blob: 042ac2dd8a5df29d786fdb64677bcdc97fe88613 [file] [log] [blame]
Benjamin Kramer43a772e2011-09-19 17:56:04 +00001//===-- MachODump.cpp - Object file dumping utility for llvm --------------===//
2//
3// The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// This file implements the MachO-specific dumper for llvm-objdump.
11//
12//===----------------------------------------------------------------------===//
13
14#include "llvm-objdump.h"
Kevin Enderby98c9acc2014-09-16 18:00:57 +000015#include "llvm-c/Disassembler.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000016#include "llvm/ADT/STLExtras.h"
Ahmed Bougachaaa790682013-05-24 01:07:04 +000017#include "llvm/ADT/StringExtras.h"
Chandler Carruth4d88a1c2012-12-04 10:44:52 +000018#include "llvm/ADT/Triple.h"
Kevin Enderby04bf6932014-10-28 23:39:46 +000019#include "llvm/Config/config.h"
Zachary Turner82af9432015-01-30 18:07:45 +000020#include "llvm/DebugInfo/DWARF/DIContext.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000021#include "llvm/MC/MCAsmInfo.h"
Lang Hamesa1bc0f52014-04-15 04:40:56 +000022#include "llvm/MC/MCContext.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000023#include "llvm/MC/MCDisassembler.h"
24#include "llvm/MC/MCInst.h"
25#include "llvm/MC/MCInstPrinter.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000026#include "llvm/MC/MCInstrDesc.h"
27#include "llvm/MC/MCInstrInfo.h"
Jim Grosbachfd93a592012-03-05 19:33:20 +000028#include "llvm/MC/MCRegisterInfo.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000029#include "llvm/MC/MCSubtargetInfo.h"
Chandler Carruth4d88a1c2012-12-04 10:44:52 +000030#include "llvm/Object/MachO.h"
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +000031#include "llvm/Object/MachOUniversal.h"
Rafael Espindola9b709252013-04-13 01:45:40 +000032#include "llvm/Support/Casting.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000033#include "llvm/Support/CommandLine.h"
34#include "llvm/Support/Debug.h"
Tim Northover4bd286a2014-08-01 13:07:19 +000035#include "llvm/Support/Endian.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000036#include "llvm/Support/Format.h"
Chandler Carruthd9903882015-01-14 11:23:27 +000037#include "llvm/Support/FormattedStream.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000038#include "llvm/Support/GraphWriter.h"
Kevin Enderby9a509442015-01-27 21:28:24 +000039#include "llvm/Support/LEB128.h"
Chandler Carruth4d88a1c2012-12-04 10:44:52 +000040#include "llvm/Support/MachO.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000041#include "llvm/Support/MemoryBuffer.h"
42#include "llvm/Support/TargetRegistry.h"
43#include "llvm/Support/TargetSelect.h"
44#include "llvm/Support/raw_ostream.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000045#include <algorithm>
46#include <cstring>
Rafael Espindolaa6e9c3e2014-06-12 17:38:55 +000047#include <system_error>
Kevin Enderby04bf6932014-10-28 23:39:46 +000048
49#if HAVE_CXXABI_H
50#include <cxxabi.h>
51#endif
52
Benjamin Kramer43a772e2011-09-19 17:56:04 +000053using namespace llvm;
54using namespace object;
55
56static cl::opt<bool>
Kevin Enderbyb28ed012014-10-29 21:28:24 +000057 UseDbg("g",
58 cl::desc("Print line information from debug info if available"));
Benjamin Kramer699128e2011-09-21 01:13:19 +000059
Kevin Enderbyb28ed012014-10-29 21:28:24 +000060static cl::opt<std::string> DSYMFile("dsym",
61 cl::desc("Use .dSYM file for debug info"));
Benjamin Kramer699128e2011-09-21 01:13:19 +000062
Kevin Enderbyb28ed012014-10-29 21:28:24 +000063static cl::opt<bool> FullLeadingAddr("full-leading-addr",
64 cl::desc("Print full leading address"));
Kevin Enderbybf246f52014-09-24 23:08:22 +000065
66static cl::opt<bool>
67 PrintImmHex("print-imm-hex",
68 cl::desc("Use hex format for immediate values"));
69
Kevin Enderby13023a12015-01-15 23:19:11 +000070cl::opt<bool> llvm::UniversalHeaders("universal-headers",
71 cl::desc("Print Mach-O universal headers "
72 "(requires -macho)"));
73
Kevin Enderby131d1772015-01-09 19:22:37 +000074cl::opt<bool>
Kevin Enderby13023a12015-01-15 23:19:11 +000075 llvm::ArchiveHeaders("archive-headers",
76 cl::desc("Print archive headers for Mach-O archives "
77 "(requires -macho)"));
Kevin Enderby131d1772015-01-09 19:22:37 +000078
Kevin Enderbya7bdc7e2015-01-22 18:55:27 +000079cl::opt<bool>
80 llvm::IndirectSymbols("indirect-symbols",
81 cl::desc("Print indirect symbol table for Mach-O "
82 "objects (requires -macho)"));
83
Kevin Enderby69fe98d2015-01-23 18:52:17 +000084cl::opt<bool>
85 llvm::DataInCode("data-in-code",
86 cl::desc("Print the data in code table for Mach-O objects "
87 "(requires -macho)"));
88
Kevin Enderby9a509442015-01-27 21:28:24 +000089cl::opt<bool>
90 llvm::LinkOptHints("link-opt-hints",
91 cl::desc("Print the linker optimization hints for "
92 "Mach-O objects (requires -macho)"));
93
Kevin Enderbyf6d25852015-01-31 00:37:11 +000094cl::list<std::string>
95 llvm::DumpSections("section",
96 cl::desc("Prints the specified segment,section for "
97 "Mach-O objects (requires -macho)"));
98
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +000099static cl::list<std::string>
100 ArchFlags("arch", cl::desc("architecture(s) from a Mach-O file to dump"),
101 cl::ZeroOrMore);
102bool ArchAll = false;
103
Kevin Enderbyec5ca032014-08-18 20:21:02 +0000104static std::string ThumbTripleName;
105
106static const Target *GetTarget(const MachOObjectFile *MachOObj,
107 const char **McpuDefault,
108 const Target **ThumbTarget) {
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000109 // Figure out the target triple.
Cameron Zwarich88cc16a2012-02-03 06:35:22 +0000110 if (TripleName.empty()) {
111 llvm::Triple TT("unknown-unknown-unknown");
Kevin Enderbyec5ca032014-08-18 20:21:02 +0000112 llvm::Triple ThumbTriple = Triple();
113 TT = MachOObj->getArch(McpuDefault, &ThumbTriple);
Cameron Zwarich88cc16a2012-02-03 06:35:22 +0000114 TripleName = TT.str();
Kevin Enderbyec5ca032014-08-18 20:21:02 +0000115 ThumbTripleName = ThumbTriple.str();
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000116 }
117
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000118 // Get the target specific parser.
119 std::string Error;
120 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
Kevin Enderbyec5ca032014-08-18 20:21:02 +0000121 if (TheTarget && ThumbTripleName.empty())
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000122 return TheTarget;
123
Kevin Enderbyec5ca032014-08-18 20:21:02 +0000124 *ThumbTarget = TargetRegistry::lookupTarget(ThumbTripleName, Error);
125 if (*ThumbTarget)
126 return TheTarget;
127
128 errs() << "llvm-objdump: error: unable to get target for '";
129 if (!TheTarget)
130 errs() << TripleName;
131 else
132 errs() << ThumbTripleName;
133 errs() << "', see --version and --triple.\n";
Craig Toppere6cb63e2014-04-25 04:24:47 +0000134 return nullptr;
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000135}
136
Owen Andersond9243c42011-10-17 21:37:35 +0000137struct SymbolSorter {
138 bool operator()(const SymbolRef &A, const SymbolRef &B) {
139 SymbolRef::Type AType, BType;
140 A.getType(AType);
141 B.getType(BType);
142
143 uint64_t AAddr, BAddr;
144 if (AType != SymbolRef::ST_Function)
145 AAddr = 0;
146 else
147 A.getAddress(AAddr);
148 if (BType != SymbolRef::ST_Function)
149 BAddr = 0;
150 else
151 B.getAddress(BAddr);
152 return AAddr < BAddr;
153 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000154};
155
Kevin Enderby273ae012013-06-06 17:20:50 +0000156// Types for the storted data in code table that is built before disassembly
157// and the predicate function to sort them.
158typedef std::pair<uint64_t, DiceRef> DiceTableEntry;
159typedef std::vector<DiceTableEntry> DiceTable;
160typedef DiceTable::iterator dice_table_iterator;
161
Kevin Enderby930fdc72014-11-06 19:00:13 +0000162// This is used to search for a data in code table entry for the PC being
163// disassembled. The j parameter has the PC in j.first. A single data in code
164// table entry can cover many bytes for each of its Kind's. So if the offset,
165// aka the i.first value, of the data in code table entry plus its Length
166// covers the PC being searched for this will return true. If not it will
167// return false.
David Majnemerea9b8ee2014-11-04 08:41:48 +0000168static bool compareDiceTableEntries(const DiceTableEntry &i,
169 const DiceTableEntry &j) {
Kevin Enderby930fdc72014-11-06 19:00:13 +0000170 uint16_t Length;
171 i.second.getLength(Length);
172
173 return j.first >= i.first && j.first < i.first + Length;
Kevin Enderby273ae012013-06-06 17:20:50 +0000174}
175
Kevin Enderby930fdc72014-11-06 19:00:13 +0000176static uint64_t DumpDataInCode(const char *bytes, uint64_t Length,
177 unsigned short Kind) {
178 uint32_t Value, Size = 1;
Kevin Enderby273ae012013-06-06 17:20:50 +0000179
180 switch (Kind) {
Kevin Enderby930fdc72014-11-06 19:00:13 +0000181 default:
Charles Davis8bdfafd2013-09-01 04:28:48 +0000182 case MachO::DICE_KIND_DATA:
Kevin Enderby930fdc72014-11-06 19:00:13 +0000183 if (Length >= 4) {
184 if (!NoShowRawInsn)
185 DumpBytes(StringRef(bytes, 4));
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000186 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
Kevin Enderby273ae012013-06-06 17:20:50 +0000187 outs() << "\t.long " << Value;
Kevin Enderby930fdc72014-11-06 19:00:13 +0000188 Size = 4;
189 } else if (Length >= 2) {
190 if (!NoShowRawInsn)
191 DumpBytes(StringRef(bytes, 2));
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000192 Value = bytes[1] << 8 | bytes[0];
Kevin Enderby273ae012013-06-06 17:20:50 +0000193 outs() << "\t.short " << Value;
Kevin Enderby930fdc72014-11-06 19:00:13 +0000194 Size = 2;
195 } else {
196 if (!NoShowRawInsn)
197 DumpBytes(StringRef(bytes, 2));
Kevin Enderby273ae012013-06-06 17:20:50 +0000198 Value = bytes[0];
199 outs() << "\t.byte " << Value;
Kevin Enderby930fdc72014-11-06 19:00:13 +0000200 Size = 1;
Kevin Enderby273ae012013-06-06 17:20:50 +0000201 }
Kevin Enderby930fdc72014-11-06 19:00:13 +0000202 if (Kind == MachO::DICE_KIND_DATA)
203 outs() << "\t@ KIND_DATA\n";
204 else
205 outs() << "\t@ data in code kind = " << Kind << "\n";
Kevin Enderby273ae012013-06-06 17:20:50 +0000206 break;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000207 case MachO::DICE_KIND_JUMP_TABLE8:
Kevin Enderby930fdc72014-11-06 19:00:13 +0000208 if (!NoShowRawInsn)
209 DumpBytes(StringRef(bytes, 1));
Kevin Enderby273ae012013-06-06 17:20:50 +0000210 Value = bytes[0];
Kevin Enderby930fdc72014-11-06 19:00:13 +0000211 outs() << "\t.byte " << format("%3u", Value) << "\t@ KIND_JUMP_TABLE8\n";
212 Size = 1;
Kevin Enderby273ae012013-06-06 17:20:50 +0000213 break;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000214 case MachO::DICE_KIND_JUMP_TABLE16:
Kevin Enderby930fdc72014-11-06 19:00:13 +0000215 if (!NoShowRawInsn)
216 DumpBytes(StringRef(bytes, 2));
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000217 Value = bytes[1] << 8 | bytes[0];
Kevin Enderby930fdc72014-11-06 19:00:13 +0000218 outs() << "\t.short " << format("%5u", Value & 0xffff)
219 << "\t@ KIND_JUMP_TABLE16\n";
220 Size = 2;
Kevin Enderby273ae012013-06-06 17:20:50 +0000221 break;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000222 case MachO::DICE_KIND_JUMP_TABLE32:
Kevin Enderby930fdc72014-11-06 19:00:13 +0000223 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
224 if (!NoShowRawInsn)
225 DumpBytes(StringRef(bytes, 4));
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000226 Value = bytes[3] << 24 | bytes[2] << 16 | bytes[1] << 8 | bytes[0];
Kevin Enderby930fdc72014-11-06 19:00:13 +0000227 outs() << "\t.long " << Value;
228 if (Kind == MachO::DICE_KIND_JUMP_TABLE32)
229 outs() << "\t@ KIND_JUMP_TABLE32\n";
230 else
231 outs() << "\t@ KIND_ABS_JUMP_TABLE32\n";
232 Size = 4;
Kevin Enderby273ae012013-06-06 17:20:50 +0000233 break;
234 }
Kevin Enderby930fdc72014-11-06 19:00:13 +0000235 return Size;
Kevin Enderby273ae012013-06-06 17:20:50 +0000236}
237
Alexey Samsonov464d2e42014-03-17 07:28:19 +0000238static void getSectionsAndSymbols(const MachO::mach_header Header,
239 MachOObjectFile *MachOObj,
240 std::vector<SectionRef> &Sections,
241 std::vector<SymbolRef> &Symbols,
242 SmallVectorImpl<uint64_t> &FoundFns,
243 uint64_t &BaseSegmentAddress) {
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +0000244 for (const SymbolRef &Symbol : MachOObj->symbols()) {
245 StringRef SymName;
246 Symbol.getName(SymName);
247 if (!SymName.startswith("ltmp"))
248 Symbols.push_back(Symbol);
249 }
Owen Andersond9243c42011-10-17 21:37:35 +0000250
Alexey Samsonov48803e52014-03-13 14:37:36 +0000251 for (const SectionRef &Section : MachOObj->sections()) {
Owen Andersond9243c42011-10-17 21:37:35 +0000252 StringRef SectName;
Alexey Samsonov48803e52014-03-13 14:37:36 +0000253 Section.getName(SectName);
254 Sections.push_back(Section);
Owen Andersond9243c42011-10-17 21:37:35 +0000255 }
256
Rafael Espindola56f976f2013-04-18 18:08:55 +0000257 MachOObjectFile::LoadCommandInfo Command =
Alexey Samsonov48803e52014-03-13 14:37:36 +0000258 MachOObj->getFirstLoadCommandInfo();
Kevin Enderby273ae012013-06-06 17:20:50 +0000259 bool BaseSegmentAddressSet = false;
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000260 for (unsigned i = 0;; ++i) {
Charles Davis8bdfafd2013-09-01 04:28:48 +0000261 if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) {
Benjamin Kramer699128e2011-09-21 01:13:19 +0000262 // We found a function starts segment, parse the addresses for later
263 // consumption.
Charles Davis8bdfafd2013-09-01 04:28:48 +0000264 MachO::linkedit_data_command LLC =
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000265 MachOObj->getLinkeditDataLoadCommand(Command);
Benjamin Kramer699128e2011-09-21 01:13:19 +0000266
Charles Davis8bdfafd2013-09-01 04:28:48 +0000267 MachOObj->ReadULEB128s(LLC.dataoff, FoundFns);
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000268 } else if (Command.C.cmd == MachO::LC_SEGMENT) {
269 MachO::segment_command SLC = MachOObj->getSegmentLoadCommand(Command);
Charles Davis8bdfafd2013-09-01 04:28:48 +0000270 StringRef SegName = SLC.segname;
Kevin Enderbyb28ed012014-10-29 21:28:24 +0000271 if (!BaseSegmentAddressSet && SegName != "__PAGEZERO") {
Kevin Enderby273ae012013-06-06 17:20:50 +0000272 BaseSegmentAddressSet = true;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000273 BaseSegmentAddress = SLC.vmaddr;
Kevin Enderby273ae012013-06-06 17:20:50 +0000274 }
275 }
Rafael Espindolafeef8c22013-04-19 11:36:47 +0000276
Charles Davis8bdfafd2013-09-01 04:28:48 +0000277 if (i == Header.ncmds - 1)
Rafael Espindolafeef8c22013-04-19 11:36:47 +0000278 break;
279 else
280 Command = MachOObj->getNextLoadCommandInfo(Command);
Benjamin Kramer8a529dc2011-09-21 22:16:43 +0000281 }
Benjamin Kramer699128e2011-09-21 01:13:19 +0000282}
283
Kevin Enderbya7bdc7e2015-01-22 18:55:27 +0000284static void PrintIndirectSymbolTable(MachOObjectFile *O, bool verbose,
285 uint32_t n, uint32_t count,
286 uint32_t stride, uint64_t addr) {
287 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
288 uint32_t nindirectsyms = Dysymtab.nindirectsyms;
289 if (n > nindirectsyms)
290 outs() << " (entries start past the end of the indirect symbol "
291 "table) (reserved1 field greater than the table size)";
292 else if (n + count > nindirectsyms)
293 outs() << " (entries extends past the end of the indirect symbol "
294 "table)";
295 outs() << "\n";
296 uint32_t cputype = O->getHeader().cputype;
297 if (cputype & MachO::CPU_ARCH_ABI64)
298 outs() << "address index";
299 else
300 outs() << "address index";
301 if (verbose)
302 outs() << " name\n";
303 else
304 outs() << "\n";
305 for (uint32_t j = 0; j < count && n + j < nindirectsyms; j++) {
306 if (cputype & MachO::CPU_ARCH_ABI64)
307 outs() << format("0x%016" PRIx64, addr + j * stride) << " ";
308 else
309 outs() << format("0x%08" PRIx32, addr + j * stride) << " ";
310 MachO::dysymtab_command Dysymtab = O->getDysymtabLoadCommand();
311 uint32_t indirect_symbol = O->getIndirectSymbolTableEntry(Dysymtab, n + j);
312 if (indirect_symbol == MachO::INDIRECT_SYMBOL_LOCAL) {
313 outs() << "LOCAL\n";
314 continue;
315 }
316 if (indirect_symbol ==
317 (MachO::INDIRECT_SYMBOL_LOCAL | MachO::INDIRECT_SYMBOL_ABS)) {
318 outs() << "LOCAL ABSOLUTE\n";
319 continue;
320 }
321 if (indirect_symbol == MachO::INDIRECT_SYMBOL_ABS) {
322 outs() << "ABSOLUTE\n";
323 continue;
324 }
325 outs() << format("%5u ", indirect_symbol);
326 MachO::symtab_command Symtab = O->getSymtabLoadCommand();
327 if (indirect_symbol < Symtab.nsyms) {
328 symbol_iterator Sym = O->getSymbolByIndex(indirect_symbol);
329 SymbolRef Symbol = *Sym;
330 StringRef SymName;
331 Symbol.getName(SymName);
332 outs() << SymName;
333 } else {
334 outs() << "?";
335 }
336 outs() << "\n";
337 }
338}
339
340static void PrintIndirectSymbols(MachOObjectFile *O, bool verbose) {
341 uint32_t LoadCommandCount = O->getHeader().ncmds;
342 MachOObjectFile::LoadCommandInfo Load = O->getFirstLoadCommandInfo();
343 for (unsigned I = 0;; ++I) {
344 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
345 MachO::segment_command_64 Seg = O->getSegment64LoadCommand(Load);
346 for (unsigned J = 0; J < Seg.nsects; ++J) {
347 MachO::section_64 Sec = O->getSection64(Load, J);
348 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
349 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
350 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
351 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
352 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
353 section_type == MachO::S_SYMBOL_STUBS) {
354 uint32_t stride;
355 if (section_type == MachO::S_SYMBOL_STUBS)
356 stride = Sec.reserved2;
357 else
358 stride = 8;
359 if (stride == 0) {
360 outs() << "Can't print indirect symbols for (" << Sec.segname << ","
361 << Sec.sectname << ") "
362 << "(size of stubs in reserved2 field is zero)\n";
363 continue;
364 }
365 uint32_t count = Sec.size / stride;
366 outs() << "Indirect symbols for (" << Sec.segname << ","
367 << Sec.sectname << ") " << count << " entries";
368 uint32_t n = Sec.reserved1;
369 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
370 }
371 }
372 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
373 MachO::segment_command Seg = O->getSegmentLoadCommand(Load);
374 for (unsigned J = 0; J < Seg.nsects; ++J) {
375 MachO::section Sec = O->getSection(Load, J);
376 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
377 if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
378 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
379 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
380 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
381 section_type == MachO::S_SYMBOL_STUBS) {
382 uint32_t stride;
383 if (section_type == MachO::S_SYMBOL_STUBS)
384 stride = Sec.reserved2;
385 else
386 stride = 4;
387 if (stride == 0) {
388 outs() << "Can't print indirect symbols for (" << Sec.segname << ","
389 << Sec.sectname << ") "
390 << "(size of stubs in reserved2 field is zero)\n";
391 continue;
392 }
393 uint32_t count = Sec.size / stride;
394 outs() << "Indirect symbols for (" << Sec.segname << ","
395 << Sec.sectname << ") " << count << " entries";
396 uint32_t n = Sec.reserved1;
397 PrintIndirectSymbolTable(O, verbose, n, count, stride, Sec.addr);
398 }
399 }
400 }
401 if (I == LoadCommandCount - 1)
402 break;
403 else
404 Load = O->getNextLoadCommandInfo(Load);
405 }
406}
407
Kevin Enderby69fe98d2015-01-23 18:52:17 +0000408static void PrintDataInCodeTable(MachOObjectFile *O, bool verbose) {
409 MachO::linkedit_data_command DIC = O->getDataInCodeLoadCommand();
410 uint32_t nentries = DIC.datasize / sizeof(struct MachO::data_in_code_entry);
411 outs() << "Data in code table (" << nentries << " entries)\n";
412 outs() << "offset length kind\n";
413 for (dice_iterator DI = O->begin_dices(), DE = O->end_dices(); DI != DE;
414 ++DI) {
415 uint32_t Offset;
416 DI->getOffset(Offset);
417 outs() << format("0x%08" PRIx32, Offset) << " ";
418 uint16_t Length;
419 DI->getLength(Length);
420 outs() << format("%6u", Length) << " ";
421 uint16_t Kind;
422 DI->getKind(Kind);
423 if (verbose) {
424 switch (Kind) {
425 case MachO::DICE_KIND_DATA:
426 outs() << "DATA";
427 break;
428 case MachO::DICE_KIND_JUMP_TABLE8:
429 outs() << "JUMP_TABLE8";
430 break;
431 case MachO::DICE_KIND_JUMP_TABLE16:
432 outs() << "JUMP_TABLE16";
433 break;
434 case MachO::DICE_KIND_JUMP_TABLE32:
435 outs() << "JUMP_TABLE32";
436 break;
437 case MachO::DICE_KIND_ABS_JUMP_TABLE32:
438 outs() << "ABS_JUMP_TABLE32";
439 break;
440 default:
441 outs() << format("0x%04" PRIx32, Kind);
442 break;
443 }
444 } else
445 outs() << format("0x%04" PRIx32, Kind);
446 outs() << "\n";
447 }
448}
449
Kevin Enderby9a509442015-01-27 21:28:24 +0000450static void PrintLinkOptHints(MachOObjectFile *O) {
451 MachO::linkedit_data_command LohLC = O->getLinkOptHintsLoadCommand();
452 const char *loh = O->getData().substr(LohLC.dataoff, 1).data();
453 uint32_t nloh = LohLC.datasize;
454 outs() << "Linker optimiztion hints (" << nloh << " total bytes)\n";
455 for (uint32_t i = 0; i < nloh;) {
456 unsigned n;
457 uint64_t identifier = decodeULEB128((const uint8_t *)(loh + i), &n);
458 i += n;
459 outs() << " identifier " << identifier << " ";
460 if (i >= nloh)
461 return;
462 switch (identifier) {
463 case 1:
464 outs() << "AdrpAdrp\n";
465 break;
466 case 2:
467 outs() << "AdrpLdr\n";
468 break;
469 case 3:
470 outs() << "AdrpAddLdr\n";
471 break;
472 case 4:
473 outs() << "AdrpLdrGotLdr\n";
474 break;
475 case 5:
476 outs() << "AdrpAddStr\n";
477 break;
478 case 6:
479 outs() << "AdrpLdrGotStr\n";
480 break;
481 case 7:
482 outs() << "AdrpAdd\n";
483 break;
484 case 8:
485 outs() << "AdrpLdrGot\n";
486 break;
487 default:
488 outs() << "Unknown identifier value\n";
489 break;
490 }
491 uint64_t narguments = decodeULEB128((const uint8_t *)(loh + i), &n);
492 i += n;
493 outs() << " narguments " << narguments << "\n";
494 if (i >= nloh)
495 return;
496
497 for (uint32_t j = 0; j < narguments; j++) {
498 uint64_t value = decodeULEB128((const uint8_t *)(loh + i), &n);
499 i += n;
500 outs() << "\tvalue " << format("0x%" PRIx64, value) << "\n";
501 if (i >= nloh)
502 return;
503 }
504 }
505}
506
Kevin Enderbyf6d25852015-01-31 00:37:11 +0000507typedef DenseMap<uint64_t, StringRef> SymbolAddressMap;
508
509static void CreateSymbolAddressMap(MachOObjectFile *O,
510 SymbolAddressMap *AddrMap) {
511 // Create a map of symbol addresses to symbol names.
512 for (const SymbolRef &Symbol : O->symbols()) {
513 SymbolRef::Type ST;
514 Symbol.getType(ST);
515 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
516 ST == SymbolRef::ST_Other) {
517 uint64_t Address;
518 Symbol.getAddress(Address);
519 StringRef SymName;
520 Symbol.getName(SymName);
521 (*AddrMap)[Address] = SymName;
522 }
523 }
524}
525
526// GuessSymbolName is passed the address of what might be a symbol and a
527// pointer to the SymbolAddressMap. It returns the name of a symbol
528// with that address or nullptr if no symbol is found with that address.
529static const char *GuessSymbolName(uint64_t value, SymbolAddressMap *AddrMap) {
530 const char *SymbolName = nullptr;
531 // A DenseMap can't lookup up some values.
532 if (value != 0xffffffffffffffffULL && value != 0xfffffffffffffffeULL) {
533 StringRef name = AddrMap->lookup(value);
534 if (!name.empty())
535 SymbolName = name.data();
536 }
537 return SymbolName;
538}
539
540static void DumpInitTermPointerSection(MachOObjectFile *O, const char *sect,
541 uint32_t sect_size, uint64_t sect_addr,
542 SymbolAddressMap *AddrMap,
543 bool verbose) {
544 uint32_t stride;
545 if (O->is64Bit())
546 stride = sizeof(uint64_t);
547 else
548 stride = sizeof(uint32_t);
549 for (uint32_t i = 0; i < sect_size; i += stride) {
550 const char *SymbolName = nullptr;
551 if (O->is64Bit()) {
552 outs() << format("0x%016" PRIx64, sect_addr + i * stride) << " ";
553 uint64_t pointer_value;
554 memcpy(&pointer_value, sect + i, stride);
555 if (O->isLittleEndian() != sys::IsLittleEndianHost)
556 sys::swapByteOrder(pointer_value);
557 outs() << format("0x%016" PRIx64, pointer_value);
558 if (verbose)
559 SymbolName = GuessSymbolName(pointer_value, AddrMap);
560 } else {
561 outs() << format("0x%08" PRIx64, sect_addr + i * stride) << " ";
562 uint32_t pointer_value;
563 memcpy(&pointer_value, sect + i, stride);
564 if (O->isLittleEndian() != sys::IsLittleEndianHost)
565 sys::swapByteOrder(pointer_value);
566 outs() << format("0x%08" PRIx32, pointer_value);
567 if (verbose)
568 SymbolName = GuessSymbolName(pointer_value, AddrMap);
569 }
570 if (SymbolName)
571 outs() << " " << SymbolName;
572 outs() << "\n";
573 }
574}
575
576static void DumpRawSectionContents(MachOObjectFile *O, const char *sect,
577 uint32_t size, uint64_t addr) {
578 uint32_t cputype = O->getHeader().cputype;
579 if (cputype == MachO::CPU_TYPE_I386 || cputype == MachO::CPU_TYPE_X86_64) {
580 uint32_t j;
581 for (uint32_t i = 0; i < size; i += j, addr += j) {
582 if (O->is64Bit())
583 outs() << format("%016" PRIx64, addr) << "\t";
584 else
585 outs() << format("%08" PRIx64, sect) << "\t";
586 for (j = 0; j < 16 && i + j < size; j++) {
587 uint8_t byte_word = *(sect + i + j);
588 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
589 }
590 outs() << "\n";
591 }
592 } else {
593 uint32_t j;
594 for (uint32_t i = 0; i < size; i += j, addr += j) {
595 if (O->is64Bit())
596 outs() << format("%016" PRIx64, addr) << "\t";
597 else
598 outs() << format("%08" PRIx64, sect) << "\t";
599 for (j = 0; j < 4 * sizeof(int32_t) && i + j < size;
600 j += sizeof(int32_t)) {
601 if (i + j + sizeof(int32_t) < size) {
602 uint32_t long_word;
603 memcpy(&long_word, sect + i + j, sizeof(int32_t));
604 if (O->isLittleEndian() != sys::IsLittleEndianHost)
605 sys::swapByteOrder(long_word);
606 outs() << format("%08" PRIx32, long_word) << " ";
607 } else {
608 for (uint32_t k = 0; i + j + k < size; k++) {
609 uint8_t byte_word = *(sect + i + j);
610 outs() << format("%02" PRIx32, (uint32_t)byte_word) << " ";
611 }
612 }
613 }
614 outs() << "\n";
615 }
616 }
617}
618
619static void DumpSectionContents(MachOObjectFile *O, bool verbose) {
620 SymbolAddressMap AddrMap;
621 if (verbose)
622 CreateSymbolAddressMap(O, &AddrMap);
623
624 for (unsigned i = 0; i < DumpSections.size(); ++i) {
625 StringRef DumpSection = DumpSections[i];
626 std::pair<StringRef, StringRef> DumpSegSectName;
627 DumpSegSectName = DumpSection.split(',');
628 StringRef DumpSegName, DumpSectName;
629 if (DumpSegSectName.second.size()) {
630 DumpSegName = DumpSegSectName.first;
631 DumpSectName = DumpSegSectName.second;
632 } else {
633 DumpSegName = "";
634 DumpSectName = DumpSegSectName.first;
635 }
636 for (const SectionRef &Section : O->sections()) {
637 StringRef SectName;
638 Section.getName(SectName);
639 DataRefImpl Ref = Section.getRawDataRefImpl();
640 StringRef SegName = O->getSectionFinalSegmentName(Ref);
641 if ((DumpSegName.empty() || SegName == DumpSegName) &&
642 (SectName == DumpSectName)) {
643 outs() << "Contents of (" << SegName << "," << SectName
644 << ") section\n";
645 uint32_t section_type;
646 if (O->is64Bit()) {
647 const MachO::section_64 Sec = O->getSection64(Ref);
648 section_type = Sec.flags & MachO::SECTION_TYPE;
649
650 } else {
651 const MachO::section Sec = O->getSection(Ref);
652 section_type = Sec.flags & MachO::SECTION_TYPE;
653 }
654
655 StringRef BytesStr;
656 Section.getContents(BytesStr);
657 const char *sect = reinterpret_cast<const char *>(BytesStr.data());
658 uint32_t sect_size = BytesStr.size();
659 uint64_t sect_addr = Section.getAddress();
660
661 if (verbose) {
662 switch (section_type) {
663 case MachO::S_REGULAR:
664 DumpRawSectionContents(O, sect, sect_size, sect_addr);
665 break;
666 case MachO::S_ZEROFILL:
667 outs() << "zerofill section and has no contents in the file\n";
668 break;
669 case MachO::S_MOD_INIT_FUNC_POINTERS:
670 case MachO::S_MOD_TERM_FUNC_POINTERS:
671 DumpInitTermPointerSection(O, sect, sect_size, sect_addr, &AddrMap,
672 verbose);
673 break;
674 default:
675 outs() << "Unknown section type ("
676 << format("0x%08" PRIx32, section_type) << ")\n";
677 DumpRawSectionContents(O, sect, sect_size, sect_addr);
678 break;
679 }
680 } else {
681 if (section_type == MachO::S_ZEROFILL)
682 outs() << "zerofill section and has no contents in the file\n";
683 else
684 DumpRawSectionContents(O, sect, sect_size, sect_addr);
685 }
686 }
687 }
688 }
689}
690
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +0000691// checkMachOAndArchFlags() checks to see if the ObjectFile is a Mach-O file
692// and if it is and there is a list of architecture flags is specified then
693// check to make sure this Mach-O file is one of those architectures or all
694// architectures were specified. If not then an error is generated and this
695// routine returns false. Else it returns true.
