Cur);
446       R.r_offset = Offset;
447       R.r_info = GroupedByInfo ? GroupRInfo : Data.getSLEB128(Cur);
448       if (GroupHasAddend && !GroupedByAddend)
449         Addend += Data.getSLEB128(Cur);
450       R.r_addend = Addend;
451       Relocs.push_back(R);
452     }
453     if (!Cur)
454       return std::move(Cur.takeError());
455   }
456 
457   return Relocs;
458 }
459 
460 template <class ELFT>
461 std::string ELFFile<ELFT>::getDynamicTagAsString(unsigned Arch,
462                                                  uint64_t Type) const {
463 #define DYNAMIC_STRINGIFY_ENUM(tag, value)                                     \
464   case value:                                                                  \
465     return #tag;
466 
467 #define DYNAMIC_TAG(n, v)
468   switch (Arch) {
469   case ELF::EM_AARCH64:
470     switch (Type) {
471 #define AARCH64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
472 #include "llvm/BinaryFormat/DynamicTags.def"
473 #undef AARCH64_DYNAMIC_TAG
474     }
475     break;
476 
477   case ELF::EM_HEXAGON:
478     switch (Type) {
479 #define HEXAGON_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
480 #include "llvm/BinaryFormat/DynamicTags.def"
481 #undef HEXAGON_DYNAMIC_TAG
482     }
483     break;
484 
485   case ELF::EM_MIPS:
486     switch (Type) {
487 #define MIPS_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
488 #include "llvm/BinaryFormat/DynamicTags.def"
489 #undef MIPS_DYNAMIC_TAG
490     }
491     break;
492 
493   case ELF::EM_PPC:
494     switch (Type) {
495 #define PPC_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
496 #include "llvm/BinaryFormat/DynamicTags.def"
497 #undef PPC_DYNAMIC_TAG
498     }
499     break;
500 
501   case ELF::EM_PPC64:
502     switch (Type) {
503 #define PPC64_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
504 #include "llvm/BinaryFormat/DynamicTags.def"
505 #undef PPC64_DYNAMIC_TAG
506     }
507     break;
508 
509   case ELF::EM_RISCV:
510     switch (Type) {
511 #define RISCV_DYNAMIC_TAG(name, value) DYNAMIC_STRINGIFY_ENUM(name, value)
512 #include "llvm/BinaryFormat/DynamicTags.def"
513 #undef RISCV_DYNAMIC_TAG
514     }
515     break;
516   }
517 #undef DYNAMIC_TAG
518   switch (Type) {
519 // Now handle all dynamic tags except the architecture specific ones
520 #define AARCH64_DYNAMIC_TAG(name, value)
521 #define MIPS_DYNAMIC_TAG(name, value)
522 #define HEXAGON_DYNAMIC_TAG(name, value)
523 #define PPC_DYNAMIC_TAG(name, value)
524 #define PPC64_DYNAMIC_TAG(name, value)
525 #define RISCV_DYNAMIC_TAG(name, value)
526 // Also ignore marker tags such as DT_HIOS (maps to DT_VERNEEDNUM), etc.
527 #define DYNAMIC_TAG_MARKER(name, value)
528 #define DYNAMIC_TAG(name, value) case value: return #name;
529 #include "llvm/BinaryFormat/DynamicTags.def"
530 #undef DYNAMIC_TAG
531 #undef AARCH64_DYNAMIC_TAG
532 #undef MIPS_DYNAMIC_TAG
533 #undef HEXAGON_DYNAMIC_TAG
534 #undef PPC_DYNAMIC_TAG
535 #undef PPC64_DYNAMIC_TAG
536 #undef RISCV_DYNAMIC_TAG
537 #undef DYNAMIC_TAG_MARKER
538 #undef DYNAMIC_STRINGIFY_ENUM
539   default:
540     return "<unknown:>0x" + utohexstr(Type, true);
541   }
542 }
543 
544 template <class ELFT>
545 std::string ELFFile<ELFT>::getDynamicTagAsString(uint64_t Type) const {
546   return getDynamicTagAsString(getHeader().e_machine, Type);
547 }
548 
549 template <class ELFT>
550 Expected<typename ELFT::DynRange> ELFFile<ELFT>::dynamicEntries() const {
551   ArrayRef<Elf_Dyn> Dyn;
552 
553   auto ProgramHeadersOrError = program_headers();
554   if (!ProgramHeadersOrError)
555     return ProgramHeadersOrError.takeError();
556 
557   for (const Elf_Phdr &Phdr : *ProgramHeadersOrError) {
558     if (Phdr.p_type == ELF::PT_DYNAMIC) {
559       Dyn = ArrayRef(reinterpret_cast<const Elf_Dyn *>(base() + Phdr.p_offset),
560                      Phdr.p_filesz / sizeof(Elf_Dyn));
561       break;
562     }
563   }
564 
565   // If we can't find the dynamic section in the program headers, we just fall
566   // back on the sections.
