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/Object/Binary.h"
22 #include "llvm/Object/Error.h"
23 #include "llvm/Object/SymbolicFile.h"
24 #include "llvm/Support/Casting.h"
25 #include "llvm/Support/Error.h"
26 #include "llvm/Support/MemoryBuffer.h"
27 #include <cassert>
28 #include <cstdint>
29 #include <memory>
30 #include <system_error>
31 
32 namespace llvm {
33 
34 class ARMAttributeParser;
35 class SubtargetFeatures;
36 
37 namespace object {
38 
39 class COFFObjectFile;
40 class MachOObjectFile;
41 class ObjectFile;
42 class SectionRef;
43 class SymbolRef;
44 class symbol_iterator;
45 class WasmObjectFile;
46 
47 using section_iterator = content_iterator<SectionRef>;
48 
49 /// This is a value type class that represents a single relocation in the list
50 /// of relocations in the object file.
51 class RelocationRef {
52   DataRefImpl RelocationPimpl;
53   const ObjectFile *OwningObject = nullptr;
54 
55 public:
56   RelocationRef() = default;
57   RelocationRef(DataRefImpl RelocationP, const ObjectFile *Owner);
58 
59   bool operator==(const RelocationRef &Other) const;
60 
61   void moveNext();
62 
63   uint64_t getOffset() const;
64   symbol_iterator getSymbol() const;
65   uint64_t getType() const;
66 
67   /// Get a string that represents the type of this relocation.
68   ///
69   /// This is for display purposes only.
70   void getTypeName(SmallVectorImpl<char> &Result) const;
71 
72   DataRefImpl getRawDataRefImpl() const;
73   const ObjectFile *getObject() const;
74 };
75 
76 using relocation_iterator = content_iterator<RelocationRef>;
77 
78 /// This is a value type class that represents a single section in the list of
79 /// sections in the object file.
80 class SectionRef {
81   friend class SymbolRef;
82 
83   DataRefImpl SectionPimpl;
84   const ObjectFile *OwningObject = nullptr;
85 
86 public:
87   SectionRef() = default;
88   SectionRef(DataRefImpl SectionP, const ObjectFile *Owner);
89 
90   bool operator==(const SectionRef &Other) const;
91   bool operator!=(const SectionRef &Other) const;
92   bool operator<(const SectionRef &Other) const;
93 
94   void moveNext();
95 
96   Expected<StringRef> getName() const;
97   uint64_t getAddress() const;
98   uint64_t getIndex() const;
99   uint64_t getSize() const;
100   Expected<StringRef> getContents() const;
101 
102   /// Get the alignment of this section as the actual value (not log 2).
103   uint64_t getAlignment() const;
104 
105   bool isCompressed() const;
106   /// Whether this section contains instructions.
107   bool isText() const;
108   /// Whether this section contains data, not instructions.
109   bool isData() const;
110   /// Whether this section contains BSS uninitialized data.
111   bool isBSS() const;
112   bool isVirtual() const;
113   bool isBitcode() const;
114   bool isStripped() const;
115 
116   /// Whether this section will be placed in the text segment, according to the
117   /// Berkeley size format. This is true if the section is allocatable, and
118   /// contains either code or readonly data.
119   bool isBerkeleyText() const;
120   /// Whether this section will be placed in the data segment, according to the
121   /// Berkeley size format. This is true if the section is allocatable and
122   /// contains data (e.g. PROGBITS), but is not text.
123   bool isBerkeleyData() const;
124 
125   /// Whether this section is a debug section.
126   bool isDebugSection(StringRef SectionName) const;
127 
128   bool containsSymbol(SymbolRef S) const;
129 
130   relocation_iterator relocation_begin() const;
131   relocation_iterator relocation_end() const;
132   iterator_range<relocation_iterator> relocations() const {
133     return make_range(relocation_begin(), relocation_end());
134   }
135   Expected<section_iterator> getRelocatedSection() const;
136 
137   DataRefImpl getRawDataRefImpl() const;
138   const ObjectFile *getObject() const;
139 };
140 
141 struct SectionedAddress {
142   const static uint64_t UndefSection = UINT64_MAX;
143 
144   uint64_t Address = 0;
145   uint64_t SectionIndex = UndefSection;
146 };
147 
148 inline bool operator<(const SectionedAddress &LHS,
149                       const SectionedAddress &RHS) {
150   return std::tie(LHS.SectionIndex, LHS.Address) <
151          std::tie(RHS.SectionIndex, RHS.Address);
152 }
153 
154 inline bool operator==(const SectionedAddress &LHS,
155                        const SectionedAddress &RHS) {
156   return std::tie(LHS.SectionIndex, LHS.Address) ==
157          std::tie(RHS.SectionIndex, RHS.Address);
158 }
159 
160 raw_ostream &operator<<(raw_ostream &OS, const SectionedAddress &Addr);
161 
162 /// This is a value type class that represents a single symbol in the list of
163 /// symbols in the object file.
