1 //===- InputFiles.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 // This file contains functions to parse Mach-O object files. In this comment,
10 // we describe the Mach-O file structure and how we parse it.
11 //
12 // Mach-O is not very different from ELF or COFF. The notion of symbols,
13 // sections and relocations exists in Mach-O as it does in ELF and COFF.
14 //
15 // Perhaps the notion that is new to those who know ELF/COFF is "subsections".
16 // In ELF/COFF, sections are an atomic unit of data copied from input files to
17 // output files. When we merge or garbage-collect sections, we treat each
18 // section as an atomic unit. In Mach-O, that's not the case. Sections can
19 // consist of multiple subsections, and subsections are a unit of merging and
20 // garbage-collecting. Therefore, Mach-O's subsections are more similar to
21 // ELF/COFF's sections than Mach-O's sections are.
22 //
23 // A section can have multiple symbols. A symbol that does not have the
24 // N_ALT_ENTRY attribute indicates a beginning of a subsection. Therefore, by
25 // definition, a symbol is always present at the beginning of each subsection. A
26 // symbol with N_ALT_ENTRY attribute does not start a new subsection and can
27 // point to a middle of a subsection.
28 //
29 // The notion of subsections also affects how relocations are represented in
30 // Mach-O. All references within a section need to be explicitly represented as
31 // relocations if they refer to different subsections, because we obviously need
32 // to fix up addresses if subsections are laid out in an output file differently
33 // than they were in object files. To represent that, Mach-O relocations can
34 // refer to an unnamed location via its address. Scattered relocations (those
35 // with the R_SCATTERED bit set) always refer to unnamed locations.
36 // Non-scattered relocations refer to an unnamed location if r_extern is not set
37 // and r_symbolnum is zero.
38 //
39 // Without the above differences, I think you can use your knowledge about ELF
40 // and COFF for Mach-O.
41 //
42 //===----------------------------------------------------------------------===//
43 
44 #include "InputFiles.h"
45 #include "InputSection.h"
46 #include "OutputSection.h"
47 #include "SymbolTable.h"
48 #include "Symbols.h"
49 #include "Target.h"
50 
51 #include "lld/Common/ErrorHandler.h"
52 #include "lld/Common/Memory.h"
53 #include "llvm/BinaryFormat/MachO.h"
54 #include "llvm/Support/Endian.h"
55 #include "llvm/Support/MemoryBuffer.h"
56 
57 using namespace llvm;
58 using namespace llvm::MachO;
59 using namespace llvm::support::endian;
60 using namespace lld;
61 using namespace lld::macho;
62 
63 std::vector<InputFile *> macho::inputFiles;
64 
65 // Open a given file path and return it as a memory-mapped file.
66 Optional<MemoryBufferRef> macho::readFile(StringRef path) {
67   // Open a file.
68   auto mbOrErr = MemoryBuffer::getFile(path);
69   if (auto ec = mbOrErr.getError()) {
70     error("cannot open " + path + ": " + ec.message());
71     return None;
72   }
73 
74   std::unique_ptr<MemoryBuffer> &mb = *mbOrErr;
75   MemoryBufferRef mbref = mb->getMemBufferRef();
76   make<std::unique_ptr<MemoryBuffer>>(std::move(mb)); // take mb ownership
77 
78   // If this is a regular non-fat file, return it.
79   const char *buf = mbref.getBufferStart();
80   auto *hdr = reinterpret_cast<const MachO::fat_header *>(buf);
81   if (read32be(&hdr->magic) != MachO::FAT_MAGIC)
82     return mbref;
83 
84   // Object files and archive files may be fat files, which contains
85   // multiple real files for different CPU ISAs. Here, we search for a
86   // file that matches with the current link target and returns it as
87   // a MemoryBufferRef.
88   auto *arch = reinterpret_cast<const MachO::fat_arch *>(buf + sizeof(*hdr));
89 
90   for (uint32_t i = 0, n = read32be(&hdr->nfat_arch); i < n; ++i) {
91     if (reinterpret_cast<const char *>(arch + i + 1) >
92         buf + mbref.getBufferSize()) {
93       error(path + ": fat_arch struct extends beyond end of file");
94       return None;
95     }
96 
97     if (read32be(&arch[i].cputype) != target->cpuType ||
98         read32be(&arch[i].cpusubtype) != target->cpuSubtype)
99       continue;
100 
101     uint32_t offset = read32be(&arch[i].offset);
102     uint32_t size = read32be(&arch[i].size);
103     if (offset + size > mbref.getBufferSize())
104       error(path + ": slice extends beyond end of file");
105     return MemoryBufferRef(StringRef(buf + offset, size), path.copy(bAlloc));
106   }
107 
108   error("unable to find matching architecture in " + path);
109   return None;
110 }
111 
112 static const load_command *findCommand(const mach_header_64 *hdr,
113                                        uint32_t type) {
114   const uint8_t *p =
115       reinterpret_cast<const uint8_t *>(hdr) + sizeof(mach_header_64);
116 
117   for (uint32_t i = 0, n = hdr->ncmds; i < n; ++i) {
118     auto *cmd = reinterpret_cast<const load_command *>(p);
119     if (cmd->cmd == type)
120       return cmd;
121     p += cmd->cmdsize;
122   }
123   return nullptr;
124 }
125 
126 std::vector<InputSection *>
127 InputFile::parseSections(ArrayRef<section_64> sections) {
128   std::vector<InputSection *> ret;
129   ret.reserve(sections.size());
130 
