1 //===- InputSection.cpp ---------------------------------------------------===//
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 "InputSection.h"
10 #include "ConcatOutputSection.h"
11 #include "Config.h"
12 #include "InputFiles.h"
13 #include "OutputSegment.h"
14 #include "Symbols.h"
15 #include "SyntheticSections.h"
16 #include "Target.h"
17 #include "UnwindInfoSection.h"
18 #include "Writer.h"
19 #include "lld/Common/Memory.h"
20 #include "llvm/Support/Endian.h"
21 #include "llvm/Support/xxhash.h"
22
23 using namespace llvm;
24 using namespace llvm::MachO;
25 using namespace llvm::support;
26 using namespace lld;
27 using namespace lld::macho;
28
29 std::vector<ConcatInputSection *> macho::inputSections;
30
getFileSize() const31 uint64_t InputSection::getFileSize() const {
32 return isZeroFill(getFlags()) ? 0 : getSize();
33 }
34
getVA(uint64_t off) const35 uint64_t InputSection::getVA(uint64_t off) const {
36 return parent->addr + getOffset(off);
37 }
38
resolveSymbolVA(const Symbol * sym,uint8_t type)39 static uint64_t resolveSymbolVA(const Symbol *sym, uint8_t type) {
40 const RelocAttrs &relocAttrs = target->getRelocAttrs(type);
41 if (relocAttrs.hasAttr(RelocAttrBits::BRANCH))
42 return sym->resolveBranchVA();
43 if (relocAttrs.hasAttr(RelocAttrBits::GOT))
44 return sym->resolveGotVA();
45 if (relocAttrs.hasAttr(RelocAttrBits::TLV))
46 return sym->resolveTlvVA();
47 return sym->getVA();
48 }
49
50 // ICF needs to hash any section that might potentially be duplicated so
51 // that it can match on content rather than identity.
isHashableForICF() const52 bool ConcatInputSection::isHashableForICF() const {
53 switch (sectionType(getFlags())) {
54 case S_REGULAR:
55 return true;
56 case S_CSTRING_LITERALS:
57 case S_4BYTE_LITERALS:
58 case S_8BYTE_LITERALS:
59 case S_16BYTE_LITERALS:
60 case S_LITERAL_POINTERS:
61 llvm_unreachable("found unexpected literal type in ConcatInputSection");
62 case S_ZEROFILL:
63 case S_GB_ZEROFILL:
64 case S_NON_LAZY_SYMBOL_POINTERS:
65 case S_LAZY_SYMBOL_POINTERS:
66 case S_SYMBOL_STUBS:
67 case S_MOD_INIT_FUNC_POINTERS:
68 case S_MOD_TERM_FUNC_POINTERS:
69 case S_COALESCED:
70 case S_INTERPOSING:
71 case S_DTRACE_DOF:
72 case S_LAZY_DYLIB_SYMBOL_POINTERS:
73 case S_THREAD_LOCAL_REGULAR:
74 case S_THREAD_LOCAL_ZEROFILL:
75 case S_THREAD_LOCAL_VARIABLES:
76 case S_THREAD_LOCAL_VARIABLE_POINTERS:
77 case S_THREAD_LOCAL_INIT_FUNCTION_POINTERS:
78 return false;
79 default:
80 llvm_unreachable("Section type");
81 }
82 }
83
hashForICF()84 void ConcatInputSection::hashForICF() {
85 assert(data.data()); // zeroFill section data has nullptr with non-zero size
86 assert(icfEqClass[0] == 0); // don't overwrite a unique ID!
87 // Turn-on the top bit to guarantee that valid hashes have no collisions
88 // with the small-integer unique IDs for ICF-ineligible sections
89 icfEqClass[0] = xxHash64(data) | (1ull << 63);
90 }
91
foldIdentical(ConcatInputSection * copy)92 void ConcatInputSection::foldIdentical(ConcatInputSection *copy) {
93 align = std::max(align, copy->align);
94 copy->live = false;
95 copy->wasCoalesced = true;
96 numRefs += copy->numRefs;
97 copy->numRefs = 0;
98 copy->replacement = this;
99 }
100
writeTo(uint8_t * buf)101 void ConcatInputSection::writeTo(uint8_t *buf) {
102 assert(!shouldOmitFromOutput());
103
104 if (getFileSize() == 0)
105 return;
106
107 memcpy(buf, data.data(), data.size());
108
109 for (size_t i = 0; i < relocs.size(); i++) {
110 const Reloc &r = relocs[i];
111 uint8_t *loc = buf + r.offset;
112 uint64_t referentVA = 0;
113 if (target->hasAttr(r.type, RelocAttrBits::SUBTRAHEND)) {
114 const Symbol *fromSym = r.referent.get<Symbol *>();
115 const Reloc &minuend = relocs[++i];
116 uint64_t minuendVA;
117 if (const Symbol *toSym = minuend.referent.dyn_cast<Symbol *>())
118 minuendVA = toSym->getVA() + minuend.addend;
119 else {
120 auto *referentIsec = minuend.referent.get<InputSection *>();
121 assert(!::shouldOmitFromOutput(referentIsec));
122 minuendVA = referentIsec->getVA(minuend.addend);
123 }
124 referentVA = minuendVA - fromSym->getVA();
125 } else if (auto *referentSym = r.referent.dyn_cast<Symbol *>()) {
126 if (target->hasAttr(r.type, RelocAttrBits::LOAD) &&
127 !referentSym->isInGot())
128 target->relaxGotLoad(loc, r.type);
129 referentVA = resolveSymbolVA(referentSym, r.type) + r.addend;
130
131 if (isThreadLocalVariables(getFlags())) {
132 // References from thread-local variable sections are treated as offsets
133 // relative to the start of the thread-local data memory area, which
134 // is initialized via copying all the TLV data sections (which are all
135 // contiguous).
