blob: 88c333a4705f13a6ac31b36677aa9cd81b162d94 [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"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000015#include "llvm/ADT/STLExtras.h"
Ahmed Bougachaaa790682013-05-24 01:07:04 +000016#include "llvm/ADT/StringExtras.h"
Chandler Carruth4d88a1c2012-12-04 10:44:52 +000017#include "llvm/ADT/Triple.h"
Benjamin Kramer699128e2011-09-21 01:13:19 +000018#include "llvm/DebugInfo/DIContext.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000019#include "llvm/MC/MCAsmInfo.h"
Lang Hamesa1bc0f52014-04-15 04:40:56 +000020#include "llvm/MC/MCContext.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000021#include "llvm/MC/MCDisassembler.h"
22#include "llvm/MC/MCInst.h"
23#include "llvm/MC/MCInstPrinter.h"
24#include "llvm/MC/MCInstrAnalysis.h"
25#include "llvm/MC/MCInstrDesc.h"
26#include "llvm/MC/MCInstrInfo.h"
Jim Grosbachfd93a592012-03-05 19:33:20 +000027#include "llvm/MC/MCRegisterInfo.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000028#include "llvm/MC/MCSubtargetInfo.h"
Chandler Carruth4d88a1c2012-12-04 10:44:52 +000029#include "llvm/Object/MachO.h"
Rafael Espindola9b709252013-04-13 01:45:40 +000030#include "llvm/Support/Casting.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000031#include "llvm/Support/CommandLine.h"
32#include "llvm/Support/Debug.h"
Tim Northover4bd286a2014-08-01 13:07:19 +000033#include "llvm/Support/Endian.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000034#include "llvm/Support/Format.h"
35#include "llvm/Support/GraphWriter.h"
Chandler Carruth4d88a1c2012-12-04 10:44:52 +000036#include "llvm/Support/MachO.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000037#include "llvm/Support/MemoryBuffer.h"
38#include "llvm/Support/TargetRegistry.h"
39#include "llvm/Support/TargetSelect.h"
40#include "llvm/Support/raw_ostream.h"
Benjamin Kramer43a772e2011-09-19 17:56:04 +000041#include <algorithm>
42#include <cstring>
Rafael Espindolaa6e9c3e2014-06-12 17:38:55 +000043#include <system_error>
Benjamin Kramer43a772e2011-09-19 17:56:04 +000044using namespace llvm;
45using namespace object;
46
47static cl::opt<bool>
Benjamin Kramer699128e2011-09-21 01:13:19 +000048 UseDbg("g", cl::desc("Print line information from debug info if available"));
49
50static cl::opt<std::string>
51 DSYMFile("dsym", cl::desc("Use .dSYM file for debug info"));
52
Rafael Espindola56f976f2013-04-18 18:08:55 +000053static const Target *GetTarget(const MachOObjectFile *MachOObj) {
Benjamin Kramer43a772e2011-09-19 17:56:04 +000054 // Figure out the target triple.
Cameron Zwarich88cc16a2012-02-03 06:35:22 +000055 if (TripleName.empty()) {
56 llvm::Triple TT("unknown-unknown-unknown");
Rafael Espindola93f4a622013-04-11 03:34:37 +000057 TT.setArch(Triple::ArchType(MachOObj->getArch()));
Cameron Zwarich88cc16a2012-02-03 06:35:22 +000058 TripleName = TT.str();
Benjamin Kramer43a772e2011-09-19 17:56:04 +000059 }
60
Benjamin Kramer43a772e2011-09-19 17:56:04 +000061 // Get the target specific parser.