696static bool checkMachOAndArchFlags(ObjectFile *O, StringRef Filename) {
697 if (isa<MachOObjectFile>(O) && !ArchAll && ArchFlags.size() != 0) {
698 MachOObjectFile *MachO = dyn_cast<MachOObjectFile>(O);
699 bool ArchFound = false;
700 MachO::mach_header H;
701 MachO::mach_header_64 H_64;
702 Triple T;
703 if (MachO->is64Bit()) {
704 H_64 = MachO->MachOObjectFile::getHeader64();
705 T = MachOObjectFile::getArch(H_64.cputype, H_64.cpusubtype);
706 } else {
707 H = MachO->MachOObjectFile::getHeader();
708 T = MachOObjectFile::getArch(H.cputype, H.cpusubtype);
709 }
710 unsigned i;
711 for (i = 0; i < ArchFlags.size(); ++i) {
712 if (ArchFlags[i] == T.getArchName())
713 ArchFound = true;
714 break;
715 }
716 if (!ArchFound) {
717 errs() << "llvm-objdump: file: " + Filename + " does not contain "
718 << "architecture: " + ArchFlags[i] + "\n";
719 return false;
720 }
721 }
722 return true;
723}
724
Kevin Enderbye2297dd2015-01-07 21:02:18 +0000725static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF);
Rafael Espindola9b709252013-04-13 01:45:40 +0000726
Kevin Enderbye2297dd2015-01-07 21:02:18 +0000727// ProcessMachO() is passed a single opened Mach-O file, which may be an
728// archive member and or in a slice of a universal file. It prints the
729// the file name and header info and then processes it according to the
730// command line options.
731static void ProcessMachO(StringRef Filename, MachOObjectFile *MachOOF,
732 StringRef ArchiveMemberName = StringRef(),
733 StringRef ArchitectureName = StringRef()) {
Kevin Enderby131d1772015-01-09 19:22:37 +0000734 // If we are doing some processing here on the Mach-O file print the header
735 // info. And don't print it otherwise like in the case of printing the
Kevin Enderbya7bdc7e2015-01-22 18:55:27 +0000736 // UniversalHeaders or ArchiveHeaders.
Kevin Enderby131d1772015-01-09 19:22:37 +0000737 if (Disassemble || PrivateHeaders || ExportsTrie || Rebase || Bind ||
Kevin Enderbyf6d25852015-01-31 00:37:11 +0000738 LazyBind || WeakBind || IndirectSymbols || DataInCode || LinkOptHints ||
739 DumpSections.size() != 0) {
Kevin Enderby131d1772015-01-09 19:22:37 +0000740 outs() << Filename;
741 if (!ArchiveMemberName.empty())
742 outs() << '(' << ArchiveMemberName << ')';
743 if (!ArchitectureName.empty())
744 outs() << " (architecture " << ArchitectureName << ")";
745 outs() << ":\n";
746 }
Kevin Enderbye2297dd2015-01-07 21:02:18 +0000747
748 if (Disassemble)
749 DisassembleMachO(Filename, MachOOF);
Kevin Enderbya7bdc7e2015-01-22 18:55:27 +0000750 if (IndirectSymbols)
751 PrintIndirectSymbols(MachOOF, true);
Kevin Enderby69fe98d2015-01-23 18:52:17 +0000752 if (DataInCode)
753 PrintDataInCodeTable(MachOOF, true);
Kevin Enderby9a509442015-01-27 21:28:24 +0000754 if (LinkOptHints)
755 PrintLinkOptHints(MachOOF);
Kevin Enderby98da6132015-01-20 21:47:46 +0000756 if (Relocations)
757 PrintRelocations(MachOOF);
758 if (SectionHeaders)
759 PrintSectionHeaders(MachOOF);
760 if (SectionContents)
761 PrintSectionContents(MachOOF);
Kevin Enderbyf6d25852015-01-31 00:37:11 +0000762 if (DumpSections.size() != 0)
763 DumpSectionContents(MachOOF, true);
Kevin Enderby98da6132015-01-20 21:47:46 +0000764 if (SymbolTable)
765 PrintSymbolTable(MachOOF);
766 if (UnwindInfo)
767 printMachOUnwindInfo(MachOOF);
Kevin Enderbye2297dd2015-01-07 21:02:18 +0000768 if (PrivateHeaders)
769 printMachOFileHeader(MachOOF);
770 if (ExportsTrie)
771 printExportsTrie(MachOOF);
772 if (Rebase)
773 printRebaseTable(MachOOF);
774 if (Bind)
775 printBindTable(MachOOF);
776 if (LazyBind)
777 printLazyBindTable(MachOOF);
778 if (WeakBind)
779 printWeakBindTable(MachOOF);
780}
781
Kevin Enderby131d1772015-01-09 19:22:37 +0000782// printUnknownCPUType() helps print_fat_headers for unknown CPU's.
783static void printUnknownCPUType(uint32_t cputype, uint32_t cpusubtype) {
784 outs() << " cputype (" << cputype << ")\n";
785 outs() << " cpusubtype (" << cpusubtype << ")\n";
786}
787
788// printCPUType() helps print_fat_headers by printing the cputype and
789// pusubtype (symbolically for the one's it knows about).
790static void printCPUType(uint32_t cputype, uint32_t cpusubtype) {
791 switch (cputype) {
792 case MachO::CPU_TYPE_I386:
793 switch (cpusubtype) {
794 case MachO::CPU_SUBTYPE_I386_ALL:
795 outs() << " cputype CPU_TYPE_I386\n";
796 outs() << " cpusubtype CPU_SUBTYPE_I386_ALL\n";
797 break;
798 default:
799 printUnknownCPUType(cputype, cpusubtype);
800 break;
801 }
802 break;
803 case MachO::CPU_TYPE_X86_64:
804 switch (cpusubtype) {
805 case MachO::CPU_SUBTYPE_X86_64_ALL:
806 outs() << " cputype CPU_TYPE_X86_64\n";
807 outs() << " cpusubtype CPU_SUBTYPE_X86_64_ALL\n";
808 break;
809 case MachO::CPU_SUBTYPE_X86_64_H:
810 outs() << " cputype CPU_TYPE_X86_64\n";
811 outs() << " cpusubtype CPU_SUBTYPE_X86_64_H\n";
812 break;
813 default:
814 printUnknownCPUType(cputype, cpusubtype);
815 break;
816 }
817 break;
818 case MachO::CPU_TYPE_ARM:
819 switch (cpusubtype) {
820 case MachO::CPU_SUBTYPE_ARM_ALL:
821 outs() << " cputype CPU_TYPE_ARM\n";
822 outs() << " cpusubtype CPU_SUBTYPE_ARM_ALL\n";
823 break;
824 case MachO::CPU_SUBTYPE_ARM_V4T:
825 outs() << " cputype CPU_TYPE_ARM\n";
826 outs() << " cpusubtype CPU_SUBTYPE_ARM_V4T\n";
827 break;
828 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
829 outs() << " cputype CPU_TYPE_ARM\n";
830 outs() << " cpusubtype CPU_SUBTYPE_ARM_V5TEJ\n";
831 break;
832 case MachO::CPU_SUBTYPE_ARM_XSCALE:
833 outs() << " cputype CPU_TYPE_ARM\n";
834 outs() << " cpusubtype CPU_SUBTYPE_ARM_XSCALE\n";
835 break;
836 case MachO::CPU_SUBTYPE_ARM_V6:
837 outs() << " cputype CPU_TYPE_ARM\n";
838 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6\n";
839 break;
840 case MachO::CPU_SUBTYPE_ARM_V6M:
841 outs() << " cputype CPU_TYPE_ARM\n";
842 outs() << " cpusubtype CPU_SUBTYPE_ARM_V6M\n";
843 break;
844 case MachO::CPU_SUBTYPE_ARM_V7:
845 outs() << " cputype CPU_TYPE_ARM\n";
846 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7\n";
847 break;
848 case MachO::CPU_SUBTYPE_ARM_V7EM:
849 outs() << " cputype CPU_TYPE_ARM\n";
850 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7EM\n";
851 break;
852 case MachO::CPU_SUBTYPE_ARM_V7K:
853 outs() << " cputype CPU_TYPE_ARM\n";
854 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7K\n";
855 break;
856 case MachO::CPU_SUBTYPE_ARM_V7M:
857 outs() << " cputype CPU_TYPE_ARM\n";
858 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7M\n";
859 break;
860 case MachO::CPU_SUBTYPE_ARM_V7S:
861 outs() << " cputype CPU_TYPE_ARM\n";
862 outs() << " cpusubtype CPU_SUBTYPE_ARM_V7S\n";
863 break;
864 default:
865 printUnknownCPUType(cputype, cpusubtype);
866 break;
867 }
868 break;
869 case MachO::CPU_TYPE_ARM64:
870 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
871 case MachO::CPU_SUBTYPE_ARM64_ALL:
872 outs() << " cputype CPU_TYPE_ARM64\n";
873 outs() << " cpusubtype CPU_SUBTYPE_ARM64_ALL\n";
874 break;
875 default:
876 printUnknownCPUType(cputype, cpusubtype);
877 break;
878 }
879 break;
880 default:
881 printUnknownCPUType(cputype, cpusubtype);
882 break;
883 }
884}
885
886static void printMachOUniversalHeaders(const object::MachOUniversalBinary *UB,
887 bool verbose) {
888 outs() << "Fat headers\n";
889 if (verbose)
890 outs() << "fat_magic FAT_MAGIC\n";
891 else
892 outs() << "fat_magic " << format("0x%" PRIx32, MachO::FAT_MAGIC) << "\n";
893
894 uint32_t nfat_arch = UB->getNumberOfObjects();
895 StringRef Buf = UB->getData();
896 uint64_t size = Buf.size();
897 uint64_t big_size = sizeof(struct MachO::fat_header) +
898 nfat_arch * sizeof(struct MachO::fat_arch);
899 outs() << "nfat_arch " << UB->getNumberOfObjects();
900 if (nfat_arch == 0)
901 outs() << " (malformed, contains zero architecture types)\n";
902 else if (big_size > size)
903 outs() << " (malformed, architectures past end of file)\n";
904 else
905 outs() << "\n";
906
907 for (uint32_t i = 0; i < nfat_arch; ++i) {
908 MachOUniversalBinary::ObjectForArch OFA(UB, i);
909 uint32_t cputype = OFA.getCPUType();
910 uint32_t cpusubtype = OFA.getCPUSubType();
911 outs() << "architecture ";
912 for (uint32_t j = 0; i != 0 && j <= i - 1; j++) {
913 MachOUniversalBinary::ObjectForArch other_OFA(UB, j);
914 uint32_t other_cputype = other_OFA.getCPUType();
915 uint32_t other_cpusubtype = other_OFA.getCPUSubType();
Kevin Enderby0512bd72015-01-09 21:55:03 +0000916 if (cputype != 0 && cpusubtype != 0 && cputype == other_cputype &&
Kevin Enderby131d1772015-01-09 19:22:37 +0000917 (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) ==
Kevin Enderby0512bd72015-01-09 21:55:03 +0000918 (other_cpusubtype & ~MachO::CPU_SUBTYPE_MASK)) {
Kevin Enderby131d1772015-01-09 19:22:37 +0000919 outs() << "(illegal duplicate architecture) ";
920 break;
Kevin Enderby0512bd72015-01-09 21:55:03 +0000921 }
Kevin Enderby131d1772015-01-09 19:22:37 +0000922 }
923 if (verbose) {
924 outs() << OFA.getArchTypeName() << "\n";
925 printCPUType(cputype, cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
926 } else {
927 outs() << i << "\n";
928 outs() << " cputype " << cputype << "\n";
929 outs() << " cpusubtype " << (cpusubtype & ~MachO::CPU_SUBTYPE_MASK)
930 << "\n";
931 }
932 if (verbose &&
933 (cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64)
934 outs() << " capabilities CPU_SUBTYPE_LIB64\n";
935 else
936 outs() << " capabilities "
937 << format("0x%" PRIx32,
938 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24) << "\n";
939 outs() << " offset " << OFA.getOffset();
940 if (OFA.getOffset() > size)
941 outs() << " (past end of file)";
942 if (OFA.getOffset() % (1 << OFA.getAlign()) != 0)
943 outs() << " (not aligned on it's alignment (2^" << OFA.getAlign() << ")";
944 outs() << "\n";
945 outs() << " size " << OFA.getSize();
946 big_size = OFA.getOffset() + OFA.getSize();
947 if (big_size > size)
948 outs() << " (past end of file)";
949 outs() << "\n";
950 outs() << " align 2^" << OFA.getAlign() << " (" << (1 << OFA.getAlign())
951 << ")\n";
952 }
953}
954
Kevin Enderby13023a12015-01-15 23:19:11 +0000955static void printArchiveChild(Archive::Child &C, bool verbose,
956 bool print_offset) {
957 if (print_offset)
958 outs() << C.getChildOffset() << "\t";
959 sys::fs::perms Mode = C.getAccessMode();
960 if (verbose) {
961 // FIXME: this first dash, "-", is for (Mode & S_IFMT) == S_IFREG.
962 // But there is nothing in sys::fs::perms for S_IFMT or S_IFREG.
963 outs() << "-";
964 if (Mode & sys::fs::owner_read)
965 outs() << "r";
966 else
967 outs() << "-";
968 if (Mode & sys::fs::owner_write)
969 outs() << "w";
970 else
971 outs() << "-";
972 if (Mode & sys::fs::owner_exe)
973 outs() << "x";
974 else
975 outs() << "-";
976 if (Mode & sys::fs::group_read)
977 outs() << "r";
978 else
979 outs() << "-";
980 if (Mode & sys::fs::group_write)
981 outs() << "w";
982 else
983 outs() << "-";
984 if (Mode & sys::fs::group_exe)
985 outs() << "x";
986 else
987 outs() << "-";
988 if (Mode & sys::fs::others_read)
989 outs() << "r";
990 else
991 outs() << "-";
992 if (Mode & sys::fs::others_write)
993 outs() << "w";
994 else
995 outs() << "-";
996 if (Mode & sys::fs::others_exe)
997 outs() << "x";
998 else
999 outs() << "-";
1000 } else {
1001 outs() << format("0%o ", Mode);
1002 }
1003
1004 unsigned UID = C.getUID();
1005 outs() << format("%3d/", UID);
1006 unsigned GID = C.getGID();
1007 outs() << format("%-3d ", GID);
Kevin Enderbyc12718932015-01-16 22:10:36 +00001008 uint64_t Size = C.getRawSize();
Kevin Enderby479ee612015-01-23 21:02:44 +00001009 outs() << format("%5" PRId64, Size) << " ";
Kevin Enderby13023a12015-01-15 23:19:11 +00001010
1011 StringRef RawLastModified = C.getRawLastModified();
1012 if (verbose) {
1013 unsigned Seconds;
1014 if (RawLastModified.getAsInteger(10, Seconds))
1015 outs() << "(date: \"%s\" contains non-decimal chars) " << RawLastModified;
1016 else {
1017 // Since cime(3) returns a 26 character string of the form:
1018 // "Sun Sep 16 01:03:52 1973\n\0"
1019 // just print 24 characters.
1020 time_t t = Seconds;
1021 outs() << format("%.24s ", ctime(&t));
1022 }
1023 } else {
1024 outs() << RawLastModified << " ";
1025 }
1026
1027 if (verbose) {
1028 ErrorOr<StringRef> NameOrErr = C.getName();
1029 if (NameOrErr.getError()) {
1030 StringRef RawName = C.getRawName();
1031 outs() << RawName << "\n";
1032 } else {
1033 StringRef Name = NameOrErr.get();
1034 outs() << Name << "\n";
1035 }
1036 } else {
1037 StringRef RawName = C.getRawName();
1038 outs() << RawName << "\n";
1039 }
1040}
1041
1042static void printArchiveHeaders(Archive *A, bool verbose, bool print_offset) {
1043 if (A->hasSymbolTable()) {
1044 Archive::child_iterator S = A->getSymbolTableChild();
1045 Archive::Child C = *S;
1046 printArchiveChild(C, verbose, print_offset);
1047 }
1048 for (Archive::child_iterator I = A->child_begin(), E = A->child_end(); I != E;
1049 ++I) {
1050 Archive::Child C = *I;
1051 printArchiveChild(C, verbose, print_offset);
1052 }
1053}
1054
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001055// ParseInputMachO() parses the named Mach-O file in Filename and handles the
1056// -arch flags selecting just those slices as specified by them and also parses
1057// archive files. Then for each individual Mach-O file ProcessMachO() is
1058// called to process the file based on the command line options.
1059void llvm::ParseInputMachO(StringRef Filename) {
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001060 // Check for -arch all and verifiy the -arch flags are valid.
1061 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1062 if (ArchFlags[i] == "all") {
1063 ArchAll = true;
1064 } else {
1065 if (!MachOObjectFile::isValidArch(ArchFlags[i])) {
1066 errs() << "llvm-objdump: Unknown architecture named '" + ArchFlags[i] +
1067 "'for the -arch option\n";
1068 return;
1069 }
1070 }
1071 }
1072
1073 // Attempt to open the binary.
1074 ErrorOr<OwningBinary<Binary>> BinaryOrErr = createBinary(Filename);
1075 if (std::error_code EC = BinaryOrErr.getError()) {
1076 errs() << "llvm-objdump: '" << Filename << "': " << EC.message() << ".\n";
Rafael Espindolade882cd2014-12-03 23:29:34 +00001077 return;
Kevin Enderby3f0ffab2014-12-03 22:29:40 +00001078 }
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001079 Binary &Bin = *BinaryOrErr.get().getBinary();
Kevin Enderby3f0ffab2014-12-03 22:29:40 +00001080
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001081 if (Archive *A = dyn_cast<Archive>(&Bin)) {
1082 outs() << "Archive : " << Filename << "\n";
Kevin Enderby13023a12015-01-15 23:19:11 +00001083 if (ArchiveHeaders)
1084 printArchiveHeaders(A, true, false);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001085 for (Archive::child_iterator I = A->child_begin(), E = A->child_end();
1086 I != E; ++I) {
1087 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = I->getAsBinary();
1088 if (ChildOrErr.getError())
1089 continue;
1090 if (MachOObjectFile *O = dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1091 if (!checkMachOAndArchFlags(O, Filename))
1092 return;
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001093 ProcessMachO(Filename, O, O->getFileName());
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001094 }
1095 }
1096 return;
1097 }
Kevin Enderby131d1772015-01-09 19:22:37 +00001098 if (UniversalHeaders) {
1099 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin))
1100 printMachOUniversalHeaders(UB, true);
1101 }
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001102 if (MachOUniversalBinary *UB = dyn_cast<MachOUniversalBinary>(&Bin)) {
1103 // If we have a list of architecture flags specified dump only those.
1104 if (!ArchAll && ArchFlags.size() != 0) {
1105 // Look for a slice in the universal binary that matches each ArchFlag.
1106 bool ArchFound;
1107 for (unsigned i = 0; i < ArchFlags.size(); ++i) {
1108 ArchFound = false;
1109 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1110 E = UB->end_objects();
1111 I != E; ++I) {
1112 if (ArchFlags[i] == I->getArchTypeName()) {
1113 ArchFound = true;
1114 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr =
1115 I->getAsObjectFile();
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001116 std::string ArchitectureName = "";
1117 if (ArchFlags.size() > 1)
1118 ArchitectureName = I->getArchTypeName();
1119 if (ObjOrErr) {
1120 ObjectFile &O = *ObjOrErr.get();
1121 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001122 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
Rafael Espindola0bfe8282014-12-09 21:05:36 +00001123 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1124 I->getAsArchive()) {
1125 std::unique_ptr<Archive> &A = *AOrErr;
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001126 outs() << "Archive : " << Filename;
1127 if (!ArchitectureName.empty())
1128 outs() << " (architecture " << ArchitectureName << ")";
1129 outs() << "\n";
Kevin Enderby13023a12015-01-15 23:19:11 +00001130 if (ArchiveHeaders)
1131 printArchiveHeaders(A.get(), true, false);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001132 for (Archive::child_iterator AI = A->child_begin(),
1133 AE = A->child_end();
1134 AI != AE; ++AI) {
1135 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1136 if (ChildOrErr.getError())
1137 continue;
1138 if (MachOObjectFile *O =
1139 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001140 ProcessMachO(Filename, O, O->getFileName(), ArchitectureName);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001141 }
1142 }
1143 }
1144 }
1145 if (!ArchFound) {
1146 errs() << "llvm-objdump: file: " + Filename + " does not contain "
1147 << "architecture: " + ArchFlags[i] + "\n";
1148 return;
1149 }
1150 }
1151 return;
1152 }
1153 // No architecture flags were specified so if this contains a slice that
1154 // matches the host architecture dump only that.
1155 if (!ArchAll) {
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001156 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1157 E = UB->end_objects();
1158 I != E; ++I) {
Kevin Enderby0512bd72015-01-09 21:55:03 +00001159 if (MachOObjectFile::getHostArch().getArchName() ==
1160 I->getArchTypeName()) {
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001161 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001162 std::string ArchiveName;
1163 ArchiveName.clear();
1164 if (ObjOrErr) {
1165 ObjectFile &O = *ObjOrErr.get();
1166 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&O))
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001167 ProcessMachO(Filename, MachOOF);
Rafael Espindola0bfe8282014-12-09 21:05:36 +00001168 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr =
1169 I->getAsArchive()) {
1170 std::unique_ptr<Archive> &A = *AOrErr;
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001171 outs() << "Archive : " << Filename << "\n";
Kevin Enderby13023a12015-01-15 23:19:11 +00001172 if (ArchiveHeaders)
1173 printArchiveHeaders(A.get(), true, false);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001174 for (Archive::child_iterator AI = A->child_begin(),
1175 AE = A->child_end();
1176 AI != AE; ++AI) {
1177 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1178 if (ChildOrErr.getError())
1179 continue;
1180 if (MachOObjectFile *O =
1181 dyn_cast<MachOObjectFile>(&*ChildOrErr.get()))
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001182 ProcessMachO(Filename, O, O->getFileName());
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001183 }
1184 }
1185 return;
1186 }
1187 }
1188 }
1189 // Either all architectures have been specified or none have been specified
1190 // and this does not contain the host architecture so dump all the slices.
1191 bool moreThanOneArch = UB->getNumberOfObjects() > 1;
1192 for (MachOUniversalBinary::object_iterator I = UB->begin_objects(),
1193 E = UB->end_objects();
1194 I != E; ++I) {
1195 ErrorOr<std::unique_ptr<ObjectFile>> ObjOrErr = I->getAsObjectFile();
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001196 std::string ArchitectureName = "";
1197 if (moreThanOneArch)
1198 ArchitectureName = I->getArchTypeName();
1199 if (ObjOrErr) {
1200 ObjectFile &Obj = *ObjOrErr.get();
1201 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&Obj))
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001202 ProcessMachO(Filename, MachOOF, "", ArchitectureName);
Rafael Espindola0bfe8282014-12-09 21:05:36 +00001203 } else if (ErrorOr<std::unique_ptr<Archive>> AOrErr = I->getAsArchive()) {
1204 std::unique_ptr<Archive> &A = *AOrErr;
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001205 outs() << "Archive : " << Filename;
1206 if (!ArchitectureName.empty())
1207 outs() << " (architecture " << ArchitectureName << ")";
1208 outs() << "\n";
Kevin Enderby13023a12015-01-15 23:19:11 +00001209 if (ArchiveHeaders)
1210 printArchiveHeaders(A.get(), true, false);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001211 for (Archive::child_iterator AI = A->child_begin(), AE = A->child_end();
1212 AI != AE; ++AI) {
1213 ErrorOr<std::unique_ptr<Binary>> ChildOrErr = AI->getAsBinary();
1214 if (ChildOrErr.getError())
1215 continue;
1216 if (MachOObjectFile *O =
1217 dyn_cast<MachOObjectFile>(&*ChildOrErr.get())) {
1218 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(O))
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001219 ProcessMachO(Filename, MachOOF, MachOOF->getFileName(),
1220 ArchitectureName);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001221 }
1222 }
1223 }
1224 }
1225 return;
1226 }
1227 if (ObjectFile *O = dyn_cast<ObjectFile>(&Bin)) {
1228 if (!checkMachOAndArchFlags(O, Filename))
1229 return;
1230 if (MachOObjectFile *MachOOF = dyn_cast<MachOObjectFile>(&*O)) {
Kevin Enderbye2297dd2015-01-07 21:02:18 +00001231 ProcessMachO(Filename, MachOOF);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00001232 } else
1233 errs() << "llvm-objdump: '" << Filename << "': "
1234 << "Object is not a Mach-O file type.\n";
1235 } else
1236 errs() << "llvm-objdump: '" << Filename << "': "
1237 << "Unrecognized file type.\n";
Rafael Espindola9b709252013-04-13 01:45:40 +00001238}
1239
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001240typedef std::pair<uint64_t, const char *> BindInfoEntry;
1241typedef std::vector<BindInfoEntry> BindTable;
1242typedef BindTable::iterator bind_table_iterator;
Kevin Enderbybf246f52014-09-24 23:08:22 +00001243
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001244// The block of info used by the Symbolizer call backs.
1245struct DisassembleInfo {
1246 bool verbose;
1247 MachOObjectFile *O;
1248 SectionRef S;
Kevin Enderbybf246f52014-09-24 23:08:22 +00001249 SymbolAddressMap *AddrMap;
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001250 std::vector<SectionRef> *Sections;
1251 const char *class_name;
1252 const char *selector_name;
1253 char *method;
Kevin Enderby04bf6932014-10-28 23:39:46 +00001254 char *demangled_name;
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001255 uint64_t adrp_addr;
1256 uint32_t adrp_inst;
Kevin Enderby078be602014-10-23 19:53:12 +00001257 BindTable *bindtable;
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001258};
1259
1260// SymbolizerGetOpInfo() is the operand information call back function.
1261// This is called to get the symbolic information for operand(s) of an
1262// instruction when it is being done. This routine does this from
1263// the relocation information, symbol table, etc. That block of information
1264// is a pointer to the struct DisassembleInfo that was passed when the
1265// disassembler context was created and passed to back to here when
1266// called back by the disassembler for instruction operands that could have
1267// relocation information. The address of the instruction containing operand is
1268// at the Pc parameter. The immediate value the operand has is passed in
1269// op_info->Value and is at Offset past the start of the instruction and has a
1270// byte Size of 1, 2 or 4. The symbolc information is returned in TagBuf is the
1271// LLVMOpInfo1 struct defined in the header "llvm-c/Disassembler.h" as symbol
1272// names and addends of the symbolic expression to add for the operand. The
1273// value of TagType is currently 1 (for the LLVMOpInfo1 struct). If symbolic
1274// information is returned then this function returns 1 else it returns 0.
1275int SymbolizerGetOpInfo(void *DisInfo, uint64_t Pc, uint64_t Offset,
1276 uint64_t Size, int TagType, void *TagBuf) {
1277 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
1278 struct LLVMOpInfo1 *op_info = (struct LLVMOpInfo1 *)TagBuf;
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001279 uint64_t value = op_info->Value;
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001280
1281 // Make sure all fields returned are zero if we don't set them.
1282 memset((void *)op_info, '\0', sizeof(struct LLVMOpInfo1));
1283 op_info->Value = value;
1284
1285 // If the TagType is not the value 1 which it code knows about or if no
1286 // verbose symbolic information is wanted then just return 0, indicating no
1287 // information is being returned.
1288 if (TagType != 1 || info->verbose == false)
1289 return 0;
1290
1291 unsigned int Arch = info->O->getArch();
1292 if (Arch == Triple::x86) {
Kevin Enderby9907d0a2014-11-04 00:43:16 +00001293 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1294 return 0;
1295 // First search the section's relocation entries (if any) for an entry
1296 // for this section offset.
1297 uint32_t sect_addr = info->S.getAddress();
1298 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1299 bool reloc_found = false;
1300 DataRefImpl Rel;
1301 MachO::any_relocation_info RE;
1302 bool isExtern = false;
1303 SymbolRef Symbol;
1304 bool r_scattered = false;
1305 uint32_t r_value, pair_r_value, r_type;
1306 for (const RelocationRef &Reloc : info->S.relocations()) {
1307 uint64_t RelocOffset;
1308 Reloc.getOffset(RelocOffset);
1309 if (RelocOffset == sect_offset) {
1310 Rel = Reloc.getRawDataRefImpl();
1311 RE = info->O->getRelocation(Rel);
Kevin Enderby3eb73e12014-11-11 19:16:45 +00001312 r_type = info->O->getAnyRelocationType(RE);
Kevin Enderby9907d0a2014-11-04 00:43:16 +00001313 r_scattered = info->O->isRelocationScattered(RE);
1314 if (r_scattered) {
1315 r_value = info->O->getScatteredRelocationValue(RE);
Kevin Enderby9907d0a2014-11-04 00:43:16 +00001316 if (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1317 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF) {
1318 DataRefImpl RelNext = Rel;
1319 info->O->moveRelocationNext(RelNext);
1320 MachO::any_relocation_info RENext;
1321 RENext = info->O->getRelocation(RelNext);
1322 if (info->O->isRelocationScattered(RENext))
Kevin Enderby930fdc72014-11-06 19:00:13 +00001323 pair_r_value = info->O->getScatteredRelocationValue(RENext);
Kevin Enderby9907d0a2014-11-04 00:43:16 +00001324 else
1325 return 0;
1326 }
1327 } else {
1328 isExtern = info->O->getPlainRelocationExternal(RE);
1329 if (isExtern) {
1330 symbol_iterator RelocSym = Reloc.getSymbol();
1331 Symbol = *RelocSym;
1332 }
1333 }
1334 reloc_found = true;
1335 break;
1336 }
1337 }
1338 if (reloc_found && isExtern) {
1339 StringRef SymName;
1340 Symbol.getName(SymName);
1341 const char *name = SymName.data();
1342 op_info->AddSymbol.Present = 1;
1343 op_info->AddSymbol.Name = name;
1344 // For i386 extern relocation entries the value in the instruction is
1345 // the offset from the symbol, and value is already set in op_info->Value.
1346 return 1;
1347 }
1348 if (reloc_found && (r_type == MachO::GENERIC_RELOC_SECTDIFF ||
1349 r_type == MachO::GENERIC_RELOC_LOCAL_SECTDIFF)) {
Kevin Enderbyf6d25852015-01-31 00:37:11 +00001350 const char *add = GuessSymbolName(r_value, info->AddrMap);
1351 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
Kevin Enderby9907d0a2014-11-04 00:43:16 +00001352 uint32_t offset = value - (r_value - pair_r_value);
1353 op_info->AddSymbol.Present = 1;
1354 if (add != nullptr)
1355 op_info->AddSymbol.Name = add;
1356 else
1357 op_info->AddSymbol.Value = r_value;
1358 op_info->SubtractSymbol.Present = 1;
1359 if (sub != nullptr)
1360 op_info->SubtractSymbol.Name = sub;
1361 else
1362 op_info->SubtractSymbol.Value = pair_r_value;
1363 op_info->Value = offset;
1364 return 1;
1365 }
1366 // TODO:
1367 // Second search the external relocation entries of a fully linked image
1368 // (if any) for an entry that matches this segment offset.
1369 // uint32_t seg_offset = (Pc + Offset);
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001370 return 0;
1371 } else if (Arch == Triple::x86_64) {
1372 if (Size != 1 && Size != 2 && Size != 4 && Size != 0)
1373 return 0;
1374 // First search the section's relocation entries (if any) for an entry
1375 // for this section offset.
Rafael Espindola80291272014-10-08 15:28:58 +00001376 uint64_t sect_addr = info->S.getAddress();
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001377 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1378 bool reloc_found = false;
1379 DataRefImpl Rel;
1380 MachO::any_relocation_info RE;
1381 bool isExtern = false;
1382 SymbolRef Symbol;
1383 for (const RelocationRef &Reloc : info->S.relocations()) {
1384 uint64_t RelocOffset;
1385 Reloc.getOffset(RelocOffset);
1386 if (RelocOffset == sect_offset) {
1387 Rel = Reloc.getRawDataRefImpl();
1388 RE = info->O->getRelocation(Rel);
1389 // NOTE: Scattered relocations don't exist on x86_64.
1390 isExtern = info->O->getPlainRelocationExternal(RE);
1391 if (isExtern) {
1392 symbol_iterator RelocSym = Reloc.getSymbol();
1393 Symbol = *RelocSym;
1394 }
1395 reloc_found = true;
1396 break;
1397 }
1398 }
1399 if (reloc_found && isExtern) {
1400 // The Value passed in will be adjusted by the Pc if the instruction
1401 // adds the Pc. But for x86_64 external relocation entries the Value
1402 // is the offset from the external symbol.
1403 if (info->O->getAnyRelocationPCRel(RE))
1404 op_info->Value -= Pc + Offset + Size;
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001405 StringRef SymName;
1406 Symbol.getName(SymName);
1407 const char *name = SymName.data();
1408 unsigned Type = info->O->getAnyRelocationType(RE);
1409 if (Type == MachO::X86_64_RELOC_SUBTRACTOR) {
1410 DataRefImpl RelNext = Rel;
1411 info->O->moveRelocationNext(RelNext);
1412 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1413 unsigned TypeNext = info->O->getAnyRelocationType(RENext);
1414 bool isExternNext = info->O->getPlainRelocationExternal(RENext);
1415 unsigned SymbolNum = info->O->getPlainRelocationSymbolNum(RENext);
1416 if (TypeNext == MachO::X86_64_RELOC_UNSIGNED && isExternNext) {
1417 op_info->SubtractSymbol.Present = 1;
1418 op_info->SubtractSymbol.Name = name;
1419 symbol_iterator RelocSymNext = info->O->getSymbolByIndex(SymbolNum);
1420 Symbol = *RelocSymNext;
1421 StringRef SymNameNext;
1422 Symbol.getName(SymNameNext);
1423 name = SymNameNext.data();
1424 }
1425 }
1426 // TODO: add the VariantKinds to op_info->VariantKind for relocation types
1427 // like: X86_64_RELOC_TLV, X86_64_RELOC_GOT_LOAD and X86_64_RELOC_GOT.
1428 op_info->AddSymbol.Present = 1;
1429 op_info->AddSymbol.Name = name;
1430 return 1;
1431 }
1432 // TODO:
1433 // Second search the external relocation entries of a fully linked image
1434 // (if any) for an entry that matches this segment offset.