567   if (Dyn.empty()) {
568     auto SectionsOrError = sections();
569     if (!SectionsOrError)
570       return SectionsOrError.takeError();
571 
572     for (const Elf_Shdr &Sec : *SectionsOrError) {
573       if (Sec.sh_type == ELF::SHT_DYNAMIC) {
574         Expected<ArrayRef<Elf_Dyn>> DynOrError =
575             getSectionContentsAsArray<Elf_Dyn>(Sec);
576         if (!DynOrError)
577           return DynOrError.takeError();
578         Dyn = *DynOrError;
579         break;
580       }
581     }
582 
583     if (!Dyn.data())
584       return ArrayRef<Elf_Dyn>();
585   }
586 
587   if (Dyn.empty())
588     return createError("invalid empty dynamic section");
589 
590   if (Dyn.back().d_tag != ELF::DT_NULL)
591     return createError("dynamic sections must be DT_NULL terminated");
592 
593   return Dyn;
594 }
595 
596 template <class ELFT>
597 Expected<const uint8_t *>
598 ELFFile<ELFT>::toMappedAddr(uint64_t VAddr, WarningHandler WarnHandler) const {
599   auto ProgramHeadersOrError = program_headers();
600   if (!ProgramHeadersOrError)
601     return ProgramHeadersOrError.takeError();
602 
603   llvm::SmallVector<Elf_Phdr *, 4> LoadSegments;
604 
605   for (const Elf_Phdr &Phdr : *ProgramHeadersOrError)
606     if (Phdr.p_type == ELF::PT_LOAD)
607       LoadSegments.push_back(const_cast<Elf_Phdr *>(&Phdr));
608 
609   auto SortPred = [](const Elf_Phdr_Impl<ELFT> *A,
610                      const Elf_Phdr_Impl<ELFT> *B) {
611     return A->p_vaddr < B->p_vaddr;
612   };
613   if (!llvm::is_sorted(LoadSegments, SortPred)) {
614     if (Error E =
615             WarnHandler("loadable segments are unsorted by virtual address"))
616       return std::move(E);
617     llvm::stable_sort(LoadSegments, SortPred);
618   }
619 
620   const Elf_Phdr *const *I = llvm::upper_bound(
621       LoadSegments, VAddr, [](uint64_t VAddr, const Elf_Phdr_Impl<ELFT> *Phdr) {
622         return VAddr < Phdr->p_vaddr;
623       });
624 
625   if (I == LoadSegments.begin())
626     return createError("virtual address is not in any segment: 0x" +
627                        Twine::utohexstr(VAddr));
628   --I;
629   const Elf_Phdr &Phdr = **I;
630   uint64_t Delta = VAddr - Phdr.p_vaddr;
631   if (Delta >= Phdr.p_filesz)
632     return createError("virtual address is not in any segment: 0x" +
633                        Twine::utohexstr(VAddr));
634 
635   uint64_t Offset = Phdr.p_offset + Delta;
636   if (Offset >= getBufSize())
637     return createError("can't map virtual address 0x" +
638                        Twine::utohexstr(VAddr) + " to the segment with index " +
639                        Twine(&Phdr - (*ProgramHeadersOrError).data() + 1) +
640                        ": the segment ends at 0x" +
641                        Twine::utohexstr(Phdr.p_offset + Phdr.p_filesz) +
642                        ", which is greater than the file size (0x" +
643                        Twine::utohexstr(getBufSize()) + ")");
644 
645   return base() + Offset;
646 }
647 
648 template <class ELFT>
649 Expected<std::vector<BBAddrMap>>
650 ELFFile<ELFT>::decodeBBAddrMap(const Elf_Shdr &Sec,
651                                const Elf_Shdr *RelaSec) const {
652   bool IsRelocatable = getHeader().e_type == ELF::ET_REL;
653 
654   // This DenseMap maps the offset of each function (the location of the
655   // reference to the function in the SHT_LLVM_BB_ADDR_MAP section) to the
656   // addend (the location of the function in the text section).