62 std::string Error;
63 const Target *TheTarget = TargetRegistry::lookupTarget(TripleName, Error);
64 if (TheTarget)
65 return TheTarget;
66
67 errs() << "llvm-objdump: error: unable to get target for '" << TripleName
68 << "', see --version and --triple.\n";
Craig Toppere6cb63e2014-04-25 04:24:47 +000069 return nullptr;
Benjamin Kramer43a772e2011-09-19 17:56:04 +000070}
71
Owen Andersond9243c42011-10-17 21:37:35 +000072struct SymbolSorter {
73 bool operator()(const SymbolRef &A, const SymbolRef &B) {
74 SymbolRef::Type AType, BType;
75 A.getType(AType);
76 B.getType(BType);
77
78 uint64_t AAddr, BAddr;
79 if (AType != SymbolRef::ST_Function)
80 AAddr = 0;
81 else
82 A.getAddress(AAddr);
83 if (BType != SymbolRef::ST_Function)
84 BAddr = 0;
85 else
86 B.getAddress(BAddr);
87 return AAddr < BAddr;
88 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +000089};
90
Kevin Enderby273ae012013-06-06 17:20:50 +000091// Types for the storted data in code table that is built before disassembly
92// and the predicate function to sort them.
93typedef std::pair<uint64_t, DiceRef> DiceTableEntry;
94typedef std::vector<DiceTableEntry> DiceTable;
95typedef DiceTable::iterator dice_table_iterator;
96
97static bool
98compareDiceTableEntries(const DiceTableEntry i,
99 const DiceTableEntry j) {
100 return i.first == j.first;
101}
102
103static void DumpDataInCode(const char *bytes, uint64_t Size,
104 unsigned short Kind) {
105 uint64_t Value;
106
107 switch (Kind) {
Charles Davis8bdfafd2013-09-01 04:28:48 +0000108 case MachO::DICE_KIND_DATA:
Kevin Enderby273ae012013-06-06 17:20:50 +0000109 switch (Size) {
110 case 4:
111 Value = bytes[3] << 24 |
112 bytes[2] << 16 |
113 bytes[1] << 8 |
114 bytes[0];
115 outs() << "\t.long " << Value;
116 break;
117 case 2:
118 Value = bytes[1] << 8 |
119 bytes[0];
120 outs() << "\t.short " << Value;
121 break;
122 case 1:
123 Value = bytes[0];
124 outs() << "\t.byte " << Value;
125 break;
126 }
127 outs() << "\t@ KIND_DATA\n";
128 break;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000129 case MachO::DICE_KIND_JUMP_TABLE8:
Kevin Enderby273ae012013-06-06 17:20:50 +0000130 Value = bytes[0];
131 outs() << "\t.byte " << Value << "\t@ KIND_JUMP_TABLE8";
132 break;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000133 case MachO::DICE_KIND_JUMP_TABLE16:
Kevin Enderby273ae012013-06-06 17:20:50 +0000134 Value = bytes[1] << 8 |
135 bytes[0];
136 outs() << "\t.short " << Value << "\t@ KIND_JUMP_TABLE16";
137 break;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000138 case MachO::DICE_KIND_JUMP_TABLE32:
Kevin Enderby273ae012013-06-06 17:20:50 +0000139 Value = bytes[3] << 24 |
140 bytes[2] << 16 |
141 bytes[1] << 8 |
142 bytes[0];
143 outs() << "\t.long " << Value << "\t@ KIND_JUMP_TABLE32";
144 break;
145 default:
146 outs() << "\t@ data in code kind = " << Kind << "\n";
147 break;
148 }
149}
150
Alexey Samsonov464d2e42014-03-17 07:28:19 +0000151static void getSectionsAndSymbols(const MachO::mach_header Header,
152 MachOObjectFile *MachOObj,
153 std::vector<SectionRef> &Sections,
154 std::vector<SymbolRef> &Symbols,
155 SmallVectorImpl<uint64_t> &FoundFns,
156 uint64_t &BaseSegmentAddress) {
157 for (const SymbolRef &Symbol : MachOObj->symbols())
158 Symbols.push_back(Symbol);
Owen Andersond9243c42011-10-17 21:37:35 +0000159
Alexey Samsonov48803e52014-03-13 14:37:36 +0000160 for (const SectionRef &Section : MachOObj->sections()) {
Owen Andersond9243c42011-10-17 21:37:35 +0000161 StringRef SectName;
Alexey Samsonov48803e52014-03-13 14:37:36 +0000162 Section.getName(SectName);
163 Sections.push_back(Section);
Owen Andersond9243c42011-10-17 21:37:35 +0000164 }
165
Rafael Espindola56f976f2013-04-18 18:08:55 +0000166 MachOObjectFile::LoadCommandInfo Command =
Alexey Samsonov48803e52014-03-13 14:37:36 +0000167 MachOObj->getFirstLoadCommandInfo();
Kevin Enderby273ae012013-06-06 17:20:50 +0000168 bool BaseSegmentAddressSet = false;
Rafael Espindolafeef8c22013-04-19 11:36:47 +0000169 for (unsigned i = 0; ; ++i) {
Charles Davis8bdfafd2013-09-01 04:28:48 +0000170 if (Command.C.cmd == MachO::LC_FUNCTION_STARTS) {
Benjamin Kramer699128e2011-09-21 01:13:19 +0000171 // We found a function starts segment, parse the addresses for later
172 // consumption.