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001435 // uint64_t seg_offset = (Pc + Offset);
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001436 return 0;
1437 } else if (Arch == Triple::arm) {
Kevin Enderby930fdc72014-11-06 19:00:13 +00001438 if (Offset != 0 || (Size != 4 && Size != 2))
1439 return 0;
1440 // First search the section's relocation entries (if any) for an entry
1441 // for this section offset.
1442 uint32_t sect_addr = info->S.getAddress();
1443 uint32_t sect_offset = (Pc + Offset) - sect_addr;
1444 bool reloc_found = false;
1445 DataRefImpl Rel;
1446 MachO::any_relocation_info RE;
1447 bool isExtern = false;
1448 SymbolRef Symbol;
1449 bool r_scattered = false;
1450 uint32_t r_value, pair_r_value, r_type, r_length, other_half;
1451 for (const RelocationRef &Reloc : info->S.relocations()) {
1452 uint64_t RelocOffset;
1453 Reloc.getOffset(RelocOffset);
1454 if (RelocOffset == sect_offset) {
1455 Rel = Reloc.getRawDataRefImpl();
1456 RE = info->O->getRelocation(Rel);
1457 r_length = info->O->getAnyRelocationLength(RE);
1458 r_scattered = info->O->isRelocationScattered(RE);
1459 if (r_scattered) {
1460 r_value = info->O->getScatteredRelocationValue(RE);
1461 r_type = info->O->getScatteredRelocationType(RE);
1462 } else {
1463 r_type = info->O->getAnyRelocationType(RE);
1464 isExtern = info->O->getPlainRelocationExternal(RE);
1465 if (isExtern) {
1466 symbol_iterator RelocSym = Reloc.getSymbol();
1467 Symbol = *RelocSym;
1468 }
1469 }
1470 if (r_type == MachO::ARM_RELOC_HALF ||
1471 r_type == MachO::ARM_RELOC_SECTDIFF ||
1472 r_type == MachO::ARM_RELOC_LOCAL_SECTDIFF ||
1473 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1474 DataRefImpl RelNext = Rel;
1475 info->O->moveRelocationNext(RelNext);
1476 MachO::any_relocation_info RENext;
1477 RENext = info->O->getRelocation(RelNext);
1478 other_half = info->O->getAnyRelocationAddress(RENext) & 0xffff;
1479 if (info->O->isRelocationScattered(RENext))
1480 pair_r_value = info->O->getScatteredRelocationValue(RENext);
1481 }
1482 reloc_found = true;
1483 break;
1484 }
1485 }
1486 if (reloc_found && isExtern) {
1487 StringRef SymName;
1488 Symbol.getName(SymName);
1489 const char *name = SymName.data();
1490 op_info->AddSymbol.Present = 1;
1491 op_info->AddSymbol.Name = name;
1492 if (value != 0) {
1493 switch (r_type) {
1494 case MachO::ARM_RELOC_HALF:
1495 if ((r_length & 0x1) == 1) {
1496 op_info->Value = value << 16 | other_half;
1497 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1498 } else {
1499 op_info->Value = other_half << 16 | value;
1500 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1501 }
1502 break;
1503 default:
1504 break;
1505 }
1506 } else {
1507 switch (r_type) {
1508 case MachO::ARM_RELOC_HALF:
1509 if ((r_length & 0x1) == 1) {
1510 op_info->Value = value << 16 | other_half;
1511 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1512 } else {
1513 op_info->Value = other_half << 16 | value;
1514 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1515 }
1516 break;
1517 default:
1518 break;
1519 }
1520 }
1521 return 1;
1522 }
1523 // If we have a branch that is not an external relocation entry then
1524 // return 0 so the code in tryAddingSymbolicOperand() can use the
1525 // SymbolLookUp call back with the branch target address to look up the
1526 // symbol and possiblity add an annotation for a symbol stub.
1527 if (reloc_found && isExtern == 0 && (r_type == MachO::ARM_RELOC_BR24 ||
1528 r_type == MachO::ARM_THUMB_RELOC_BR22))
1529 return 0;
1530
1531 uint32_t offset = 0;
1532 if (reloc_found) {
1533 if (r_type == MachO::ARM_RELOC_HALF ||
1534 r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1535 if ((r_length & 0x1) == 1)
1536 value = value << 16 | other_half;
1537 else
1538 value = other_half << 16 | value;
1539 }
1540 if (r_scattered && (r_type != MachO::ARM_RELOC_HALF &&
1541 r_type != MachO::ARM_RELOC_HALF_SECTDIFF)) {
1542 offset = value - r_value;
1543 value = r_value;
1544 }
1545 }
1546
1547 if (reloc_found && r_type == MachO::ARM_RELOC_HALF_SECTDIFF) {
1548 if ((r_length & 0x1) == 1)
1549 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1550 else
1551 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
Kevin Enderbyf6d25852015-01-31 00:37:11 +00001552 const char *add = GuessSymbolName(r_value, info->AddrMap);
1553 const char *sub = GuessSymbolName(pair_r_value, info->AddrMap);
Kevin Enderby930fdc72014-11-06 19:00:13 +00001554 int32_t offset = value - (r_value - pair_r_value);
1555 op_info->AddSymbol.Present = 1;
1556 if (add != nullptr)
1557 op_info->AddSymbol.Name = add;
1558 else
1559 op_info->AddSymbol.Value = r_value;
1560 op_info->SubtractSymbol.Present = 1;
1561 if (sub != nullptr)
1562 op_info->SubtractSymbol.Name = sub;
1563 else
1564 op_info->SubtractSymbol.Value = pair_r_value;
1565 op_info->Value = offset;
1566 return 1;
1567 }
1568
1569 if (reloc_found == false)
1570 return 0;
1571
1572 op_info->AddSymbol.Present = 1;
1573 op_info->Value = offset;
1574 if (reloc_found) {
1575 if (r_type == MachO::ARM_RELOC_HALF) {
1576 if ((r_length & 0x1) == 1)
1577 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_HI16;
1578 else
1579 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM_LO16;
1580 }
1581 }
Kevin Enderbyf6d25852015-01-31 00:37:11 +00001582 const char *add = GuessSymbolName(value, info->AddrMap);
Kevin Enderby930fdc72014-11-06 19:00:13 +00001583 if (add != nullptr) {
1584 op_info->AddSymbol.Name = add;
1585 return 1;
1586 }
1587 op_info->AddSymbol.Value = value;
1588 return 1;
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001589 } else if (Arch == Triple::aarch64) {
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001590 if (Offset != 0 || Size != 4)
1591 return 0;
1592 // First search the section's relocation entries (if any) for an entry
1593 // for this section offset.
1594 uint64_t sect_addr = info->S.getAddress();
1595 uint64_t sect_offset = (Pc + Offset) - sect_addr;
1596 bool reloc_found = false;
1597 DataRefImpl Rel;
1598 MachO::any_relocation_info RE;
1599 bool isExtern = false;
1600 SymbolRef Symbol;
1601 uint32_t r_type = 0;
1602 for (const RelocationRef &Reloc : info->S.relocations()) {
1603 uint64_t RelocOffset;
1604 Reloc.getOffset(RelocOffset);
1605 if (RelocOffset == sect_offset) {
1606 Rel = Reloc.getRawDataRefImpl();
1607 RE = info->O->getRelocation(Rel);
1608 r_type = info->O->getAnyRelocationType(RE);
1609 if (r_type == MachO::ARM64_RELOC_ADDEND) {
1610 DataRefImpl RelNext = Rel;
1611 info->O->moveRelocationNext(RelNext);
1612 MachO::any_relocation_info RENext = info->O->getRelocation(RelNext);
1613 if (value == 0) {
1614 value = info->O->getPlainRelocationSymbolNum(RENext);
1615 op_info->Value = value;
1616 }
1617 }
1618 // NOTE: Scattered relocations don't exist on arm64.
1619 isExtern = info->O->getPlainRelocationExternal(RE);
1620 if (isExtern) {
1621 symbol_iterator RelocSym = Reloc.getSymbol();
1622 Symbol = *RelocSym;
1623 }
1624 reloc_found = true;
1625 break;
1626 }
1627 }
1628 if (reloc_found && isExtern) {
1629 StringRef SymName;
1630 Symbol.getName(SymName);
1631 const char *name = SymName.data();
1632 op_info->AddSymbol.Present = 1;
1633 op_info->AddSymbol.Name = name;
1634
1635 switch (r_type) {
1636 case MachO::ARM64_RELOC_PAGE21:
1637 /* @page */
1638 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGE;
1639 break;
1640 case MachO::ARM64_RELOC_PAGEOFF12:
1641 /* @pageoff */
1642 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_PAGEOFF;
1643 break;
1644 case MachO::ARM64_RELOC_GOT_LOAD_PAGE21:
1645 /* @gotpage */
1646 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGE;
1647 break;
1648 case MachO::ARM64_RELOC_GOT_LOAD_PAGEOFF12:
1649 /* @gotpageoff */
1650 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_GOTPAGEOFF;
1651 break;
1652 case MachO::ARM64_RELOC_TLVP_LOAD_PAGE21:
1653 /* @tvlppage is not implemented in llvm-mc */
1654 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVP;
1655 break;
1656 case MachO::ARM64_RELOC_TLVP_LOAD_PAGEOFF12:
1657 /* @tvlppageoff is not implemented in llvm-mc */
1658 op_info->VariantKind = LLVMDisassembler_VariantKind_ARM64_TLVOFF;
1659 break;
1660 default:
1661 case MachO::ARM64_RELOC_BRANCH26:
1662 op_info->VariantKind = LLVMDisassembler_VariantKind_None;
1663 break;
1664 }
1665 return 1;
1666 }
Kevin Enderby98c9acc2014-09-16 18:00:57 +00001667 return 0;
1668 } else {
1669 return 0;
1670 }
1671}
1672
Kevin Enderbybf246f52014-09-24 23:08:22 +00001673// GuessCstringPointer is passed the address of what might be a pointer to a
1674// literal string in a cstring section. If that address is in a cstring section
1675// it returns a pointer to that string. Else it returns nullptr.
1676const char *GuessCstringPointer(uint64_t ReferenceValue,
1677 struct DisassembleInfo *info) {
1678 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1679 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1680 for (unsigned I = 0;; ++I) {
1681 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1682 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1683 for (unsigned J = 0; J < Seg.nsects; ++J) {
1684 MachO::section_64 Sec = info->O->getSection64(Load, J);
1685 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1686 if (section_type == MachO::S_CSTRING_LITERALS &&
1687 ReferenceValue >= Sec.addr &&
1688 ReferenceValue < Sec.addr + Sec.size) {
1689 uint64_t sect_offset = ReferenceValue - Sec.addr;
1690 uint64_t object_offset = Sec.offset + sect_offset;
1691 StringRef MachOContents = info->O->getData();
1692 uint64_t object_size = MachOContents.size();
1693 const char *object_addr = (const char *)MachOContents.data();
1694 if (object_offset < object_size) {
1695 const char *name = object_addr + object_offset;
1696 return name;
1697 } else {
1698 return nullptr;
1699 }
1700 }
1701 }
1702 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1703 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1704 for (unsigned J = 0; J < Seg.nsects; ++J) {
1705 MachO::section Sec = info->O->getSection(Load, J);
1706 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1707 if (section_type == MachO::S_CSTRING_LITERALS &&
1708 ReferenceValue >= Sec.addr &&
1709 ReferenceValue < Sec.addr + Sec.size) {
1710 uint64_t sect_offset = ReferenceValue - Sec.addr;
1711 uint64_t object_offset = Sec.offset + sect_offset;
1712 StringRef MachOContents = info->O->getData();
1713 uint64_t object_size = MachOContents.size();
1714 const char *object_addr = (const char *)MachOContents.data();
1715 if (object_offset < object_size) {
1716 const char *name = object_addr + object_offset;
1717 return name;
1718 } else {
1719 return nullptr;
1720 }
1721 }
1722 }
1723 }
1724 if (I == LoadCommandCount - 1)
1725 break;
1726 else
1727 Load = info->O->getNextLoadCommandInfo(Load);
1728 }
1729 return nullptr;
1730}
1731
Kevin Enderby85974882014-09-26 22:20:44 +00001732// GuessIndirectSymbol returns the name of the indirect symbol for the
1733// ReferenceValue passed in or nullptr. This is used when ReferenceValue maybe
1734// an address of a symbol stub or a lazy or non-lazy pointer to associate the
1735// symbol name being referenced by the stub or pointer.
1736static const char *GuessIndirectSymbol(uint64_t ReferenceValue,
1737 struct DisassembleInfo *info) {
1738 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1739 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1740 MachO::dysymtab_command Dysymtab = info->O->getDysymtabLoadCommand();
1741 MachO::symtab_command Symtab = info->O->getSymtabLoadCommand();
1742 for (unsigned I = 0;; ++I) {
1743 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1744 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1745 for (unsigned J = 0; J < Seg.nsects; ++J) {
1746 MachO::section_64 Sec = info->O->getSection64(Load, J);
1747 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1748 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1749 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1750 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1751 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1752 section_type == MachO::S_SYMBOL_STUBS) &&
1753 ReferenceValue >= Sec.addr &&
1754 ReferenceValue < Sec.addr + Sec.size) {
1755 uint32_t stride;
1756 if (section_type == MachO::S_SYMBOL_STUBS)
1757 stride = Sec.reserved2;
1758 else
1759 stride = 8;
1760 if (stride == 0)
1761 return nullptr;
1762 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1763 if (index < Dysymtab.nindirectsyms) {
1764 uint32_t indirect_symbol =
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001765 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
Kevin Enderby85974882014-09-26 22:20:44 +00001766 if (indirect_symbol < Symtab.nsyms) {
1767 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1768 SymbolRef Symbol = *Sym;
1769 StringRef SymName;
1770 Symbol.getName(SymName);
1771 const char *name = SymName.data();
1772 return name;
1773 }
1774 }
1775 }
1776 }
1777 } else if (Load.C.cmd == MachO::LC_SEGMENT) {
1778 MachO::segment_command Seg = info->O->getSegmentLoadCommand(Load);
1779 for (unsigned J = 0; J < Seg.nsects; ++J) {
1780 MachO::section Sec = info->O->getSection(Load, J);
1781 uint32_t section_type = Sec.flags & MachO::SECTION_TYPE;
1782 if ((section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
1783 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
1784 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
1785 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS ||
1786 section_type == MachO::S_SYMBOL_STUBS) &&
1787 ReferenceValue >= Sec.addr &&
1788 ReferenceValue < Sec.addr + Sec.size) {
1789 uint32_t stride;
1790 if (section_type == MachO::S_SYMBOL_STUBS)
1791 stride = Sec.reserved2;
1792 else
1793 stride = 4;
1794 if (stride == 0)
1795 return nullptr;
1796 uint32_t index = Sec.reserved1 + (ReferenceValue - Sec.addr) / stride;
1797 if (index < Dysymtab.nindirectsyms) {
1798 uint32_t indirect_symbol =
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001799 info->O->getIndirectSymbolTableEntry(Dysymtab, index);
Kevin Enderby85974882014-09-26 22:20:44 +00001800 if (indirect_symbol < Symtab.nsyms) {
1801 symbol_iterator Sym = info->O->getSymbolByIndex(indirect_symbol);
1802 SymbolRef Symbol = *Sym;
1803 StringRef SymName;
1804 Symbol.getName(SymName);
1805 const char *name = SymName.data();
1806 return name;
1807 }
1808 }
1809 }
1810 }
1811 }
1812 if (I == LoadCommandCount - 1)
1813 break;
1814 else
1815 Load = info->O->getNextLoadCommandInfo(Load);
1816 }
1817 return nullptr;
1818}
1819
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001820// method_reference() is called passing it the ReferenceName that might be
1821// a reference it to an Objective-C method call. If so then it allocates and
1822// assembles a method call string with the values last seen and saved in
1823// the DisassembleInfo's class_name and selector_name fields. This is saved
1824// into the method field of the info and any previous string is free'ed.
1825// Then the class_name field in the info is set to nullptr. The method call
1826// string is set into ReferenceName and ReferenceType is set to
1827// LLVMDisassembler_ReferenceType_Out_Objc_Message. If this not a method call
1828// then both ReferenceType and ReferenceName are left unchanged.
1829static void method_reference(struct DisassembleInfo *info,
1830 uint64_t *ReferenceType,
1831 const char **ReferenceName) {
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001832 unsigned int Arch = info->O->getArch();
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001833 if (*ReferenceName != nullptr) {
1834 if (strcmp(*ReferenceName, "_objc_msgSend") == 0) {
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001835 if (info->selector_name != nullptr) {
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001836 if (info->method != nullptr)
1837 free(info->method);
1838 if (info->class_name != nullptr) {
1839 info->method = (char *)malloc(5 + strlen(info->class_name) +
1840 strlen(info->selector_name));
1841 if (info->method != nullptr) {
1842 strcpy(info->method, "+[");
1843 strcat(info->method, info->class_name);
1844 strcat(info->method, " ");
1845 strcat(info->method, info->selector_name);
1846 strcat(info->method, "]");
1847 *ReferenceName = info->method;
1848 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1849 }
1850 } else {
1851 info->method = (char *)malloc(9 + strlen(info->selector_name));
1852 if (info->method != nullptr) {
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001853 if (Arch == Triple::x86_64)
1854 strcpy(info->method, "-[%rdi ");
1855 else if (Arch == Triple::aarch64)
1856 strcpy(info->method, "-[x0 ");
1857 else
1858 strcpy(info->method, "-[r? ");
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001859 strcat(info->method, info->selector_name);
1860 strcat(info->method, "]");
1861 *ReferenceName = info->method;
1862 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1863 }
1864 }
1865 info->class_name = nullptr;
1866 }
1867 } else if (strcmp(*ReferenceName, "_objc_msgSendSuper2") == 0) {
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001868 if (info->selector_name != nullptr) {
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001869 if (info->method != nullptr)
1870 free(info->method);
1871 info->method = (char *)malloc(17 + strlen(info->selector_name));
1872 if (info->method != nullptr) {
Kevin Enderbyae3c1262014-11-14 21:52:18 +00001873 if (Arch == Triple::x86_64)
1874 strcpy(info->method, "-[[%rdi super] ");
1875 else if (Arch == Triple::aarch64)
1876 strcpy(info->method, "-[[x0 super] ");
1877 else
1878 strcpy(info->method, "-[[r? super] ");
Kevin Enderby6f326ce2014-10-23 19:37:31 +00001879 strcat(info->method, info->selector_name);
1880 strcat(info->method, "]");
1881 *ReferenceName = info->method;
1882 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message;
1883 }
1884 info->class_name = nullptr;
1885 }
1886 }
1887 }
1888}
1889
1890// GuessPointerPointer() is passed the address of what might be a pointer to
1891// a reference to an Objective-C class, selector, message ref or cfstring.
1892// If so the value of the pointer is returned and one of the booleans are set
1893// to true. If not zero is returned and all the booleans are set to false.
1894static uint64_t GuessPointerPointer(uint64_t ReferenceValue,
1895 struct DisassembleInfo *info,
1896 bool &classref, bool &selref, bool &msgref,
1897 bool &cfstring) {
1898 classref = false;
1899 selref = false;
1900 msgref = false;
1901 cfstring = false;
1902 uint32_t LoadCommandCount = info->O->getHeader().ncmds;
1903 MachOObjectFile::LoadCommandInfo Load = info->O->getFirstLoadCommandInfo();
1904 for (unsigned I = 0;; ++I) {
1905 if (Load.C.cmd == MachO::LC_SEGMENT_64) {
1906 MachO::segment_command_64 Seg = info->O->getSegment64LoadCommand(Load);
1907 for (unsigned J = 0; J < Seg.nsects; ++J) {
1908 MachO::section_64 Sec = info->O->getSection64(Load, J);
1909 if ((strncmp(Sec.sectname, "__objc_selrefs", 16) == 0 ||
1910 strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1911 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0 ||
1912 strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 ||
1913 strncmp(Sec.sectname, "__cfstring", 16) == 0) &&
1914 ReferenceValue >= Sec.addr &&
1915 ReferenceValue < Sec.addr + Sec.size) {
1916 uint64_t sect_offset = ReferenceValue - Sec.addr;
1917 uint64_t object_offset = Sec.offset + sect_offset;
1918 StringRef MachOContents = info->O->getData();
1919 uint64_t object_size = MachOContents.size();
1920 const char *object_addr = (const char *)MachOContents.data();
1921 if (object_offset < object_size) {
1922 uint64_t pointer_value;
1923 memcpy(&pointer_value, object_addr + object_offset,
1924 sizeof(uint64_t));
1925 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1926 sys::swapByteOrder(pointer_value);
1927 if (strncmp(Sec.sectname, "__objc_selrefs", 16) == 0)
1928 selref = true;
1929 else if (strncmp(Sec.sectname, "__objc_classrefs", 16) == 0 ||
1930 strncmp(Sec.sectname, "__objc_superrefs", 16) == 0)
1931 classref = true;
1932 else if (strncmp(Sec.sectname, "__objc_msgrefs", 16) == 0 &&
1933 ReferenceValue + 8 < Sec.addr + Sec.size) {
1934 msgref = true;
1935 memcpy(&pointer_value, object_addr + object_offset + 8,
1936 sizeof(uint64_t));
1937 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
1938 sys::swapByteOrder(pointer_value);
1939 } else if (strncmp(Sec.sectname, "__cfstring", 16) == 0)
1940 cfstring = true;
1941 return pointer_value;
1942 } else {
1943 return 0;
1944 }
1945 }
1946 }
1947 }
1948 // TODO: Look for LC_SEGMENT for 32-bit Mach-O files.
1949 if (I == LoadCommandCount - 1)
1950 break;
1951 else
1952 Load = info->O->getNextLoadCommandInfo(Load);
1953 }
1954 return 0;
1955}
1956
1957// get_pointer_64 returns a pointer to the bytes in the object file at the
1958// Address from a section in the Mach-O file. And indirectly returns the
1959// offset into the section, number of bytes left in the section past the offset
1960// and which section is was being referenced. If the Address is not in a
1961// section nullptr is returned.
1962const char *get_pointer_64(uint64_t Address, uint32_t &offset, uint32_t &left,
1963 SectionRef &S, DisassembleInfo *info) {
1964 offset = 0;
1965 left = 0;
1966 S = SectionRef();
1967 for (unsigned SectIdx = 0; SectIdx != info->Sections->size(); SectIdx++) {
1968 uint64_t SectAddress = ((*(info->Sections))[SectIdx]).getAddress();
1969 uint64_t SectSize = ((*(info->Sections))[SectIdx]).getSize();
1970 if (Address >= SectAddress && Address < SectAddress + SectSize) {
1971 S = (*(info->Sections))[SectIdx];
1972 offset = Address - SectAddress;
1973 left = SectSize - offset;
1974 StringRef SectContents;
1975 ((*(info->Sections))[SectIdx]).getContents(SectContents);
1976 return SectContents.data() + offset;
1977 }
1978 }
1979 return nullptr;
1980}
1981
1982// get_symbol_64() returns the name of a symbol (or nullptr) and the address of
1983// the symbol indirectly through n_value. Based on the relocation information
1984// for the specified section offset in the specified section reference.
1985const char *get_symbol_64(uint32_t sect_offset, SectionRef S,
1986 DisassembleInfo *info, uint64_t &n_value) {
1987 n_value = 0;
1988 if (info->verbose == false)
1989 return nullptr;
1990
1991 // See if there is an external relocation entry at the sect_offset.
1992 bool reloc_found = false;
1993 DataRefImpl Rel;
1994 MachO::any_relocation_info RE;
1995 bool isExtern = false;
1996 SymbolRef Symbol;
1997 for (const RelocationRef &Reloc : S.relocations()) {
1998 uint64_t RelocOffset;
1999 Reloc.getOffset(RelocOffset);
2000 if (RelocOffset == sect_offset) {
2001 Rel = Reloc.getRawDataRefImpl();
2002 RE = info->O->getRelocation(Rel);
2003 if (info->O->isRelocationScattered(RE))
2004 continue;
2005 isExtern = info->O->getPlainRelocationExternal(RE);
2006 if (isExtern) {
2007 symbol_iterator RelocSym = Reloc.getSymbol();
2008 Symbol = *RelocSym;
2009 }
2010 reloc_found = true;
2011 break;
2012 }
2013 }
2014 // If there is an external relocation entry for a symbol in this section
2015 // at this section_offset then use that symbol's value for the n_value
2016 // and return its name.
2017 const char *SymbolName = nullptr;
2018 if (reloc_found && isExtern) {
2019 Symbol.getAddress(n_value);
2020 StringRef name;
2021 Symbol.getName(name);
2022 if (!name.empty()) {
2023 SymbolName = name.data();
2024 return SymbolName;
2025 }
2026 }
2027
2028 // TODO: For fully linked images, look through the external relocation
2029 // entries off the dynamic symtab command. For these the r_offset is from the
2030 // start of the first writeable segment in the Mach-O file. So the offset
2031 // to this section from that segment is passed to this routine by the caller,
2032 // as the database_offset. Which is the difference of the section's starting
2033 // address and the first writable segment.
2034 //
2035 // NOTE: need add passing the database_offset to this routine.
2036
2037 // TODO: We did not find an external relocation entry so look up the
2038 // ReferenceValue as an address of a symbol and if found return that symbol's
2039 // name.
2040 //
2041 // NOTE: need add passing the ReferenceValue to this routine. Then that code
2042 // would simply be this:
Kevin Enderbyf6d25852015-01-31 00:37:11 +00002043 // SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002044
2045 return SymbolName;
2046}
2047
2048// These are structs in the Objective-C meta data and read to produce the
2049// comments for disassembly. While these are part of the ABI they are no
2050// public defintions. So the are here not in include/llvm/Support/MachO.h .
2051
2052// The cfstring object in a 64-bit Mach-O file.
2053struct cfstring64_t {
2054 uint64_t isa; // class64_t * (64-bit pointer)
2055 uint64_t flags; // flag bits
2056 uint64_t characters; // char * (64-bit pointer)
2057 uint64_t length; // number of non-NULL characters in above
2058};
2059
2060// The class object in a 64-bit Mach-O file.
2061struct class64_t {
2062 uint64_t isa; // class64_t * (64-bit pointer)
2063 uint64_t superclass; // class64_t * (64-bit pointer)
2064 uint64_t cache; // Cache (64-bit pointer)
2065 uint64_t vtable; // IMP * (64-bit pointer)
2066 uint64_t data; // class_ro64_t * (64-bit pointer)
2067};
2068
2069struct class_ro64_t {
2070 uint32_t flags;
2071 uint32_t instanceStart;
2072 uint32_t instanceSize;
2073 uint32_t reserved;
2074 uint64_t ivarLayout; // const uint8_t * (64-bit pointer)
2075 uint64_t name; // const char * (64-bit pointer)
2076 uint64_t baseMethods; // const method_list_t * (64-bit pointer)
2077 uint64_t baseProtocols; // const protocol_list_t * (64-bit pointer)
2078 uint64_t ivars; // const ivar_list_t * (64-bit pointer)
2079 uint64_t weakIvarLayout; // const uint8_t * (64-bit pointer)
2080 uint64_t baseProperties; // const struct objc_property_list (64-bit pointer)
2081};
2082
2083inline void swapStruct(struct cfstring64_t &cfs) {
2084 sys::swapByteOrder(cfs.isa);
2085 sys::swapByteOrder(cfs.flags);
2086 sys::swapByteOrder(cfs.characters);
2087 sys::swapByteOrder(cfs.length);
2088}
2089
2090inline void swapStruct(struct class64_t &c) {
2091 sys::swapByteOrder(c.isa);
2092 sys::swapByteOrder(c.superclass);
2093 sys::swapByteOrder(c.cache);
2094 sys::swapByteOrder(c.vtable);
2095 sys::swapByteOrder(c.data);
2096}
2097
2098inline void swapStruct(struct class_ro64_t &cro) {
2099 sys::swapByteOrder(cro.flags);
2100 sys::swapByteOrder(cro.instanceStart);
2101 sys::swapByteOrder(cro.instanceSize);
2102 sys::swapByteOrder(cro.reserved);
2103 sys::swapByteOrder(cro.ivarLayout);
2104 sys::swapByteOrder(cro.name);
2105 sys::swapByteOrder(cro.baseMethods);
2106 sys::swapByteOrder(cro.baseProtocols);
2107 sys::swapByteOrder(cro.ivars);
2108 sys::swapByteOrder(cro.weakIvarLayout);
2109 sys::swapByteOrder(cro.baseProperties);
2110}
2111
2112static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
2113 struct DisassembleInfo *info);
2114
2115// get_objc2_64bit_class_name() is used for disassembly and is passed a pointer
2116// to an Objective-C class and returns the class name. It is also passed the
2117// address of the pointer, so when the pointer is zero as it can be in an .o
2118// file, that is used to look for an external relocation entry with a symbol
2119// name.
2120const char *get_objc2_64bit_class_name(uint64_t pointer_value,
2121 uint64_t ReferenceValue,
2122 struct DisassembleInfo *info) {
2123 const char *r;
2124 uint32_t offset, left;
2125 SectionRef S;
2126
2127 // The pointer_value can be 0 in an object file and have a relocation
2128 // entry for the class symbol at the ReferenceValue (the address of the
2129 // pointer).
2130 if (pointer_value == 0) {
2131 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2132 if (r == nullptr || left < sizeof(uint64_t))
2133 return nullptr;
2134 uint64_t n_value;
2135 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2136 if (symbol_name == nullptr)
2137 return nullptr;
Hans Wennborgdb53e302014-10-23 21:59:17 +00002138 const char *class_name = strrchr(symbol_name, '$');
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002139 if (class_name != nullptr && class_name[1] == '_' && class_name[2] != '\0')
2140 return class_name + 2;
2141 else
2142 return nullptr;
2143 }
2144
2145 // The case were the pointer_value is non-zero and points to a class defined
2146 // in this Mach-O file.
2147 r = get_pointer_64(pointer_value, offset, left, S, info);
2148 if (r == nullptr || left < sizeof(struct class64_t))
2149 return nullptr;
2150 struct class64_t c;
2151 memcpy(&c, r, sizeof(struct class64_t));
2152 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2153 swapStruct(c);
2154 if (c.data == 0)
2155 return nullptr;
2156 r = get_pointer_64(c.data, offset, left, S, info);
2157 if (r == nullptr || left < sizeof(struct class_ro64_t))
2158 return nullptr;
2159 struct class_ro64_t cro;
2160 memcpy(&cro, r, sizeof(struct class_ro64_t));
2161 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2162 swapStruct(cro);
2163 if (cro.name == 0)
2164 return nullptr;
2165 const char *name = get_pointer_64(cro.name, offset, left, S, info);
2166 return name;
2167}
2168
2169// get_objc2_64bit_cfstring_name is used for disassembly and is passed a
2170// pointer to a cfstring and returns its name or nullptr.
2171const char *get_objc2_64bit_cfstring_name(uint64_t ReferenceValue,
2172 struct DisassembleInfo *info) {
2173 const char *r, *name;
2174 uint32_t offset, left;
2175 SectionRef S;
2176 struct cfstring64_t cfs;
2177 uint64_t cfs_characters;
2178
2179 r = get_pointer_64(ReferenceValue, offset, left, S, info);
2180 if (r == nullptr || left < sizeof(struct cfstring64_t))
2181 return nullptr;
2182 memcpy(&cfs, r, sizeof(struct cfstring64_t));
2183 if (info->O->isLittleEndian() != sys::IsLittleEndianHost)
2184 swapStruct(cfs);
2185 if (cfs.characters == 0) {
2186 uint64_t n_value;
2187 const char *symbol_name = get_symbol_64(
2188 offset + offsetof(struct cfstring64_t, characters), S, info, n_value);
2189 if (symbol_name == nullptr)
2190 return nullptr;
2191 cfs_characters = n_value;
2192 } else
2193 cfs_characters = cfs.characters;
2194 name = get_pointer_64(cfs_characters, offset, left, S, info);
2195
2196 return name;
2197}
2198
2199// get_objc2_64bit_selref() is used for disassembly and is passed a the address
2200// of a pointer to an Objective-C selector reference when the pointer value is
2201// zero as in a .o file and is likely to have a external relocation entry with
2202// who's symbol's n_value is the real pointer to the selector name. If that is
2203// the case the real pointer to the selector name is returned else 0 is
2204// returned
2205uint64_t get_objc2_64bit_selref(uint64_t ReferenceValue,
2206 struct DisassembleInfo *info) {
2207 uint32_t offset, left;
2208 SectionRef S;
2209
2210 const char *r = get_pointer_64(ReferenceValue, offset, left, S, info);
2211 if (r == nullptr || left < sizeof(uint64_t))
2212 return 0;
2213 uint64_t n_value;
2214 const char *symbol_name = get_symbol_64(offset, S, info, n_value);
2215 if (symbol_name == nullptr)
2216 return 0;
2217 return n_value;
2218}
2219
Kevin Enderbybf246f52014-09-24 23:08:22 +00002220// GuessLiteralPointer returns a string which for the item in the Mach-O file
2221// for the address passed in as ReferenceValue for printing as a comment with
2222// the instruction and also returns the corresponding type of that item
2223// indirectly through ReferenceType.
2224//
2225// If ReferenceValue is an address of literal cstring then a pointer to the
2226// cstring is returned and ReferenceType is set to
2227// LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr .
2228//
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002229// If ReferenceValue is an address of an Objective-C CFString, Selector ref or
2230// Class ref that name is returned and the ReferenceType is set accordingly.
2231//
2232// Lastly, literals which are Symbol address in a literal pool are looked for
2233// and if found the symbol name is returned and ReferenceType is set to
2234// LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr .
2235//
2236// If there is no item in the Mach-O file for the address passed in as
2237// ReferenceValue nullptr is returned and ReferenceType is unchanged.
Kevin Enderbybf246f52014-09-24 23:08:22 +00002238const char *GuessLiteralPointer(uint64_t ReferenceValue, uint64_t ReferencePC,
2239 uint64_t *ReferenceType,
2240 struct DisassembleInfo *info) {
Kevin Enderbybf246f52014-09-24 23:08:22 +00002241 // First see if there is an external relocation entry at the ReferencePC.
