1 //===- Chunks.h -------------------------------------------------*- C++ -*-===//
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 #ifndef LLD_COFF_CHUNKS_H
11 #define LLD_COFF_CHUNKS_H
14 #include "InputFiles.h"
15 #include "lld/Core/LLVM.h"
16 #include "llvm/ADT/ArrayRef.h"
17 #include "llvm/ADT/iterator.h"
18 #include "llvm/ADT/iterator_range.h"
19 #include "llvm/Object/COFF.h"
26 using llvm::COFF::ImportDirectoryTableEntry;
27 using llvm::object::COFFSymbolRef;
28 using llvm::object::SectionRef;
29 using llvm::object::coff_relocation;
30 using llvm::object::coff_section;
34 class DefinedImportData;
40 // Mask for section types (code, data, bss, disacardable, etc.)
41 // and permissions (writable, readable or executable).
42 const uint32_t PermMask = 0xFF0000F0;
44 // A Chunk represents a chunk of data that will occupy space in the
45 // output (if the resolver chose that). It may or may not be backed by
46 // a section of an input file. It could be linker-created data, or
47 // doesn't even have actual data (if common or bss).
50 enum Kind { SectionKind, OtherKind };
51 Kind kind() const { return ChunkKind; }
52 virtual ~Chunk() = default;
54 // Returns the size of this chunk (even if this is a common or BSS.)
55 virtual size_t getSize() const = 0;
57 // Write this chunk to a mmap'ed file, assuming Buf is pointing to
58 // beginning of the file. Because this function may use RVA values
59 // of other chunks for relocations, you need to set them properly
60 // before calling this function.
61 virtual void writeTo(uint8_t *Buf) const {}
63 // The writer sets and uses the addresses.
64 uint64_t getRVA() const { return RVA; }
65 uint32_t getAlign() const { return Align; }
66 void setRVA(uint64_t V) { RVA = V; }
68 // Returns true if this has non-zero data. BSS chunks return
69 // false. If false is returned, the space occupied by this chunk
70 // will be filled with zeros.
71 virtual bool hasData() const { return true; }
73 // Returns readable/writable/executable bits.
74 virtual uint32_t getPermissions() const { return 0; }
76 // Returns the section name if this is a section chunk.
77 // It is illegal to call this function on non-section chunks.
78 virtual StringRef getSectionName() const {
79 llvm_unreachable("unimplemented getSectionName");
82 // An output section has pointers to chunks in the section, and each
83 // chunk has a back pointer to an output section.
84 void setOutputSection(OutputSection *O) { Out = O; }
85 OutputSection *getOutputSection() { return Out; }
88 // Collect all locations that contain absolute addresses for base relocations.
89 virtual void getBaserels(std::vector<Baserel> *Res) {}
91 // Returns a human-readable name of this chunk. Chunks are unnamed chunks of
92 // bytes, so this is used only for logging or debugging.
93 virtual StringRef getDebugName() { return ""; }
96 Chunk(Kind K = OtherKind) : ChunkKind(K) {}
99 // The alignment of this chunk. The writer uses the value.
102 // The RVA of this chunk in the output. The writer sets a value.
106 // The offset from beginning of the output section. The writer sets a value.
107 uint64_t OutputSectionOff = 0;
110 // The output section for this chunk.
111 OutputSection *Out = nullptr;
114 // A chunk corresponding a section of an input file.
115 class SectionChunk : public Chunk {
116 // Identical COMDAT Folding feature accesses section internal data.
120 class symbol_iterator : public llvm::iterator_adaptor_base<
121 symbol_iterator, const coff_relocation *,
122 std::random_access_iterator_tag, SymbolBody *> {
127 symbol_iterator(ObjectFile *File, const coff_relocation *I)
128 : symbol_iterator::iterator_adaptor_base(I), File(File) {}
131 symbol_iterator() = default;
133 SymbolBody *operator*() const {
134 return File->getSymbolBody(I->SymbolTableIndex);
138 SectionChunk(ObjectFile *File, const coff_section *Header);
139 static bool classof(const Chunk *C) { return C->kind() == SectionKind; }
140 size_t getSize() const override { return Header->SizeOfRawData; }
141 ArrayRef<uint8_t> getContents() const;
142 void writeTo(uint8_t *Buf) const override;
143 bool hasData() const override;
144 uint32_t getPermissions() const override;
145 StringRef getSectionName() const override { return SectionName; }
146 void getBaserels(std::vector<Baserel> *Res) override;
147 bool isCOMDAT() const;
148 void applyRelX64(uint8_t *Off, uint16_t Type, OutputSection *OS, uint64_t S,
150 void applyRelX86(uint8_t *Off, uint16_t Type, OutputSection *OS, uint64_t S,
152 void applyRelARM(uint8_t *Off, uint16_t Type, OutputSection *OS, uint64_t S,
155 // Called if the garbage collector decides to not include this chunk
156 // in a final output. It's supposed to print out a log message to stdout.
157 void printDiscardedMessage() const;
159 // Adds COMDAT associative sections to this COMDAT section. A chunk
160 // and its children are treated as a group by the garbage collector.
161 void addAssociative(SectionChunk *Child);
163 StringRef getDebugName() override;
164 void setSymbol(DefinedRegular *S) { if (!Sym) Sym = S; }
166 // Returns true if the chunk was not dropped by GC or COMDAT deduplication.
167 bool isLive() { return Live && !Discarded; }
169 // Used by the garbage collector.