Charles Davis8bdfafd2013-09-01 04:28:48 +0000173 MachO::linkedit_data_command LLC =
Rafael Espindola56f976f2013-04-18 18:08:55 +0000174 MachOObj->getLinkeditDataLoadCommand(Command);
Benjamin Kramer699128e2011-09-21 01:13:19 +0000175
Charles Davis8bdfafd2013-09-01 04:28:48 +0000176 MachOObj->ReadULEB128s(LLC.dataoff, FoundFns);
Benjamin Kramer8a529dc2011-09-21 22:16:43 +0000177 }
Charles Davis8bdfafd2013-09-01 04:28:48 +0000178 else if (Command.C.cmd == MachO::LC_SEGMENT) {
179 MachO::segment_command SLC =
Kevin Enderby273ae012013-06-06 17:20:50 +0000180 MachOObj->getSegmentLoadCommand(Command);
Charles Davis8bdfafd2013-09-01 04:28:48 +0000181 StringRef SegName = SLC.segname;
Kevin Enderby273ae012013-06-06 17:20:50 +0000182 if(!BaseSegmentAddressSet && SegName != "__PAGEZERO") {
183 BaseSegmentAddressSet = true;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000184 BaseSegmentAddress = SLC.vmaddr;
Kevin Enderby273ae012013-06-06 17:20:50 +0000185 }
186 }
Rafael Espindolafeef8c22013-04-19 11:36:47 +0000187
Charles Davis8bdfafd2013-09-01 04:28:48 +0000188 if (i == Header.ncmds - 1)
Rafael Espindolafeef8c22013-04-19 11:36:47 +0000189 break;
190 else
191 Command = MachOObj->getNextLoadCommandInfo(Command);
Benjamin Kramer8a529dc2011-09-21 22:16:43 +0000192 }
Benjamin Kramer699128e2011-09-21 01:13:19 +0000193}
194
Rafael Espindola9b709252013-04-13 01:45:40 +0000195static void DisassembleInputMachO2(StringRef Filename,
Rafael Espindola56f976f2013-04-18 18:08:55 +0000196 MachOObjectFile *MachOOF);
Rafael Espindola9b709252013-04-13 01:45:40 +0000197
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000198void llvm::DisassembleInputMachO(StringRef Filename) {
Rafael Espindolaadf21f22014-07-06 17:43:13 +0000199 ErrorOr<std::unique_ptr<MemoryBuffer>> Buff =
200 MemoryBuffer::getFileOrSTDIN(Filename);
201 if (std::error_code EC = Buff.getError()) {
202 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << "\n";
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000203 return;
204 }
205
Rafael Espindola437b0d52014-07-31 03:12:45 +0000206 std::unique_ptr<MachOObjectFile> MachOOF =
207 std::move(ObjectFile::createMachOObjectFile(Buff.get()).get());
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000208
Rafael Espindola56f976f2013-04-18 18:08:55 +0000209 DisassembleInputMachO2(Filename, MachOOF.get());
Rafael Espindola9b709252013-04-13 01:45:40 +0000210}
211
Rafael Espindola9b709252013-04-13 01:45:40 +0000212static void DisassembleInputMachO2(StringRef Filename,
Rafael Espindola56f976f2013-04-18 18:08:55 +0000213 MachOObjectFile *MachOOF) {
Rafael Espindola9b709252013-04-13 01:45:40 +0000214 const Target *TheTarget = GetTarget(MachOOF);
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000215 if (!TheTarget) {
216 // GetTarget prints out stuff.