657   llvm::DenseMap<uint64_t, uint64_t> FunctionOffsetTranslations;
658   if (IsRelocatable && RelaSec) {
659     assert(RelaSec &&
660            "Can't read a SHT_LLVM_BB_ADDR_MAP section in a relocatable "
661            "object file without providing a relocation section.");
662     Expected<Elf_Rela_Range> Relas = this->relas(*RelaSec);
663     if (!Relas)
664       return createError("unable to read relocations for section " +
665                          describe(*this, Sec) + ": " +
666                          toString(Relas.takeError()));
667     for (Elf_Rela Rela : *Relas)
668       FunctionOffsetTranslations[Rela.r_offset] = Rela.r_addend;
669   }
670   Expected<ArrayRef<uint8_t>> ContentsOrErr = getSectionContents(Sec);
671   if (!ContentsOrErr)
672     return ContentsOrErr.takeError();
673   ArrayRef<uint8_t> Content = *ContentsOrErr;
674   DataExtractor Data(Content, isLE(), ELFT::Is64Bits ? 8 : 4);
675   std::vector<BBAddrMap> FunctionEntries;
676 
677   DataExtractor::Cursor Cur(0);
678   Error ULEBSizeErr = Error::success();
679   Error MetadataDecodeErr = Error::success();
680   // Helper to extract and decode the next ULEB128 value as uint32_t.
681   // Returns zero and sets ULEBSizeErr if the ULEB128 value exceeds the uint32_t
682   // limit.
683   // Also returns zero if ULEBSizeErr is already in an error state.
684   auto ReadULEB128AsUInt32 = [&Data, &Cur, &ULEBSizeErr]() -> uint32_t {
685     // Bail out and do not extract data if ULEBSizeErr is already set.
686     if (ULEBSizeErr)
687       return 0;
688     uint64_t Offset = Cur.tell();
689     uint64_t Value = Data.getULEB128(Cur);
690     if (Value > UINT32_MAX) {
691       ULEBSizeErr = createError(
692           "ULEB128 value at offset 0x" + Twine::utohexstr(Offset) +
693           " exceeds UINT32_MAX (0x" + Twine::utohexstr(Value) + ")");
694       return 0;
695     }
696     return static_cast<uint32_t>(Value);
697   };
698 
699   uint8_t Version = 0;
700   while (!ULEBSizeErr && !MetadataDecodeErr && Cur &&
701          Cur.tell() < Content.size()) {
702     if (Sec.sh_type == ELF::SHT_LLVM_BB_ADDR_MAP) {
703       Version = Data.getU8(Cur);
704       if (!Cur)
705         break;
706       if (Version > 2)
707         return createError("unsupported SHT_LLVM_BB_ADDR_MAP version: " +
708                            Twine(static_cast<int>(Version)));
709       Data.getU8(Cur); // Feature byte
710     }
711     uint64_t SectionOffset = Cur.tell();
712     uintX_t Address = static_cast<uintX_t>(Data.getAddress(Cur));
713     if (!Cur)
714       return Cur.takeError();
715     if (IsRelocatable) {
716       assert(Address == 0);
717       auto FOTIterator = FunctionOffsetTranslations.find(SectionOffset);
718       if (FOTIterator == FunctionOffsetTranslations.end()) {
719         return createError("failed to get relocation data for offset: " +
720                            Twine::utohexstr(SectionOffset) + " in section " +
721                            describe(*this, Sec));
722       }
723       Address = FOTIterator->second;
724     }
725     uint32_t NumBlocks = ReadULEB128AsUInt32();
726     std::vector<BBAddrMap::BBEntry> BBEntries;
727     uint32_t PrevBBEndOffset = 0;
728     for (uint32_t BlockIndex = 0;
729          !MetadataDecodeErr && !ULEBSizeErr && Cur && (BlockIndex < NumBlocks);
730          ++BlockIndex) {
731       uint32_t ID = Version >= 2 ? ReadULEB128AsUInt32() : BlockIndex;
732       uint32_t Offset = ReadULEB128AsUInt32();
733       uint32_t Size = ReadULEB128AsUInt32();
734       uint32_t MD = ReadULEB128AsUInt32();
735       if (Version >= 1) {
736         // Offset is calculated relative to the end of the previous BB.