171 assert(Config->DoGC && "should only mark things live from GC");
172 assert(!isLive() && "Cannot mark an already live section!");
176 // Returns true if this chunk was dropped by COMDAT deduplication.
177 bool isDiscarded() const { return Discarded; }
179 // Used by the SymbolTable when discarding unused comdat sections. This is
180 // redundant when GC is enabled, as all comdat sections will start out dead.
181 void markDiscarded() { Discarded = true; }
183 // True if this is a codeview debug info chunk. These will not be laid out in
184 // the image. Instead they will end up in the PDB, if one is requested.
185 bool isCodeView() const {
186 return SectionName == ".debug" || SectionName.startswith(".debug$");
189 // Allow iteration over the bodies of this chunk's relocated symbols.
190 llvm::iterator_range<symbol_iterator> symbols() const {
191 return llvm::make_range(symbol_iterator(File, Relocs.begin()),
192 symbol_iterator(File, Relocs.end()));
195 // Allow iteration over the associated child chunks for this section.
196 ArrayRef<SectionChunk *> children() const { return AssocChildren; }
198 // A pointer pointing to a replacement for this chunk.
199 // Initially it points to "this" object. If this chunk is merged
200 // with other chunk by ICF, it points to another chunk,
201 // and this chunk is considrered as dead.
204 // The CRC of the contents as described in the COFF spec 4.5.5.
205 // Auxiliary Format 5: Section Definitions. Used for ICF.
206 uint32_t Checksum = 0;
208 const coff_section *Header;
210 // The file that this chunk was created from.
214 StringRef SectionName;
215 std::vector<SectionChunk *> AssocChildren;
216 llvm::iterator_range<const coff_relocation *> Relocs;
219 // True if this chunk was discarded because it was a duplicate comdat section.
222 // Used by the garbage collector.
225 // Used for ICF (Identical COMDAT Folding)
226 void replace(SectionChunk *Other);
227 uint32_t Class[2] = {0, 0};
229 // Sym points to a section symbol if this is a COMDAT chunk.
230 DefinedRegular *Sym = nullptr;
233 // A chunk for common symbols. Common chunks don't have actual data.
234 class CommonChunk : public Chunk {
236 CommonChunk(const COFFSymbolRef Sym);
237 size_t getSize() const override { return Sym.getValue(); }
238 bool hasData() const override { return false; }
239 uint32_t getPermissions() const override;
240 StringRef getSectionName() const override { return ".bss"; }
243 const COFFSymbolRef Sym;
246 // A chunk for linker-created strings.
247 class StringChunk : public Chunk {
249 explicit StringChunk(StringRef S) : Str(S) {}
250 size_t getSize() const override { return Str.size() + 1; }
251 void writeTo(uint8_t *Buf) const override;
257 static const uint8_t ImportThunkX86[] = {
258 0xff, 0x25, 0x00, 0x00, 0x00, 0x00, // JMP *0x0
261 static const uint8_t ImportThunkARM[] = {
262 0x40, 0xf2, 0x00, 0x0c, // mov.w ip, #0
263 0xc0, 0xf2, 0x00, 0x0c, // mov.t ip, #0
264 0xdc, 0xf8, 0x00, 0xf0, // ldr.w pc, [ip]
268 // A chunk for DLL import jump table entry. In a final output, it's
269 // contents will be a JMP instruction to some __imp_ symbol.
270 class ImportThunkChunkX64 : public Chunk {
272 explicit ImportThunkChunkX64(Defined *S);
273 size_t getSize() const override { return sizeof(ImportThunkX86); }
274 void writeTo(uint8_t *Buf) const override;
280 class ImportThunkChunkX86 : public Chunk {
282 explicit ImportThunkChunkX86(Defined *S) : ImpSymbol(S) {}
283 size_t getSize() const override { return sizeof(ImportThunkX86); }
284 void getBaserels(std::vector<Baserel> *Res) override;
285 void writeTo(uint8_t *Buf) const override;
291 class ImportThunkChunkARM : public Chunk {
293 explicit ImportThunkChunkARM(Defined *S) : ImpSymbol(S) {}
294 size_t getSize() const override { return sizeof(ImportThunkARM); }
295 void getBaserels(std::vector<Baserel> *Res) override;
296 void writeTo(uint8_t *Buf) const override;
303 // See comments for DefinedLocalImport class.
304 class LocalImportChunk : public Chunk {
306 explicit LocalImportChunk(Defined *S) : Sym(S) {}
307 size_t getSize() const override;
308 void getBaserels(std::vector<Baserel> *Res) override;
309 void writeTo(uint8_t *Buf) const override;
316 // A chunk for SEH table which contains RVAs of safe exception handler
317 // functions. x86-only.
318 class SEHTableChunk : public Chunk {
320 explicit SEHTableChunk(std::set<Defined *> S) : Syms(std::move(S)) {}
321 size_t getSize() const override { return Syms.size() * 4; }
322 void writeTo(uint8_t *Buf) const override;
325 std::set<Defined *> Syms;
329 // This class represents a block in .reloc section.
330 // See the PE/COFF spec 5.6 for details.
331 class BaserelChunk : public Chunk {
333 BaserelChunk(uint32_t Page, Baserel *Begin, Baserel *End);
334 size_t getSize() const override { return Data.size(); }
335 void writeTo(uint8_t *Buf) const override;
338 std::vector<uint8_t> Data;
343 Baserel(uint32_t V, uint8_t Ty) : RVA(V), Type(Ty) {}
344 explicit Baserel(uint32_t V) : Baserel(V, getDefaultType()) {}
345 uint8_t getDefaultType();