131   auto *buf = reinterpret_cast<const uint8_t *>(mb.getBufferStart());
132 
133   for (const section_64 &sec : sections) {
134     InputSection *isec = make<InputSection>();
135     isec->file = this;
136     isec->name = StringRef(sec.sectname, strnlen(sec.sectname, 16));
137     isec->segname = StringRef(sec.segname, strnlen(sec.segname, 16));
138     isec->data = {buf + sec.offset, static_cast<size_t>(sec.size)};
139     if (sec.align >= 32)
140       error("alignment " + std::to_string(sec.align) + " of section " +
141             isec->name + " is too large");
142     else
143       isec->align = 1 << sec.align;
144     isec->flags = sec.flags;
145     ret.push_back(isec);
146   }
147 
148   return ret;
149 }
150 
151 void InputFile::parseRelocations(const section_64 &sec,
152                                  std::vector<Reloc> &relocs) {
153   auto *buf = reinterpret_cast<const uint8_t *>(mb.getBufferStart());
154   ArrayRef<any_relocation_info> relInfos(
155       reinterpret_cast<const any_relocation_info *>(buf + sec.reloff),
156       sec.nreloc);
157 
158   for (const any_relocation_info &anyRel : relInfos) {
159     Reloc r;
160     if (anyRel.r_word0 & R_SCATTERED) {
161       error("TODO: Scattered relocations not supported");
162     } else {
163       auto rel = reinterpret_cast<const relocation_info &>(anyRel);
164       r.type = rel.r_type;
165       r.offset = rel.r_address;
166       r.addend = target->getImplicitAddend(buf + sec.offset + r.offset, r.type);
167       if (rel.r_extern)
168         r.target = symbols[rel.r_symbolnum];
169       else {
170         error("TODO: Non-extern relocations are not supported");
171         continue;
172       }
173     }
174     relocs.push_back(r);
175   }
176 }
177 
178 ObjFile::ObjFile(MemoryBufferRef mb) : InputFile(ObjKind, mb) {
179   auto *buf = reinterpret_cast<const uint8_t *>(mb.getBufferStart());
180   auto *hdr = reinterpret_cast<const mach_header_64 *>(mb.getBufferStart());
181   ArrayRef<section_64> objSections;
182 
183   if (const load_command *cmd = findCommand(hdr, LC_SEGMENT_64)) {
184     auto *c = reinterpret_cast<const segment_command_64 *>(cmd);
185     objSections = ArrayRef<section_64>{
186         reinterpret_cast<const section_64 *>(c + 1), c->nsects};
187     sections = parseSections(objSections);
188   }
189 
190   // TODO: Error on missing LC_SYMTAB?
191   if (const load_command *cmd = findCommand(hdr, LC_SYMTAB)) {
192     auto *c = reinterpret_cast<const symtab_command *>(cmd);
193     const char *strtab = reinterpret_cast<const char *>(buf) + c->stroff;
194     ArrayRef<const nlist_64> nList(
195         reinterpret_cast<const nlist_64 *>(buf + c->symoff), c->nsyms);
196 
197     symbols.reserve(c->nsyms);
198 
199     for (const nlist_64 &sym : nList) {
200       StringRef name = strtab + sym.n_strx;
201 
202       // Undefined symbol
203       if (!sym.n_sect) {
204         symbols.push_back(symtab->addUndefined(name));
205         continue;
206       }
207 
208       InputSection *isec = sections[sym.n_sect - 1];
209       const section_64 &objSec = objSections[sym.n_sect - 1];
210       uint64_t value = sym.n_value - objSec.addr;
211 
212       // Global defined symbol
213       if (sym.n_type & N_EXT) {
214         symbols.push_back(symtab->addDefined(name, isec, value));
215         continue;
216       }
217 
218       // Local defined symbol
219       symbols.push_back(make<Defined>(name, isec, value));
220     }
221   }
222 
223   // The relocations may refer to the symbols, so we parse them after we have
224   // the symbols loaded.
225   if (!sections.empty()) {
226     auto it = sections.begin();
227     for (const section_64 &sec : objSections) {
228       parseRelocations(sec, (*it)->relocs);
229       ++it;
230     }
231   }
232 }
233 
234 DylibFile::DylibFile(MemoryBufferRef mb) : InputFile(DylibKind, mb) {
235   auto *buf = reinterpret_cast<const uint8_t *>(mb.getBufferStart());
236   auto *hdr = reinterpret_cast<const mach_header_64 *>(mb.getBufferStart());
237 
238   // Initialize dylibName.
239   if (const load_command *cmd = findCommand(hdr, LC_ID_DYLIB)) {
240     auto *c = reinterpret_cast<const dylib_command *>(cmd);
241     dylibName = reinterpret_cast<const char *>(cmd) + read32le(&c->dylib.name);
242   } else {
243     error("dylib " + getName() + " missing LC_ID_DYLIB load command");
244     return;
245   }
246 
247   // Initialize symbols.
248   if (const load_command *cmd = findCommand(hdr, LC_SYMTAB)) {
249     auto *c = reinterpret_cast<const symtab_command *>(cmd);
250     const char *strtab = reinterpret_cast<const char *>(buf + c->stroff);
251     ArrayRef<const nlist_64> nList(
252         reinterpret_cast<const nlist_64 *>(buf + c->symoff), c->nsyms);
253 
254     symbols.reserve(c->nsyms);
255 
256     for (const nlist_64 &sym : nList) {
257       StringRef name = strtab + sym.n_strx;
258       // TODO: Figure out what to do about undefined symbols: ignore or warn
259       // if unsatisfied? Also make sure we handle re-exported symbols
260       // correctly.
261       symbols.push_back(symtab->addDylib(name, this));
262     }
263   }
264 }
265 
266 // Returns "<internal>" or "baz.o".
267 std::string lld::toString(const InputFile *file) {
268   return file ? std::string(file->getName()) : "<internal>";
269 }
270