Rafael Espindola80291272014-10-08 15:28:58 +00002242 uint64_t sect_addr = info->S.getAddress();
Kevin Enderbybf246f52014-09-24 23:08:22 +00002243 uint64_t sect_offset = ReferencePC - sect_addr;
2244 bool reloc_found = false;
2245 DataRefImpl Rel;
2246 MachO::any_relocation_info RE;
2247 bool isExtern = false;
2248 SymbolRef Symbol;
2249 for (const RelocationRef &Reloc : info->S.relocations()) {
2250 uint64_t RelocOffset;
2251 Reloc.getOffset(RelocOffset);
2252 if (RelocOffset == sect_offset) {
2253 Rel = Reloc.getRawDataRefImpl();
2254 RE = info->O->getRelocation(Rel);
2255 if (info->O->isRelocationScattered(RE))
2256 continue;
2257 isExtern = info->O->getPlainRelocationExternal(RE);
2258 if (isExtern) {
2259 symbol_iterator RelocSym = Reloc.getSymbol();
2260 Symbol = *RelocSym;
2261 }
2262 reloc_found = true;
2263 break;
2264 }
2265 }
2266 // If there is an external relocation entry for a symbol in a section
2267 // then used that symbol's value for the value of the reference.
2268 if (reloc_found && isExtern) {
2269 if (info->O->getAnyRelocationPCRel(RE)) {
2270 unsigned Type = info->O->getAnyRelocationType(RE);
2271 if (Type == MachO::X86_64_RELOC_SIGNED) {
2272 Symbol.getAddress(ReferenceValue);
2273 }
2274 }
2275 }
2276
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002277 // Look for literals such as Objective-C CFStrings refs, Selector refs,
2278 // Message refs and Class refs.
2279 bool classref, selref, msgref, cfstring;
2280 uint64_t pointer_value = GuessPointerPointer(ReferenceValue, info, classref,
2281 selref, msgref, cfstring);
2282 if (classref == true && pointer_value == 0) {
2283 // Note the ReferenceValue is a pointer into the __objc_classrefs section.
2284 // And the pointer_value in that section is typically zero as it will be
2285 // set by dyld as part of the "bind information".
2286 const char *name = get_dyld_bind_info_symbolname(ReferenceValue, info);
2287 if (name != nullptr) {
2288 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
Hans Wennborgdb53e302014-10-23 21:59:17 +00002289 const char *class_name = strrchr(name, '$');
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002290 if (class_name != nullptr && class_name[1] == '_' &&
2291 class_name[2] != '\0') {
2292 info->class_name = class_name + 2;
2293 return name;
2294 }
2295 }
2296 }
Kevin Enderbybf246f52014-09-24 23:08:22 +00002297
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002298 if (classref == true) {
2299 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Class_Ref;
2300 const char *name =
2301 get_objc2_64bit_class_name(pointer_value, ReferenceValue, info);
2302 if (name != nullptr)
2303 info->class_name = name;
2304 else
2305 name = "bad class ref";
Kevin Enderbybf246f52014-09-24 23:08:22 +00002306 return name;
2307 }
2308
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002309 if (cfstring == true) {
2310 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_CFString_Ref;
2311 const char *name = get_objc2_64bit_cfstring_name(ReferenceValue, info);
2312 return name;
2313 }
2314
2315 if (selref == true && pointer_value == 0)
2316 pointer_value = get_objc2_64bit_selref(ReferenceValue, info);
2317
2318 if (pointer_value != 0)
2319 ReferenceValue = pointer_value;
2320
2321 const char *name = GuessCstringPointer(ReferenceValue, info);
2322 if (name) {
2323 if (pointer_value != 0 && selref == true) {
2324 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Selector_Ref;
2325 info->selector_name = name;
2326 } else if (pointer_value != 0 && msgref == true) {
2327 info->class_name = nullptr;
2328 *ReferenceType = LLVMDisassembler_ReferenceType_Out_Objc_Message_Ref;
2329 info->selector_name = name;
2330 } else
2331 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_CstrAddr;
2332 return name;
2333 }
2334
2335 // Lastly look for an indirect symbol with this ReferenceValue which is in
2336 // a literal pool. If found return that symbol name.
2337 name = GuessIndirectSymbol(ReferenceValue, info);
2338 if (name) {
2339 *ReferenceType = LLVMDisassembler_ReferenceType_Out_LitPool_SymAddr;
2340 return name;
2341 }
Kevin Enderbybf246f52014-09-24 23:08:22 +00002342
2343 return nullptr;
2344}
2345
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002346// SymbolizerSymbolLookUp is the symbol lookup function passed when creating
Kevin Enderbybf246f52014-09-24 23:08:22 +00002347// the Symbolizer. It looks up the ReferenceValue using the info passed via the
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002348// pointer to the struct DisassembleInfo that was passed when MCSymbolizer
2349// is created and returns the symbol name that matches the ReferenceValue or
2350// nullptr if none. The ReferenceType is passed in for the IN type of
2351// reference the instruction is making from the values in defined in the header
2352// "llvm-c/Disassembler.h". On return the ReferenceType can set to a specific
2353// Out type and the ReferenceName will also be set which is added as a comment
2354// to the disassembled instruction.
2355//
Kevin Enderby04bf6932014-10-28 23:39:46 +00002356#if HAVE_CXXABI_H
2357// If the symbol name is a C++ mangled name then the demangled name is
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002358// returned through ReferenceName and ReferenceType is set to
2359// LLVMDisassembler_ReferenceType_DeMangled_Name .
Kevin Enderby04bf6932014-10-28 23:39:46 +00002360#endif
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002361//
2362// When this is called to get a symbol name for a branch target then the
2363// ReferenceType will be LLVMDisassembler_ReferenceType_In_Branch and then
2364// SymbolValue will be looked for in the indirect symbol table to determine if
2365// it is an address for a symbol stub. If so then the symbol name for that
2366// stub is returned indirectly through ReferenceName and then ReferenceType is
2367// set to LLVMDisassembler_ReferenceType_Out_SymbolStub.
2368//
Kevin Enderbybf246f52014-09-24 23:08:22 +00002369// When this is called with an value loaded via a PC relative load then
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002370// ReferenceType will be LLVMDisassembler_ReferenceType_In_PCrel_Load then the
2371// SymbolValue is checked to be an address of literal pointer, symbol pointer,
2372// or an Objective-C meta data reference. If so the output ReferenceType is
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002373// set to correspond to that as well as setting the ReferenceName.
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002374const char *SymbolizerSymbolLookUp(void *DisInfo, uint64_t ReferenceValue,
2375 uint64_t *ReferenceType,
2376 uint64_t ReferencePC,
2377 const char **ReferenceName) {
2378 struct DisassembleInfo *info = (struct DisassembleInfo *)DisInfo;
Kevin Enderbybf246f52014-09-24 23:08:22 +00002379 // If no verbose symbolic information is wanted then just return nullptr.
2380 if (info->verbose == false) {
2381 *ReferenceName = nullptr;
2382 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002383 return nullptr;
2384 }
Kevin Enderbybf246f52014-09-24 23:08:22 +00002385
Kevin Enderbyf6d25852015-01-31 00:37:11 +00002386 const char *SymbolName = GuessSymbolName(ReferenceValue, info->AddrMap);
Kevin Enderbybf246f52014-09-24 23:08:22 +00002387
Kevin Enderby85974882014-09-26 22:20:44 +00002388 if (*ReferenceType == LLVMDisassembler_ReferenceType_In_Branch) {
2389 *ReferenceName = GuessIndirectSymbol(ReferenceValue, info);
Kevin Enderby04bf6932014-10-28 23:39:46 +00002390 if (*ReferenceName != nullptr) {
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002391 method_reference(info, ReferenceType, ReferenceName);
2392 if (*ReferenceType != LLVMDisassembler_ReferenceType_Out_Objc_Message)
2393 *ReferenceType = LLVMDisassembler_ReferenceType_Out_SymbolStub;
2394 } else
Kevin Enderby04bf6932014-10-28 23:39:46 +00002395#if HAVE_CXXABI_H
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002396 if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
Kevin Enderby04bf6932014-10-28 23:39:46 +00002397 if (info->demangled_name != nullptr)
2398 free(info->demangled_name);
2399 int status;
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002400 info->demangled_name =
2401 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
Kevin Enderby04bf6932014-10-28 23:39:46 +00002402 if (info->demangled_name != nullptr) {
2403 *ReferenceName = info->demangled_name;
2404 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2405 } else
2406 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2407 } else
2408#endif
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002409 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2410 } else if (*ReferenceType == LLVMDisassembler_ReferenceType_In_PCrel_Load) {
2411 *ReferenceName =
2412 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
Kevin Enderby85974882014-09-26 22:20:44 +00002413 if (*ReferenceName)
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002414 method_reference(info, ReferenceType, ReferenceName);
Kevin Enderby85974882014-09-26 22:20:44 +00002415 else
2416 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
Kevin Enderbyae3c1262014-11-14 21:52:18 +00002417 // If this is arm64 and the reference is an adrp instruction save the
2418 // instruction, passed in ReferenceValue and the address of the instruction
2419 // for use later if we see and add immediate instruction.
2420 } else if (info->O->getArch() == Triple::aarch64 &&
2421 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADRP) {
2422 info->adrp_inst = ReferenceValue;
2423 info->adrp_addr = ReferencePC;
2424 SymbolName = nullptr;
2425 *ReferenceName = nullptr;
2426 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2427 // If this is arm64 and reference is an add immediate instruction and we
2428 // have
2429 // seen an adrp instruction just before it and the adrp's Xd register
2430 // matches
2431 // this add's Xn register reconstruct the value being referenced and look to
2432 // see if it is a literal pointer. Note the add immediate instruction is
2433 // passed in ReferenceValue.
2434 } else if (info->O->getArch() == Triple::aarch64 &&
2435 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADDXri &&
2436 ReferencePC - 4 == info->adrp_addr &&
2437 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2438 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2439 uint32_t addxri_inst;
2440 uint64_t adrp_imm, addxri_imm;
2441
2442 adrp_imm =
2443 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2444 if (info->adrp_inst & 0x0200000)
2445 adrp_imm |= 0xfffffffffc000000LL;
2446
2447 addxri_inst = ReferenceValue;
2448 addxri_imm = (addxri_inst >> 10) & 0xfff;
2449 if (((addxri_inst >> 22) & 0x3) == 1)
2450 addxri_imm <<= 12;
2451
2452 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2453 (adrp_imm << 12) + addxri_imm;
2454
2455 *ReferenceName =
2456 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2457 if (*ReferenceName == nullptr)
2458 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2459 // If this is arm64 and the reference is a load register instruction and we
2460 // have seen an adrp instruction just before it and the adrp's Xd register
2461 // matches this add's Xn register reconstruct the value being referenced and
2462 // look to see if it is a literal pointer. Note the load register
2463 // instruction is passed in ReferenceValue.
2464 } else if (info->O->getArch() == Triple::aarch64 &&
2465 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXui &&
2466 ReferencePC - 4 == info->adrp_addr &&
2467 (info->adrp_inst & 0x9f000000) == 0x90000000 &&
2468 (info->adrp_inst & 0x1f) == ((ReferenceValue >> 5) & 0x1f)) {
2469 uint32_t ldrxui_inst;
2470 uint64_t adrp_imm, ldrxui_imm;
2471
2472 adrp_imm =
2473 ((info->adrp_inst & 0x00ffffe0) >> 3) | ((info->adrp_inst >> 29) & 0x3);
2474 if (info->adrp_inst & 0x0200000)
2475 adrp_imm |= 0xfffffffffc000000LL;
2476
2477 ldrxui_inst = ReferenceValue;
2478 ldrxui_imm = (ldrxui_inst >> 10) & 0xfff;
2479
2480 ReferenceValue = (info->adrp_addr & 0xfffffffffffff000LL) +
2481 (adrp_imm << 12) + (ldrxui_imm << 3);
2482
2483 *ReferenceName =
2484 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2485 if (*ReferenceName == nullptr)
2486 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2487 }
2488 // If this arm64 and is an load register (PC-relative) instruction the
2489 // ReferenceValue is the PC plus the immediate value.
2490 else if (info->O->getArch() == Triple::aarch64 &&
2491 (*ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_LDRXl ||
2492 *ReferenceType == LLVMDisassembler_ReferenceType_In_ARM64_ADR)) {
2493 *ReferenceName =
2494 GuessLiteralPointer(ReferenceValue, ReferencePC, ReferenceType, info);
2495 if (*ReferenceName == nullptr)
2496 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
Kevin Enderby85974882014-09-26 22:20:44 +00002497 }
Kevin Enderby04bf6932014-10-28 23:39:46 +00002498#if HAVE_CXXABI_H
2499 else if (SymbolName != nullptr && strncmp(SymbolName, "__Z", 3) == 0) {
2500 if (info->demangled_name != nullptr)
2501 free(info->demangled_name);
2502 int status;
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002503 info->demangled_name =
2504 abi::__cxa_demangle(SymbolName + 1, nullptr, nullptr, &status);
Kevin Enderby04bf6932014-10-28 23:39:46 +00002505 if (info->demangled_name != nullptr) {
2506 *ReferenceName = info->demangled_name;
2507 *ReferenceType = LLVMDisassembler_ReferenceType_DeMangled_Name;
2508 }
2509 }
2510#endif
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002511 else {
Kevin Enderbybf246f52014-09-24 23:08:22 +00002512 *ReferenceName = nullptr;
2513 *ReferenceType = LLVMDisassembler_ReferenceType_InOut_None;
2514 }
2515
2516 return SymbolName;
2517}
2518
Kevin Enderbybf246f52014-09-24 23:08:22 +00002519/// \brief Emits the comments that are stored in the CommentStream.
2520/// Each comment in the CommentStream must end with a newline.
2521static void emitComments(raw_svector_ostream &CommentStream,
2522 SmallString<128> &CommentsToEmit,
2523 formatted_raw_ostream &FormattedOS,
2524 const MCAsmInfo &MAI) {
2525 // Flush the stream before taking its content.
2526 CommentStream.flush();
2527 StringRef Comments = CommentsToEmit.str();
2528 // Get the default information for printing a comment.
2529 const char *CommentBegin = MAI.getCommentString();
2530 unsigned CommentColumn = MAI.getCommentColumn();
2531 bool IsFirst = true;
2532 while (!Comments.empty()) {
2533 if (!IsFirst)
2534 FormattedOS << '\n';
2535 // Emit a line of comments.
2536 FormattedOS.PadToColumn(CommentColumn);
2537 size_t Position = Comments.find('\n');
2538 FormattedOS << CommentBegin << ' ' << Comments.substr(0, Position);
2539 // Move after the newline character.
2540 Comments = Comments.substr(Position + 1);
2541 IsFirst = false;
2542 }
2543 FormattedOS.flush();
2544
2545 // Tell the comment stream that the vector changed underneath it.
2546 CommentsToEmit.clear();
2547 CommentStream.resync();
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002548}
2549
Kevin Enderbye2297dd2015-01-07 21:02:18 +00002550static void DisassembleMachO(StringRef Filename, MachOObjectFile *MachOOF) {
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002551 const char *McpuDefault = nullptr;
2552 const Target *ThumbTarget = nullptr;
2553 const Target *TheTarget = GetTarget(MachOOF, &McpuDefault, &ThumbTarget);
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002554 if (!TheTarget) {
2555 // GetTarget prints out stuff.
2556 return;
2557 }
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002558 if (MCPU.empty() && McpuDefault)
2559 MCPU = McpuDefault;
2560
Ahmed Charles56440fd2014-03-06 05:51:42 +00002561 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002562 std::unique_ptr<const MCInstrInfo> ThumbInstrInfo;
Kevin Enderbyae3c1262014-11-14 21:52:18 +00002563 if (ThumbTarget)
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002564 ThumbInstrInfo.reset(ThumbTarget->createMCInstrInfo());
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002565
Kevin Enderbyc9595622014-08-06 23:24:41 +00002566 // Package up features to be passed to target/subtarget
2567 std::string FeaturesStr;
2568 if (MAttrs.size()) {
2569 SubtargetFeatures Features;
2570 for (unsigned i = 0; i != MAttrs.size(); ++i)
2571 Features.AddFeature(MAttrs[i]);
2572 FeaturesStr = Features.getString();
2573 }
2574
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002575 // Set up disassembler.
Ahmed Charles56440fd2014-03-06 05:51:42 +00002576 std::unique_ptr<const MCRegisterInfo> MRI(
2577 TheTarget->createMCRegInfo(TripleName));
2578 std::unique_ptr<const MCAsmInfo> AsmInfo(
Rafael Espindola227144c2013-05-13 01:16:13 +00002579 TheTarget->createMCAsmInfo(*MRI, TripleName));
Ahmed Charles56440fd2014-03-06 05:51:42 +00002580 std::unique_ptr<const MCSubtargetInfo> STI(
Kevin Enderbyc9595622014-08-06 23:24:41 +00002581 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
Craig Toppere6cb63e2014-04-25 04:24:47 +00002582 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002583 std::unique_ptr<MCDisassembler> DisAsm(
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002584 TheTarget->createMCDisassembler(*STI, Ctx));
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002585 std::unique_ptr<MCSymbolizer> Symbolizer;
2586 struct DisassembleInfo SymbolizerInfo;
2587 std::unique_ptr<MCRelocationInfo> RelInfo(
2588 TheTarget->createMCRelocationInfo(TripleName, Ctx));
2589 if (RelInfo) {
2590 Symbolizer.reset(TheTarget->createMCSymbolizer(
2591 TripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
David Blaikie186db432015-01-18 20:45:48 +00002592 &SymbolizerInfo, &Ctx, std::move(RelInfo)));
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002593 DisAsm->setSymbolizer(std::move(Symbolizer));
2594 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002595 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
Ahmed Charles56440fd2014-03-06 05:51:42 +00002596 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
2597 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
Kevin Enderbybf246f52014-09-24 23:08:22 +00002598 // Set the display preference for hex vs. decimal immediates.
2599 IP->setPrintImmHex(PrintImmHex);
2600 // Comment stream and backing vector.
2601 SmallString<128> CommentsToEmit;
2602 raw_svector_ostream CommentStream(CommentsToEmit);
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00002603 // FIXME: Setting the CommentStream in the InstPrinter is problematic in that
2604 // if it is done then arm64 comments for string literals don't get printed
2605 // and some constant get printed instead and not setting it causes intel
2606 // (32-bit and 64-bit) comments printed with different spacing before the
2607 // comment causing different diffs with the 'C' disassembler library API.
2608 // IP->setCommentStream(CommentStream);
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002609
Kevin Enderbyae3c1262014-11-14 21:52:18 +00002610 if (!AsmInfo || !STI || !DisAsm || !IP) {
Michael J. Spencerc1363cf2011-10-07 19:25:47 +00002611 errs() << "error: couldn't initialize disassembler for target "
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002612 << TripleName << '\n';
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002613 return;
2614 }
2615
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002616 // Set up thumb disassembler.
2617 std::unique_ptr<const MCRegisterInfo> ThumbMRI;
2618 std::unique_ptr<const MCAsmInfo> ThumbAsmInfo;
2619 std::unique_ptr<const MCSubtargetInfo> ThumbSTI;
Kevin Enderby930fdc72014-11-06 19:00:13 +00002620 std::unique_ptr<MCDisassembler> ThumbDisAsm;
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002621 std::unique_ptr<MCInstPrinter> ThumbIP;
2622 std::unique_ptr<MCContext> ThumbCtx;
Kevin Enderby930fdc72014-11-06 19:00:13 +00002623 std::unique_ptr<MCSymbolizer> ThumbSymbolizer;
2624 struct DisassembleInfo ThumbSymbolizerInfo;
2625 std::unique_ptr<MCRelocationInfo> ThumbRelInfo;
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002626 if (ThumbTarget) {
2627 ThumbMRI.reset(ThumbTarget->createMCRegInfo(ThumbTripleName));
2628 ThumbAsmInfo.reset(
2629 ThumbTarget->createMCAsmInfo(*ThumbMRI, ThumbTripleName));
2630 ThumbSTI.reset(
2631 ThumbTarget->createMCSubtargetInfo(ThumbTripleName, MCPU, FeaturesStr));
2632 ThumbCtx.reset(new MCContext(ThumbAsmInfo.get(), ThumbMRI.get(), nullptr));
2633 ThumbDisAsm.reset(ThumbTarget->createMCDisassembler(*ThumbSTI, *ThumbCtx));
Kevin Enderby930fdc72014-11-06 19:00:13 +00002634 MCContext *PtrThumbCtx = ThumbCtx.get();
2635 ThumbRelInfo.reset(
2636 ThumbTarget->createMCRelocationInfo(ThumbTripleName, *PtrThumbCtx));
2637 if (ThumbRelInfo) {
2638 ThumbSymbolizer.reset(ThumbTarget->createMCSymbolizer(
2639 ThumbTripleName, SymbolizerGetOpInfo, SymbolizerSymbolLookUp,
David Blaikie186db432015-01-18 20:45:48 +00002640 &ThumbSymbolizerInfo, PtrThumbCtx, std::move(ThumbRelInfo)));
Kevin Enderby930fdc72014-11-06 19:00:13 +00002641 ThumbDisAsm->setSymbolizer(std::move(ThumbSymbolizer));
2642 }
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002643 int ThumbAsmPrinterVariant = ThumbAsmInfo->getAssemblerDialect();
2644 ThumbIP.reset(ThumbTarget->createMCInstPrinter(
2645 ThumbAsmPrinterVariant, *ThumbAsmInfo, *ThumbInstrInfo, *ThumbMRI,
2646 *ThumbSTI));
Kevin Enderbybf246f52014-09-24 23:08:22 +00002647 // Set the display preference for hex vs. decimal immediates.
2648 ThumbIP->setPrintImmHex(PrintImmHex);
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002649 }
2650
Kevin Enderbyae3c1262014-11-14 21:52:18 +00002651 if (ThumbTarget && (!ThumbAsmInfo || !ThumbSTI || !ThumbDisAsm || !ThumbIP)) {
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002652 errs() << "error: couldn't initialize disassembler for target "
2653 << ThumbTripleName << '\n';
2654 return;
2655 }
2656
Charles Davis8bdfafd2013-09-01 04:28:48 +00002657 MachO::mach_header Header = MachOOF->getHeader();
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002658
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002659 // FIXME: Using the -cfg command line option, this code used to be able to
2660 // annotate relocations with the referenced symbol's name, and if this was
2661 // inside a __[cf]string section, the data it points to. This is now replaced
2662 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
Owen Andersond9243c42011-10-17 21:37:35 +00002663 std::vector<SectionRef> Sections;
2664 std::vector<SymbolRef> Symbols;
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002665 SmallVector<uint64_t, 8> FoundFns;
Kevin Enderby273ae012013-06-06 17:20:50 +00002666 uint64_t BaseSegmentAddress;
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002667
Kevin Enderby273ae012013-06-06 17:20:50 +00002668 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
2669 BaseSegmentAddress);
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002670
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002671 // Sort the symbols by address, just in case they didn't come in that way.
Owen Andersond9243c42011-10-17 21:37:35 +00002672 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002673
Kevin Enderby273ae012013-06-06 17:20:50 +00002674 // Build a data in code table that is sorted on by the address of each entry.
2675 uint64_t BaseAddress = 0;
Charles Davis8bdfafd2013-09-01 04:28:48 +00002676 if (Header.filetype == MachO::MH_OBJECT)
Rafael Espindola80291272014-10-08 15:28:58 +00002677 BaseAddress = Sections[0].getAddress();
Kevin Enderby273ae012013-06-06 17:20:50 +00002678 else
2679 BaseAddress = BaseSegmentAddress;
2680 DiceTable Dices;
Kevin Enderby273ae012013-06-06 17:20:50 +00002681 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
Rafael Espindola5e812af2014-01-30 02:49:50 +00002682 DI != DE; ++DI) {
Kevin Enderby273ae012013-06-06 17:20:50 +00002683 uint32_t Offset;
2684 DI->getOffset(Offset);
2685 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
2686 }
2687 array_pod_sort(Dices.begin(), Dices.end());
2688
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002689#ifndef NDEBUG
2690 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
2691#else
2692 raw_ostream &DebugOut = nulls();
2693#endif
2694
Ahmed Charles56440fd2014-03-06 05:51:42 +00002695 std::unique_ptr<DIContext> diContext;
Rafael Espindola9b709252013-04-13 01:45:40 +00002696 ObjectFile *DbgObj = MachOOF;
Benjamin Kramer699128e2011-09-21 01:13:19 +00002697 // Try to find debug info and set up the DIContext for it.
2698 if (UseDbg) {
Benjamin Kramer699128e2011-09-21 01:13:19 +00002699 // A separate DSym file path was specified, parse it as a macho file,
2700 // get the sections and supply it to the section name parsing machinery.
2701 if (!DSYMFile.empty()) {
Rafael Espindola48af1c22014-08-19 18:44:46 +00002702 ErrorOr<std::unique_ptr<MemoryBuffer>> BufOrErr =
Rafael Espindolaadf21f22014-07-06 17:43:13 +00002703 MemoryBuffer::getFileOrSTDIN(DSYMFile);
Rafael Espindola48af1c22014-08-19 18:44:46 +00002704 if (std::error_code EC = BufOrErr.getError()) {
Rafael Espindolaadf21f22014-07-06 17:43:13 +00002705 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
Benjamin Kramer699128e2011-09-21 01:13:19 +00002706 return;
2707 }
Rafael Espindola48af1c22014-08-19 18:44:46 +00002708 DbgObj =
2709 ObjectFile::createMachOObjectFile(BufOrErr.get()->getMemBufferRef())
2710 .get()
2711 .release();
Benjamin Kramer699128e2011-09-21 01:13:19 +00002712 }
2713
Eric Christopher7370b552012-11-12 21:40:38 +00002714 // Setup the DIContext
Rafael Espindolaa04bb5b2014-07-31 20:19:36 +00002715 diContext.reset(DIContext::getDWARFContext(*DbgObj));
Benjamin Kramer699128e2011-09-21 01:13:19 +00002716 }
2717
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00002718 // TODO: For now this only disassembles the (__TEXT,__text) section (see the
2719 // checks in the code below at the top of this loop). It should allow a
2720 // darwin otool(1) like -s option to disassemble any named segment & section
2721 // that is marked as containing instructions with the attributes
2722 // S_ATTR_PURE_INSTRUCTIONS or S_ATTR_SOME_INSTRUCTIONS in the flags field of
2723 // the section structure.
2724 outs() << "(__TEXT,__text) section\n";
2725
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002726 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002727
Rafael Espindola80291272014-10-08 15:28:58 +00002728 bool SectIsText = Sections[SectIdx].isText();
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002729 if (SectIsText == false)
2730 continue;
2731
Owen Andersond9243c42011-10-17 21:37:35 +00002732 StringRef SectName;
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002733 if (Sections[SectIdx].getName(SectName) || SectName != "__text")
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002734 continue; // Skip non-text sections
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002735
Rafael Espindolaa9f810b2012-12-21 03:47:03 +00002736 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002737
Rafael Espindolab0f76a42013-04-05 15:15:22 +00002738 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
2739 if (SegmentName != "__TEXT")
Rafael Espindolaa9f810b2012-12-21 03:47:03 +00002740 continue;
2741
Rafael Espindola7fc5b872014-11-12 02:04:27 +00002742 StringRef BytesStr;
2743 Sections[SectIdx].getContents(BytesStr);
Aaron Ballman106fd7b2014-11-12 14:01:17 +00002744 ArrayRef<uint8_t> Bytes(reinterpret_cast<const uint8_t *>(BytesStr.data()),
2745 BytesStr.size());
Rafael Espindola80291272014-10-08 15:28:58 +00002746 uint64_t SectAddress = Sections[SectIdx].getAddress();
Rafael Espindolabd604f22014-11-07 00:52:15 +00002747
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002748 bool symbolTableWorked = false;
2749
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002750 // Parse relocations.
Alexey Samsonovaa4d2952014-03-14 14:22:49 +00002751 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
2752 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
Rafael Espindola80291272014-10-08 15:28:58 +00002753 uint64_t RelocOffset;
Alexey Samsonovaa4d2952014-03-14 14:22:49 +00002754 Reloc.getOffset(RelocOffset);
Rafael Espindola80291272014-10-08 15:28:58 +00002755 uint64_t SectionAddress = Sections[SectIdx].getAddress();
Owen Andersond9243c42011-10-17 21:37:35 +00002756 RelocOffset -= SectionAddress;
2757
Alexey Samsonovaa4d2952014-03-14 14:22:49 +00002758 symbol_iterator RelocSym = Reloc.getSymbol();
Owen Andersond9243c42011-10-17 21:37:35 +00002759
Rafael Espindola806f0062013-06-05 01:33:53 +00002760 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002761 }
2762 array_pod_sort(Relocs.begin(), Relocs.end());
2763
Kevin Enderbybf246f52014-09-24 23:08:22 +00002764 // Create a map of symbol addresses to symbol names for use by
2765 // the SymbolizerSymbolLookUp() routine.
2766 SymbolAddressMap AddrMap;
2767 for (const SymbolRef &Symbol : MachOOF->symbols()) {
2768 SymbolRef::Type ST;
2769 Symbol.getType(ST);
2770 if (ST == SymbolRef::ST_Function || ST == SymbolRef::ST_Data ||
2771 ST == SymbolRef::ST_Other) {
2772 uint64_t Address;
2773 Symbol.getAddress(Address);
2774 StringRef SymName;
2775 Symbol.getName(SymName);
2776 AddrMap[Address] = SymName;
2777 }
2778 }
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002779 // Set up the block of info used by the Symbolizer call backs.
2780 SymbolizerInfo.verbose = true;
2781 SymbolizerInfo.O = MachOOF;
2782 SymbolizerInfo.S = Sections[SectIdx];
Kevin Enderbybf246f52014-09-24 23:08:22 +00002783 SymbolizerInfo.AddrMap = &AddrMap;
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002784 SymbolizerInfo.Sections = &Sections;
2785 SymbolizerInfo.class_name = nullptr;
2786 SymbolizerInfo.selector_name = nullptr;
2787 SymbolizerInfo.method = nullptr;
Kevin Enderby04bf6932014-10-28 23:39:46 +00002788 SymbolizerInfo.demangled_name = nullptr;
Kevin Enderby078be602014-10-23 19:53:12 +00002789 SymbolizerInfo.bindtable = nullptr;
Kevin Enderby10738222014-11-19 20:20:16 +00002790 SymbolizerInfo.adrp_addr = 0;
2791 SymbolizerInfo.adrp_inst = 0;
Kevin Enderby930fdc72014-11-06 19:00:13 +00002792 // Same for the ThumbSymbolizer
2793 ThumbSymbolizerInfo.verbose = true;
2794 ThumbSymbolizerInfo.O = MachOOF;
2795 ThumbSymbolizerInfo.S = Sections[SectIdx];
2796 ThumbSymbolizerInfo.AddrMap = &AddrMap;
2797 ThumbSymbolizerInfo.Sections = &Sections;
2798 ThumbSymbolizerInfo.class_name = nullptr;
2799 ThumbSymbolizerInfo.selector_name = nullptr;
2800 ThumbSymbolizerInfo.method = nullptr;
2801 ThumbSymbolizerInfo.demangled_name = nullptr;
2802 ThumbSymbolizerInfo.bindtable = nullptr;
Kevin Enderby10738222014-11-19 20:20:16 +00002803 ThumbSymbolizerInfo.adrp_addr = 0;
2804 ThumbSymbolizerInfo.adrp_inst = 0;
Kevin Enderby98c9acc2014-09-16 18:00:57 +00002805
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002806 // Disassemble symbol by symbol.
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002807 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
Owen Andersond9243c42011-10-17 21:37:35 +00002808 StringRef SymName;
2809 Symbols[SymIdx].getName(SymName);
2810
2811 SymbolRef::Type ST;
2812 Symbols[SymIdx].getType(ST);
2813 if (ST != SymbolRef::ST_Function)
2814 continue;
2815
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002816 // Make sure the symbol is defined in this section.
Rafael Espindola80291272014-10-08 15:28:58 +00002817 bool containsSym = Sections[SectIdx].containsSymbol(Symbols[SymIdx]);
Owen Andersond9243c42011-10-17 21:37:35 +00002818 if (!containsSym)
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002819 continue;
2820
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002821 // Start at the address of the symbol relative to the section's address.
Owen Andersond9243c42011-10-17 21:37:35 +00002822 uint64_t Start = 0;
Rafael Espindola80291272014-10-08 15:28:58 +00002823 uint64_t SectionAddress = Sections[SectIdx].getAddress();
Danil Malyshevcbe72fc2011-11-29 17:40:10 +00002824 Symbols[SymIdx].getAddress(Start);
Cameron Zwarich54478a52012-02-03 05:42:17 +00002825 Start -= SectionAddress;
Owen Andersond9243c42011-10-17 21:37:35 +00002826
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +00002827 // Stop disassembling either at the beginning of the next symbol or at
2828 // the end of the section.