217 return;
218 }
Ahmed Charles56440fd2014-03-06 05:51:42 +0000219 std::unique_ptr<const MCInstrInfo> InstrInfo(TheTarget->createMCInstrInfo());
220 std::unique_ptr<MCInstrAnalysis> InstrAnalysis(
221 TheTarget->createMCInstrAnalysis(InstrInfo.get()));
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000222
Kevin Enderbyc9595622014-08-06 23:24:41 +0000223 // Package up features to be passed to target/subtarget
224 std::string FeaturesStr;
225 if (MAttrs.size()) {
226 SubtargetFeatures Features;
227 for (unsigned i = 0; i != MAttrs.size(); ++i)
228 Features.AddFeature(MAttrs[i]);
229 FeaturesStr = Features.getString();
230 }
231
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000232 // Set up disassembler.
Ahmed Charles56440fd2014-03-06 05:51:42 +0000233 std::unique_ptr<const MCRegisterInfo> MRI(
234 TheTarget->createMCRegInfo(TripleName));
235 std::unique_ptr<const MCAsmInfo> AsmInfo(
Rafael Espindola227144c2013-05-13 01:16:13 +0000236 TheTarget->createMCAsmInfo(*MRI, TripleName));
Ahmed Charles56440fd2014-03-06 05:51:42 +0000237 std::unique_ptr<const MCSubtargetInfo> STI(
Kevin Enderbyc9595622014-08-06 23:24:41 +0000238 TheTarget->createMCSubtargetInfo(TripleName, MCPU, FeaturesStr));
Craig Toppere6cb63e2014-04-25 04:24:47 +0000239 MCContext Ctx(AsmInfo.get(), MRI.get(), nullptr);
Ahmed Charles56440fd2014-03-06 05:51:42 +0000240 std::unique_ptr<const MCDisassembler> DisAsm(
Lang Hamesa1bc0f52014-04-15 04:40:56 +0000241 TheTarget->createMCDisassembler(*STI, Ctx));
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000242 int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
Ahmed Charles56440fd2014-03-06 05:51:42 +0000243 std::unique_ptr<MCInstPrinter> IP(TheTarget->createMCInstPrinter(
244 AsmPrinterVariant, *AsmInfo, *InstrInfo, *MRI, *STI));
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000245
246 if (!InstrAnalysis || !AsmInfo || !STI || !DisAsm || !IP) {
Michael J. Spencerc1363cf2011-10-07 19:25:47 +0000247 errs() << "error: couldn't initialize disassembler for target "
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000248 << TripleName << '\n';
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000249 return;
250 }
251
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000252 outs() << '\n' << Filename << ":\n\n";
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000253
Charles Davis8bdfafd2013-09-01 04:28:48 +0000254 MachO::mach_header Header = MachOOF->getHeader();
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000255
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000256 // FIXME: FoundFns isn't used anymore. Using symbols/LC_FUNCTION_STARTS to
257 // determine function locations will eventually go in MCObjectDisassembler.
258 // FIXME: Using the -cfg command line option, this code used to be able to
259 // annotate relocations with the referenced symbol's name, and if this was
260 // inside a __[cf]string section, the data it points to. This is now replaced
261 // by the upcoming MCSymbolizer, which needs the appropriate setup done above.
Owen Andersond9243c42011-10-17 21:37:35 +0000262 std::vector<SectionRef> Sections;
263 std::vector<SymbolRef> Symbols;
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000264 SmallVector<uint64_t, 8> FoundFns;
Kevin Enderby273ae012013-06-06 17:20:50 +0000265 uint64_t BaseSegmentAddress;
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000266
Kevin Enderby273ae012013-06-06 17:20:50 +0000267 getSectionsAndSymbols(Header, MachOOF, Sections, Symbols, FoundFns,
268 BaseSegmentAddress);
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000269
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000270 // Sort the symbols by address, just in case they didn't come in that way.
Owen Andersond9243c42011-10-17 21:37:35 +0000271 std::sort(Symbols.begin(), Symbols.end(), SymbolSorter());
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000272
Kevin Enderby273ae012013-06-06 17:20:50 +0000273 // Build a data in code table that is sorted on by the address of each entry.