164 class SymbolRef : public BasicSymbolRef {
165   friend class SectionRef;
166 
167 public:
168   enum Type {
169     ST_Unknown, // Type not specified
170     ST_Data,
171     ST_Debug,
172     ST_File,
173     ST_Function,
174     ST_Other
175   };
176 
177   SymbolRef() = default;
178   SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner);
179   SymbolRef(const BasicSymbolRef &B) : BasicSymbolRef(B) {
180     assert(isa<ObjectFile>(BasicSymbolRef::getObject()));
181   }
182 
183   Expected<StringRef> getName() const;
184   /// Returns the symbol virtual address (i.e. address at which it will be
185   /// mapped).
186   Expected<uint64_t> getAddress() const;
187 
188   /// Return the value of the symbol depending on the object this can be an
189   /// offset or a virtual address.
190   Expected<uint64_t> getValue() const;
191 
192   /// Get the alignment of this symbol as the actual value (not log 2).
193   uint32_t getAlignment() const;
194   uint64_t getCommonSize() const;
195   Expected<SymbolRef::Type> getType() const;
196 
197   /// Get section this symbol is defined in reference to. Result is
198   /// end_sections() if it is undefined or is an absolute symbol.
199   Expected<section_iterator> getSection() const;
200 
201   const ObjectFile *getObject() const;
202 };
203 
204 class symbol_iterator : public basic_symbol_iterator {
205 public:
206   symbol_iterator(SymbolRef Sym) : basic_symbol_iterator(Sym) {}
207   symbol_iterator(const basic_symbol_iterator &B)
208       : basic_symbol_iterator(SymbolRef(B->getRawDataRefImpl(),
209                                         cast<ObjectFile>(B->getObject()))) {}
210 
211   const SymbolRef *operator->() const {
212     const BasicSymbolRef &P = basic_symbol_iterator::operator *();
213     return static_cast<const SymbolRef*>(&P);
214   }
215 
216   const SymbolRef &operator*() const {
217     const BasicSymbolRef &P = basic_symbol_iterator::operator *();
218     return static_cast<const SymbolRef&>(P);
219   }
220 };
221 
222 /// This class is the base class for all object file types. Concrete instances
223 /// of this object are created by createObjectFile, which figures out which type
224 /// to create.
225 class ObjectFile : public SymbolicFile {
226   virtual void anchor();
227 
228 protected:
229   ObjectFile(unsigned int Type, MemoryBufferRef Source);
230 
231   const uint8_t *base() const {
232     return reinterpret_cast<const uint8_t *>(Data.getBufferStart());
233   }
234 
235   // These functions are for SymbolRef to call internally. The main goal of
236   // this is to allow SymbolRef::SymbolPimpl to point directly to the symbol
237   // entry in the memory mapped object file. SymbolPimpl cannot contain any
238   // virtual functions because then it could not point into the memory mapped
239   // file.
240   //
241   // Implementations assume that the DataRefImpl is valid and has not been
242   // modified externally. It's UB otherwise.
243   friend class SymbolRef;
244 
245   virtual Expected<StringRef> getSymbolName(DataRefImpl Symb) const = 0;
246   Error printSymbolName(raw_ostream &OS,
247                                   DataRefImpl Symb) const override;
248   virtual Expected<uint64_t> getSymbolAddress(DataRefImpl Symb) const = 0;
249   virtual uint64_t getSymbolValueImpl(DataRefImpl Symb) const = 0;
250   virtual uint32_t getSymbolAlignment(DataRefImpl Symb) const;
251   virtual uint64_t getCommonSymbolSizeImpl(DataRefImpl Symb) const = 0;
252   virtual Expected<SymbolRef::Type> getSymbolType(DataRefImpl Symb) const = 0;
253   virtual Expected<section_iterator>
254   getSymbolSection(DataRefImpl Symb) const = 0;
255 
256   // Same as above for SectionRef.