136 if (isa<Defined>(referentSym))
137 referentVA -= firstTLVDataSection->addr;
138 }
139 } else if (auto *referentIsec = r.referent.dyn_cast<InputSection *>()) {
140 assert(!::shouldOmitFromOutput(referentIsec));
141 referentVA = referentIsec->getVA(r.addend);
142 }
143 target->relocateOne(loc, r, referentVA, getVA() + r.offset);
144 }
145 }
146
splitIntoPieces()147 void CStringInputSection::splitIntoPieces() {
148 size_t off = 0;
149 StringRef s = toStringRef(data);
150 while (!s.empty()) {
151 size_t end = s.find(0);
152 if (end == StringRef::npos)
153 fatal(toString(this) + ": string is not null terminated");
154 size_t size = end + 1;
155 uint32_t hash = config->dedupLiterals ? xxHash64(s.substr(0, size)) : 0;
156 pieces.emplace_back(off, hash);
157 s = s.substr(size);
158 off += size;
159 }
160 }
161
getStringPiece(uint64_t off)162 StringPiece &CStringInputSection::getStringPiece(uint64_t off) {
163 if (off >= data.size())
164 fatal(toString(this) + ": offset is outside the section");
165
166 auto it =
167 partition_point(pieces, [=](StringPiece p) { return p.inSecOff <= off; });
168 return it[-1];
169 }
170
getStringPiece(uint64_t off) const171 const StringPiece &CStringInputSection::getStringPiece(uint64_t off) const {
172 return const_cast<CStringInputSection *>(this)->getStringPiece(off);
173 }
174
getOffset(uint64_t off) const175 uint64_t CStringInputSection::getOffset(uint64_t off) const {
176 const StringPiece &piece = getStringPiece(off);
177 uint64_t addend = off - piece.inSecOff;
178 return piece.outSecOff + addend;
179 }
180
WordLiteralInputSection(StringRef segname,StringRef name,InputFile * file,ArrayRef<uint8_t> data,uint32_t align,uint32_t flags)181 WordLiteralInputSection::WordLiteralInputSection(StringRef segname,
182 StringRef name,
183 InputFile *file,
184 ArrayRef<uint8_t> data,
185 uint32_t align, uint32_t flags)
186 : InputSection(WordLiteralKind, segname, name, file, data, align, flags) {
187 switch (sectionType(flags)) {
188 case S_4BYTE_LITERALS:
189 power2LiteralSize = 2;
190 break;
191 case S_8BYTE_LITERALS:
192 power2LiteralSize = 3;
193 break;
194 case S_16BYTE_LITERALS:
195 power2LiteralSize = 4;
196 break;
197 default:
198 llvm_unreachable("invalid literal section type");
199 }
200
201 live.resize(data.size() >> power2LiteralSize, !config->deadStrip);
202 }
203
getOffset(uint64_t off) const204 uint64_t WordLiteralInputSection::getOffset(uint64_t off) const {
205 auto *osec = cast<WordLiteralSection>(parent);
206 const uint8_t *buf = data.data();
207 switch (sectionType(getFlags())) {
208 case S_4BYTE_LITERALS:
209 return osec->getLiteral4Offset(buf + off);
210 case S_8BYTE_LITERALS:
211 return osec->getLiteral8Offset(buf + off);
212 case S_16BYTE_LITERALS:
213 return osec->getLiteral16Offset(buf + off);
214 default:
215 llvm_unreachable("invalid literal section type");
216 }
217 }
218
isCodeSection(const InputSection * isec)219 bool macho::isCodeSection(const InputSection *isec) {
220 uint32_t type = sectionType(isec->getFlags());
221 if (type != S_REGULAR && type != S_COALESCED)
222 return false;
223
224 uint32_t attr = isec->getFlags() & SECTION_ATTRIBUTES_USR;
225 if (attr == S_ATTR_PURE_INSTRUCTIONS)
226 return true;
227
228 if (isec->getSegName() == segment_names::text)
229 return StringSwitch<bool>(isec->getName())
230 .Cases(section_names::textCoalNt, section_names::staticInit, true)
231 .Default(false);
232
233 return false;
234 }
235
isCfStringSection(const InputSection * isec)236 bool macho::isCfStringSection(const InputSection *isec) {
237 return isec->getName() == section_names::cfString &&
238 isec->getSegName() == segment_names::data;
239 }
240
toString(const InputSection * isec)241 std::string lld::toString(const InputSection *isec) {
242 return (toString(isec->getFile()) + ":(" + isec->getName() + ")").str();
243 }
244