Kevin Enderbyedd58722012-05-15 18:57:14 +00002829 bool containsNextSym = false;
Owen Andersond9243c42011-10-17 21:37:35 +00002830 uint64_t NextSym = 0;
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002831 uint64_t NextSymIdx = SymIdx + 1;
Owen Andersond9243c42011-10-17 21:37:35 +00002832 while (Symbols.size() > NextSymIdx) {
2833 SymbolRef::Type NextSymType;
2834 Symbols[NextSymIdx].getType(NextSymType);
2835 if (NextSymType == SymbolRef::ST_Function) {
Rafael Espindola80291272014-10-08 15:28:58 +00002836 containsNextSym =
2837 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx]);
Danil Malyshevcbe72fc2011-11-29 17:40:10 +00002838 Symbols[NextSymIdx].getAddress(NextSym);
Cameron Zwarich54478a52012-02-03 05:42:17 +00002839 NextSym -= SectionAddress;
Owen Andersond9243c42011-10-17 21:37:35 +00002840 break;
2841 }
2842 ++NextSymIdx;
2843 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002844
Rafael Espindola80291272014-10-08 15:28:58 +00002845 uint64_t SectSize = Sections[SectIdx].getSize();
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002846 uint64_t End = containsNextSym ? NextSym : SectSize;
Owen Andersond9243c42011-10-17 21:37:35 +00002847 uint64_t Size;
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002848
2849 symbolTableWorked = true;
Rafael Espindolabd604f22014-11-07 00:52:15 +00002850
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002851 DataRefImpl Symb = Symbols[SymIdx].getRawDataRefImpl();
2852 bool isThumb =
2853 (MachOOF->getSymbolFlags(Symb) & SymbolRef::SF_Thumb) && ThumbTarget;
2854
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002855 outs() << SymName << ":\n";
2856 DILineInfo lastLine;
2857 for (uint64_t Index = Start; Index < End; Index += Size) {
2858 MCInst Inst;
Owen Andersond9243c42011-10-17 21:37:35 +00002859
Kevin Enderbybf246f52014-09-24 23:08:22 +00002860 uint64_t PC = SectAddress + Index;
2861 if (FullLeadingAddr) {
2862 if (MachOOF->is64Bit())
2863 outs() << format("%016" PRIx64, PC);
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002864 else
Kevin Enderbybf246f52014-09-24 23:08:22 +00002865 outs() << format("%08" PRIx64, PC);
2866 } else {
2867 outs() << format("%8" PRIx64 ":", PC);
2868 }
2869 if (!NoShowRawInsn)
2870 outs() << "\t";
Kevin Enderby273ae012013-06-06 17:20:50 +00002871
2872 // Check the data in code table here to see if this is data not an
2873 // instruction to be disassembled.
2874 DiceTable Dice;
Kevin Enderbybf246f52014-09-24 23:08:22 +00002875 Dice.push_back(std::make_pair(PC, DiceRef()));
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002876 dice_table_iterator DTI =
2877 std::search(Dices.begin(), Dices.end(), Dice.begin(), Dice.end(),
2878 compareDiceTableEntries);
2879 if (DTI != Dices.end()) {
Kevin Enderby273ae012013-06-06 17:20:50 +00002880 uint16_t Length;
2881 DTI->second.getLength(Length);
Kevin Enderby273ae012013-06-06 17:20:50 +00002882 uint16_t Kind;
2883 DTI->second.getKind(Kind);
Aaron Ballman106fd7b2014-11-12 14:01:17 +00002884 Size = DumpDataInCode(reinterpret_cast<const char *>(Bytes.data()) +
2885 Index,
2886 Length, Kind);
Kevin Enderby930fdc72014-11-06 19:00:13 +00002887 if ((Kind == MachO::DICE_KIND_JUMP_TABLE8) &&
2888 (PC == (DTI->first + Length - 1)) && (Length & 1))
2889 Size++;
Kevin Enderby273ae012013-06-06 17:20:50 +00002890 continue;
2891 }
2892
Kevin Enderbybf246f52014-09-24 23:08:22 +00002893 SmallVector<char, 64> AnnotationsBytes;
2894 raw_svector_ostream Annotations(AnnotationsBytes);
2895
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002896 bool gotInst;
2897 if (isThumb)
Rafael Espindola7fc5b872014-11-12 02:04:27 +00002898 gotInst = ThumbDisAsm->getInstruction(Inst, Size, Bytes.slice(Index),
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002899 PC, DebugOut, Annotations);
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002900 else
Rafael Espindola7fc5b872014-11-12 02:04:27 +00002901 gotInst = DisAsm->getInstruction(Inst, Size, Bytes.slice(Index), PC,
Kevin Enderbybf246f52014-09-24 23:08:22 +00002902 DebugOut, Annotations);
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002903 if (gotInst) {
Kevin Enderbybf246f52014-09-24 23:08:22 +00002904 if (!NoShowRawInsn) {
Aaron Ballman106fd7b2014-11-12 14:01:17 +00002905 DumpBytes(StringRef(
2906 reinterpret_cast<const char *>(Bytes.data()) + Index, Size));
Kevin Enderbybf246f52014-09-24 23:08:22 +00002907 }
2908 formatted_raw_ostream FormattedOS(outs());
2909 Annotations.flush();
2910 StringRef AnnotationsStr = Annotations.str();
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002911 if (isThumb)
Kevin Enderbybf246f52014-09-24 23:08:22 +00002912 ThumbIP->printInst(&Inst, FormattedOS, AnnotationsStr);
Kevin Enderbyec5ca032014-08-18 20:21:02 +00002913 else
Kevin Enderbybf246f52014-09-24 23:08:22 +00002914 IP->printInst(&Inst, FormattedOS, AnnotationsStr);
2915 emitComments(CommentStream, CommentsToEmit, FormattedOS, *AsmInfo);
Owen Andersond9243c42011-10-17 21:37:35 +00002916
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002917 // Print debug info.
2918 if (diContext) {
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002919 DILineInfo dli = diContext->getLineInfoForAddress(PC);
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002920 // Print valid line info if it changed.
Alexey Samsonovd0109992014-04-18 21:36:39 +00002921 if (dli != lastLine && dli.Line != 0)
2922 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
2923 << dli.Column;
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002924 lastLine = dli;
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002925 }
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002926 outs() << "\n";
2927 } else {
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002928 unsigned int Arch = MachOOF->getArch();
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002929 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002930 outs() << format("\t.byte 0x%02x #bad opcode\n",
2931 *(Bytes.data() + Index) & 0xff);
2932 Size = 1; // skip exactly one illegible byte and move on.
Kevin Enderbyae3c1262014-11-14 21:52:18 +00002933 } else if (Arch == Triple::aarch64) {
2934 uint32_t opcode = (*(Bytes.data() + Index) & 0xff) |
2935 (*(Bytes.data() + Index + 1) & 0xff) << 8 |
2936 (*(Bytes.data() + Index + 2) & 0xff) << 16 |
2937 (*(Bytes.data() + Index + 3) & 0xff) << 24;
2938 outs() << format("\t.long\t0x%08x\n", opcode);
2939 Size = 4;
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002940 } else {
2941 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2942 if (Size == 0)
2943 Size = 1; // skip illegible bytes
2944 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002945 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +00002946 }
2947 }
Ahmed Bougachaaa790682013-05-24 01:07:04 +00002948 if (!symbolTableWorked) {
Rafael Espindola80291272014-10-08 15:28:58 +00002949 // Reading the symbol table didn't work, disassemble the whole section.
2950 uint64_t SectAddress = Sections[SectIdx].getAddress();
2951 uint64_t SectSize = Sections[SectIdx].getSize();
Kevin Enderbybadd1002012-05-18 00:13:56 +00002952 uint64_t InstSize;
2953 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
Bill Wendling4e68e062012-07-19 00:17:40 +00002954 MCInst Inst;
Kevin Enderbybadd1002012-05-18 00:13:56 +00002955
Kevin Enderbybf246f52014-09-24 23:08:22 +00002956 uint64_t PC = SectAddress + Index;
Rafael Espindola7fc5b872014-11-12 02:04:27 +00002957 if (DisAsm->getInstruction(Inst, InstSize, Bytes.slice(Index), PC,
2958 DebugOut, nulls())) {
Kevin Enderbybf246f52014-09-24 23:08:22 +00002959 if (FullLeadingAddr) {
2960 if (MachOOF->is64Bit())
2961 outs() << format("%016" PRIx64, PC);
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002962 else
Kevin Enderbybf246f52014-09-24 23:08:22 +00002963 outs() << format("%08" PRIx64, PC);
2964 } else {
2965 outs() << format("%8" PRIx64 ":", PC);
2966 }
2967 if (!NoShowRawInsn) {
2968 outs() << "\t";
Aaron Ballman106fd7b2014-11-12 14:01:17 +00002969 DumpBytes(
2970 StringRef(reinterpret_cast<const char *>(Bytes.data()) + Index,
2971 InstSize));
Kevin Enderbybf246f52014-09-24 23:08:22 +00002972 }
Bill Wendling4e68e062012-07-19 00:17:40 +00002973 IP->printInst(&Inst, outs(), "");
2974 outs() << "\n";
2975 } else {
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002976 unsigned int Arch = MachOOF->getArch();
Kevin Enderbyb28ed012014-10-29 21:28:24 +00002977 if (Arch == Triple::x86_64 || Arch == Triple::x86) {
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002978 outs() << format("\t.byte 0x%02x #bad opcode\n",
2979 *(Bytes.data() + Index) & 0xff);
2980 InstSize = 1; // skip exactly one illegible byte and move on.
2981 } else {
2982 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
2983 if (InstSize == 0)
2984 InstSize = 1; // skip illegible bytes
2985 }
Bill Wendling4e68e062012-07-19 00:17:40 +00002986 }
Kevin Enderbybadd1002012-05-18 00:13:56 +00002987 }
2988 }
Kevin Enderbyef3ad2f2014-12-04 23:56:27 +00002989 // The TripleName's need to be reset if we are called again for a different
2990 // archtecture.
2991 TripleName = "";
2992 ThumbTripleName = "";
2993
Kevin Enderby6f326ce2014-10-23 19:37:31 +00002994 if (SymbolizerInfo.method != nullptr)
2995 free(SymbolizerInfo.method);
Kevin Enderby04bf6932014-10-28 23:39:46 +00002996 if (SymbolizerInfo.demangled_name != nullptr)
2997 free(SymbolizerInfo.demangled_name);
Kevin Enderby078be602014-10-23 19:53:12 +00002998 if (SymbolizerInfo.bindtable != nullptr)
2999 delete SymbolizerInfo.bindtable;
Kevin Enderby930fdc72014-11-06 19:00:13 +00003000 if (ThumbSymbolizerInfo.method != nullptr)
3001 free(ThumbSymbolizerInfo.method);
3002 if (ThumbSymbolizerInfo.demangled_name != nullptr)
3003 free(ThumbSymbolizerInfo.demangled_name);
3004 if (ThumbSymbolizerInfo.bindtable != nullptr)
3005 delete ThumbSymbolizerInfo.bindtable;
Benjamin Kramer43a772e2011-09-19 17:56:04 +00003006 }
3007}
Tim Northover4bd286a2014-08-01 13:07:19 +00003008
Tim Northover39c70bb2014-08-12 11:52:59 +00003009//===----------------------------------------------------------------------===//
3010// __compact_unwind section dumping
3011//===----------------------------------------------------------------------===//
3012
Tim Northover4bd286a2014-08-01 13:07:19 +00003013namespace {
Tim Northover39c70bb2014-08-12 11:52:59 +00003014
3015template <typename T> static uint64_t readNext(const char *&Buf) {
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003016 using llvm::support::little;
3017 using llvm::support::unaligned;
Tim Northover39c70bb2014-08-12 11:52:59 +00003018
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003019 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
3020 Buf += sizeof(T);
3021 return Val;
3022}
Tim Northover39c70bb2014-08-12 11:52:59 +00003023
Tim Northover4bd286a2014-08-01 13:07:19 +00003024struct CompactUnwindEntry {
3025 uint32_t OffsetInSection;
3026
3027 uint64_t FunctionAddr;
3028 uint32_t Length;
3029 uint32_t CompactEncoding;
3030 uint64_t PersonalityAddr;
3031 uint64_t LSDAAddr;
3032
3033 RelocationRef FunctionReloc;
3034 RelocationRef PersonalityReloc;
3035 RelocationRef LSDAReloc;
3036
3037 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003038 : OffsetInSection(Offset) {
Tim Northover4bd286a2014-08-01 13:07:19 +00003039 if (Is64)
3040 read<uint64_t>(Contents.data() + Offset);
3041 else
3042 read<uint32_t>(Contents.data() + Offset);
3043 }
3044
3045private:
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003046 template <typename UIntPtr> void read(const char *Buf) {
Tim Northover4bd286a2014-08-01 13:07:19 +00003047 FunctionAddr = readNext<UIntPtr>(Buf);
3048 Length = readNext<uint32_t>(Buf);
3049 CompactEncoding = readNext<uint32_t>(Buf);
3050 PersonalityAddr = readNext<UIntPtr>(Buf);
3051 LSDAAddr = readNext<UIntPtr>(Buf);
3052 }
3053};
3054}
3055
3056/// Given a relocation from __compact_unwind, consisting of the RelocationRef
3057/// and data being relocated, determine the best base Name and Addend to use for
3058/// display purposes.
3059///
3060/// 1. An Extern relocation will directly reference a symbol (and the data is
3061/// then already an addend), so use that.
3062/// 2. Otherwise the data is an offset in the object file's layout; try to find
3063// a symbol before it in the same section, and use the offset from there.
3064/// 3. Finally, if all that fails, fall back to an offset from the start of the
3065/// referenced section.
3066static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
3067 std::map<uint64_t, SymbolRef> &Symbols,
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003068 const RelocationRef &Reloc, uint64_t Addr,
Tim Northover4bd286a2014-08-01 13:07:19 +00003069 StringRef &Name, uint64_t &Addend) {
3070 if (Reloc.getSymbol() != Obj->symbol_end()) {
3071 Reloc.getSymbol()->getName(Name);
3072 Addend = Addr;
3073 return;
3074 }
3075
3076 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
3077 SectionRef RelocSection = Obj->getRelocationSection(RE);
3078
Rafael Espindola80291272014-10-08 15:28:58 +00003079 uint64_t SectionAddr = RelocSection.getAddress();
Tim Northover4bd286a2014-08-01 13:07:19 +00003080
3081 auto Sym = Symbols.upper_bound(Addr);
3082 if (Sym == Symbols.begin()) {
3083 // The first symbol in the object is after this reference, the best we can
3084 // do is section-relative notation.
3085 RelocSection.getName(Name);
3086 Addend = Addr - SectionAddr;
3087 return;
3088 }
3089
3090 // Go back one so that SymbolAddress <= Addr.
3091 --Sym;
3092
3093 section_iterator SymSection = Obj->section_end();
3094 Sym->second.getSection(SymSection);
3095 if (RelocSection == *SymSection) {
3096 // There's a valid symbol in the same section before this reference.
3097 Sym->second.getName(Name);
3098 Addend = Addr - Sym->first;
3099 return;
3100 }
3101
3102 // There is a symbol before this reference, but it's in a different
3103 // section. Probably not helpful to mention it, so use the section name.
3104 RelocSection.getName(Name);
3105 Addend = Addr - SectionAddr;
3106}
3107
3108static void printUnwindRelocDest(const MachOObjectFile *Obj,
3109 std::map<uint64_t, SymbolRef> &Symbols,
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003110 const RelocationRef &Reloc, uint64_t Addr) {
Tim Northover4bd286a2014-08-01 13:07:19 +00003111 StringRef Name;
3112 uint64_t Addend;
3113
Tim Northover0b0add52014-09-09 10:45:06 +00003114 if (!Reloc.getObjectFile())
3115 return;
3116
Tim Northover4bd286a2014-08-01 13:07:19 +00003117 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
3118
3119 outs() << Name;
3120 if (Addend)
Tim Northover63a25622014-08-11 09:14:06 +00003121 outs() << " + " << format("0x%" PRIx64, Addend);
Tim Northover4bd286a2014-08-01 13:07:19 +00003122}
3123
3124static void
3125printMachOCompactUnwindSection(const MachOObjectFile *Obj,
3126 std::map<uint64_t, SymbolRef> &Symbols,
3127 const SectionRef &CompactUnwind) {
3128
3129 assert(Obj->isLittleEndian() &&
3130 "There should not be a big-endian .o with __compact_unwind");
3131
3132 bool Is64 = Obj->is64Bit();
3133 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
3134 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
3135
3136 StringRef Contents;
3137 CompactUnwind.getContents(Contents);
3138
3139 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
3140
3141 // First populate the initial raw offsets, encodings and so on from the entry.
3142 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
3143 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
3144 CompactUnwinds.push_back(Entry);
3145 }
3146
3147 // Next we need to look at the relocations to find out what objects are
3148 // actually being referred to.
3149 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
3150 uint64_t RelocAddress;
3151 Reloc.getOffset(RelocAddress);
3152
3153 uint32_t EntryIdx = RelocAddress / EntrySize;
3154 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
3155 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
3156
3157 if (OffsetInEntry == 0)
3158 Entry.FunctionReloc = Reloc;
3159 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
3160 Entry.PersonalityReloc = Reloc;
3161 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
3162 Entry.LSDAReloc = Reloc;
3163 else
3164 llvm_unreachable("Unexpected relocation in __compact_unwind section");
3165 }
3166
3167 // Finally, we're ready to print the data we've gathered.
3168 outs() << "Contents of __compact_unwind section:\n";
3169 for (auto &Entry : CompactUnwinds) {
Tim Northover06af2602014-08-08 12:08:51 +00003170 outs() << " Entry at offset "
3171 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
Tim Northover4bd286a2014-08-01 13:07:19 +00003172
3173 // 1. Start of the region this entry applies to.
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003174 outs() << " start: " << format("0x%" PRIx64,
3175 Entry.FunctionAddr) << ' ';
3176 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc, Entry.FunctionAddr);
Tim Northover4bd286a2014-08-01 13:07:19 +00003177 outs() << '\n';
3178
3179 // 2. Length of the region this entry applies to.
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003180 outs() << " length: " << format("0x%" PRIx32, Entry.Length)
3181 << '\n';
Tim Northover4bd286a2014-08-01 13:07:19 +00003182 // 3. The 32-bit compact encoding.
3183 outs() << " compact encoding: "
Tim Northoverb911bf82014-08-08 12:00:09 +00003184 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
Tim Northover4bd286a2014-08-01 13:07:19 +00003185
3186 // 4. The personality function, if present.
3187 if (Entry.PersonalityReloc.getObjectFile()) {
3188 outs() << " personality function: "
Tim Northoverb911bf82014-08-08 12:00:09 +00003189 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
Tim Northover4bd286a2014-08-01 13:07:19 +00003190 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
3191 Entry.PersonalityAddr);
3192 outs() << '\n';
3193 }
3194
3195 // 5. This entry's language-specific data area.
3196 if (Entry.LSDAReloc.getObjectFile()) {
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003197 outs() << " LSDA: " << format("0x%" PRIx64,
3198 Entry.LSDAAddr) << ' ';
Tim Northover4bd286a2014-08-01 13:07:19 +00003199 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
3200 outs() << '\n';
3201 }
3202 }
3203}
3204
Tim Northover39c70bb2014-08-12 11:52:59 +00003205//===----------------------------------------------------------------------===//
3206// __unwind_info section dumping
3207//===----------------------------------------------------------------------===//
3208
3209static void printRegularSecondLevelUnwindPage(const char *PageStart) {
3210 const char *Pos = PageStart;
3211 uint32_t Kind = readNext<uint32_t>(Pos);
3212 (void)Kind;
3213 assert(Kind == 2 && "kind for a regular 2nd level index should be 2");
3214
3215 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3216 uint16_t NumEntries = readNext<uint16_t>(Pos);
3217
3218 Pos = PageStart + EntriesStart;
3219 for (unsigned i = 0; i < NumEntries; ++i) {
3220 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3221 uint32_t Encoding = readNext<uint32_t>(Pos);
3222
3223 outs() << " [" << i << "]: "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003224 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3225 << ", "
3226 << "encoding=" << format("0x%08" PRIx32, Encoding) << '\n';
Tim Northover39c70bb2014-08-12 11:52:59 +00003227 }
3228}
3229
3230static void printCompressedSecondLevelUnwindPage(
3231 const char *PageStart, uint32_t FunctionBase,
3232 const SmallVectorImpl<uint32_t> &CommonEncodings) {
3233 const char *Pos = PageStart;
3234 uint32_t Kind = readNext<uint32_t>(Pos);
3235 (void)Kind;
3236 assert(Kind == 3 && "kind for a compressed 2nd level index should be 3");
3237
3238 uint16_t EntriesStart = readNext<uint16_t>(Pos);
3239 uint16_t NumEntries = readNext<uint16_t>(Pos);
3240
3241 uint16_t EncodingsStart = readNext<uint16_t>(Pos);
3242 readNext<uint16_t>(Pos);
Aaron Ballman80930af2014-08-14 13:53:19 +00003243 const auto *PageEncodings = reinterpret_cast<const support::ulittle32_t *>(
3244 PageStart + EncodingsStart);
Tim Northover39c70bb2014-08-12 11:52:59 +00003245
3246 Pos = PageStart + EntriesStart;
3247 for (unsigned i = 0; i < NumEntries; ++i) {
3248 uint32_t Entry = readNext<uint32_t>(Pos);
3249 uint32_t FunctionOffset = FunctionBase + (Entry & 0xffffff);
3250 uint32_t EncodingIdx = Entry >> 24;
3251
3252 uint32_t Encoding;
3253 if (EncodingIdx < CommonEncodings.size())
3254 Encoding = CommonEncodings[EncodingIdx];
3255 else
3256 Encoding = PageEncodings[EncodingIdx - CommonEncodings.size()];
3257
3258 outs() << " [" << i << "]: "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003259 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3260 << ", "
3261 << "encoding[" << EncodingIdx
3262 << "]=" << format("0x%08" PRIx32, Encoding) << '\n';
Tim Northover39c70bb2014-08-12 11:52:59 +00003263 }
3264}
3265
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003266static void printMachOUnwindInfoSection(const MachOObjectFile *Obj,
3267 std::map<uint64_t, SymbolRef> &Symbols,
3268 const SectionRef &UnwindInfo) {
Tim Northover39c70bb2014-08-12 11:52:59 +00003269
3270 assert(Obj->isLittleEndian() &&
3271 "There should not be a big-endian .o with __unwind_info");
3272
3273 outs() << "Contents of __unwind_info section:\n";
3274
3275 StringRef Contents;
3276 UnwindInfo.getContents(Contents);
3277 const char *Pos = Contents.data();
3278
3279 //===----------------------------------
3280 // Section header
3281 //===----------------------------------
3282
3283 uint32_t Version = readNext<uint32_t>(Pos);
3284 outs() << " Version: "
3285 << format("0x%" PRIx32, Version) << '\n';
3286 assert(Version == 1 && "only understand version 1");
3287
3288 uint32_t CommonEncodingsStart = readNext<uint32_t>(Pos);
3289 outs() << " Common encodings array section offset: "
3290 << format("0x%" PRIx32, CommonEncodingsStart) << '\n';
3291 uint32_t NumCommonEncodings = readNext<uint32_t>(Pos);
3292 outs() << " Number of common encodings in array: "
3293 << format("0x%" PRIx32, NumCommonEncodings) << '\n';
3294
3295 uint32_t PersonalitiesStart = readNext<uint32_t>(Pos);
3296 outs() << " Personality function array section offset: "
3297 << format("0x%" PRIx32, PersonalitiesStart) << '\n';
3298 uint32_t NumPersonalities = readNext<uint32_t>(Pos);
3299 outs() << " Number of personality functions in array: "
3300 << format("0x%" PRIx32, NumPersonalities) << '\n';
3301
3302 uint32_t IndicesStart = readNext<uint32_t>(Pos);
3303 outs() << " Index array section offset: "
3304 << format("0x%" PRIx32, IndicesStart) << '\n';
3305 uint32_t NumIndices = readNext<uint32_t>(Pos);
3306 outs() << " Number of indices in array: "
3307 << format("0x%" PRIx32, NumIndices) << '\n';
3308
3309 //===----------------------------------
3310 // A shared list of common encodings
3311 //===----------------------------------
3312
3313 // These occupy indices in the range [0, N] whenever an encoding is referenced
3314 // from a compressed 2nd level index table. In practice the linker only
3315 // creates ~128 of these, so that indices are available to embed encodings in
3316 // the 2nd level index.
3317
3318 SmallVector<uint32_t, 64> CommonEncodings;
3319 outs() << " Common encodings: (count = " << NumCommonEncodings << ")\n";
3320 Pos = Contents.data() + CommonEncodingsStart;
3321 for (unsigned i = 0; i < NumCommonEncodings; ++i) {
3322 uint32_t Encoding = readNext<uint32_t>(Pos);
3323 CommonEncodings.push_back(Encoding);
3324
3325 outs() << " encoding[" << i << "]: " << format("0x%08" PRIx32, Encoding)
3326 << '\n';
3327 }
3328
Tim Northover39c70bb2014-08-12 11:52:59 +00003329 //===----------------------------------
3330 // Personality functions used in this executable
3331 //===----------------------------------
3332
3333 // There should be only a handful of these (one per source language,
3334 // roughly). Particularly since they only get 2 bits in the compact encoding.
3335
3336 outs() << " Personality functions: (count = " << NumPersonalities << ")\n";
3337 Pos = Contents.data() + PersonalitiesStart;
3338 for (unsigned i = 0; i < NumPersonalities; ++i) {
3339 uint32_t PersonalityFn = readNext<uint32_t>(Pos);
3340 outs() << " personality[" << i + 1
3341 << "]: " << format("0x%08" PRIx32, PersonalityFn) << '\n';
3342 }
3343
3344 //===----------------------------------
3345 // The level 1 index entries
3346 //===----------------------------------
3347
3348 // These specify an approximate place to start searching for the more detailed
3349 // information, sorted by PC.
3350
3351 struct IndexEntry {
3352 uint32_t FunctionOffset;
3353 uint32_t SecondLevelPageStart;
3354 uint32_t LSDAStart;
3355 };
3356
3357 SmallVector<IndexEntry, 4> IndexEntries;
3358
3359 outs() << " Top level indices: (count = " << NumIndices << ")\n";
3360 Pos = Contents.data() + IndicesStart;
3361 for (unsigned i = 0; i < NumIndices; ++i) {
3362 IndexEntry Entry;
3363
3364 Entry.FunctionOffset = readNext<uint32_t>(Pos);
3365 Entry.SecondLevelPageStart = readNext<uint32_t>(Pos);
3366 Entry.LSDAStart = readNext<uint32_t>(Pos);
3367 IndexEntries.push_back(Entry);
3368
3369 outs() << " [" << i << "]: "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003370 << "function offset=" << format("0x%08" PRIx32, Entry.FunctionOffset)
3371 << ", "
Tim Northover39c70bb2014-08-12 11:52:59 +00003372 << "2nd level page offset="
3373 << format("0x%08" PRIx32, Entry.SecondLevelPageStart) << ", "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003374 << "LSDA offset=" << format("0x%08" PRIx32, Entry.LSDAStart) << '\n';
Tim Northover39c70bb2014-08-12 11:52:59 +00003375 }
3376
Tim Northover39c70bb2014-08-12 11:52:59 +00003377 //===----------------------------------
3378 // Next come the LSDA tables
3379 //===----------------------------------
3380
3381 // The LSDA layout is rather implicit: it's a contiguous array of entries from
3382 // the first top-level index's LSDAOffset to the last (sentinel).
3383
3384 outs() << " LSDA descriptors:\n";
3385 Pos = Contents.data() + IndexEntries[0].LSDAStart;
3386 int NumLSDAs = (IndexEntries.back().LSDAStart - IndexEntries[0].LSDAStart) /
3387 (2 * sizeof(uint32_t));
3388 for (int i = 0; i < NumLSDAs; ++i) {
3389 uint32_t FunctionOffset = readNext<uint32_t>(Pos);
3390 uint32_t LSDAOffset = readNext<uint32_t>(Pos);
3391 outs() << " [" << i << "]: "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00003392 << "function offset=" << format("0x%08" PRIx32, FunctionOffset)
3393 << ", "
3394 << "LSDA offset=" << format("0x%08" PRIx32, LSDAOffset) << '\n';
Tim Northover39c70bb2014-08-12 11:52:59 +00003395 }
3396
3397 //===----------------------------------
3398 // Finally, the 2nd level indices
3399 //===----------------------------------
3400
3401 // Generally these are 4K in size, and have 2 possible forms:
3402 // + Regular stores up to 511 entries with disparate encodings
3403 // + Compressed stores up to 1021 entries if few enough compact encoding
3404 // values are used.
3405 outs() << " Second level indices:\n";
3406 for (unsigned i = 0; i < IndexEntries.size() - 1; ++i) {
3407 // The final sentinel top-level index has no associated 2nd level page
3408 if (IndexEntries[i].SecondLevelPageStart == 0)
3409 break;
3410
3411 outs() << " Second level index[" << i << "]: "
3412 << "offset in section="
3413 << format("0x%08" PRIx32, IndexEntries[i].SecondLevelPageStart)
3414 << ", "
3415 << "base function offset="
3416 << format("0x%08" PRIx32, IndexEntries[i].FunctionOffset) << '\n';
3417
3418 Pos = Contents.data() + IndexEntries[i].SecondLevelPageStart;
Aaron Ballman80930af2014-08-14 13:53:19 +00003419 uint32_t Kind = *reinterpret_cast<const support::ulittle32_t *>(Pos);
Tim Northover39c70bb2014-08-12 11:52:59 +00003420 if (Kind == 2)
3421 printRegularSecondLevelUnwindPage(Pos);
3422 else if (Kind == 3)
3423 printCompressedSecondLevelUnwindPage(Pos, IndexEntries[i].FunctionOffset,
3424 CommonEncodings);
3425 else
3426 llvm_unreachable("Do not know how to print this kind of 2nd level page");
Tim Northover39c70bb2014-08-12 11:52:59 +00003427 }
3428}
3429
Tim Northover4bd286a2014-08-01 13:07:19 +00003430void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
3431 std::map<uint64_t, SymbolRef> Symbols;
3432 for (const SymbolRef &SymRef : Obj->symbols()) {
3433 // Discard any undefined or absolute symbols. They're not going to take part
3434 // in the convenience lookup for unwind info and just take up resources.