257   friend class SectionRef;
258 
259   virtual void moveSectionNext(DataRefImpl &Sec) const = 0;
260   virtual Expected<StringRef> getSectionName(DataRefImpl Sec) const = 0;
261   virtual uint64_t getSectionAddress(DataRefImpl Sec) const = 0;
262   virtual uint64_t getSectionIndex(DataRefImpl Sec) const = 0;
263   virtual uint64_t getSectionSize(DataRefImpl Sec) const = 0;
264   virtual Expected<ArrayRef<uint8_t>>
265   getSectionContents(DataRefImpl Sec) const = 0;
266   virtual uint64_t getSectionAlignment(DataRefImpl Sec) const = 0;
267   virtual bool isSectionCompressed(DataRefImpl Sec) const = 0;
268   virtual bool isSectionText(DataRefImpl Sec) const = 0;
269   virtual bool isSectionData(DataRefImpl Sec) const = 0;
270   virtual bool isSectionBSS(DataRefImpl Sec) const = 0;
271   // A section is 'virtual' if its contents aren't present in the object image.
272   virtual bool isSectionVirtual(DataRefImpl Sec) const = 0;
273   virtual bool isSectionBitcode(DataRefImpl Sec) const;
274   virtual bool isSectionStripped(DataRefImpl Sec) const;
275   virtual bool isBerkeleyText(DataRefImpl Sec) const;
276   virtual bool isBerkeleyData(DataRefImpl Sec) const;
277   virtual bool isDebugSection(StringRef SectionName) const;
278   virtual relocation_iterator section_rel_begin(DataRefImpl Sec) const = 0;
279   virtual relocation_iterator section_rel_end(DataRefImpl Sec) const = 0;
280   virtual Expected<section_iterator> getRelocatedSection(DataRefImpl Sec) const;
281 
282   // Same as above for RelocationRef.
283   friend class RelocationRef;
284   virtual void moveRelocationNext(DataRefImpl &Rel) const = 0;
285   virtual uint64_t getRelocationOffset(DataRefImpl Rel) const = 0;
286   virtual symbol_iterator getRelocationSymbol(DataRefImpl Rel) const = 0;
287   virtual uint64_t getRelocationType(DataRefImpl Rel) const = 0;
288   virtual void getRelocationTypeName(DataRefImpl Rel,
289                                      SmallVectorImpl<char> &Result) const = 0;
290 
291   Expected<uint64_t> getSymbolValue(DataRefImpl Symb) const;
292 
293 public:
294   ObjectFile() = delete;
295   ObjectFile(const ObjectFile &other) = delete;
296 
297   uint64_t getCommonSymbolSize(DataRefImpl Symb) const {
298     Expected<uint32_t> SymbolFlagsOrErr = getSymbolFlags(Symb);
299     if (!SymbolFlagsOrErr)
300       // TODO: Actually report errors helpfully.
301       report_fatal_error(SymbolFlagsOrErr.takeError());
302     assert(*SymbolFlagsOrErr & SymbolRef::SF_Common);
303     return getCommonSymbolSizeImpl(Symb);
304   }
305 
306   virtual std::vector<SectionRef> dynamic_relocation_sections() const {
307     return std::vector<SectionRef>();
308   }
309 
310   using symbol_iterator_range = iterator_range<symbol_iterator>;
311   symbol_iterator_range symbols() const {
312     return symbol_iterator_range(symbol_begin(), symbol_end());
313   }
314 
315   virtual section_iterator section_begin() const = 0;
316   virtual section_iterator section_end() const = 0;
317 
318   using section_iterator_range = iterator_range<section_iterator>;
319   section_iterator_range sections() const {
320     return section_iterator_range(section_begin(), section_end());
321   }
322 
323   /// The number of bytes used to represent an address in this object
324   ///        file format.
325   virtual uint8_t getBytesInAddress() const = 0;
326 
327   virtual StringRef getFileFormatName() const = 0;
328   virtual Triple::ArchType getArch() const = 0;
329   virtual SubtargetFeatures getFeatures() const = 0;
330   virtual Optional<StringRef> tryGetCPUName() const { return None; };
331   virtual void setARMSubArch(Triple &TheTriple) const { }
332   virtual Expected<uint64_t> getStartAddress() const {
333     return errorCodeToError(object_error::parse_failed);
334   };
335 
336   /// Create a triple from the data in this object file.
337   Triple makeTriple() const;
338 
339   /// Maps a debug section name to a standard DWARF section name.
340   virtual StringRef mapDebugSectionName(StringRef Name) const { return Name; }
341 
342   /// True if this is a relocatable object (.o/.obj).
343   virtual bool isRelocatableObject() const = 0;
344 
345   /// @returns Pointer to ObjectFile subclass to handle this type of object.
346   /// @param ObjectPath The path to the object file. ObjectPath.isObject must
347   ///        return true.
348   /// Create ObjectFile from path.