737         Offset += PrevBBEndOffset;
738         PrevBBEndOffset = Offset + Size;
739       }
740       Expected<BBAddrMap::BBEntry::Metadata> MetadataOrErr =
741           BBAddrMap::BBEntry::Metadata::decode(MD);
742       if (!MetadataOrErr) {
743         MetadataDecodeErr = MetadataOrErr.takeError();
744         break;
745       }
746       BBEntries.push_back({ID, Offset, Size, *MetadataOrErr});
747     }
748     FunctionEntries.push_back({Address, std::move(BBEntries)});
749   }
750   // Either Cur is in the error state, or we have an error in ULEBSizeErr or
751   // MetadataDecodeErr (but not both), but we join all errors here to be safe.
752   if (!Cur || ULEBSizeErr || MetadataDecodeErr)
753     return joinErrors(joinErrors(Cur.takeError(), std::move(ULEBSizeErr)),
754                       std::move(MetadataDecodeErr));
755   return FunctionEntries;
756 }
757 
758 template <class ELFT>
759 Expected<
760     MapVector<const typename ELFT::Shdr *, const typename ELFT::Shdr *>>
761 ELFFile<ELFT>::getSectionAndRelocations(
762     std::function<Expected<bool>(const Elf_Shdr &)> IsMatch) const {
763   MapVector<const Elf_Shdr *, const Elf_Shdr *> SecToRelocMap;
764   Error Errors = Error::success();
765   for (const Elf_Shdr &Sec : cantFail(this->sections())) {
766     Expected<bool> DoesSectionMatch = IsMatch(Sec);
767     if (!DoesSectionMatch) {
768       Errors = joinErrors(std::move(Errors), DoesSectionMatch.takeError());
769       continue;
770     }
771     if (*DoesSectionMatch) {
772       if (SecToRelocMap.insert(std::make_pair(&Sec, (const Elf_Shdr *)nullptr))
773               .second)
774         continue;
775     }
776 
777     if (Sec.sh_type != ELF::SHT_RELA && Sec.sh_type != ELF::SHT_REL)
778       continue;
779 
780     Expected<const Elf_Shdr *> RelSecOrErr = this->getSection(Sec.sh_info);
781     if (!RelSecOrErr) {
782       Errors = joinErrors(std::move(Errors),
783                           createError(describe(*this, Sec) +
784                                       ": failed to get a relocated section: " +
785                                       toString(RelSecOrErr.takeError())));
786       continue;
787     }
788     const Elf_Shdr *ContentsSec = *RelSecOrErr;
789     Expected<bool> DoesRelTargetMatch = IsMatch(*ContentsSec);
790     if (!DoesRelTargetMatch) {
791       Errors = joinErrors(std::move(Errors), DoesRelTargetMatch.takeError());
792       continue;
793     }
794     if (*DoesRelTargetMatch)
795       SecToRelocMap[ContentsSec] = &Sec;
796   }
797   if(Errors)
798     return std::move(Errors);
799   return SecToRelocMap;
800 }
801 
802 template class llvm::object::ELFFile<ELF32LE>;
803 template class llvm::object::ELFFile<ELF32BE>;
804 template class llvm::object::ELFFile<ELF64LE>;
805 template class llvm::object::ELFFile<ELF64BE>;
806