274 uint64_t BaseAddress = 0;
Charles Davis8bdfafd2013-09-01 04:28:48 +0000275 if (Header.filetype == MachO::MH_OBJECT)
Kevin Enderby273ae012013-06-06 17:20:50 +0000276 Sections[0].getAddress(BaseAddress);
277 else
278 BaseAddress = BaseSegmentAddress;
279 DiceTable Dices;
Kevin Enderby273ae012013-06-06 17:20:50 +0000280 for (dice_iterator DI = MachOOF->begin_dices(), DE = MachOOF->end_dices();
Rafael Espindola5e812af2014-01-30 02:49:50 +0000281 DI != DE; ++DI) {
Kevin Enderby273ae012013-06-06 17:20:50 +0000282 uint32_t Offset;
283 DI->getOffset(Offset);
284 Dices.push_back(std::make_pair(BaseAddress + Offset, *DI));
285 }
286 array_pod_sort(Dices.begin(), Dices.end());
287
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000288#ifndef NDEBUG
289 raw_ostream &DebugOut = DebugFlag ? dbgs() : nulls();
290#else
291 raw_ostream &DebugOut = nulls();
292#endif
293
Ahmed Charles56440fd2014-03-06 05:51:42 +0000294 std::unique_ptr<DIContext> diContext;
Rafael Espindola9b709252013-04-13 01:45:40 +0000295 ObjectFile *DbgObj = MachOOF;
Benjamin Kramer699128e2011-09-21 01:13:19 +0000296 // Try to find debug info and set up the DIContext for it.
297 if (UseDbg) {
Benjamin Kramer699128e2011-09-21 01:13:19 +0000298 // A separate DSym file path was specified, parse it as a macho file,
299 // get the sections and supply it to the section name parsing machinery.
300 if (!DSYMFile.empty()) {
Rafael Espindolaadf21f22014-07-06 17:43:13 +0000301 ErrorOr<std::unique_ptr<MemoryBuffer>> Buf =
302 MemoryBuffer::getFileOrSTDIN(DSYMFile);
303 if (std::error_code EC = Buf.getError()) {
304 errs() << "llvm-objdump: " << Filename << ": " << EC.message() << '\n';
Benjamin Kramer699128e2011-09-21 01:13:19 +0000305 return;
306 }
Rafael Espindola437b0d52014-07-31 03:12:45 +0000307 DbgObj = ObjectFile::createMachOObjectFile(Buf.get()).get().release();
Benjamin Kramer699128e2011-09-21 01:13:19 +0000308 }
309
Eric Christopher7370b552012-11-12 21:40:38 +0000310 // Setup the DIContext
Rafael Espindolaa04bb5b2014-07-31 20:19:36 +0000311 diContext.reset(DIContext::getDWARFContext(*DbgObj));
Benjamin Kramer699128e2011-09-21 01:13:19 +0000312 }
313
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000314 for (unsigned SectIdx = 0; SectIdx != Sections.size(); SectIdx++) {
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000315
316 bool SectIsText = false;
317 Sections[SectIdx].isText(SectIsText);
318 if (SectIsText == false)
319 continue;
320
Owen Andersond9243c42011-10-17 21:37:35 +0000321 StringRef SectName;
322 if (Sections[SectIdx].getName(SectName) ||
Rafael Espindolaa9f810b2012-12-21 03:47:03 +0000323 SectName != "__text")
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000324 continue; // Skip non-text sections
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000325
Rafael Espindolaa9f810b2012-12-21 03:47:03 +0000326 DataRefImpl DR = Sections[SectIdx].getRawDataRefImpl();
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000327
Rafael Espindolab0f76a42013-04-05 15:15:22 +0000328 StringRef SegmentName = MachOOF->getSectionFinalSegmentName(DR);
329 if (SegmentName != "__TEXT")
Rafael Espindolaa9f810b2012-12-21 03:47:03 +0000330 continue;
331
Owen Andersond9243c42011-10-17 21:37:35 +0000332 StringRef Bytes;
333 Sections[SectIdx].getContents(Bytes);
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000334 StringRefMemoryObject memoryObject(Bytes);
335 bool symbolTableWorked = false;
336
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000337 // Parse relocations.