3435 section_iterator Section = Obj->section_end();
3436 SymRef.getSection(Section);
3437 if (Section == Obj->section_end())
3438 continue;
3439
3440 uint64_t Addr;
3441 SymRef.getAddress(Addr);
3442 Symbols.insert(std::make_pair(Addr, SymRef));
3443 }
3444
3445 for (const SectionRef &Section : Obj->sections()) {
3446 StringRef SectName;
3447 Section.getName(SectName);
3448 if (SectName == "__compact_unwind")
3449 printMachOCompactUnwindSection(Obj, Symbols, Section);
3450 else if (SectName == "__unwind_info")
Tim Northover39c70bb2014-08-12 11:52:59 +00003451 printMachOUnwindInfoSection(Obj, Symbols, Section);
Tim Northover4bd286a2014-08-01 13:07:19 +00003452 else if (SectName == "__eh_frame")
3453 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
Tim Northover4bd286a2014-08-01 13:07:19 +00003454 }
3455}
Kevin Enderbyb76d3862014-08-22 20:35:18 +00003456
3457static void PrintMachHeader(uint32_t magic, uint32_t cputype,
3458 uint32_t cpusubtype, uint32_t filetype,
3459 uint32_t ncmds, uint32_t sizeofcmds, uint32_t flags,
3460 bool verbose) {
3461 outs() << "Mach header\n";
3462 outs() << " magic cputype cpusubtype caps filetype ncmds "
3463 "sizeofcmds flags\n";
3464 if (verbose) {
3465 if (magic == MachO::MH_MAGIC)
3466 outs() << " MH_MAGIC";
3467 else if (magic == MachO::MH_MAGIC_64)
3468 outs() << "MH_MAGIC_64";
3469 else
3470 outs() << format(" 0x%08" PRIx32, magic);
3471 switch (cputype) {
3472 case MachO::CPU_TYPE_I386:
3473 outs() << " I386";
3474 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3475 case MachO::CPU_SUBTYPE_I386_ALL:
3476 outs() << " ALL";
3477 break;
3478 default:
3479 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3480 break;
3481 }
3482 break;
3483 case MachO::CPU_TYPE_X86_64:
3484 outs() << " X86_64";
Kevin Enderby131d1772015-01-09 19:22:37 +00003485 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3486 case MachO::CPU_SUBTYPE_X86_64_ALL:
3487 outs() << " ALL";
3488 break;
3489 case MachO::CPU_SUBTYPE_X86_64_H:
3490 outs() << " Haswell";
3491 break;
3492 default:
3493 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3494 break;
3495 }
Kevin Enderbyb76d3862014-08-22 20:35:18 +00003496 break;
3497 case MachO::CPU_TYPE_ARM:
3498 outs() << " ARM";
3499 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3500 case MachO::CPU_SUBTYPE_ARM_ALL:
3501 outs() << " ALL";
3502 break;
3503 case MachO::CPU_SUBTYPE_ARM_V4T:
3504 outs() << " V4T";
3505 break;
3506 case MachO::CPU_SUBTYPE_ARM_V5TEJ:
3507 outs() << " V5TEJ";
3508 break;
3509 case MachO::CPU_SUBTYPE_ARM_XSCALE:
3510 outs() << " XSCALE";
3511 break;
3512 case MachO::CPU_SUBTYPE_ARM_V6:
3513 outs() << " V6";
3514 break;
3515 case MachO::CPU_SUBTYPE_ARM_V6M:
3516 outs() << " V6M";
3517 break;
3518 case MachO::CPU_SUBTYPE_ARM_V7:
3519 outs() << " V7";
3520 break;
3521 case MachO::CPU_SUBTYPE_ARM_V7EM:
3522 outs() << " V7EM";
3523 break;
3524 case MachO::CPU_SUBTYPE_ARM_V7K:
3525 outs() << " V7K";
3526 break;
3527 case MachO::CPU_SUBTYPE_ARM_V7M:
3528 outs() << " V7M";
3529 break;
3530 case MachO::CPU_SUBTYPE_ARM_V7S:
3531 outs() << " V7S";
3532 break;
3533 default:
3534 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3535 break;
3536 }
3537 break;
3538 case MachO::CPU_TYPE_ARM64:
3539 outs() << " ARM64";
3540 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3541 case MachO::CPU_SUBTYPE_ARM64_ALL:
3542 outs() << " ALL";
3543 break;
3544 default:
3545 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3546 break;
3547 }
3548 break;
3549 case MachO::CPU_TYPE_POWERPC:
3550 outs() << " PPC";
3551 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3552 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3553 outs() << " ALL";
3554 break;
3555 default:
3556 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3557 break;
3558 }
3559 break;
3560 case MachO::CPU_TYPE_POWERPC64:
3561 outs() << " PPC64";
3562 switch (cpusubtype & ~MachO::CPU_SUBTYPE_MASK) {
3563 case MachO::CPU_SUBTYPE_POWERPC_ALL:
3564 outs() << " ALL";
3565 break;
3566 default:
3567 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3568 break;
3569 }
3570 break;
3571 }
3572 if ((cpusubtype & MachO::CPU_SUBTYPE_MASK) == MachO::CPU_SUBTYPE_LIB64) {
Kevin Enderby8ae63c12014-09-04 16:54:47 +00003573 outs() << " LIB64";
Kevin Enderbyb76d3862014-08-22 20:35:18 +00003574 } else {
3575 outs() << format(" 0x%02" PRIx32,
3576 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3577 }
3578 switch (filetype) {
3579 case MachO::MH_OBJECT:
3580 outs() << " OBJECT";
3581 break;
3582 case MachO::MH_EXECUTE:
3583 outs() << " EXECUTE";
3584 break;
3585 case MachO::MH_FVMLIB:
3586 outs() << " FVMLIB";
3587 break;
3588 case MachO::MH_CORE:
3589 outs() << " CORE";
3590 break;
3591 case MachO::MH_PRELOAD:
3592 outs() << " PRELOAD";
3593 break;
3594 case MachO::MH_DYLIB:
3595 outs() << " DYLIB";
3596 break;
3597 case MachO::MH_DYLIB_STUB:
3598 outs() << " DYLIB_STUB";
3599 break;
3600 case MachO::MH_DYLINKER:
3601 outs() << " DYLINKER";
3602 break;
3603 case MachO::MH_BUNDLE:
3604 outs() << " BUNDLE";
3605 break;
3606 case MachO::MH_DSYM:
3607 outs() << " DSYM";
3608 break;
3609 case MachO::MH_KEXT_BUNDLE:
3610 outs() << " KEXTBUNDLE";
3611 break;
3612 default:
3613 outs() << format(" %10u", filetype);
3614 break;
3615 }
3616 outs() << format(" %5u", ncmds);
3617 outs() << format(" %10u", sizeofcmds);
3618 uint32_t f = flags;
3619 if (f & MachO::MH_NOUNDEFS) {
3620 outs() << " NOUNDEFS";
3621 f &= ~MachO::MH_NOUNDEFS;
3622 }
3623 if (f & MachO::MH_INCRLINK) {
3624 outs() << " INCRLINK";
3625 f &= ~MachO::MH_INCRLINK;
3626 }
3627 if (f & MachO::MH_DYLDLINK) {
3628 outs() << " DYLDLINK";
3629 f &= ~MachO::MH_DYLDLINK;
3630 }
3631 if (f & MachO::MH_BINDATLOAD) {
3632 outs() << " BINDATLOAD";
3633 f &= ~MachO::MH_BINDATLOAD;
3634 }
3635 if (f & MachO::MH_PREBOUND) {
3636 outs() << " PREBOUND";
3637 f &= ~MachO::MH_PREBOUND;
3638 }
3639 if (f & MachO::MH_SPLIT_SEGS) {
3640 outs() << " SPLIT_SEGS";
3641 f &= ~MachO::MH_SPLIT_SEGS;
3642 }
3643 if (f & MachO::MH_LAZY_INIT) {
3644 outs() << " LAZY_INIT";
3645 f &= ~MachO::MH_LAZY_INIT;
3646 }
3647 if (f & MachO::MH_TWOLEVEL) {
3648 outs() << " TWOLEVEL";
3649 f &= ~MachO::MH_TWOLEVEL;
3650 }
3651 if (f & MachO::MH_FORCE_FLAT) {
3652 outs() << " FORCE_FLAT";
3653 f &= ~MachO::MH_FORCE_FLAT;
3654 }
3655 if (f & MachO::MH_NOMULTIDEFS) {
3656 outs() << " NOMULTIDEFS";
3657 f &= ~MachO::MH_NOMULTIDEFS;
3658 }
3659 if (f & MachO::MH_NOFIXPREBINDING) {
3660 outs() << " NOFIXPREBINDING";
3661 f &= ~MachO::MH_NOFIXPREBINDING;
3662 }
3663 if (f & MachO::MH_PREBINDABLE) {
3664 outs() << " PREBINDABLE";
3665 f &= ~MachO::MH_PREBINDABLE;
3666 }
3667 if (f & MachO::MH_ALLMODSBOUND) {
3668 outs() << " ALLMODSBOUND";
3669 f &= ~MachO::MH_ALLMODSBOUND;
3670 }
3671 if (f & MachO::MH_SUBSECTIONS_VIA_SYMBOLS) {
3672 outs() << " SUBSECTIONS_VIA_SYMBOLS";
3673 f &= ~MachO::MH_SUBSECTIONS_VIA_SYMBOLS;
3674 }
3675 if (f & MachO::MH_CANONICAL) {
3676 outs() << " CANONICAL";
3677 f &= ~MachO::MH_CANONICAL;
3678 }
3679 if (f & MachO::MH_WEAK_DEFINES) {
3680 outs() << " WEAK_DEFINES";
3681 f &= ~MachO::MH_WEAK_DEFINES;
3682 }
3683 if (f & MachO::MH_BINDS_TO_WEAK) {
3684 outs() << " BINDS_TO_WEAK";
3685 f &= ~MachO::MH_BINDS_TO_WEAK;
3686 }
3687 if (f & MachO::MH_ALLOW_STACK_EXECUTION) {
3688 outs() << " ALLOW_STACK_EXECUTION";
3689 f &= ~MachO::MH_ALLOW_STACK_EXECUTION;
3690 }
3691 if (f & MachO::MH_DEAD_STRIPPABLE_DYLIB) {
3692 outs() << " DEAD_STRIPPABLE_DYLIB";
3693 f &= ~MachO::MH_DEAD_STRIPPABLE_DYLIB;
3694 }
3695 if (f & MachO::MH_PIE) {
3696 outs() << " PIE";
3697 f &= ~MachO::MH_PIE;
3698 }
3699 if (f & MachO::MH_NO_REEXPORTED_DYLIBS) {
3700 outs() << " NO_REEXPORTED_DYLIBS";
3701 f &= ~MachO::MH_NO_REEXPORTED_DYLIBS;
3702 }
3703 if (f & MachO::MH_HAS_TLV_DESCRIPTORS) {
3704 outs() << " MH_HAS_TLV_DESCRIPTORS";
3705 f &= ~MachO::MH_HAS_TLV_DESCRIPTORS;
3706 }
3707 if (f & MachO::MH_NO_HEAP_EXECUTION) {
3708 outs() << " MH_NO_HEAP_EXECUTION";
3709 f &= ~MachO::MH_NO_HEAP_EXECUTION;
3710 }
3711 if (f & MachO::MH_APP_EXTENSION_SAFE) {
3712 outs() << " APP_EXTENSION_SAFE";
3713 f &= ~MachO::MH_APP_EXTENSION_SAFE;
3714 }
3715 if (f != 0 || flags == 0)
3716 outs() << format(" 0x%08" PRIx32, f);
3717 } else {
3718 outs() << format(" 0x%08" PRIx32, magic);
3719 outs() << format(" %7d", cputype);
3720 outs() << format(" %10d", cpusubtype & ~MachO::CPU_SUBTYPE_MASK);
3721 outs() << format(" 0x%02" PRIx32,
3722 (cpusubtype & MachO::CPU_SUBTYPE_MASK) >> 24);
3723 outs() << format(" %10u", filetype);
3724 outs() << format(" %5u", ncmds);
3725 outs() << format(" %10u", sizeofcmds);
3726 outs() << format(" 0x%08" PRIx32, flags);
3727 }
3728 outs() << "\n";
3729}
3730
Kevin Enderby956366c2014-08-29 22:30:52 +00003731static void PrintSegmentCommand(uint32_t cmd, uint32_t cmdsize,
3732 StringRef SegName, uint64_t vmaddr,
3733 uint64_t vmsize, uint64_t fileoff,
3734 uint64_t filesize, uint32_t maxprot,
3735 uint32_t initprot, uint32_t nsects,
3736 uint32_t flags, uint32_t object_size,
3737 bool verbose) {
3738 uint64_t expected_cmdsize;
3739 if (cmd == MachO::LC_SEGMENT) {
3740 outs() << " cmd LC_SEGMENT\n";
3741 expected_cmdsize = nsects;
3742 expected_cmdsize *= sizeof(struct MachO::section);
3743 expected_cmdsize += sizeof(struct MachO::segment_command);
3744 } else {
3745 outs() << " cmd LC_SEGMENT_64\n";
3746 expected_cmdsize = nsects;
3747 expected_cmdsize *= sizeof(struct MachO::section_64);
3748 expected_cmdsize += sizeof(struct MachO::segment_command_64);
3749 }
3750 outs() << " cmdsize " << cmdsize;
3751 if (cmdsize != expected_cmdsize)
3752 outs() << " Inconsistent size\n";
3753 else
3754 outs() << "\n";
3755 outs() << " segname " << SegName << "\n";
3756 if (cmd == MachO::LC_SEGMENT_64) {
3757 outs() << " vmaddr " << format("0x%016" PRIx64, vmaddr) << "\n";
3758 outs() << " vmsize " << format("0x%016" PRIx64, vmsize) << "\n";
3759 } else {
Kevin Enderbyadb7c432014-12-16 18:58:11 +00003760 outs() << " vmaddr " << format("0x%08" PRIx64, vmaddr) << "\n";
3761 outs() << " vmsize " << format("0x%08" PRIx64, vmsize) << "\n";
Kevin Enderby956366c2014-08-29 22:30:52 +00003762 }
3763 outs() << " fileoff " << fileoff;
3764 if (fileoff > object_size)
3765 outs() << " (past end of file)\n";
3766 else
3767 outs() << "\n";
3768 outs() << " filesize " << filesize;
3769 if (fileoff + filesize > object_size)
3770 outs() << " (past end of file)\n";
3771 else
3772 outs() << "\n";
3773 if (verbose) {
3774 if ((maxprot &
3775 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3776 MachO::VM_PROT_EXECUTE)) != 0)
3777 outs() << " maxprot ?" << format("0x%08" PRIx32, maxprot) << "\n";
3778 else {
3779 if (maxprot & MachO::VM_PROT_READ)
3780 outs() << " maxprot r";
3781 else
3782 outs() << " maxprot -";
3783 if (maxprot & MachO::VM_PROT_WRITE)
3784 outs() << "w";
3785 else
3786 outs() << "-";
3787 if (maxprot & MachO::VM_PROT_EXECUTE)
3788 outs() << "x\n";
3789 else
3790 outs() << "-\n";
3791 }
3792 if ((initprot &
3793 ~(MachO::VM_PROT_READ | MachO::VM_PROT_WRITE |
3794 MachO::VM_PROT_EXECUTE)) != 0)
3795 outs() << " initprot ?" << format("0x%08" PRIx32, initprot) << "\n";
3796 else {
3797 if (initprot & MachO::VM_PROT_READ)
3798 outs() << " initprot r";
3799 else
3800 outs() << " initprot -";
3801 if (initprot & MachO::VM_PROT_WRITE)
3802 outs() << "w";
3803 else
3804 outs() << "-";
3805 if (initprot & MachO::VM_PROT_EXECUTE)
3806 outs() << "x\n";
3807 else
3808 outs() << "-\n";
3809 }
3810 } else {
3811 outs() << " maxprot " << format("0x%08" PRIx32, maxprot) << "\n";
3812 outs() << " initprot " << format("0x%08" PRIx32, initprot) << "\n";
3813 }
3814 outs() << " nsects " << nsects << "\n";
3815 if (verbose) {
3816 outs() << " flags";
3817 if (flags == 0)
3818 outs() << " (none)\n";
3819 else {
3820 if (flags & MachO::SG_HIGHVM) {
3821 outs() << " HIGHVM";
3822 flags &= ~MachO::SG_HIGHVM;
3823 }
3824 if (flags & MachO::SG_FVMLIB) {
3825 outs() << " FVMLIB";
3826 flags &= ~MachO::SG_FVMLIB;
3827 }
3828 if (flags & MachO::SG_NORELOC) {
3829 outs() << " NORELOC";
3830 flags &= ~MachO::SG_NORELOC;
3831 }
3832 if (flags & MachO::SG_PROTECTED_VERSION_1) {
3833 outs() << " PROTECTED_VERSION_1";
3834 flags &= ~MachO::SG_PROTECTED_VERSION_1;
3835 }
3836 if (flags)
3837 outs() << format(" 0x%08" PRIx32, flags) << " (unknown flags)\n";
3838 else
3839 outs() << "\n";
3840 }
3841 } else {
3842 outs() << " flags " << format("0x%" PRIx32, flags) << "\n";
3843 }
3844}
3845
3846static void PrintSection(const char *sectname, const char *segname,
3847 uint64_t addr, uint64_t size, uint32_t offset,
3848 uint32_t align, uint32_t reloff, uint32_t nreloc,
3849 uint32_t flags, uint32_t reserved1, uint32_t reserved2,
3850 uint32_t cmd, const char *sg_segname,
3851 uint32_t filetype, uint32_t object_size,
3852 bool verbose) {
3853 outs() << "Section\n";
3854 outs() << " sectname " << format("%.16s\n", sectname);
3855 outs() << " segname " << format("%.16s", segname);
3856 if (filetype != MachO::MH_OBJECT && strncmp(sg_segname, segname, 16) != 0)
3857 outs() << " (does not match segment)\n";
3858 else
3859 outs() << "\n";
3860 if (cmd == MachO::LC_SEGMENT_64) {
3861 outs() << " addr " << format("0x%016" PRIx64, addr) << "\n";
3862 outs() << " size " << format("0x%016" PRIx64, size);
3863 } else {
Kevin Enderby75594b62014-12-16 21:00:25 +00003864 outs() << " addr " << format("0x%08" PRIx64, addr) << "\n";
3865 outs() << " size " << format("0x%08" PRIx64, size);
Kevin Enderby956366c2014-08-29 22:30:52 +00003866 }
3867 if ((flags & MachO::S_ZEROFILL) != 0 && offset + size > object_size)
3868 outs() << " (past end of file)\n";
3869 else
3870 outs() << "\n";
3871 outs() << " offset " << offset;
3872 if (offset > object_size)
3873 outs() << " (past end of file)\n";
3874 else
3875 outs() << "\n";
3876 uint32_t align_shifted = 1 << align;
3877 outs() << " align 2^" << align << " (" << align_shifted << ")\n";
3878 outs() << " reloff " << reloff;
3879 if (reloff > object_size)
3880 outs() << " (past end of file)\n";
3881 else
3882 outs() << "\n";
3883 outs() << " nreloc " << nreloc;
3884 if (reloff + nreloc * sizeof(struct MachO::relocation_info) > object_size)
3885 outs() << " (past end of file)\n";
3886 else
3887 outs() << "\n";
3888 uint32_t section_type = flags & MachO::SECTION_TYPE;
3889 if (verbose) {
3890 outs() << " type";
3891 if (section_type == MachO::S_REGULAR)
3892 outs() << " S_REGULAR\n";
3893 else if (section_type == MachO::S_ZEROFILL)
3894 outs() << " S_ZEROFILL\n";
3895 else if (section_type == MachO::S_CSTRING_LITERALS)
3896 outs() << " S_CSTRING_LITERALS\n";
3897 else if (section_type == MachO::S_4BYTE_LITERALS)
3898 outs() << " S_4BYTE_LITERALS\n";
3899 else if (section_type == MachO::S_8BYTE_LITERALS)
3900 outs() << " S_8BYTE_LITERALS\n";
3901 else if (section_type == MachO::S_16BYTE_LITERALS)
3902 outs() << " S_16BYTE_LITERALS\n";
3903 else if (section_type == MachO::S_LITERAL_POINTERS)
3904 outs() << " S_LITERAL_POINTERS\n";
3905 else if (section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS)
3906 outs() << " S_NON_LAZY_SYMBOL_POINTERS\n";
3907 else if (section_type == MachO::S_LAZY_SYMBOL_POINTERS)
3908 outs() << " S_LAZY_SYMBOL_POINTERS\n";
3909 else if (section_type == MachO::S_SYMBOL_STUBS)
3910 outs() << " S_SYMBOL_STUBS\n";
3911 else if (section_type == MachO::S_MOD_INIT_FUNC_POINTERS)
3912 outs() << " S_MOD_INIT_FUNC_POINTERS\n";
3913 else if (section_type == MachO::S_MOD_TERM_FUNC_POINTERS)
3914 outs() << " S_MOD_TERM_FUNC_POINTERS\n";
3915 else if (section_type == MachO::S_COALESCED)
3916 outs() << " S_COALESCED\n";
3917 else if (section_type == MachO::S_INTERPOSING)
3918 outs() << " S_INTERPOSING\n";
3919 else if (section_type == MachO::S_DTRACE_DOF)
3920 outs() << " S_DTRACE_DOF\n";
3921 else if (section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS)
3922 outs() << " S_LAZY_DYLIB_SYMBOL_POINTERS\n";
3923 else if (section_type == MachO::S_THREAD_LOCAL_REGULAR)
3924 outs() << " S_THREAD_LOCAL_REGULAR\n";
3925 else if (section_type == MachO::S_THREAD_LOCAL_ZEROFILL)
3926 outs() << " S_THREAD_LOCAL_ZEROFILL\n";
3927 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLES)
3928 outs() << " S_THREAD_LOCAL_VARIABLES\n";
3929 else if (section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3930 outs() << " S_THREAD_LOCAL_VARIABLE_POINTERS\n";
3931 else if (section_type == MachO::S_THREAD_LOCAL_INIT_FUNCTION_POINTERS)
3932 outs() << " S_THREAD_LOCAL_INIT_FUNCTION_POINTERS\n";
3933 else
3934 outs() << format("0x%08" PRIx32, section_type) << "\n";
3935 outs() << "attributes";
3936 uint32_t section_attributes = flags & MachO::SECTION_ATTRIBUTES;
3937 if (section_attributes & MachO::S_ATTR_PURE_INSTRUCTIONS)
3938 outs() << " PURE_INSTRUCTIONS";
3939 if (section_attributes & MachO::S_ATTR_NO_TOC)
3940 outs() << " NO_TOC";
3941 if (section_attributes & MachO::S_ATTR_STRIP_STATIC_SYMS)
3942 outs() << " STRIP_STATIC_SYMS";
3943 if (section_attributes & MachO::S_ATTR_NO_DEAD_STRIP)
3944 outs() << " NO_DEAD_STRIP";
3945 if (section_attributes & MachO::S_ATTR_LIVE_SUPPORT)
3946 outs() << " LIVE_SUPPORT";
3947 if (section_attributes & MachO::S_ATTR_SELF_MODIFYING_CODE)
3948 outs() << " SELF_MODIFYING_CODE";
3949 if (section_attributes & MachO::S_ATTR_DEBUG)
3950 outs() << " DEBUG";
3951 if (section_attributes & MachO::S_ATTR_SOME_INSTRUCTIONS)
3952 outs() << " SOME_INSTRUCTIONS";
3953 if (section_attributes & MachO::S_ATTR_EXT_RELOC)
3954 outs() << " EXT_RELOC";
3955 if (section_attributes & MachO::S_ATTR_LOC_RELOC)
3956 outs() << " LOC_RELOC";
3957 if (section_attributes == 0)
3958 outs() << " (none)";
3959 outs() << "\n";
3960 } else
3961 outs() << " flags " << format("0x%08" PRIx32, flags) << "\n";
3962 outs() << " reserved1 " << reserved1;
3963 if (section_type == MachO::S_SYMBOL_STUBS ||
3964 section_type == MachO::S_LAZY_SYMBOL_POINTERS ||
3965 section_type == MachO::S_LAZY_DYLIB_SYMBOL_POINTERS ||
3966 section_type == MachO::S_NON_LAZY_SYMBOL_POINTERS ||
3967 section_type == MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
3968 outs() << " (index into indirect symbol table)\n";
3969 else
3970 outs() << "\n";
3971 outs() << " reserved2 " << reserved2;
3972 if (section_type == MachO::S_SYMBOL_STUBS)
3973 outs() << " (size of stubs)\n";
3974 else
3975 outs() << "\n";
3976}
3977
David Majnemer73cc6ff2014-11-13 19:48:56 +00003978static void PrintSymtabLoadCommand(MachO::symtab_command st, bool Is64Bit,
Kevin Enderby956366c2014-08-29 22:30:52 +00003979 uint32_t object_size) {
3980 outs() << " cmd LC_SYMTAB\n";
3981 outs() << " cmdsize " << st.cmdsize;
3982 if (st.cmdsize != sizeof(struct MachO::symtab_command))
3983 outs() << " Incorrect size\n";
3984 else
3985 outs() << "\n";
3986 outs() << " symoff " << st.symoff;
3987 if (st.symoff > object_size)
3988 outs() << " (past end of file)\n";
3989 else
3990 outs() << "\n";
3991 outs() << " nsyms " << st.nsyms;
3992 uint64_t big_size;
David Majnemer73cc6ff2014-11-13 19:48:56 +00003993 if (Is64Bit) {
Kevin Enderby956366c2014-08-29 22:30:52 +00003994 big_size = st.nsyms;
3995 big_size *= sizeof(struct MachO::nlist_64);
3996 big_size += st.symoff;
3997 if (big_size > object_size)
3998 outs() << " (past end of file)\n";
3999 else
4000 outs() << "\n";
4001 } else {
4002 big_size = st.nsyms;
4003 big_size *= sizeof(struct MachO::nlist);
4004 big_size += st.symoff;
4005 if (big_size > object_size)
4006 outs() << " (past end of file)\n";
4007 else
4008 outs() << "\n";
4009 }
4010 outs() << " stroff " << st.stroff;
4011 if (st.stroff > object_size)
4012 outs() << " (past end of file)\n";
4013 else
4014 outs() << "\n";
4015 outs() << " strsize " << st.strsize;
4016 big_size = st.stroff;
4017 big_size += st.strsize;
4018 if (big_size > object_size)
4019 outs() << " (past end of file)\n";
4020 else
4021 outs() << "\n";
4022}
4023
4024static void PrintDysymtabLoadCommand(MachO::dysymtab_command dyst,
4025 uint32_t nsyms, uint32_t object_size,
David Majnemer73cc6ff2014-11-13 19:48:56 +00004026 bool Is64Bit) {
Kevin Enderby956366c2014-08-29 22:30:52 +00004027 outs() << " cmd LC_DYSYMTAB\n";
4028 outs() << " cmdsize " << dyst.cmdsize;
4029 if (dyst.cmdsize != sizeof(struct MachO::dysymtab_command))
4030 outs() << " Incorrect size\n";
4031 else
4032 outs() << "\n";
4033 outs() << " ilocalsym " << dyst.ilocalsym;
4034 if (dyst.ilocalsym > nsyms)
4035 outs() << " (greater than the number of symbols)\n";
4036 else
4037 outs() << "\n";
4038 outs() << " nlocalsym " << dyst.nlocalsym;
4039 uint64_t big_size;
4040 big_size = dyst.ilocalsym;
4041 big_size += dyst.nlocalsym;
4042 if (big_size > nsyms)
4043 outs() << " (past the end of the symbol table)\n";
4044 else
4045 outs() << "\n";
4046 outs() << " iextdefsym " << dyst.iextdefsym;
4047 if (dyst.iextdefsym > nsyms)
4048 outs() << " (greater than the number of symbols)\n";
4049 else
4050 outs() << "\n";
4051 outs() << " nextdefsym " << dyst.nextdefsym;
4052 big_size = dyst.iextdefsym;
4053 big_size += dyst.nextdefsym;
4054 if (big_size > nsyms)
4055 outs() << " (past the end of the symbol table)\n";
4056 else
4057 outs() << "\n";
4058 outs() << " iundefsym " << dyst.iundefsym;
4059 if (dyst.iundefsym > nsyms)
4060 outs() << " (greater than the number of symbols)\n";
4061 else
4062 outs() << "\n";
4063 outs() << " nundefsym " << dyst.nundefsym;
4064 big_size = dyst.iundefsym;
4065 big_size += dyst.nundefsym;
4066 if (big_size > nsyms)
4067 outs() << " (past the end of the symbol table)\n";
4068 else
4069 outs() << "\n";
4070 outs() << " tocoff " << dyst.tocoff;
4071 if (dyst.tocoff > object_size)
4072 outs() << " (past end of file)\n";
4073 else
4074 outs() << "\n";
4075 outs() << " ntoc " << dyst.ntoc;
4076 big_size = dyst.ntoc;
4077 big_size *= sizeof(struct MachO::dylib_table_of_contents);
4078 big_size += dyst.tocoff;
4079 if (big_size > object_size)
4080 outs() << " (past end of file)\n";
4081 else
4082 outs() << "\n";
4083 outs() << " modtaboff " << dyst.modtaboff;
4084 if (dyst.modtaboff > object_size)
4085 outs() << " (past end of file)\n";
4086 else
4087 outs() << "\n";
4088 outs() << " nmodtab " << dyst.nmodtab;
4089 uint64_t modtabend;
David Majnemer73cc6ff2014-11-13 19:48:56 +00004090 if (Is64Bit) {
Kevin Enderby956366c2014-08-29 22:30:52 +00004091 modtabend = dyst.nmodtab;
4092 modtabend *= sizeof(struct MachO::dylib_module_64);
4093 modtabend += dyst.modtaboff;
4094 } else {
4095 modtabend = dyst.nmodtab;
4096 modtabend *= sizeof(struct MachO::dylib_module);
4097 modtabend += dyst.modtaboff;
4098 }
4099 if (modtabend > object_size)
4100 outs() << " (past end of file)\n";
4101 else
4102 outs() << "\n";
4103 outs() << " extrefsymoff " << dyst.extrefsymoff;
4104 if (dyst.extrefsymoff > object_size)
4105 outs() << " (past end of file)\n";
4106 else
4107 outs() << "\n";
4108 outs() << " nextrefsyms " << dyst.nextrefsyms;
4109 big_size = dyst.nextrefsyms;
4110 big_size *= sizeof(struct MachO::dylib_reference);
4111 big_size += dyst.extrefsymoff;
4112 if (big_size > object_size)
4113 outs() << " (past end of file)\n";
4114 else
4115 outs() << "\n";
4116 outs() << " indirectsymoff " << dyst.indirectsymoff;
4117 if (dyst.indirectsymoff > object_size)
4118 outs() << " (past end of file)\n";
4119 else
4120 outs() << "\n";
4121 outs() << " nindirectsyms " << dyst.nindirectsyms;
4122 big_size = dyst.nindirectsyms;
4123 big_size *= sizeof(uint32_t);
4124 big_size += dyst.indirectsymoff;
4125 if (big_size > object_size)
4126 outs() << " (past end of file)\n";
4127 else
4128 outs() << "\n";
4129 outs() << " extreloff " << dyst.extreloff;
4130 if (dyst.extreloff > object_size)
4131 outs() << " (past end of file)\n";
4132 else
4133 outs() << "\n";
4134 outs() << " nextrel " << dyst.nextrel;
4135 big_size = dyst.nextrel;
4136 big_size *= sizeof(struct MachO::relocation_info);
4137 big_size += dyst.extreloff;
4138 if (big_size > object_size)
4139 outs() << " (past end of file)\n";
4140 else
4141 outs() << "\n";
4142 outs() << " locreloff " << dyst.locreloff;
4143 if (dyst.locreloff > object_size)
4144 outs() << " (past end of file)\n";
4145 else
4146 outs() << "\n";
4147 outs() << " nlocrel " << dyst.nlocrel;
4148 big_size = dyst.nlocrel;
4149 big_size *= sizeof(struct MachO::relocation_info);
4150 big_size += dyst.locreloff;
4151 if (big_size > object_size)
4152 outs() << " (past end of file)\n";
4153 else
4154 outs() << "\n";
4155}
4156
Kevin Enderby8ae63c12014-09-04 16:54:47 +00004157static void PrintDyldInfoLoadCommand(MachO::dyld_info_command dc,
4158 uint32_t object_size) {
4159 if (dc.cmd == MachO::LC_DYLD_INFO)
4160 outs() << " cmd LC_DYLD_INFO\n";
4161 else
4162 outs() << " cmd LC_DYLD_INFO_ONLY\n";
4163 outs() << " cmdsize " << dc.cmdsize;
4164 if (dc.cmdsize != sizeof(struct MachO::dyld_info_command))
4165 outs() << " Incorrect size\n";
4166 else
4167 outs() << "\n";
4168 outs() << " rebase_off " << dc.rebase_off;
4169 if (dc.rebase_off > object_size)
4170 outs() << " (past end of file)\n";
4171 else
4172 outs() << "\n";
4173 outs() << " rebase_size " << dc.rebase_size;
4174 uint64_t big_size;
4175 big_size = dc.rebase_off;
4176 big_size += dc.rebase_size;
4177 if (big_size > object_size)
4178 outs() << " (past end of file)\n";
4179 else
4180 outs() << "\n";
4181 outs() << " bind_off " << dc.bind_off;
4182 if (dc.bind_off > object_size)
4183 outs() << " (past end of file)\n";
4184 else
4185 outs() << "\n";
4186 outs() << " bind_size " << dc.bind_size;
4187 big_size = dc.bind_off;
4188 big_size += dc.bind_size;
4189 if (big_size > object_size)
4190 outs() << " (past end of file)\n";
4191 else
4192 outs() << "\n";
4193 outs() << " weak_bind_off " << dc.weak_bind_off;
4194 if (dc.weak_bind_off > object_size)
4195 outs() << " (past end of file)\n";
4196 else
4197 outs() << "\n";
4198 outs() << " weak_bind_size " << dc.weak_bind_size;
4199 big_size = dc.weak_bind_off;
4200 big_size += dc.weak_bind_size;
4201 if (big_size > object_size)
4202 outs() << " (past end of file)\n";
4203 else
4204 outs() << "\n";
4205 outs() << " lazy_bind_off " << dc.lazy_bind_off;
4206 if (dc.lazy_bind_off > object_size)
4207 outs() << " (past end of file)\n";
4208 else
4209 outs() << "\n";
4210 outs() << " lazy_bind_size " << dc.lazy_bind_size;
4211 big_size = dc.lazy_bind_off;
4212 big_size += dc.lazy_bind_size;
4213 if (big_size > object_size)
4214 outs() << " (past end of file)\n";
4215 else
4216 outs() << "\n";
4217 outs() << " export_off " << dc.export_off;
4218 if (dc.export_off > object_size)
4219 outs() << " (past end of file)\n";
4220 else
4221 outs() << "\n";
4222 outs() << " export_size " << dc.export_size;
4223 big_size = dc.export_off;
4224 big_size += dc.export_size;
4225 if (big_size > object_size)
4226 outs() << " (past end of file)\n";
4227 else
4228 outs() << "\n";
4229}
4230
4231static void PrintDyldLoadCommand(MachO::dylinker_command dyld,
4232 const char *Ptr) {
4233 if (dyld.cmd == MachO::LC_ID_DYLINKER)
4234 outs() << " cmd LC_ID_DYLINKER\n";
4235 else if (dyld.cmd == MachO::LC_LOAD_DYLINKER)
4236 outs() << " cmd LC_LOAD_DYLINKER\n";
4237 else if (dyld.cmd == MachO::LC_DYLD_ENVIRONMENT)
4238 outs() << " cmd LC_DYLD_ENVIRONMENT\n";
4239 else
4240 outs() << " cmd ?(" << dyld.cmd << ")\n";
4241 outs() << " cmdsize " << dyld.cmdsize;
4242 if (dyld.cmdsize < sizeof(struct MachO::dylinker_command))
4243 outs() << " Incorrect size\n";
4244 else
4245 outs() << "\n";
4246 if (dyld.name >= dyld.cmdsize)
4247 outs() << " name ?(bad offset " << dyld.name << ")\n";
4248 else {
Kevin Enderbyb28ed012014-10-29 21:28:24 +00004249 const char *P = (const char *)(Ptr) + dyld.name;
Kevin Enderby8ae63c12014-09-04 16:54:47 +00004250 outs() << " name " << P << " (offset " << dyld.name << ")\n";
4251 }
4252}
4253
4254static void PrintUuidLoadCommand(MachO::uuid_command uuid) {
4255 outs() << " cmd LC_UUID\n";
4256 outs() << " cmdsize " << uuid.cmdsize;
4257 if (uuid.cmdsize != sizeof(struct MachO::uuid_command))
4258 outs() << " Incorrect size\n";
4259 else
4260 outs() << "\n";
4261 outs() << " uuid ";
4262 outs() << format("%02" PRIX32, uuid.uuid[0]);
4263 outs() << format("%02" PRIX32, uuid.uuid[1]);
4264 outs() << format("%02" PRIX32, uuid.uuid[2]);
4265 outs() << format("%02" PRIX32, uuid.uuid[3]);
4266 outs() << "-";
4267 outs() << format("%02" PRIX32, uuid.uuid[4]);
4268 outs() << format("%02" PRIX32, uuid.uuid[5]);
4269 outs() << "-";
4270 outs() << format("%02" PRIX32, uuid.uuid[6]);
4271 outs() << format("%02" PRIX32, uuid.uuid[7]);
4272 outs() << "-";
4273 outs() << format("%02" PRIX32, uuid.uuid[8]);
4274 outs() << format("%02" PRIX32, uuid.uuid[9]);
4275 outs() << "-";
4276 outs() << format("%02" PRIX32, uuid.uuid[10]);
4277 outs() << format("%02" PRIX32, uuid.uuid[11]);
4278 outs() << format("%02" PRIX32, uuid.uuid[12]);
4279 outs() << format("%02" PRIX32, uuid.uuid[13]);
4280 outs() << format("%02" PRIX32, uuid.uuid[14]);
4281 outs() << format("%02" PRIX32, uuid.uuid[15]);
4282 outs() << "\n";
4283}
4284
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004285static void PrintRpathLoadCommand(MachO::rpath_command rpath, const char *Ptr) {
Jean-Daniel Dupas00cc1f52014-12-04 07:37:02 +00004286 outs() << " cmd LC_RPATH\n";
4287 outs() << " cmdsize " << rpath.cmdsize;
4288 if (rpath.cmdsize < sizeof(struct MachO::rpath_command))
4289 outs() << " Incorrect size\n";
4290 else
4291 outs() << "\n";
4292 if (rpath.path >= rpath.cmdsize)
4293 outs() << " path ?(bad offset " << rpath.path << ")\n";
4294 else {
4295 const char *P = (const char *)(Ptr) + rpath.path;
4296 outs() << " path " << P << " (offset " << rpath.path << ")\n";
4297 }
4298}
4299
Kevin Enderby8ae63c12014-09-04 16:54:47 +00004300static void PrintVersionMinLoadCommand(MachO::version_min_command vd) {
4301 if (vd.cmd == MachO::LC_VERSION_MIN_MACOSX)
4302 outs() << " cmd LC_VERSION_MIN_MACOSX\n";
4303 else if (vd.cmd == MachO::LC_VERSION_MIN_IPHONEOS)
4304 outs() << " cmd LC_VERSION_MIN_IPHONEOS\n";
4305 else
4306 outs() << " cmd " << vd.cmd << " (?)\n";
4307 outs() << " cmdsize " << vd.cmdsize;
4308 if (vd.cmdsize != sizeof(struct MachO::version_min_command))
4309 outs() << " Incorrect size\n";
4310 else
4311 outs() << "\n";
4312 outs() << " version " << ((vd.version >> 16) & 0xffff) << "."
