1 //===- ObjectFile.h - File format independent object file -------*- C++ -*-===//
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 // This file declares a file format independent ObjectFile class.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #ifndef LLVM_OBJECT_OBJECTFILE_H
14 #define LLVM_OBJECT_OBJECTFILE_H
15 
16 #include "llvm/ADT/DenseMapInfo.h"
17 #include "llvm/ADT/StringRef.h"
18 #include "llvm/ADT/Triple.h"
19 #include "llvm/ADT/iterator_range.h"
20 #include "llvm/BinaryFormat/Magic.h"
21 #include "llvm/MC/SubtargetFeature.h"
22 #include "llvm/Object/Binary.h"
23 #include "llvm/Object/Error.h"
24 #include "llvm/Object/SymbolicFile.h"
25 #include "llvm/Support/Casting.h"
26 #include "llvm/Support/Error.h"
27 #include "llvm/Support/FileSystem.h"
28 #include "llvm/Support/MemoryBuffer.h"
29 #include <cassert>
30 #include <cstdint>
31 #include <memory>
32 #include <system_error>
33 
34 namespace llvm {
35 
36 class ARMAttributeParser;
37 
38 namespace object {
39 
40 class COFFObjectFile;
41 class MachOObjectFile;
42 class ObjectFile;
43 class SectionRef;
44 class SymbolRef;
45 class symbol_iterator;
46 class WasmObjectFile;
47 
48 using section_iterator = content_iterator<SectionRef>;
49 
50 /// This is a value type class that represents a single relocation in the list
51 /// of relocations in the object file.
52 class RelocationRef {
53   DataRefImpl RelocationPimpl;
54   const ObjectFile *OwningObject = nullptr;
55 
56 public:
57   RelocationRef() = default;
58   RelocationRef(DataRefImpl RelocationP, const ObjectFile *Owner);
59 
60   bool operator==(const RelocationRef &Other) const;
61 
62   void moveNext();
63 
64   uint64_t getOffset() const;
65   symbol_iterator getSymbol() const;
66   uint64_t getType() const;
67 
68   /// Get a string that represents the type of this relocation.
69   ///
70   /// This is for display purposes only.
71   void getTypeName(SmallVectorImpl<char> &Result) const;
72 
73   DataRefImpl getRawDataRefImpl() const;
74   const ObjectFile *getObject() const;
75 };
76 
77 using relocation_iterator = content_iterator<RelocationRef>;
78 
79 /// This is a value type class that represents a single section in the list of
80 /// sections in the object file.
81 class SectionRef {
82   friend class SymbolRef;
83 
84   DataRefImpl SectionPimpl;
85   const ObjectFile *OwningObject = nullptr;
86 
87 public:
88   SectionRef() = default;
89   SectionRef(DataRefImpl SectionP, const ObjectFile *Owner);
90 
91   bool operator==(const SectionRef &Other) const;
92   bool operator!=(const SectionRef &Other) const;
93   bool operator<(const SectionRef &Other) const;
94 
95   void moveNext();
96 
97   Expected<StringRef> getName() const;
98   uint64_t getAddress() const;
99   uint64_t getIndex() const;
100   uint64_t getSize() const;
101   Expected<StringRef> getContents() const;
102 
103   /// Get the alignment of this section as the actual value (not log 2).
104   uint64_t getAlignment() const;
105 
106   bool isCompressed() const;
107   /// Whether this section contains instructions.
108   bool isText() const;
109   /// Whether this section contains data, not instructions.
110   bool isData() const;
111   /// Whether this section contains BSS uninitialized data.
112   bool isBSS() const;
113   bool isVirtual() const;
114   bool isBitcode() const;
115   bool isStripped() const;
116 
117   /// Whether this section will be placed in the text segment, according to the
118   /// Berkeley size format. This is true if the section is allocatable, and
119   /// contains either code or readonly data.
120   bool isBerkeleyText() const;
121   /// Whether this section will be placed in the data segment, according to the
122   /// Berkeley size format. This is true if the section is allocatable and
123   /// contains data (e.g. PROGBITS), but is not text.