349   static Expected<OwningBinary<ObjectFile>>
350   createObjectFile(StringRef ObjectPath);
351 
352   static Expected<std::unique_ptr<ObjectFile>>
353   createObjectFile(MemoryBufferRef Object, llvm::file_magic Type,
354                    bool InitContent = true);
355   static Expected<std::unique_ptr<ObjectFile>>
356   createObjectFile(MemoryBufferRef Object) {
357     return createObjectFile(Object, llvm::file_magic::unknown);
358   }
359 
360   static bool classof(const Binary *v) {
361     return v->isObject();
362   }
363 
364   static Expected<std::unique_ptr<COFFObjectFile>>
365   createCOFFObjectFile(MemoryBufferRef Object);
366 
367   static Expected<std::unique_ptr<ObjectFile>>
368   createXCOFFObjectFile(MemoryBufferRef Object, unsigned FileType);
369 
370   static Expected<std::unique_ptr<ObjectFile>>
371   createELFObjectFile(MemoryBufferRef Object, bool InitContent = true);
372 
373   static Expected<std::unique_ptr<MachOObjectFile>>
374   createMachOObjectFile(MemoryBufferRef Object,
375                         uint32_t UniversalCputype = 0,
376                         uint32_t UniversalIndex = 0);
377 
378   static Expected<std::unique_ptr<WasmObjectFile>>
379   createWasmObjectFile(MemoryBufferRef Object);
380 };
381 
382 // Inline function definitions.
383 inline SymbolRef::SymbolRef(DataRefImpl SymbolP, const ObjectFile *Owner)
384     : BasicSymbolRef(SymbolP, Owner) {}
385 
386 inline Expected<StringRef> SymbolRef::getName() const {
387   return getObject()->getSymbolName(getRawDataRefImpl());
388 }
389 
390 inline Expected<uint64_t> SymbolRef::getAddress() const {
391   return getObject()->getSymbolAddress(getRawDataRefImpl());
392 }
393 
394 inline Expected<uint64_t> SymbolRef::getValue() const {
395   return getObject()->getSymbolValue(getRawDataRefImpl());
396 }
397 
398 inline uint32_t SymbolRef::getAlignment() const {
399   return getObject()->getSymbolAlignment(getRawDataRefImpl());
400 }
401 
402 inline uint64_t SymbolRef::getCommonSize() const {
403   return getObject()->getCommonSymbolSize(getRawDataRefImpl());
404 }
405 
406 inline Expected<section_iterator> SymbolRef::getSection() const {
407   return getObject()->getSymbolSection(getRawDataRefImpl());
408 }
409 
410 inline Expected<SymbolRef::Type> SymbolRef::getType() const {
411   return getObject()->getSymbolType(getRawDataRefImpl());
412 }
413 
414 inline const ObjectFile *SymbolRef::getObject() const {
415   const SymbolicFile *O = BasicSymbolRef::getObject();
416   return cast<ObjectFile>(O);
417 }
418 
419 /// SectionRef
420 inline SectionRef::SectionRef(DataRefImpl SectionP,
421                               const ObjectFile *Owner)
422   : SectionPimpl(SectionP)
423   , OwningObject(Owner) {}
424 
425 inline bool SectionRef::operator==(const SectionRef &Other) const {
426   return OwningObject == Other.OwningObject &&
427          SectionPimpl == Other.SectionPimpl;
428 }
429 
430 inline bool SectionRef::operator!=(const SectionRef &Other) const {
431   return !(*this == Other);
432 }
433 
434 inline bool SectionRef::operator<(const SectionRef &Other) const {
435   assert(OwningObject == Other.OwningObject);
436   return SectionPimpl < Other.SectionPimpl;
437 }
438 
439 inline void SectionRef::moveNext() {
440   return OwningObject->moveSectionNext(SectionPimpl);
441 }
442 
443 inline Expected<StringRef> SectionRef::getName() const {
444   return OwningObject->getSectionName(SectionPimpl);
445 }
446 
447 inline uint64_t SectionRef::getAddress() const {
448   return OwningObject->getSectionAddress(SectionPimpl);
449 }
450 
451 inline uint64_t SectionRef::getIndex() const {
452   return OwningObject->getSectionIndex(SectionPimpl);
453 }
454 
455 inline uint64_t SectionRef::getSize() const {
456   return OwningObject->getSectionSize(SectionPimpl);
457 }
458 
459 inline Expected<StringRef> SectionRef::getContents() const {
460   Expected<ArrayRef<uint8_t>> Res =
461       OwningObject->getSectionContents(SectionPimpl);
462   if (!Res)
463     return Res.takeError();