4313 << ((vd.version >> 8) & 0xff);
4314 if ((vd.version & 0xff) != 0)
4315 outs() << "." << (vd.version & 0xff);
4316 outs() << "\n";
4317 if (vd.sdk == 0)
Kevin Enderby57538292014-12-17 01:01:30 +00004318 outs() << " sdk n/a";
Kevin Enderby8ae63c12014-09-04 16:54:47 +00004319 else {
4320 outs() << " sdk " << ((vd.sdk >> 16) & 0xffff) << "."
4321 << ((vd.sdk >> 8) & 0xff);
4322 }
4323 if ((vd.sdk & 0xff) != 0)
4324 outs() << "." << (vd.sdk & 0xff);
4325 outs() << "\n";
4326}
4327
4328static void PrintSourceVersionCommand(MachO::source_version_command sd) {
4329 outs() << " cmd LC_SOURCE_VERSION\n";
4330 outs() << " cmdsize " << sd.cmdsize;
4331 if (sd.cmdsize != sizeof(struct MachO::source_version_command))
4332 outs() << " Incorrect size\n";
4333 else
4334 outs() << "\n";
4335 uint64_t a = (sd.version >> 40) & 0xffffff;
4336 uint64_t b = (sd.version >> 30) & 0x3ff;
4337 uint64_t c = (sd.version >> 20) & 0x3ff;
4338 uint64_t d = (sd.version >> 10) & 0x3ff;
4339 uint64_t e = sd.version & 0x3ff;
4340 outs() << " version " << a << "." << b;
4341 if (e != 0)
4342 outs() << "." << c << "." << d << "." << e;
4343 else if (d != 0)
4344 outs() << "." << c << "." << d;
4345 else if (c != 0)
4346 outs() << "." << c;
4347 outs() << "\n";
4348}
4349
4350static void PrintEntryPointCommand(MachO::entry_point_command ep) {
4351 outs() << " cmd LC_MAIN\n";
4352 outs() << " cmdsize " << ep.cmdsize;
4353 if (ep.cmdsize != sizeof(struct MachO::entry_point_command))
4354 outs() << " Incorrect size\n";
4355 else
4356 outs() << "\n";
4357 outs() << " entryoff " << ep.entryoff << "\n";
4358 outs() << " stacksize " << ep.stacksize << "\n";
4359}
4360
Kevin Enderby0804f4672014-12-16 23:25:52 +00004361static void PrintEncryptionInfoCommand(MachO::encryption_info_command ec,
4362 uint32_t object_size) {
4363 outs() << " cmd LC_ENCRYPTION_INFO\n";
4364 outs() << " cmdsize " << ec.cmdsize;
4365 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command))
4366 outs() << " Incorrect size\n";
4367 else
4368 outs() << "\n";
4369 outs() << " cryptoff " << ec.cryptoff;
4370 if (ec.cryptoff > object_size)
4371 outs() << " (past end of file)\n";
4372 else
4373 outs() << "\n";
4374 outs() << " cryptsize " << ec.cryptsize;
4375 if (ec.cryptsize > object_size)
4376 outs() << " (past end of file)\n";
4377 else
4378 outs() << "\n";
4379 outs() << " cryptid " << ec.cryptid << "\n";
4380}
4381
Kevin Enderby57538292014-12-17 01:01:30 +00004382static void PrintEncryptionInfoCommand64(MachO::encryption_info_command_64 ec,
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004383 uint32_t object_size) {
Kevin Enderby57538292014-12-17 01:01:30 +00004384 outs() << " cmd LC_ENCRYPTION_INFO_64\n";
4385 outs() << " cmdsize " << ec.cmdsize;
4386 if (ec.cmdsize != sizeof(struct MachO::encryption_info_command_64))
4387 outs() << " Incorrect size\n";
4388 else
4389 outs() << "\n";
4390 outs() << " cryptoff " << ec.cryptoff;
4391 if (ec.cryptoff > object_size)
4392 outs() << " (past end of file)\n";
4393 else
4394 outs() << "\n";
4395 outs() << " cryptsize " << ec.cryptsize;
4396 if (ec.cryptsize > object_size)
4397 outs() << " (past end of file)\n";
4398 else
4399 outs() << "\n";
4400 outs() << " cryptid " << ec.cryptid << "\n";
4401 outs() << " pad " << ec.pad << "\n";
4402}
4403
Kevin Enderbyd0b6b7f2014-12-18 00:53:40 +00004404static void PrintLinkerOptionCommand(MachO::linker_option_command lo,
4405 const char *Ptr) {
4406 outs() << " cmd LC_LINKER_OPTION\n";
4407 outs() << " cmdsize " << lo.cmdsize;
4408 if (lo.cmdsize < sizeof(struct MachO::linker_option_command))
4409 outs() << " Incorrect size\n";
4410 else
4411 outs() << "\n";
4412 outs() << " count " << lo.count << "\n";
4413 const char *string = Ptr + sizeof(struct MachO::linker_option_command);
4414 uint32_t left = lo.cmdsize - sizeof(struct MachO::linker_option_command);
4415 uint32_t i = 0;
4416 while (left > 0) {
4417 while (*string == '\0' && left > 0) {
4418 string++;
4419 left--;
4420 }
4421 if (left > 0) {
4422 i++;
4423 outs() << " string #" << i << " " << format("%.*s\n", left, string);
David Majnemerd4449ed2014-12-20 08:24:43 +00004424 uint32_t NullPos = StringRef(string, left).find('\0');
4425 uint32_t len = std::min(NullPos, left) + 1;
Kevin Enderbyd0b6b7f2014-12-18 00:53:40 +00004426 string += len;
4427 left -= len;
4428 }
4429 }
4430 if (lo.count != i)
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004431 outs() << " count " << lo.count << " does not match number of strings "
4432 << i << "\n";
Kevin Enderbyd0b6b7f2014-12-18 00:53:40 +00004433}
4434
Kevin Enderbyb4b79312014-12-18 19:24:35 +00004435static void PrintSubFrameworkCommand(MachO::sub_framework_command sub,
4436 const char *Ptr) {
4437 outs() << " cmd LC_SUB_FRAMEWORK\n";
4438 outs() << " cmdsize " << sub.cmdsize;
4439 if (sub.cmdsize < sizeof(struct MachO::sub_framework_command))
4440 outs() << " Incorrect size\n";
4441 else
4442 outs() << "\n";
4443 if (sub.umbrella < sub.cmdsize) {
4444 const char *P = Ptr + sub.umbrella;
4445 outs() << " umbrella " << P << " (offset " << sub.umbrella << ")\n";
4446 } else {
4447 outs() << " umbrella ?(bad offset " << sub.umbrella << ")\n";
4448 }
4449}
4450
Kevin Enderbya2bd8d92014-12-18 23:13:26 +00004451static void PrintSubUmbrellaCommand(MachO::sub_umbrella_command sub,
4452 const char *Ptr) {
4453 outs() << " cmd LC_SUB_UMBRELLA\n";
4454 outs() << " cmdsize " << sub.cmdsize;
4455 if (sub.cmdsize < sizeof(struct MachO::sub_umbrella_command))
4456 outs() << " Incorrect size\n";
4457 else
4458 outs() << "\n";
4459 if (sub.sub_umbrella < sub.cmdsize) {
4460 const char *P = Ptr + sub.sub_umbrella;
4461 outs() << " sub_umbrella " << P << " (offset " << sub.sub_umbrella << ")\n";
4462 } else {
4463 outs() << " sub_umbrella ?(bad offset " << sub.sub_umbrella << ")\n";
4464 }
4465}
4466
Kevin Enderby36c8d3a2014-12-19 19:48:16 +00004467static void PrintSubLibraryCommand(MachO::sub_library_command sub,
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004468 const char *Ptr) {
Kevin Enderby36c8d3a2014-12-19 19:48:16 +00004469 outs() << " cmd LC_SUB_LIBRARY\n";
4470 outs() << " cmdsize " << sub.cmdsize;
4471 if (sub.cmdsize < sizeof(struct MachO::sub_library_command))
4472 outs() << " Incorrect size\n";
4473 else
4474 outs() << "\n";
4475 if (sub.sub_library < sub.cmdsize) {
4476 const char *P = Ptr + sub.sub_library;
4477 outs() << " sub_library " << P << " (offset " << sub.sub_library << ")\n";
4478 } else {
4479 outs() << " sub_library ?(bad offset " << sub.sub_library << ")\n";
4480 }
4481}
4482
Kevin Enderby186eac32014-12-19 21:06:24 +00004483static void PrintSubClientCommand(MachO::sub_client_command sub,
4484 const char *Ptr) {
4485 outs() << " cmd LC_SUB_CLIENT\n";
4486 outs() << " cmdsize " << sub.cmdsize;
4487 if (sub.cmdsize < sizeof(struct MachO::sub_client_command))
4488 outs() << " Incorrect size\n";
4489 else
4490 outs() << "\n";
4491 if (sub.client < sub.cmdsize) {
4492 const char *P = Ptr + sub.client;
4493 outs() << " client " << P << " (offset " << sub.client << ")\n";
4494 } else {
4495 outs() << " client ?(bad offset " << sub.client << ")\n";
4496 }
4497}
Kevin Enderby36c8d3a2014-12-19 19:48:16 +00004498
Kevin Enderby52e4ce42014-12-19 22:25:22 +00004499static void PrintRoutinesCommand(MachO::routines_command r) {
4500 outs() << " cmd LC_ROUTINES\n";
4501 outs() << " cmdsize " << r.cmdsize;
4502 if (r.cmdsize != sizeof(struct MachO::routines_command))
4503 outs() << " Incorrect size\n";
4504 else
4505 outs() << "\n";
4506 outs() << " init_address " << format("0x%08" PRIx32, r.init_address) << "\n";
4507 outs() << " init_module " << r.init_module << "\n";
4508 outs() << " reserved1 " << r.reserved1 << "\n";
4509 outs() << " reserved2 " << r.reserved2 << "\n";
4510 outs() << " reserved3 " << r.reserved3 << "\n";
4511 outs() << " reserved4 " << r.reserved4 << "\n";
4512 outs() << " reserved5 " << r.reserved5 << "\n";
4513 outs() << " reserved6 " << r.reserved6 << "\n";
4514}
4515
4516static void PrintRoutinesCommand64(MachO::routines_command_64 r) {
4517 outs() << " cmd LC_ROUTINES_64\n";
4518 outs() << " cmdsize " << r.cmdsize;
4519 if (r.cmdsize != sizeof(struct MachO::routines_command_64))
4520 outs() << " Incorrect size\n";
4521 else
4522 outs() << "\n";
4523 outs() << " init_address " << format("0x%016" PRIx64, r.init_address) << "\n";
4524 outs() << " init_module " << r.init_module << "\n";
4525 outs() << " reserved1 " << r.reserved1 << "\n";
4526 outs() << " reserved2 " << r.reserved2 << "\n";
4527 outs() << " reserved3 " << r.reserved3 << "\n";
4528 outs() << " reserved4 " << r.reserved4 << "\n";
4529 outs() << " reserved5 " << r.reserved5 << "\n";
4530 outs() << " reserved6 " << r.reserved6 << "\n";
4531}
4532
Kevin Enderby48ef5342014-12-23 22:56:39 +00004533static void Print_x86_thread_state64_t(MachO::x86_thread_state64_t &cpu64) {
4534 outs() << " rax " << format("0x%016" PRIx64, cpu64.rax);
4535 outs() << " rbx " << format("0x%016" PRIx64, cpu64.rbx);
4536 outs() << " rcx " << format("0x%016" PRIx64, cpu64.rcx) << "\n";
4537 outs() << " rdx " << format("0x%016" PRIx64, cpu64.rdx);
4538 outs() << " rdi " << format("0x%016" PRIx64, cpu64.rdi);
4539 outs() << " rsi " << format("0x%016" PRIx64, cpu64.rsi) << "\n";
4540 outs() << " rbp " << format("0x%016" PRIx64, cpu64.rbp);
4541 outs() << " rsp " << format("0x%016" PRIx64, cpu64.rsp);
4542 outs() << " r8 " << format("0x%016" PRIx64, cpu64.r8) << "\n";
4543 outs() << " r9 " << format("0x%016" PRIx64, cpu64.r9);
4544 outs() << " r10 " << format("0x%016" PRIx64, cpu64.r10);
4545 outs() << " r11 " << format("0x%016" PRIx64, cpu64.r11) << "\n";
4546 outs() << " r12 " << format("0x%016" PRIx64, cpu64.r12);
4547 outs() << " r13 " << format("0x%016" PRIx64, cpu64.r13);
4548 outs() << " r14 " << format("0x%016" PRIx64, cpu64.r14) << "\n";
4549 outs() << " r15 " << format("0x%016" PRIx64, cpu64.r15);
4550 outs() << " rip " << format("0x%016" PRIx64, cpu64.rip) << "\n";
4551 outs() << "rflags " << format("0x%016" PRIx64, cpu64.rflags);
4552 outs() << " cs " << format("0x%016" PRIx64, cpu64.cs);
4553 outs() << " fs " << format("0x%016" PRIx64, cpu64.fs) << "\n";
4554 outs() << " gs " << format("0x%016" PRIx64, cpu64.gs) << "\n";
4555}
4556
Kevin Enderby227df342014-12-23 23:43:59 +00004557static void Print_mmst_reg(MachO::mmst_reg_t &r) {
Kevin Enderby48ef5342014-12-23 22:56:39 +00004558 uint32_t f;
4559 outs() << "\t mmst_reg ";
4560 for (f = 0; f < 10; f++)
4561 outs() << format("%02" PRIx32, (r.mmst_reg[f] & 0xff)) << " ";
4562 outs() << "\n";
4563 outs() << "\t mmst_rsrv ";
4564 for (f = 0; f < 6; f++)
4565 outs() << format("%02" PRIx32, (r.mmst_rsrv[f] & 0xff)) << " ";
4566 outs() << "\n";
4567}
4568
Kevin Enderbyaefb0032014-12-24 00:16:51 +00004569static void Print_xmm_reg(MachO::xmm_reg_t &r) {
Kevin Enderby48ef5342014-12-23 22:56:39 +00004570 uint32_t f;
4571 outs() << "\t xmm_reg ";
4572 for (f = 0; f < 16; f++)
4573 outs() << format("%02" PRIx32, (r.xmm_reg[f] & 0xff)) << " ";
4574 outs() << "\n";
4575}
4576
4577static void Print_x86_float_state_t(MachO::x86_float_state64_t &fpu) {
4578 outs() << "\t fpu_reserved[0] " << fpu.fpu_reserved[0];
4579 outs() << " fpu_reserved[1] " << fpu.fpu_reserved[1] << "\n";
4580 outs() << "\t control: invalid " << fpu.fpu_fcw.invalid;
4581 outs() << " denorm " << fpu.fpu_fcw.denorm;
4582 outs() << " zdiv " << fpu.fpu_fcw.zdiv;
4583 outs() << " ovrfl " << fpu.fpu_fcw.ovrfl;
4584 outs() << " undfl " << fpu.fpu_fcw.undfl;
4585 outs() << " precis " << fpu.fpu_fcw.precis << "\n";
4586 outs() << "\t\t pc ";
4587 if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_24B)
4588 outs() << "FP_PREC_24B ";
4589 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_53B)
4590 outs() << "FP_PREC_53B ";
4591 else if (fpu.fpu_fcw.pc == MachO::x86_FP_PREC_64B)
4592 outs() << "FP_PREC_64B ";
4593 else
4594 outs() << fpu.fpu_fcw.pc << " ";
4595 outs() << "rc ";
4596 if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_NEAR)
4597 outs() << "FP_RND_NEAR ";
4598 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_DOWN)
4599 outs() << "FP_RND_DOWN ";
4600 else if (fpu.fpu_fcw.rc == MachO::x86_FP_RND_UP)
4601 outs() << "FP_RND_UP ";
4602 else if (fpu.fpu_fcw.rc == MachO::x86_FP_CHOP)
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004603 outs() << "FP_CHOP ";
Kevin Enderby48ef5342014-12-23 22:56:39 +00004604 outs() << "\n";
4605 outs() << "\t status: invalid " << fpu.fpu_fsw.invalid;
4606 outs() << " denorm " << fpu.fpu_fsw.denorm;
4607 outs() << " zdiv " << fpu.fpu_fsw.zdiv;
4608 outs() << " ovrfl " << fpu.fpu_fsw.ovrfl;
4609 outs() << " undfl " << fpu.fpu_fsw.undfl;
4610 outs() << " precis " << fpu.fpu_fsw.precis;
4611 outs() << " stkflt " << fpu.fpu_fsw.stkflt << "\n";
4612 outs() << "\t errsumm " << fpu.fpu_fsw.errsumm;
4613 outs() << " c0 " << fpu.fpu_fsw.c0;
4614 outs() << " c1 " << fpu.fpu_fsw.c1;
4615 outs() << " c2 " << fpu.fpu_fsw.c2;
4616 outs() << " tos " << fpu.fpu_fsw.tos;
4617 outs() << " c3 " << fpu.fpu_fsw.c3;
4618 outs() << " busy " << fpu.fpu_fsw.busy << "\n";
4619 outs() << "\t fpu_ftw " << format("0x%02" PRIx32, fpu.fpu_ftw);
4620 outs() << " fpu_rsrv1 " << format("0x%02" PRIx32, fpu.fpu_rsrv1);
4621 outs() << " fpu_fop " << format("0x%04" PRIx32, fpu.fpu_fop);
4622 outs() << " fpu_ip " << format("0x%08" PRIx32, fpu.fpu_ip) << "\n";
4623 outs() << "\t fpu_cs " << format("0x%04" PRIx32, fpu.fpu_cs);
4624 outs() << " fpu_rsrv2 " << format("0x%04" PRIx32, fpu.fpu_rsrv2);
4625 outs() << " fpu_dp " << format("0x%08" PRIx32, fpu.fpu_dp);
4626 outs() << " fpu_ds " << format("0x%04" PRIx32, fpu.fpu_ds) << "\n";
4627 outs() << "\t fpu_rsrv3 " << format("0x%04" PRIx32, fpu.fpu_rsrv3);
4628 outs() << " fpu_mxcsr " << format("0x%08" PRIx32, fpu.fpu_mxcsr);
4629 outs() << " fpu_mxcsrmask " << format("0x%08" PRIx32, fpu.fpu_mxcsrmask);
4630 outs() << "\n";
4631 outs() << "\t fpu_stmm0:\n";
4632 Print_mmst_reg(fpu.fpu_stmm0);
4633 outs() << "\t fpu_stmm1:\n";
4634 Print_mmst_reg(fpu.fpu_stmm1);
4635 outs() << "\t fpu_stmm2:\n";
4636 Print_mmst_reg(fpu.fpu_stmm2);
4637 outs() << "\t fpu_stmm3:\n";
4638 Print_mmst_reg(fpu.fpu_stmm3);
4639 outs() << "\t fpu_stmm4:\n";
4640 Print_mmst_reg(fpu.fpu_stmm4);
4641 outs() << "\t fpu_stmm5:\n";
4642 Print_mmst_reg(fpu.fpu_stmm5);
4643 outs() << "\t fpu_stmm6:\n";
4644 Print_mmst_reg(fpu.fpu_stmm6);
4645 outs() << "\t fpu_stmm7:\n";
4646 Print_mmst_reg(fpu.fpu_stmm7);
4647 outs() << "\t fpu_xmm0:\n";
4648 Print_xmm_reg(fpu.fpu_xmm0);
4649 outs() << "\t fpu_xmm1:\n";
4650 Print_xmm_reg(fpu.fpu_xmm1);
4651 outs() << "\t fpu_xmm2:\n";
4652 Print_xmm_reg(fpu.fpu_xmm2);
4653 outs() << "\t fpu_xmm3:\n";
4654 Print_xmm_reg(fpu.fpu_xmm3);
4655 outs() << "\t fpu_xmm4:\n";
4656 Print_xmm_reg(fpu.fpu_xmm4);
4657 outs() << "\t fpu_xmm5:\n";
4658 Print_xmm_reg(fpu.fpu_xmm5);
4659 outs() << "\t fpu_xmm6:\n";
4660 Print_xmm_reg(fpu.fpu_xmm6);
4661 outs() << "\t fpu_xmm7:\n";
4662 Print_xmm_reg(fpu.fpu_xmm7);
4663 outs() << "\t fpu_xmm8:\n";
4664 Print_xmm_reg(fpu.fpu_xmm8);
4665 outs() << "\t fpu_xmm9:\n";
4666 Print_xmm_reg(fpu.fpu_xmm9);
4667 outs() << "\t fpu_xmm10:\n";
4668 Print_xmm_reg(fpu.fpu_xmm10);
4669 outs() << "\t fpu_xmm11:\n";
4670 Print_xmm_reg(fpu.fpu_xmm11);
4671 outs() << "\t fpu_xmm12:\n";
4672 Print_xmm_reg(fpu.fpu_xmm12);
4673 outs() << "\t fpu_xmm13:\n";
4674 Print_xmm_reg(fpu.fpu_xmm13);
4675 outs() << "\t fpu_xmm14:\n";
4676 Print_xmm_reg(fpu.fpu_xmm14);
4677 outs() << "\t fpu_xmm15:\n";
4678 Print_xmm_reg(fpu.fpu_xmm15);
4679 outs() << "\t fpu_rsrv4:\n";
4680 for (uint32_t f = 0; f < 6; f++) {
4681 outs() << "\t ";
4682 for (uint32_t g = 0; g < 16; g++)
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004683 outs() << format("%02" PRIx32, fpu.fpu_rsrv4[f * g]) << " ";
Kevin Enderby48ef5342014-12-23 22:56:39 +00004684 outs() << "\n";
4685 }
4686 outs() << "\t fpu_reserved1 " << format("0x%08" PRIx32, fpu.fpu_reserved1);
4687 outs() << "\n";
4688}
4689
4690static void Print_x86_exception_state_t(MachO::x86_exception_state64_t &exc64) {
4691 outs() << "\t trapno " << format("0x%08" PRIx32, exc64.trapno);
4692 outs() << " err " << format("0x%08" PRIx32, exc64.err);
4693 outs() << " faultvaddr " << format("0x%016" PRIx64, exc64.faultvaddr) << "\n";
4694}
4695
4696static void PrintThreadCommand(MachO::thread_command t, const char *Ptr,
4697 bool isLittleEndian, uint32_t cputype) {
4698 if (t.cmd == MachO::LC_THREAD)
4699 outs() << " cmd LC_THREAD\n";
4700 else if (t.cmd == MachO::LC_UNIXTHREAD)
4701 outs() << " cmd LC_UNIXTHREAD\n";
4702 else
4703 outs() << " cmd " << t.cmd << " (unknown)\n";
4704 outs() << " cmdsize " << t.cmdsize;
4705 if (t.cmdsize < sizeof(struct MachO::thread_command) + 2 * sizeof(uint32_t))
4706 outs() << " Incorrect size\n";
4707 else
4708 outs() << "\n";
4709
4710 const char *begin = Ptr + sizeof(struct MachO::thread_command);
4711 const char *end = Ptr + t.cmdsize;
4712 uint32_t flavor, count, left;
4713 if (cputype == MachO::CPU_TYPE_X86_64) {
4714 while (begin < end) {
4715 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4716 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4717 begin += sizeof(uint32_t);
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004718 } else {
Kevin Enderby48ef5342014-12-23 22:56:39 +00004719 flavor = 0;
4720 begin = end;
4721 }
4722 if (isLittleEndian != sys::IsLittleEndianHost)
4723 sys::swapByteOrder(flavor);
4724 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4725 memcpy((char *)&count, begin, sizeof(uint32_t));
4726 begin += sizeof(uint32_t);
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004727 } else {
Kevin Enderby48ef5342014-12-23 22:56:39 +00004728 count = 0;
4729 begin = end;
4730 }
4731 if (isLittleEndian != sys::IsLittleEndianHost)
4732 sys::swapByteOrder(count);
4733 if (flavor == MachO::x86_THREAD_STATE64) {
4734 outs() << " flavor x86_THREAD_STATE64\n";
4735 if (count == MachO::x86_THREAD_STATE64_COUNT)
4736 outs() << " count x86_THREAD_STATE64_COUNT\n";
4737 else
4738 outs() << " count " << count
4739 << " (not x86_THREAD_STATE64_COUNT)\n";
4740 MachO::x86_thread_state64_t cpu64;
4741 left = end - begin;
4742 if (left >= sizeof(MachO::x86_thread_state64_t)) {
4743 memcpy(&cpu64, begin, sizeof(MachO::x86_thread_state64_t));
4744 begin += sizeof(MachO::x86_thread_state64_t);
4745 } else {
4746 memset(&cpu64, '\0', sizeof(MachO::x86_thread_state64_t));
4747 memcpy(&cpu64, begin, left);
4748 begin += left;
4749 }
4750 if (isLittleEndian != sys::IsLittleEndianHost)
4751 swapStruct(cpu64);
4752 Print_x86_thread_state64_t(cpu64);
4753 } else if (flavor == MachO::x86_THREAD_STATE) {
4754 outs() << " flavor x86_THREAD_STATE\n";
4755 if (count == MachO::x86_THREAD_STATE_COUNT)
4756 outs() << " count x86_THREAD_STATE_COUNT\n";
4757 else
4758 outs() << " count " << count
4759 << " (not x86_THREAD_STATE_COUNT)\n";
4760 struct MachO::x86_thread_state_t ts;
4761 left = end - begin;
4762 if (left >= sizeof(MachO::x86_thread_state_t)) {
4763 memcpy(&ts, begin, sizeof(MachO::x86_thread_state_t));
4764 begin += sizeof(MachO::x86_thread_state_t);
4765 } else {
4766 memset(&ts, '\0', sizeof(MachO::x86_thread_state_t));
4767 memcpy(&ts, begin, left);
4768 begin += left;
4769 }
4770 if (isLittleEndian != sys::IsLittleEndianHost)
4771 swapStruct(ts);
4772 if (ts.tsh.flavor == MachO::x86_THREAD_STATE64) {
4773 outs() << "\t tsh.flavor x86_THREAD_STATE64 ";
4774 if (ts.tsh.count == MachO::x86_THREAD_STATE64_COUNT)
4775 outs() << "tsh.count x86_THREAD_STATE64_COUNT\n";
4776 else
4777 outs() << "tsh.count " << ts.tsh.count
4778 << " (not x86_THREAD_STATE64_COUNT\n";
4779 Print_x86_thread_state64_t(ts.uts.ts64);
4780 } else {
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004781 outs() << "\t tsh.flavor " << ts.tsh.flavor << " tsh.count "
4782 << ts.tsh.count << "\n";
Kevin Enderby48ef5342014-12-23 22:56:39 +00004783 }
4784 } else if (flavor == MachO::x86_FLOAT_STATE) {
4785 outs() << " flavor x86_FLOAT_STATE\n";
4786 if (count == MachO::x86_FLOAT_STATE_COUNT)
4787 outs() << " count x86_FLOAT_STATE_COUNT\n";
4788 else
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004789 outs() << " count " << count << " (not x86_FLOAT_STATE_COUNT)\n";
Kevin Enderby48ef5342014-12-23 22:56:39 +00004790 struct MachO::x86_float_state_t fs;
4791 left = end - begin;
4792 if (left >= sizeof(MachO::x86_float_state_t)) {
4793 memcpy(&fs, begin, sizeof(MachO::x86_float_state_t));
4794 begin += sizeof(MachO::x86_float_state_t);
4795 } else {
4796 memset(&fs, '\0', sizeof(MachO::x86_float_state_t));
4797 memcpy(&fs, begin, left);
4798 begin += left;
4799 }
4800 if (isLittleEndian != sys::IsLittleEndianHost)
4801 swapStruct(fs);
4802 if (fs.fsh.flavor == MachO::x86_FLOAT_STATE64) {
4803 outs() << "\t fsh.flavor x86_FLOAT_STATE64 ";
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004804 if (fs.fsh.count == MachO::x86_FLOAT_STATE64_COUNT)
Kevin Enderby48ef5342014-12-23 22:56:39 +00004805 outs() << "fsh.count x86_FLOAT_STATE64_COUNT\n";
4806 else
4807 outs() << "fsh.count " << fs.fsh.count
4808 << " (not x86_FLOAT_STATE64_COUNT\n";
4809 Print_x86_float_state_t(fs.ufs.fs64);
4810 } else {
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004811 outs() << "\t fsh.flavor " << fs.fsh.flavor << " fsh.count "
4812 << fs.fsh.count << "\n";
Kevin Enderby48ef5342014-12-23 22:56:39 +00004813 }
4814 } else if (flavor == MachO::x86_EXCEPTION_STATE) {
4815 outs() << " flavor x86_EXCEPTION_STATE\n";
4816 if (count == MachO::x86_EXCEPTION_STATE_COUNT)
4817 outs() << " count x86_EXCEPTION_STATE_COUNT\n";
4818 else
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004819 outs() << " count " << count
Kevin Enderby48ef5342014-12-23 22:56:39 +00004820 << " (not x86_EXCEPTION_STATE_COUNT)\n";
4821 struct MachO::x86_exception_state_t es;
4822 left = end - begin;
4823 if (left >= sizeof(MachO::x86_exception_state_t)) {
4824 memcpy(&es, begin, sizeof(MachO::x86_exception_state_t));
4825 begin += sizeof(MachO::x86_exception_state_t);
4826 } else {
4827 memset(&es, '\0', sizeof(MachO::x86_exception_state_t));
4828 memcpy(&es, begin, left);
4829 begin += left;
4830 }
4831 if (isLittleEndian != sys::IsLittleEndianHost)
4832 swapStruct(es);
4833 if (es.esh.flavor == MachO::x86_EXCEPTION_STATE64) {
4834 outs() << "\t esh.flavor x86_EXCEPTION_STATE64\n";
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004835 if (es.esh.count == MachO::x86_EXCEPTION_STATE64_COUNT)
Kevin Enderby48ef5342014-12-23 22:56:39 +00004836 outs() << "\t esh.count x86_EXCEPTION_STATE64_COUNT\n";
4837 else
4838 outs() << "\t esh.count " << es.esh.count
4839 << " (not x86_EXCEPTION_STATE64_COUNT\n";
4840 Print_x86_exception_state_t(es.ues.es64);
4841 } else {
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004842 outs() << "\t esh.flavor " << es.esh.flavor << " esh.count "
4843 << es.esh.count << "\n";
Kevin Enderby48ef5342014-12-23 22:56:39 +00004844 }
4845 } else {
4846 outs() << " flavor " << flavor << " (unknown)\n";
4847 outs() << " count " << count << "\n";
4848 outs() << " state (unknown)\n";
4849 begin += count * sizeof(uint32_t);
4850 }
4851 }
4852 } else {
4853 while (begin < end) {
4854 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4855 memcpy((char *)&flavor, begin, sizeof(uint32_t));
4856 begin += sizeof(uint32_t);
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004857 } else {
Kevin Enderby48ef5342014-12-23 22:56:39 +00004858 flavor = 0;
4859 begin = end;
4860 }
4861 if (isLittleEndian != sys::IsLittleEndianHost)
4862 sys::swapByteOrder(flavor);
4863 if (end - begin > (ptrdiff_t)sizeof(uint32_t)) {
4864 memcpy((char *)&count, begin, sizeof(uint32_t));
4865 begin += sizeof(uint32_t);
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004866 } else {
Kevin Enderby48ef5342014-12-23 22:56:39 +00004867 count = 0;
4868 begin = end;
4869 }
4870 if (isLittleEndian != sys::IsLittleEndianHost)
4871 sys::swapByteOrder(count);
4872 outs() << " flavor " << flavor << "\n";
4873 outs() << " count " << count << "\n";
4874 outs() << " state (Unknown cputype/cpusubtype)\n";
4875 begin += count * sizeof(uint32_t);
4876 }
4877 }
4878}
Kevin Enderby66d51fc2015-01-08 00:25:24 +00004879
Kevin Enderby8ae63c12014-09-04 16:54:47 +00004880static void PrintDylibCommand(MachO::dylib_command dl, const char *Ptr) {
4881 if (dl.cmd == MachO::LC_ID_DYLIB)
4882 outs() << " cmd LC_ID_DYLIB\n";
4883 else if (dl.cmd == MachO::LC_LOAD_DYLIB)
4884 outs() << " cmd LC_LOAD_DYLIB\n";
4885 else if (dl.cmd == MachO::LC_LOAD_WEAK_DYLIB)
4886 outs() << " cmd LC_LOAD_WEAK_DYLIB\n";
4887 else if (dl.cmd == MachO::LC_REEXPORT_DYLIB)
4888 outs() << " cmd LC_REEXPORT_DYLIB\n";
4889 else if (dl.cmd == MachO::LC_LAZY_LOAD_DYLIB)
4890 outs() << " cmd LC_LAZY_LOAD_DYLIB\n";
4891 else if (dl.cmd == MachO::LC_LOAD_UPWARD_DYLIB)
4892 outs() << " cmd LC_LOAD_UPWARD_DYLIB\n";
4893 else
4894 outs() << " cmd " << dl.cmd << " (unknown)\n";
4895 outs() << " cmdsize " << dl.cmdsize;
4896 if (dl.cmdsize < sizeof(struct MachO::dylib_command))
4897 outs() << " Incorrect size\n";
4898 else
4899 outs() << "\n";
4900 if (dl.dylib.name < dl.cmdsize) {
Kevin Enderbyb28ed012014-10-29 21:28:24 +00004901 const char *P = (const char *)(Ptr) + dl.dylib.name;
Kevin Enderby8ae63c12014-09-04 16:54:47 +00004902 outs() << " name " << P << " (offset " << dl.dylib.name << ")\n";
4903 } else {
4904 outs() << " name ?(bad offset " << dl.dylib.name << ")\n";
4905 }
4906 outs() << " time stamp " << dl.dylib.timestamp << " ";
4907 time_t t = dl.dylib.timestamp;
4908 outs() << ctime(&t);
4909 outs() << " current version ";
4910 if (dl.dylib.current_version == 0xffffffff)
4911 outs() << "n/a\n";
4912 else
4913 outs() << ((dl.dylib.current_version >> 16) & 0xffff) << "."