Alexey Samsonovaa4d2952014-03-14 14:22:49 +0000338 std::vector<std::pair<uint64_t, SymbolRef>> Relocs;
339 for (const RelocationRef &Reloc : Sections[SectIdx].relocations()) {
Owen Andersond9243c42011-10-17 21:37:35 +0000340 uint64_t RelocOffset, SectionAddress;
Alexey Samsonovaa4d2952014-03-14 14:22:49 +0000341 Reloc.getOffset(RelocOffset);
Owen Andersond9243c42011-10-17 21:37:35 +0000342 Sections[SectIdx].getAddress(SectionAddress);
343 RelocOffset -= SectionAddress;
344
Alexey Samsonovaa4d2952014-03-14 14:22:49 +0000345 symbol_iterator RelocSym = Reloc.getSymbol();
Owen Andersond9243c42011-10-17 21:37:35 +0000346
Rafael Espindola806f0062013-06-05 01:33:53 +0000347 Relocs.push_back(std::make_pair(RelocOffset, *RelocSym));
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000348 }
349 array_pod_sort(Relocs.begin(), Relocs.end());
350
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000351 // Disassemble symbol by symbol.
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000352 for (unsigned SymIdx = 0; SymIdx != Symbols.size(); SymIdx++) {
Owen Andersond9243c42011-10-17 21:37:35 +0000353 StringRef SymName;
354 Symbols[SymIdx].getName(SymName);
355
356 SymbolRef::Type ST;
357 Symbols[SymIdx].getType(ST);
358 if (ST != SymbolRef::ST_Function)
359 continue;
360
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000361 // Make sure the symbol is defined in this section.
Owen Andersond9243c42011-10-17 21:37:35 +0000362 bool containsSym = false;
363 Sections[SectIdx].containsSymbol(Symbols[SymIdx], containsSym);
364 if (!containsSym)
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000365 continue;
366
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000367 // Start at the address of the symbol relative to the section's address.
Cameron Zwarich54478a52012-02-03 05:42:17 +0000368 uint64_t SectionAddress = 0;
Owen Andersond9243c42011-10-17 21:37:35 +0000369 uint64_t Start = 0;
Cameron Zwarich54478a52012-02-03 05:42:17 +0000370 Sections[SectIdx].getAddress(SectionAddress);
Danil Malyshevcbe72fc2011-11-29 17:40:10 +0000371 Symbols[SymIdx].getAddress(Start);
Cameron Zwarich54478a52012-02-03 05:42:17 +0000372 Start -= SectionAddress;
Owen Andersond9243c42011-10-17 21:37:35 +0000373
Benjamin Kramer2ad2eb52011-09-20 17:53:01 +0000374 // Stop disassembling either at the beginning of the next symbol or at
375 // the end of the section.
Kevin Enderbyedd58722012-05-15 18:57:14 +0000376 bool containsNextSym = false;
Owen Andersond9243c42011-10-17 21:37:35 +0000377 uint64_t NextSym = 0;
378 uint64_t NextSymIdx = SymIdx+1;
379 while (Symbols.size() > NextSymIdx) {
380 SymbolRef::Type NextSymType;
381 Symbols[NextSymIdx].getType(NextSymType);
382 if (NextSymType == SymbolRef::ST_Function) {
383 Sections[SectIdx].containsSymbol(Symbols[NextSymIdx],
384 containsNextSym);
Danil Malyshevcbe72fc2011-11-29 17:40:10 +0000385 Symbols[NextSymIdx].getAddress(NextSym);
Cameron Zwarich54478a52012-02-03 05:42:17 +0000386 NextSym -= SectionAddress;
Owen Andersond9243c42011-10-17 21:37:35 +0000387 break;
388 }
389 ++NextSymIdx;
390 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000391
Owen Andersond9243c42011-10-17 21:37:35 +0000392 uint64_t SectSize;
393 Sections[SectIdx].getSize(SectSize);
394 uint64_t End = containsNextSym ? NextSym : SectSize;
395 uint64_t Size;
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000396
397 symbolTableWorked = true;
398
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000399 outs() << SymName << ":\n";
400 DILineInfo lastLine;
401 for (uint64_t Index = Start; Index < End; Index += Size) {
402 MCInst Inst;
Owen Andersond9243c42011-10-17 21:37:35 +0000403
Kevin Enderby273ae012013-06-06 17:20:50 +0000404 uint64_t SectAddress = 0;
405 Sections[SectIdx].getAddress(SectAddress);
406 outs() << format("%8" PRIx64 ":\t", SectAddress + Index);
407
408 // Check the data in code table here to see if this is data not an
409 // instruction to be disassembled.