124   bool isBerkeleyData() const;
125 
126   bool containsSymbol(SymbolRef S) const;
127 
128   relocation_iterator relocation_begin() const;
129   relocation_iterator relocation_end() const;
130   iterator_range<relocation_iterator> relocations() const {
131     return make_range(relocation_begin(), relocation_end());
132   }
133   Expected<section_iterator> getRelocatedSection() const;
134 
135   DataRefImpl getRawDataRefImpl() const;
136   const ObjectFile *getObject() const;
137 };
138 
139 struct SectionedAddress {
140   const static uint64_t UndefSection = UINT64_MAX;
141 
142   uint64_t Address = 0;
143   uint64_t SectionIndex = UndefSection;
144 };
145 
146 inline bool operator<(const SectionedAddress &LHS,
147                       const SectionedAddress &RHS) {
148   return std::tie(LHS.SectionIndex, LHS.Address) <
149          std::tie(RHS.SectionIndex, RHS.Address);
150 }
151 
152 inline bool operator==(const SectionedAddress &LHS,
153                        const SectionedAddress &RHS) {
154   return std::tie(LHS.SectionIndex, LHS.Address) ==
155          std::tie(RHS.SectionIndex, RHS.Address);
156 }
157 
158 raw_ostream &operator<<(raw_ostream &OS, const SectionedAddress &Addr);
159 
160 /// This is a value type class that represents a single symbol in the list of
161 /// symbols in the object file.
162 class SymbolRef : public BasicSymbolRef {
163   friend class SectionRef;
164 
165 public:
166   enum Type {
167     ST_Unknown, // Type not specified
168     ST_Data,
169     ST_Debug,
170     ST_File,
171     ST_Function,
172     ST_Other
173   };
174 
175   SymbolRef() = default;
176   SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner);
177   SymbolRef(const BasicSymbolRef &B) : BasicSymbolRef(B) {
178     assert(isa<ObjectFile>(BasicSymbolRef::getObject()));
179   }
180 
181   Expected<StringRef> getName() const;
182   /// Returns the symbol virtual address (i.e. address at which it will be
183   /// mapped).
184   Expected<uint64_t> getAddress() const;
185 
186   /// Return the value of the symbol depending on the object this can be an
187   /// offset or a virtual address.
188   uint64_t getValue() const;
189 
190   /// Get the alignment of this symbol as the actual value (not log 2).
191   uint32_t getAlignment() const;
192   uint64_t getCommonSize() const;
193   Expected<SymbolRef::Type> getType() const;
194 
195   /// Get section this symbol is defined in reference to. Result is
196   /// end_sections() if it is undefined or is an absolute symbol.
197   Expected<section_iterator> getSection() const;
198 
199   const ObjectFile *getObject() const;
200 };
201 
202 class symbol_iterator : public basic_symbol_iterator {
203 public:
204   symbol_iterator(SymbolRef Sym) : basic_symbol_iterator(Sym) {}
205   symbol_iterator(const basic_symbol_iterator &B)
206       : basic_symbol_iterator(SymbolRef(B->getRawDataRefImpl(),
207                                         cast<ObjectFile>(B->getObject()))) {}
208 
209   const SymbolRef *operator->() const {
210     const BasicSymbolRef &P = basic_symbol_iterator::operator *();
211     return static_cast<const SymbolRef*>(&P);
212   }
213 
214   const SymbolRef &operator*() const {
215     const BasicSymbolRef &P = basic_symbol_iterator::operator *();
216     return static_cast<const SymbolRef&>(P);
217   }
218 };
219 
220 /// This class is the base class for all object file types. Concrete instances
221 /// of this object are created by createObjectFile, which figures out which type
222 /// to create.
223 class ObjectFile : public SymbolicFile {
224   virtual void anchor();
225 
226 protected:
227   ObjectFile(unsigned int Type, MemoryBufferRef Source);
228 
229   const uint8_t *base() const {
230     return reinterpret_cast<const uint8_t *>(Data.getBufferStart());
231   }
232 
233   // These functions are for SymbolRef to call internally. The main goal of
234   // this is to allow SymbolRef::SymbolPimpl to point directly to the symbol
235   // entry in the memory mapped object file. SymbolPimpl cannot contain any
236   // virtual functions because then it could not point into the memory mapped
237   // file.
238   //
239   // Implementations assume that the DataRefImpl is valid and has not been
240   // modified externally. It's UB otherwise.
241   friend class SymbolRef;
242 
243   virtual Expected<StringRef> getSymbolName(DataRefImpl Symb) const = 0;
244   Error printSymbolName(raw_ostream &OS,
245                                   DataRefImpl Symb) const override;
246   virtual Expected<uint64_t> getSymbolAddress(DataRefImpl Symb) const = 0;
247   virtual uint64_t getSymbolValueImpl(DataRefImpl Symb) const = 0;
248   virtual uint32_t getSymbolAlignment(DataRefImpl Symb) const;
249   virtual uint64_t getCommonSymbolSizeImpl(DataRefImpl Symb) const = 0;
250   virtual Expected<SymbolRef::Type> getSymbolType(DataRefImpl Symb) const = 0;
251   virtual Expected<section_iterator>
252   getSymbolSection(DataRefImpl Symb) const = 0;
253 
254   // Same as above for SectionRef.