4914 << ((dl.dylib.current_version >> 8) & 0xff) << "."
4915 << (dl.dylib.current_version & 0xff) << "\n";
4916 outs() << "compatibility version ";
4917 if (dl.dylib.compatibility_version == 0xffffffff)
4918 outs() << "n/a\n";
4919 else
4920 outs() << ((dl.dylib.compatibility_version >> 16) & 0xffff) << "."
4921 << ((dl.dylib.compatibility_version >> 8) & 0xff) << "."
4922 << (dl.dylib.compatibility_version & 0xff) << "\n";
4923}
4924
4925static void PrintLinkEditDataCommand(MachO::linkedit_data_command ld,
4926 uint32_t object_size) {
4927 if (ld.cmd == MachO::LC_CODE_SIGNATURE)
4928 outs() << " cmd LC_FUNCTION_STARTS\n";
4929 else if (ld.cmd == MachO::LC_SEGMENT_SPLIT_INFO)
4930 outs() << " cmd LC_SEGMENT_SPLIT_INFO\n";
4931 else if (ld.cmd == MachO::LC_FUNCTION_STARTS)
4932 outs() << " cmd LC_FUNCTION_STARTS\n";
4933 else if (ld.cmd == MachO::LC_DATA_IN_CODE)
4934 outs() << " cmd LC_DATA_IN_CODE\n";
4935 else if (ld.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS)
4936 outs() << " cmd LC_DYLIB_CODE_SIGN_DRS\n";
4937 else if (ld.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT)
4938 outs() << " cmd LC_LINKER_OPTIMIZATION_HINT\n";
4939 else
4940 outs() << " cmd " << ld.cmd << " (?)\n";
4941 outs() << " cmdsize " << ld.cmdsize;
4942 if (ld.cmdsize != sizeof(struct MachO::linkedit_data_command))
4943 outs() << " Incorrect size\n";
4944 else
4945 outs() << "\n";
4946 outs() << " dataoff " << ld.dataoff;
4947 if (ld.dataoff > object_size)
4948 outs() << " (past end of file)\n";
4949 else
4950 outs() << "\n";
4951 outs() << " datasize " << ld.datasize;
4952 uint64_t big_size = ld.dataoff;
4953 big_size += ld.datasize;
4954 if (big_size > object_size)
4955 outs() << " (past end of file)\n";
4956 else
4957 outs() << "\n";
4958}
4959
Kevin Enderby956366c2014-08-29 22:30:52 +00004960static void PrintLoadCommands(const MachOObjectFile *Obj, uint32_t ncmds,
4961 uint32_t filetype, uint32_t cputype,
4962 bool verbose) {
Filipe Cabecinhase71bd0c2015-01-06 17:08:26 +00004963 if (ncmds == 0)
4964 return;
Kevin Enderby956366c2014-08-29 22:30:52 +00004965 StringRef Buf = Obj->getData();
4966 MachOObjectFile::LoadCommandInfo Command = Obj->getFirstLoadCommandInfo();
4967 for (unsigned i = 0;; ++i) {
4968 outs() << "Load command " << i << "\n";
4969 if (Command.C.cmd == MachO::LC_SEGMENT) {
4970 MachO::segment_command SLC = Obj->getSegmentLoadCommand(Command);
4971 const char *sg_segname = SLC.segname;
4972 PrintSegmentCommand(SLC.cmd, SLC.cmdsize, SLC.segname, SLC.vmaddr,
4973 SLC.vmsize, SLC.fileoff, SLC.filesize, SLC.maxprot,
4974 SLC.initprot, SLC.nsects, SLC.flags, Buf.size(),
4975 verbose);
4976 for (unsigned j = 0; j < SLC.nsects; j++) {
Kevin Enderbyc9713382014-12-16 01:14:45 +00004977 MachO::section S = Obj->getSection(Command, j);
Kevin Enderby956366c2014-08-29 22:30:52 +00004978 PrintSection(S.sectname, S.segname, S.addr, S.size, S.offset, S.align,
4979 S.reloff, S.nreloc, S.flags, S.reserved1, S.reserved2,
4980 SLC.cmd, sg_segname, filetype, Buf.size(), verbose);
4981 }
4982 } else if (Command.C.cmd == MachO::LC_SEGMENT_64) {
4983 MachO::segment_command_64 SLC_64 = Obj->getSegment64LoadCommand(Command);
4984 const char *sg_segname = SLC_64.segname;
4985 PrintSegmentCommand(SLC_64.cmd, SLC_64.cmdsize, SLC_64.segname,
4986 SLC_64.vmaddr, SLC_64.vmsize, SLC_64.fileoff,
4987 SLC_64.filesize, SLC_64.maxprot, SLC_64.initprot,
4988 SLC_64.nsects, SLC_64.flags, Buf.size(), verbose);
4989 for (unsigned j = 0; j < SLC_64.nsects; j++) {
4990 MachO::section_64 S_64 = Obj->getSection64(Command, j);
4991 PrintSection(S_64.sectname, S_64.segname, S_64.addr, S_64.size,
4992 S_64.offset, S_64.align, S_64.reloff, S_64.nreloc,
4993 S_64.flags, S_64.reserved1, S_64.reserved2, SLC_64.cmd,
4994 sg_segname, filetype, Buf.size(), verbose);
4995 }
4996 } else if (Command.C.cmd == MachO::LC_SYMTAB) {
4997 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
David Majnemer73cc6ff2014-11-13 19:48:56 +00004998 PrintSymtabLoadCommand(Symtab, Obj->is64Bit(), Buf.size());
Kevin Enderby956366c2014-08-29 22:30:52 +00004999 } else if (Command.C.cmd == MachO::LC_DYSYMTAB) {
5000 MachO::dysymtab_command Dysymtab = Obj->getDysymtabLoadCommand();
5001 MachO::symtab_command Symtab = Obj->getSymtabLoadCommand();
David Majnemer73cc6ff2014-11-13 19:48:56 +00005002 PrintDysymtabLoadCommand(Dysymtab, Symtab.nsyms, Buf.size(),
5003 Obj->is64Bit());
Kevin Enderby8ae63c12014-09-04 16:54:47 +00005004 } else if (Command.C.cmd == MachO::LC_DYLD_INFO ||
5005 Command.C.cmd == MachO::LC_DYLD_INFO_ONLY) {
5006 MachO::dyld_info_command DyldInfo = Obj->getDyldInfoLoadCommand(Command);
5007 PrintDyldInfoLoadCommand(DyldInfo, Buf.size());
5008 } else if (Command.C.cmd == MachO::LC_LOAD_DYLINKER ||
5009 Command.C.cmd == MachO::LC_ID_DYLINKER ||
5010 Command.C.cmd == MachO::LC_DYLD_ENVIRONMENT) {
5011 MachO::dylinker_command Dyld = Obj->getDylinkerCommand(Command);
5012 PrintDyldLoadCommand(Dyld, Command.Ptr);
5013 } else if (Command.C.cmd == MachO::LC_UUID) {
5014 MachO::uuid_command Uuid = Obj->getUuidCommand(Command);
5015 PrintUuidLoadCommand(Uuid);
Jean-Daniel Dupas00cc1f52014-12-04 07:37:02 +00005016 } else if (Command.C.cmd == MachO::LC_RPATH) {
5017 MachO::rpath_command Rpath = Obj->getRpathCommand(Command);
5018 PrintRpathLoadCommand(Rpath, Command.Ptr);
Kevin Enderby1ff0ecc2014-12-16 21:48:27 +00005019 } else if (Command.C.cmd == MachO::LC_VERSION_MIN_MACOSX ||
5020 Command.C.cmd == MachO::LC_VERSION_MIN_IPHONEOS) {
Kevin Enderby8ae63c12014-09-04 16:54:47 +00005021 MachO::version_min_command Vd = Obj->getVersionMinLoadCommand(Command);
5022 PrintVersionMinLoadCommand(Vd);
5023 } else if (Command.C.cmd == MachO::LC_SOURCE_VERSION) {
5024 MachO::source_version_command Sd = Obj->getSourceVersionCommand(Command);
5025 PrintSourceVersionCommand(Sd);
5026 } else if (Command.C.cmd == MachO::LC_MAIN) {
5027 MachO::entry_point_command Ep = Obj->getEntryPointCommand(Command);
5028 PrintEntryPointCommand(Ep);
Kevin Enderby0804f4672014-12-16 23:25:52 +00005029 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO) {
Kevin Enderby66d51fc2015-01-08 00:25:24 +00005030 MachO::encryption_info_command Ei =
5031 Obj->getEncryptionInfoCommand(Command);
Kevin Enderby0804f4672014-12-16 23:25:52 +00005032 PrintEncryptionInfoCommand(Ei, Buf.size());
Kevin Enderby57538292014-12-17 01:01:30 +00005033 } else if (Command.C.cmd == MachO::LC_ENCRYPTION_INFO_64) {
Kevin Enderby66d51fc2015-01-08 00:25:24 +00005034 MachO::encryption_info_command_64 Ei =
5035 Obj->getEncryptionInfoCommand64(Command);
Kevin Enderby57538292014-12-17 01:01:30 +00005036 PrintEncryptionInfoCommand64(Ei, Buf.size());
Kevin Enderbyd0b6b7f2014-12-18 00:53:40 +00005037 } else if (Command.C.cmd == MachO::LC_LINKER_OPTION) {
Kevin Enderby66d51fc2015-01-08 00:25:24 +00005038 MachO::linker_option_command Lo =
5039 Obj->getLinkerOptionLoadCommand(Command);
Kevin Enderbyd0b6b7f2014-12-18 00:53:40 +00005040 PrintLinkerOptionCommand(Lo, Command.Ptr);
Kevin Enderbyb4b79312014-12-18 19:24:35 +00005041 } else if (Command.C.cmd == MachO::LC_SUB_FRAMEWORK) {
5042 MachO::sub_framework_command Sf = Obj->getSubFrameworkCommand(Command);
5043 PrintSubFrameworkCommand(Sf, Command.Ptr);
Kevin Enderbya2bd8d92014-12-18 23:13:26 +00005044 } else if (Command.C.cmd == MachO::LC_SUB_UMBRELLA) {
5045 MachO::sub_umbrella_command Sf = Obj->getSubUmbrellaCommand(Command);
5046 PrintSubUmbrellaCommand(Sf, Command.Ptr);
Kevin Enderby36c8d3a2014-12-19 19:48:16 +00005047 } else if (Command.C.cmd == MachO::LC_SUB_LIBRARY) {
5048 MachO::sub_library_command Sl = Obj->getSubLibraryCommand(Command);
5049 PrintSubLibraryCommand(Sl, Command.Ptr);
Kevin Enderby186eac32014-12-19 21:06:24 +00005050 } else if (Command.C.cmd == MachO::LC_SUB_CLIENT) {
5051 MachO::sub_client_command Sc = Obj->getSubClientCommand(Command);
5052 PrintSubClientCommand(Sc, Command.Ptr);
Kevin Enderby52e4ce42014-12-19 22:25:22 +00005053 } else if (Command.C.cmd == MachO::LC_ROUTINES) {
5054 MachO::routines_command Rc = Obj->getRoutinesCommand(Command);
5055 PrintRoutinesCommand(Rc);
5056 } else if (Command.C.cmd == MachO::LC_ROUTINES_64) {
5057 MachO::routines_command_64 Rc = Obj->getRoutinesCommand64(Command);
5058 PrintRoutinesCommand64(Rc);
Kevin Enderby48ef5342014-12-23 22:56:39 +00005059 } else if (Command.C.cmd == MachO::LC_THREAD ||
5060 Command.C.cmd == MachO::LC_UNIXTHREAD) {
5061 MachO::thread_command Tc = Obj->getThreadCommand(Command);
5062 PrintThreadCommand(Tc, Command.Ptr, Obj->isLittleEndian(), cputype);
Nick Kledzik15558912014-10-16 18:58:20 +00005063 } else if (Command.C.cmd == MachO::LC_LOAD_DYLIB ||
5064 Command.C.cmd == MachO::LC_ID_DYLIB ||
5065 Command.C.cmd == MachO::LC_LOAD_WEAK_DYLIB ||
5066 Command.C.cmd == MachO::LC_REEXPORT_DYLIB ||
5067 Command.C.cmd == MachO::LC_LAZY_LOAD_DYLIB ||
5068 Command.C.cmd == MachO::LC_LOAD_UPWARD_DYLIB) {
Kevin Enderby8ae63c12014-09-04 16:54:47 +00005069 MachO::dylib_command Dl = Obj->getDylibIDLoadCommand(Command);
5070 PrintDylibCommand(Dl, Command.Ptr);
5071 } else if (Command.C.cmd == MachO::LC_CODE_SIGNATURE ||
5072 Command.C.cmd == MachO::LC_SEGMENT_SPLIT_INFO ||
5073 Command.C.cmd == MachO::LC_FUNCTION_STARTS ||
5074 Command.C.cmd == MachO::LC_DATA_IN_CODE ||
5075 Command.C.cmd == MachO::LC_DYLIB_CODE_SIGN_DRS ||
5076 Command.C.cmd == MachO::LC_LINKER_OPTIMIZATION_HINT) {
5077 MachO::linkedit_data_command Ld =
5078 Obj->getLinkeditDataLoadCommand(Command);
5079 PrintLinkEditDataCommand(Ld, Buf.size());
Kevin Enderby956366c2014-08-29 22:30:52 +00005080 } else {
5081 outs() << " cmd ?(" << format("0x%08" PRIx32, Command.C.cmd)
5082 << ")\n";
5083 outs() << " cmdsize " << Command.C.cmdsize << "\n";
5084 // TODO: get and print the raw bytes of the load command.
5085 }
5086 // TODO: print all the other kinds of load commands.
5087 if (i == ncmds - 1)
5088 break;
5089 else
5090 Command = Obj->getNextLoadCommandInfo(Command);
5091 }
5092}
5093
5094static void getAndPrintMachHeader(const MachOObjectFile *Obj, uint32_t &ncmds,
5095 uint32_t &filetype, uint32_t &cputype,
5096 bool verbose) {
Kevin Enderbyb76d3862014-08-22 20:35:18 +00005097 if (Obj->is64Bit()) {
5098 MachO::mach_header_64 H_64;
5099 H_64 = Obj->getHeader64();
5100 PrintMachHeader(H_64.magic, H_64.cputype, H_64.cpusubtype, H_64.filetype,
5101 H_64.ncmds, H_64.sizeofcmds, H_64.flags, verbose);
Kevin Enderby956366c2014-08-29 22:30:52 +00005102 ncmds = H_64.ncmds;
5103 filetype = H_64.filetype;
5104 cputype = H_64.cputype;
Kevin Enderbyb76d3862014-08-22 20:35:18 +00005105 } else {
5106 MachO::mach_header H;
5107 H = Obj->getHeader();
5108 PrintMachHeader(H.magic, H.cputype, H.cpusubtype, H.filetype, H.ncmds,
5109 H.sizeofcmds, H.flags, verbose);
Kevin Enderby956366c2014-08-29 22:30:52 +00005110 ncmds = H.ncmds;
5111 filetype = H.filetype;
5112 cputype = H.cputype;
Kevin Enderbyb76d3862014-08-22 20:35:18 +00005113 }
5114}
5115
5116void llvm::printMachOFileHeader(const object::ObjectFile *Obj) {
5117 const MachOObjectFile *file = dyn_cast<const MachOObjectFile>(Obj);
Kevin Enderby956366c2014-08-29 22:30:52 +00005118 uint32_t ncmds = 0;
5119 uint32_t filetype = 0;
5120 uint32_t cputype = 0;
5121 getAndPrintMachHeader(file, ncmds, filetype, cputype, true);
5122 PrintLoadCommands(file, ncmds, filetype, cputype, true);
Kevin Enderbyb76d3862014-08-22 20:35:18 +00005123}
Nick Kledzikd04bc352014-08-30 00:20:14 +00005124
5125//===----------------------------------------------------------------------===//
5126// export trie dumping
5127//===----------------------------------------------------------------------===//
5128
5129void llvm::printMachOExportsTrie(const object::MachOObjectFile *Obj) {
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005130 for (const llvm::object::ExportEntry &Entry : Obj->exports()) {
5131 uint64_t Flags = Entry.flags();
Nick Kledzikd04bc352014-08-30 00:20:14 +00005132 bool ReExport = (Flags & MachO::EXPORT_SYMBOL_FLAGS_REEXPORT);
5133 bool WeakDef = (Flags & MachO::EXPORT_SYMBOL_FLAGS_WEAK_DEFINITION);
5134 bool ThreadLocal = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5135 MachO::EXPORT_SYMBOL_FLAGS_KIND_THREAD_LOCAL);
5136 bool Abs = ((Flags & MachO::EXPORT_SYMBOL_FLAGS_KIND_MASK) ==
5137 MachO::EXPORT_SYMBOL_FLAGS_KIND_ABSOLUTE);
5138 bool Resolver = (Flags & MachO::EXPORT_SYMBOL_FLAGS_STUB_AND_RESOLVER);
5139 if (ReExport)
5140 outs() << "[re-export] ";
5141 else
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005142 outs() << format("0x%08llX ",
5143 Entry.address()); // FIXME:add in base address
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005144 outs() << Entry.name();
Nick Kledzikd04bc352014-08-30 00:20:14 +00005145 if (WeakDef || ThreadLocal || Resolver || Abs) {
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005146 bool NeedsComma = false;
Nick Kledzik1d1ac4b2014-09-03 01:12:52 +00005147 outs() << " [";
Nick Kledzikd04bc352014-08-30 00:20:14 +00005148 if (WeakDef) {
5149 outs() << "weak_def";
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005150 NeedsComma = true;
Nick Kledzikd04bc352014-08-30 00:20:14 +00005151 }
5152 if (ThreadLocal) {
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005153 if (NeedsComma)
Nick Kledzikd04bc352014-08-30 00:20:14 +00005154 outs() << ", ";
5155 outs() << "per-thread";
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005156 NeedsComma = true;
Nick Kledzikd04bc352014-08-30 00:20:14 +00005157 }
5158 if (Abs) {
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005159 if (NeedsComma)
Nick Kledzikd04bc352014-08-30 00:20:14 +00005160 outs() << ", ";
5161 outs() << "absolute";
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005162 NeedsComma = true;
Nick Kledzikd04bc352014-08-30 00:20:14 +00005163 }
5164 if (Resolver) {
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005165 if (NeedsComma)
Nick Kledzikd04bc352014-08-30 00:20:14 +00005166 outs() << ", ";
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005167 outs() << format("resolver=0x%08llX", Entry.other());
5168 NeedsComma = true;
Nick Kledzikd04bc352014-08-30 00:20:14 +00005169 }
5170 outs() << "]";
5171 }
5172 if (ReExport) {
5173 StringRef DylibName = "unknown";
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005174 int Ordinal = Entry.other() - 1;
5175 Obj->getLibraryShortNameByIndex(Ordinal, DylibName);
5176 if (Entry.otherName().empty())
Nick Kledzikd04bc352014-08-30 00:20:14 +00005177 outs() << " (from " << DylibName << ")";
5178 else
Nick Kledzikac7cbdc2014-09-02 18:50:24 +00005179 outs() << " (" << Entry.otherName() << " from " << DylibName << ")";
Nick Kledzikd04bc352014-08-30 00:20:14 +00005180 }
5181 outs() << "\n";
5182 }
5183}
Nick Kledzikac431442014-09-12 21:34:15 +00005184
Nick Kledzikac431442014-09-12 21:34:15 +00005185//===----------------------------------------------------------------------===//
5186// rebase table dumping
5187//===----------------------------------------------------------------------===//
5188
5189namespace {
5190class SegInfo {
5191public:
5192 SegInfo(const object::MachOObjectFile *Obj);
5193
5194 StringRef segmentName(uint32_t SegIndex);
5195 StringRef sectionName(uint32_t SegIndex, uint64_t SegOffset);
5196 uint64_t address(uint32_t SegIndex, uint64_t SegOffset);
5197
5198private:
5199 struct SectionInfo {
5200 uint64_t Address;
5201 uint64_t Size;
5202 StringRef SectionName;
5203 StringRef SegmentName;
5204 uint64_t OffsetInSegment;
5205 uint64_t SegmentStartAddress;
5206 uint32_t SegmentIndex;
5207 };
5208 const SectionInfo &findSection(uint32_t SegIndex, uint64_t SegOffset);
5209 SmallVector<SectionInfo, 32> Sections;
5210};
5211}
5212
5213SegInfo::SegInfo(const object::MachOObjectFile *Obj) {
5214 // Build table of sections so segIndex/offset pairs can be translated.
Nick Kledzik56ebef42014-09-16 01:41:51 +00005215 uint32_t CurSegIndex = Obj->hasPageZeroSegment() ? 1 : 0;
Nick Kledzikac431442014-09-12 21:34:15 +00005216 StringRef CurSegName;
5217 uint64_t CurSegAddress;
5218 for (const SectionRef &Section : Obj->sections()) {
5219 SectionInfo Info;
5220 if (error(Section.getName(Info.SectionName)))
5221 return;
Rafael Espindola80291272014-10-08 15:28:58 +00005222 Info.Address = Section.getAddress();
5223 Info.Size = Section.getSize();
Nick Kledzikac431442014-09-12 21:34:15 +00005224 Info.SegmentName =
5225 Obj->getSectionFinalSegmentName(Section.getRawDataRefImpl());
5226 if (!Info.SegmentName.equals(CurSegName)) {
5227 ++CurSegIndex;
5228 CurSegName = Info.SegmentName;
5229 CurSegAddress = Info.Address;
5230 }
5231 Info.SegmentIndex = CurSegIndex - 1;
5232 Info.OffsetInSegment = Info.Address - CurSegAddress;
5233 Info.SegmentStartAddress = CurSegAddress;
5234 Sections.push_back(Info);
5235 }
5236}
5237
5238StringRef SegInfo::segmentName(uint32_t SegIndex) {
5239 for (const SectionInfo &SI : Sections) {
5240 if (SI.SegmentIndex == SegIndex)
5241 return SI.SegmentName;
5242 }
5243 llvm_unreachable("invalid segIndex");
5244}
5245
5246const SegInfo::SectionInfo &SegInfo::findSection(uint32_t SegIndex,
5247 uint64_t OffsetInSeg) {
5248 for (const SectionInfo &SI : Sections) {
5249 if (SI.SegmentIndex != SegIndex)
5250 continue;
5251 if (SI.OffsetInSegment > OffsetInSeg)
5252 continue;
5253 if (OffsetInSeg >= (SI.OffsetInSegment + SI.Size))
5254 continue;
5255 return SI;
5256 }
5257 llvm_unreachable("segIndex and offset not in any section");
5258}
5259
5260StringRef SegInfo::sectionName(uint32_t SegIndex, uint64_t OffsetInSeg) {
5261 return findSection(SegIndex, OffsetInSeg).SectionName;
5262}
5263
5264uint64_t SegInfo::address(uint32_t SegIndex, uint64_t OffsetInSeg) {
5265 const SectionInfo &SI = findSection(SegIndex, OffsetInSeg);
5266 return SI.SegmentStartAddress + OffsetInSeg;
5267}
5268
5269void llvm::printMachORebaseTable(const object::MachOObjectFile *Obj) {
5270 // Build table of sections so names can used in final output.
5271 SegInfo sectionTable(Obj);
5272
5273 outs() << "segment section address type\n";
5274 for (const llvm::object::MachORebaseEntry &Entry : Obj->rebaseTable()) {
5275 uint32_t SegIndex = Entry.segmentIndex();
5276 uint64_t OffsetInSeg = Entry.segmentOffset();
5277 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5278 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5279 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5280
5281 // Table lines look like: __DATA __nl_symbol_ptr 0x0000F00C pointer
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005282 outs() << format("%-8s %-18s 0x%08" PRIX64 " %s\n",
5283 SegmentName.str().c_str(), SectionName.str().c_str(),
5284 Address, Entry.typeName().str().c_str());
Nick Kledzikac431442014-09-12 21:34:15 +00005285 }
5286}
Nick Kledzik56ebef42014-09-16 01:41:51 +00005287
5288static StringRef ordinalName(const object::MachOObjectFile *Obj, int Ordinal) {
5289 StringRef DylibName;
5290 switch (Ordinal) {
5291 case MachO::BIND_SPECIAL_DYLIB_SELF:
5292 return "this-image";
5293 case MachO::BIND_SPECIAL_DYLIB_MAIN_EXECUTABLE:
5294 return "main-executable";
5295 case MachO::BIND_SPECIAL_DYLIB_FLAT_LOOKUP:
5296 return "flat-namespace";
5297 default:
Nick Kledzikabd29872014-09-16 22:03:13 +00005298 if (Ordinal > 0) {
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005299 std::error_code EC =
5300 Obj->getLibraryShortNameByIndex(Ordinal - 1, DylibName);
Nick Kledzikabd29872014-09-16 22:03:13 +00005301 if (EC)
Nick Kledzik51d2c2b2014-10-14 23:29:38 +00005302 return "<<bad library ordinal>>";
Nick Kledzikabd29872014-09-16 22:03:13 +00005303 return DylibName;
5304 }
Nick Kledzik56ebef42014-09-16 01:41:51 +00005305 }
Nick Kledzikabd29872014-09-16 22:03:13 +00005306 return "<<unknown special ordinal>>";
Nick Kledzik56ebef42014-09-16 01:41:51 +00005307}
5308
5309//===----------------------------------------------------------------------===//
5310// bind table dumping
5311//===----------------------------------------------------------------------===//
5312
5313void llvm::printMachOBindTable(const object::MachOObjectFile *Obj) {
5314 // Build table of sections so names can used in final output.
5315 SegInfo sectionTable(Obj);
5316
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005317 outs() << "segment section address type "
5318 "addend dylib symbol\n";
Nick Kledzik56ebef42014-09-16 01:41:51 +00005319 for (const llvm::object::MachOBindEntry &Entry : Obj->bindTable()) {
5320 uint32_t SegIndex = Entry.segmentIndex();
5321 uint64_t OffsetInSeg = Entry.segmentOffset();
5322 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5323 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5324 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5325
5326 // Table lines look like:
5327 // __DATA __got 0x00012010 pointer 0 libSystem ___stack_chk_guard
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005328 StringRef Attr;
Nick Kledzik56ebef42014-09-16 01:41:51 +00005329 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_WEAK_IMPORT)
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005330 Attr = " (weak_import)";
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005331 outs() << left_justify(SegmentName, 8) << " "
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005332 << left_justify(SectionName, 18) << " "
5333 << format_hex(Address, 10, true) << " "
5334 << left_justify(Entry.typeName(), 8) << " "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005335 << format_decimal(Entry.addend(), 8) << " "
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005336 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005337 << Entry.symbolName() << Attr << "\n";
Nick Kledzik56ebef42014-09-16 01:41:51 +00005338 }
5339}
5340
5341//===----------------------------------------------------------------------===//
5342// lazy bind table dumping
5343//===----------------------------------------------------------------------===//
5344
5345void llvm::printMachOLazyBindTable(const object::MachOObjectFile *Obj) {
5346 // Build table of sections so names can used in final output.
5347 SegInfo sectionTable(Obj);
5348
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005349 outs() << "segment section address "
5350 "dylib symbol\n";
Nick Kledzik56ebef42014-09-16 01:41:51 +00005351 for (const llvm::object::MachOBindEntry &Entry : Obj->lazyBindTable()) {
5352 uint32_t SegIndex = Entry.segmentIndex();
5353 uint64_t OffsetInSeg = Entry.segmentOffset();
5354 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5355 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5356 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5357
5358 // Table lines look like:
5359 // __DATA __got 0x00012010 libSystem ___stack_chk_guard
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005360 outs() << left_justify(SegmentName, 8) << " "
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005361 << left_justify(SectionName, 18) << " "
5362 << format_hex(Address, 10, true) << " "
5363 << left_justify(ordinalName(Obj, Entry.ordinal()), 16) << " "
Nick Kledzik56ebef42014-09-16 01:41:51 +00005364 << Entry.symbolName() << "\n";
5365 }
5366}
5367
Nick Kledzik56ebef42014-09-16 01:41:51 +00005368//===----------------------------------------------------------------------===//
5369// weak bind table dumping
5370//===----------------------------------------------------------------------===//
5371
5372void llvm::printMachOWeakBindTable(const object::MachOObjectFile *Obj) {
5373 // Build table of sections so names can used in final output.
5374 SegInfo sectionTable(Obj);
5375
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005376 outs() << "segment section address "
5377 "type addend symbol\n";
Nick Kledzik56ebef42014-09-16 01:41:51 +00005378 for (const llvm::object::MachOBindEntry &Entry : Obj->weakBindTable()) {
5379 // Strong symbols don't have a location to update.
5380 if (Entry.flags() & MachO::BIND_SYMBOL_FLAGS_NON_WEAK_DEFINITION) {
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005381 outs() << " strong "
Nick Kledzik56ebef42014-09-16 01:41:51 +00005382 << Entry.symbolName() << "\n";
5383 continue;
5384 }
5385 uint32_t SegIndex = Entry.segmentIndex();
5386 uint64_t OffsetInSeg = Entry.segmentOffset();
5387 StringRef SegmentName = sectionTable.segmentName(SegIndex);
5388 StringRef SectionName = sectionTable.sectionName(SegIndex, OffsetInSeg);
5389 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5390
5391 // Table lines look like:
5392 // __DATA __data 0x00001000 pointer 0 _foo
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005393 outs() << left_justify(SegmentName, 8) << " "
Nick Kledzik5ffacc12014-09-30 00:19:58 +00005394 << left_justify(SectionName, 18) << " "
5395 << format_hex(Address, 10, true) << " "
5396 << left_justify(Entry.typeName(), 8) << " "
Kevin Enderbyb28ed012014-10-29 21:28:24 +00005397 << format_decimal(Entry.addend(), 8) << " " << Entry.symbolName()
5398 << "\n";
Nick Kledzik56ebef42014-09-16 01:41:51 +00005399 }
5400}
5401
Kevin Enderby6f326ce2014-10-23 19:37:31 +00005402// get_dyld_bind_info_symbolname() is used for disassembly and passed an
5403// address, ReferenceValue, in the Mach-O file and looks in the dyld bind
5404// information for that address. If the address is found its binding symbol
5405// name is returned. If not nullptr is returned.
5406static const char *get_dyld_bind_info_symbolname(uint64_t ReferenceValue,
5407 struct DisassembleInfo *info) {
Kevin Enderby078be602014-10-23 19:53:12 +00005408 if (info->bindtable == nullptr) {
5409 info->bindtable = new (BindTable);
Kevin Enderby6f326ce2014-10-23 19:37:31 +00005410 SegInfo sectionTable(info->O);
5411 for (const llvm::object::MachOBindEntry &Entry : info->O->bindTable()) {
5412 uint32_t SegIndex = Entry.segmentIndex();
5413 uint64_t OffsetInSeg = Entry.segmentOffset();
5414 uint64_t Address = sectionTable.address(SegIndex, OffsetInSeg);
5415 const char *SymbolName = nullptr;
5416 StringRef name = Entry.symbolName();
5417 if (!name.empty())
5418 SymbolName = name.data();
Kevin Enderby078be602014-10-23 19:53:12 +00005419 info->bindtable->push_back(std::make_pair(Address, SymbolName));
Kevin Enderby6f326ce2014-10-23 19:37:31 +00005420 }
5421 }
Kevin Enderby078be602014-10-23 19:53:12 +00005422 for (bind_table_iterator BI = info->bindtable->begin(),
5423 BE = info->bindtable->end();
Kevin Enderby6f326ce2014-10-23 19:37:31 +00005424 BI != BE; ++BI) {
5425 uint64_t Address = BI->first;
5426 if (ReferenceValue == Address) {
5427 const char *SymbolName = BI->second;
5428 return SymbolName;
5429 }
5430 }
5431 return nullptr;
5432}