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 "OutputSegment.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   return mbref;
78 }
79 
80 static const load_command *findCommand(const mach_header_64 *hdr,
81                                        uint32_t type) {
82   const uint8_t *p =
83       reinterpret_cast<const uint8_t *>(hdr) + sizeof(mach_header_64);
84 
85   for (uint32_t i = 0, n = hdr->ncmds; i < n; ++i) {
86     auto *cmd = reinterpret_cast<const load_command *>(p);
87     if (cmd->cmd == type)
88       return cmd;
89     p += cmd->cmdsize;
90   }
91   return nullptr;
92 }
93 
94 std::vector<InputSection *>
95 InputFile::parseSections(ArrayRef<section_64> sections) {
96   std::vector<InputSection *> ret;
97   ret.reserve(sections.size());
98 
99   auto *buf = reinterpret_cast<const uint8_t *>(mb.getBufferStart());
100 
101   for (const section_64 &sec : sections) {
102     InputSection *isec = make<InputSection>();
103     isec->file = this;
104     isec->name = StringRef(sec.sectname, strnlen(sec.sectname, 16));
105     isec->segname = StringRef(sec.segname, strnlen(sec.segname, 16));
106     isec->data = {buf + sec.offset, static_cast<size_t>(sec.size)};
107     if (sec.align >= 32)
108       error("alignment " + std::to_string(sec.align) + " of section " +
109             isec->name + " is too large");
110     else
111       isec->align = 1 << sec.align;
112     isec->flags = sec.flags;
113     ret.push_back(isec);
114   }
115 
116   return ret;
117 }
118 
119 void InputFile::parseRelocations(const section_64 &sec,
120                                  std::vector<Reloc> &relocs) {
121   auto *buf = reinterpret_cast<const uint8_t *>(mb.getBufferStart());
122   ArrayRef<any_relocation_info> relInfos(
123       reinterpret_cast<const any_relocation_info *>(buf + sec.reloff),
124       sec.nreloc);
125 
126   for (const any_relocation_info &anyRel : relInfos) {
127     Reloc r;
128     if (anyRel.r_word0 & R_SCATTERED) {
129       error("TODO: Scattered relocations not supported");
130     } else {
131       auto rel = reinterpret_cast<const relocation_info &>(anyRel);
132       r.type = rel.r_type;
133       r.offset = rel.r_address;
134       r.addend = target->getImplicitAddend(buf + sec.offset + r.offset, r.type);
135       if (rel.r_extern)
136         r.target = symbols[rel.r_symbolnum];
137       else {
138         error("TODO: Non-extern relocations are not supported");
139         continue;
140       }
141     }
142     relocs.push_back(r);
143   }
144 }
145 
146 ObjFile::ObjFile(MemoryBufferRef mb) : InputFile(ObjKind, mb) {
147   auto *buf = reinterpret_cast<const uint8_t *>(mb.getBufferStart());
148   auto *hdr = reinterpret_cast<const mach_header_64 *>(mb.getBufferStart());
149   ArrayRef<section_64> objSections;
150 
151   if (const load_command *cmd = findCommand(hdr, LC_SEGMENT_64)) {
152     auto *c = reinterpret_cast<const segment_command_64 *>(cmd);
153     objSections = ArrayRef<section_64>{
154         reinterpret_cast<const section_64 *>(c + 1), c->nsects};
155     sections = parseSections(objSections);
156   }
157 
158   if (const load_command *cmd = findCommand(hdr, LC_SYMTAB)) {
159     auto *c = reinterpret_cast<const symtab_command *>(cmd);
160     const char *strtab = reinterpret_cast<const char *>(buf) + c->stroff;
161     ArrayRef<const nlist_64> nList(
162         reinterpret_cast<const nlist_64 *>(buf + c->symoff), c->nsyms);
163 
164     symbols.reserve(c->nsyms);
165 
166     for (const nlist_64 &sym : nList) {
167       StringRef name = strtab + sym.n_strx;
168 
169       // Undefined symbol
170       if (!sym.n_sect) {
171         error("TODO: Support undefined symbols");
172         continue;
173       }
174 
175       InputSection *isec = sections[sym.n_sect - 1];
176       const section_64 &objSec = objSections[sym.n_sect - 1];
177       uint64_t value = sym.n_value - objSec.addr;
178 
179       // Global defined symbol
180       if (sym.n_type & N_EXT) {
181         symbols.push_back(symtab->addDefined(name, isec, value));
182         continue;
183       }
184 
185       // Local defined symbol
186       symbols.push_back(make<Defined>(name, isec, value));
187     }
188   }
189 
190   // The relocations may refer to the symbols, so we parse them after we have
191   // the symbols loaded.
192   if (!sections.empty()) {
193     auto it = sections.begin();
194     for (const section_64 &sec : objSections) {
195       parseRelocations(sec, (*it)->relocs);
196       ++it;
197     }
198   }
199 }
200 
201 // Returns "<internal>" or "baz.o".
202 std::string lld::toString(const InputFile *file) {
203   return file ? std::string(file->getName()) : "<internal>";
204 }
205