255   friend class SectionRef;
256 
257   virtual void moveSectionNext(DataRefImpl &Sec) const = 0;
258   virtual Expected<StringRef> getSectionName(DataRefImpl Sec) const = 0;
259   virtual uint64_t getSectionAddress(DataRefImpl Sec) const = 0;
260   virtual uint64_t getSectionIndex(DataRefImpl Sec) const = 0;
261   virtual uint64_t getSectionSize(DataRefImpl Sec) const = 0;
262   virtual Expected<ArrayRef<uint8_t>>
263   getSectionContents(DataRefImpl Sec) const = 0;
264   virtual uint64_t getSectionAlignment(DataRefImpl Sec) const = 0;
265   virtual bool isSectionCompressed(DataRefImpl Sec) const = 0;
266   virtual bool isSectionText(DataRefImpl Sec) const = 0;
267   virtual bool isSectionData(DataRefImpl Sec) const = 0;
268   virtual bool isSectionBSS(DataRefImpl Sec) const = 0;
269   // A section is 'virtual' if its contents aren't present in the object image.
270   virtual bool isSectionVirtual(DataRefImpl Sec) const = 0;
271   virtual bool isSectionBitcode(DataRefImpl Sec) const;
272   virtual bool isSectionStripped(DataRefImpl Sec) const;
273   virtual bool isBerkeleyText(DataRefImpl Sec) const;
274   virtual bool isBerkeleyData(DataRefImpl Sec) const;
275   virtual relocation_iterator section_rel_begin(DataRefImpl Sec) const = 0;
276   virtual relocation_iterator section_rel_end(DataRefImpl Sec) const = 0;
277   virtual Expected<section_iterator> getRelocatedSection(DataRefImpl Sec) const;
278 
279   // Same as above for RelocationRef.
280   friend class RelocationRef;
281   virtual void moveRelocationNext(DataRefImpl &Rel) const = 0;
282   virtual uint64_t getRelocationOffset(DataRefImpl Rel) const = 0;
283   virtual symbol_iterator getRelocationSymbol(DataRefImpl Rel) const = 0;
284   virtual uint64_t getRelocationType(DataRefImpl Rel) const = 0;
285   virtual void getRelocationTypeName(DataRefImpl Rel,
286                                      SmallVectorImpl<char> &Result) const = 0;
287 
288   uint64_t getSymbolValue(DataRefImpl Symb) const;
289 
290 public:
291   ObjectFile() = delete;
292   ObjectFile(const ObjectFile &other) = delete;
293 
294   uint64_t getCommonSymbolSize(DataRefImpl Symb) const {
295     assert(getSymbolFlags(Symb) & SymbolRef::SF_Common);
296     return getCommonSymbolSizeImpl(Symb);
297   }
298 
299   virtual std::vector<SectionRef> dynamic_relocation_sections() const {
300     return std::vector<SectionRef>();
301   }
302 
303   using symbol_iterator_range = iterator_range<symbol_iterator>;
304   symbol_iterator_range symbols() const {
305     return symbol_iterator_range(symbol_begin(), symbol_end());
306   }
307 
308   virtual section_iterator section_begin() const = 0;
309   virtual section_iterator section_end() const = 0;
310 
311   using section_iterator_range = iterator_range<section_iterator>;
312   section_iterator_range sections() const {
313     return section_iterator_range(section_begin(), section_end());
314   }
315 
316   /// The number of bytes used to represent an address in this object
317   ///        file format.
318   virtual uint8_t getBytesInAddress() const = 0;
319 
320   virtual StringRef getFileFormatName() const = 0;
321   virtual Triple::ArchType getArch() const = 0;
322   virtual SubtargetFeatures getFeatures() const = 0;
323   virtual void setARMSubArch(Triple &TheTriple) const { }
324   virtual Expected<uint64_t> getStartAddress() const {
325     return errorCodeToError(object_error::parse_failed);
326   };
327 
328   /// Create a triple from the data in this object file.
329   Triple makeTriple() const;
330 
331   /// Maps a debug section name to a standard DWARF section name.
332   virtual StringRef mapDebugSectionName(StringRef Name) const { return Name; }
333 
334   /// True if this is a relocatable object (.o/.obj).
335   virtual bool isRelocatableObject() const = 0;
336 
337   /// @returns Pointer to ObjectFile subclass to handle this type of object.