464   return StringRef(reinterpret_cast<const char *>(Res->data()), Res->size());
465 }
466 
467 inline uint64_t SectionRef::getAlignment() const {
468   return OwningObject->getSectionAlignment(SectionPimpl);
469 }
470 
471 inline bool SectionRef::isCompressed() const {
472   return OwningObject->isSectionCompressed(SectionPimpl);
473 }
474 
475 inline bool SectionRef::isText() const {
476   return OwningObject->isSectionText(SectionPimpl);
477 }
478 
479 inline bool SectionRef::isData() const {
480   return OwningObject->isSectionData(SectionPimpl);
481 }
482 
483 inline bool SectionRef::isBSS() const {
484   return OwningObject->isSectionBSS(SectionPimpl);
485 }
486 
487 inline bool SectionRef::isVirtual() const {
488   return OwningObject->isSectionVirtual(SectionPimpl);
489 }
490 
491 inline bool SectionRef::isBitcode() const {
492   return OwningObject->isSectionBitcode(SectionPimpl);
493 }
494 
495 inline bool SectionRef::isStripped() const {
496   return OwningObject->isSectionStripped(SectionPimpl);
497 }
498 
499 inline bool SectionRef::isBerkeleyText() const {
500   return OwningObject->isBerkeleyText(SectionPimpl);
501 }
502 
503 inline bool SectionRef::isBerkeleyData() const {
504   return OwningObject->isBerkeleyData(SectionPimpl);
505 }
506 
507 inline bool SectionRef::isDebugSection(StringRef SectionName) const {
508   return OwningObject->isDebugSection(SectionName);
509 }
510 
511 inline relocation_iterator SectionRef::relocation_begin() const {
512   return OwningObject->section_rel_begin(SectionPimpl);
513 }
514 
515 inline relocation_iterator SectionRef::relocation_end() const {
516   return OwningObject->section_rel_end(SectionPimpl);
517 }
518 
519 inline Expected<section_iterator> SectionRef::getRelocatedSection() const {
520   return OwningObject->getRelocatedSection(SectionPimpl);
521 }
522 
523 inline DataRefImpl SectionRef::getRawDataRefImpl() const {
524   return SectionPimpl;
525 }
526 
527 inline const ObjectFile *SectionRef::getObject() const {
528   return OwningObject;
529 }
530 
531 /// RelocationRef
532 inline RelocationRef::RelocationRef(DataRefImpl RelocationP,
533                               const ObjectFile *Owner)
534   : RelocationPimpl(RelocationP)
535   , OwningObject(Owner) {}
536 
537 inline bool RelocationRef::operator==(const RelocationRef &Other) const {
538   return RelocationPimpl == Other.RelocationPimpl;
539 }
540 
541 inline void RelocationRef::moveNext() {
542   return OwningObject->moveRelocationNext(RelocationPimpl);
543 }
544 
545 inline uint64_t RelocationRef::getOffset() const {
546   return OwningObject->getRelocationOffset(RelocationPimpl);
547 }
548 
549 inline symbol_iterator RelocationRef::getSymbol() const {
550   return OwningObject->getRelocationSymbol(RelocationPimpl);
551 }
552 
553 inline uint64_t RelocationRef::getType() const {
554   return OwningObject->getRelocationType(RelocationPimpl);
555 }
556 
557 inline void RelocationRef::getTypeName(SmallVectorImpl<char> &Result) const {
558   return OwningObject->getRelocationTypeName(RelocationPimpl, Result);
559 }
560 
561 inline DataRefImpl RelocationRef::getRawDataRefImpl() const {
562   return RelocationPimpl;
563 }
564 
565 inline const ObjectFile *RelocationRef::getObject() const {
566   return OwningObject;
567 }
568 
569 } // end namespace object
570 
571 template <> struct DenseMapInfo<object::SectionRef> {
572   static bool isEqual(const object::SectionRef &A,
573                       const object::SectionRef &B) {
574     return A == B;
575   }
576   static object::SectionRef getEmptyKey() {
577     return object::SectionRef({}, nullptr);
578   }
579   static object::SectionRef getTombstoneKey() {
580     object::DataRefImpl TS;
581     TS.p = (uintptr_t)-1;
582     return object::SectionRef(TS, nullptr);
583   }
584   static unsigned getHashValue(const object::SectionRef &Sec) {
585     object::DataRefImpl Raw = Sec.getRawDataRefImpl();
586     return hash_combine(Raw.p, Raw.d.a, Raw.d.b);
587   }
588 };
589 
590 } // end namespace llvm
591 
592 #endif // LLVM_OBJECT_OBJECTFILE_H
593