338   /// @param ObjectPath The path to the object file. ObjectPath.isObject must
339   ///        return true.
340   /// Create ObjectFile from path.
341   static Expected<OwningBinary<ObjectFile>>
342   createObjectFile(StringRef ObjectPath);
343 
344   static Expected<std::unique_ptr<ObjectFile>>
345   createObjectFile(MemoryBufferRef Object, llvm::file_magic Type);
346   static Expected<std::unique_ptr<ObjectFile>>
347   createObjectFile(MemoryBufferRef Object) {
348     return createObjectFile(Object, llvm::file_magic::unknown);
349   }
350 
351   static bool classof(const Binary *v) {
352     return v->isObject();
353   }
354 
355   static Expected<std::unique_ptr<COFFObjectFile>>
356   createCOFFObjectFile(MemoryBufferRef Object);
357 
358   static Expected<std::unique_ptr<ObjectFile>>
359   createXCOFFObjectFile(MemoryBufferRef Object, unsigned FileType);
360 
361   static Expected<std::unique_ptr<ObjectFile>>
362   createELFObjectFile(MemoryBufferRef Object);
363 
364   static Expected<std::unique_ptr<MachOObjectFile>>
365   createMachOObjectFile(MemoryBufferRef Object,
366                         uint32_t UniversalCputype = 0,
367                         uint32_t UniversalIndex = 0);
368 
369   static Expected<std::unique_ptr<WasmObjectFile>>
370   createWasmObjectFile(MemoryBufferRef Object);
371 };
372 
373 // Inline function definitions.
374 inline SymbolRef::SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner)
375     : BasicSymbolRef(SymbolP, Owner) {}
376 
377 inline Expected<StringRef> SymbolRef::getName() const {
378   return getObject()->getSymbolName(getRawDataRefImpl());
379 }
380 
381 inline Expected<uint64_t> SymbolRef::getAddress() const {
382   return getObject()->getSymbolAddress(getRawDataRefImpl());
383 }
384 
385 inline uint64_t SymbolRef::getValue() const {
386   return getObject()->getSymbolValue(getRawDataRefImpl());
387 }
388 
389 inline uint32_t SymbolRef::getAlignment() const {
390   return getObject()->getSymbolAlignment(getRawDataRefImpl());
391 }
392 
393 inline uint64_t SymbolRef::getCommonSize() const {
394   return getObject()->getCommonSymbolSize(getRawDataRefImpl());
395 }
396 
397 inline Expected<section_iterator> SymbolRef::getSection() const {
398   return getObject()->getSymbolSection(getRawDataRefImpl());
399 }
400 
401 inline Expected<SymbolRef::Type> SymbolRef::getType() const {
402   return getObject()->getSymbolType(getRawDataRefImpl());
403 }
404 
405 inline const ObjectFile *SymbolRef::getObject() const {
406   const SymbolicFile *O = BasicSymbolRef::getObject();
407   return cast<ObjectFile>(O);
408 }
409 
410 /// SectionRef
411 inline SectionRef::SectionRef(DataRefImpl SectionP,
412                               const ObjectFile *Owner)
413   : SectionPimpl(SectionP)
414   , OwningObject(Owner) {}
415 
416 inline bool SectionRef::operator==(const SectionRef &Other) const {
417   return OwningObject == Other.OwningObject &&
418          SectionPimpl == Other.SectionPimpl;
419 }
420 
421 inline bool SectionRef::operator!=(const SectionRef &Other) const {
422   return !(*this == Other);
423 }
424 
425 inline bool SectionRef::operator<(const SectionRef &Other) const {
426   assert(OwningObject == Other.OwningObject);
427   return SectionPimpl < Other.SectionPimpl;
428 }
429 
430 inline void SectionRef::moveNext() {
431   return OwningObject->moveSectionNext(SectionPimpl);
432 }
433 
434 inline Expected<StringRef> SectionRef::getName() const {
435   return OwningObject->getSectionName(SectionPimpl);
436 }
437 
438 inline uint64_t SectionRef::getAddress() const {
439   return OwningObject->getSectionAddress(SectionPimpl);
440 }
441 
442 inline uint64_t SectionRef::getIndex() const {
443   return OwningObject->getSectionIndex(SectionPimpl);
444 }
445 
446 inline uint64_t SectionRef::getSize() const {
447   return OwningObject->getSectionSize(SectionPimpl);
448 }
449 
450 inline Expected<StringRef> SectionRef::getContents() const {
451   Expected<ArrayRef<uint8_t>> Res =
452       OwningObject->getSectionContents(SectionPimpl);
453   if (!Res)
454     return Res.takeError();
455   return StringRef(reinterpret_cast<const char *>(Res->data()), Res->size());
456 }
457 
458 inline uint64_t SectionRef::getAlignment() const {
459   return OwningObject->getSectionAlignment(SectionPimpl);
460 }
461 
462 inline bool SectionRef::isCompressed() const {
463   return OwningObject->isSectionCompressed(SectionPimpl);
464 }
465 
466 inline bool SectionRef::isText() const {
467   return OwningObject->isSectionText(SectionPimpl);
468 }
469 
470 inline bool SectionRef::isData() const {
471   return OwningObject->isSectionData(SectionPimpl);
472 }
473 
474 inline bool SectionRef::isBSS() const {
475   return OwningObject->isSectionBSS(SectionPimpl);
476 }
477 
478 inline bool SectionRef::isVirtual() const {
479   return OwningObject->isSectionVirtual(SectionPimpl);
480 }
481 
482 inline bool SectionRef::isBitcode() const {
483   return OwningObject->isSectionBitcode(SectionPimpl);
484 }
485 
486 inline bool SectionRef::isStripped() const {
487   return OwningObject->isSectionStripped(SectionPimpl);
488 }
489 
490 inline bool SectionRef::isBerkeleyText() const {
491   return OwningObject->isBerkeleyText(SectionPimpl);
492 }
493 
494 inline bool SectionRef::isBerkeleyData() const {
495   return OwningObject->isBerkeleyData(SectionPimpl);
496 }
497 
498 inline relocation_iterator SectionRef::relocation_begin() const {
499   return OwningObject->section_rel_begin(SectionPimpl);
500 }
501 
502 inline relocation_iterator SectionRef::relocation_end() const {
503   return OwningObject->section_rel_end(SectionPimpl);
504 }
505 
506 inline Expected<section_iterator> SectionRef::getRelocatedSection() const {
507   return OwningObject->getRelocatedSection(SectionPimpl);
508 }
509 
510 inline DataRefImpl SectionRef::getRawDataRefImpl() const {
511   return SectionPimpl;
512 }
513 
514 inline const ObjectFile *SectionRef::getObject() const {
515   return OwningObject;
516 }
517 
518 /// RelocationRef
519 inline RelocationRef::RelocationRef(DataRefImpl RelocationP,
520                               const ObjectFile *Owner)
521   : RelocationPimpl(RelocationP)
522   , OwningObject(Owner) {}
523 
524 inline bool RelocationRef::operator==(const RelocationRef &Other) const {
525   return RelocationPimpl == Other.RelocationPimpl;
526 }
527 
528 inline void RelocationRef::moveNext() {
529   return OwningObject->moveRelocationNext(RelocationPimpl);
530 }
531 
532 inline uint64_t RelocationRef::getOffset() const {
533   return OwningObject->getRelocationOffset(RelocationPimpl);
534 }
535 
536 inline symbol_iterator RelocationRef::getSymbol() const {
537   return OwningObject->getRelocationSymbol(RelocationPimpl);
538 }
539 
540 inline uint64_t RelocationRef::getType() const {
541   return OwningObject->getRelocationType(RelocationPimpl);
542 }
543 
544 inline void RelocationRef::getTypeName(SmallVectorImpl<char> &Result) const {
545   return OwningObject->getRelocationTypeName(RelocationPimpl, Result);
546 }
547 
548 inline DataRefImpl RelocationRef::getRawDataRefImpl() const {
549   return RelocationPimpl;
550 }
551 
552 inline const ObjectFile *RelocationRef::getObject() const {
553   return OwningObject;
554 }
555 
556 } // end namespace object
557 
558 template <> struct DenseMapInfo<object::SectionRef> {
559   static bool isEqual(const object::SectionRef &A,
560                       const object::SectionRef &B) {
561     return A == B;
562   }
563   static object::SectionRef getEmptyKey() {
564     return object::SectionRef({}, nullptr);
565   }
566   static object::SectionRef getTombstoneKey() {
567     object::DataRefImpl TS;
568     TS.p = (uintptr_t)-1;
569     return object::SectionRef(TS, nullptr);
570   }
571   static unsigned getHashValue(const object::SectionRef &Sec) {
572     object::DataRefImpl Raw = Sec.getRawDataRefImpl();
573     return hash_combine(Raw.p, Raw.d.a, Raw.d.b);
574   }
575 };
576 
577 } // end namespace llvm
578 
579 #endif // LLVM_OBJECT_OBJECTFILE_H
580