410 DiceTable Dice;
411 Dice.push_back(std::make_pair(SectAddress + Index, DiceRef()));
412 dice_table_iterator DTI = std::search(Dices.begin(), Dices.end(),
413 Dice.begin(), Dice.end(),
414 compareDiceTableEntries);
415 if (DTI != Dices.end()){
416 uint16_t Length;
417 DTI->second.getLength(Length);
418 DumpBytes(StringRef(Bytes.data() + Index, Length));
419 uint16_t Kind;
420 DTI->second.getKind(Kind);
421 DumpDataInCode(Bytes.data() + Index, Length, Kind);
422 continue;
423 }
424
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000425 if (DisAsm->getInstruction(Inst, Size, memoryObject, Index,
426 DebugOut, nulls())) {
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000427 DumpBytes(StringRef(Bytes.data() + Index, Size));
428 IP->printInst(&Inst, outs(), "");
Owen Andersond9243c42011-10-17 21:37:35 +0000429
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000430 // Print debug info.
431 if (diContext) {
432 DILineInfo dli =
433 diContext->getLineInfoForAddress(SectAddress + Index);
434 // Print valid line info if it changed.
Alexey Samsonovd0109992014-04-18 21:36:39 +0000435 if (dli != lastLine && dli.Line != 0)
436 outs() << "\t## " << dli.FileName << ':' << dli.Line << ':'
437 << dli.Column;
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000438 lastLine = dli;
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000439 }
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000440 outs() << "\n";
441 } else {
442 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
443 if (Size == 0)
444 Size = 1; // skip illegible bytes
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000445 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000446 }
447 }
Ahmed Bougachaaa790682013-05-24 01:07:04 +0000448 if (!symbolTableWorked) {
Kevin Enderbybadd1002012-05-18 00:13:56 +0000449 // Reading the symbol table didn't work, disassemble the whole section.
450 uint64_t SectAddress;
451 Sections[SectIdx].getAddress(SectAddress);
452 uint64_t SectSize;
453 Sections[SectIdx].getSize(SectSize);
454 uint64_t InstSize;
455 for (uint64_t Index = 0; Index < SectSize; Index += InstSize) {
Bill Wendling4e68e062012-07-19 00:17:40 +0000456 MCInst Inst;
Kevin Enderbybadd1002012-05-18 00:13:56 +0000457
Bill Wendling4e68e062012-07-19 00:17:40 +0000458 if (DisAsm->getInstruction(Inst, InstSize, memoryObject, Index,
459 DebugOut, nulls())) {
460 outs() << format("%8" PRIx64 ":\t", SectAddress + Index);
461 DumpBytes(StringRef(Bytes.data() + Index, InstSize));
462 IP->printInst(&Inst, outs(), "");
463 outs() << "\n";
464 } else {
465 errs() << "llvm-objdump: warning: invalid instruction encoding\n";
466 if (InstSize == 0)
467 InstSize = 1; // skip illegible bytes
468 }
Kevin Enderbybadd1002012-05-18 00:13:56 +0000469 }
470 }
Benjamin Kramer43a772e2011-09-19 17:56:04 +0000471 }
472}
Tim Northover4bd286a2014-08-01 13:07:19 +0000473
474namespace {
475struct CompactUnwindEntry {
476 uint32_t OffsetInSection;
477
478 uint64_t FunctionAddr;
479 uint32_t Length;
480 uint32_t CompactEncoding;
481 uint64_t PersonalityAddr;
482 uint64_t LSDAAddr;
483
484 RelocationRef FunctionReloc;
485 RelocationRef PersonalityReloc;
486 RelocationRef LSDAReloc;
487
488 CompactUnwindEntry(StringRef Contents, unsigned Offset, bool Is64)
489 : OffsetInSection(Offset) {
490 if (Is64)
491 read<uint64_t>(Contents.data() + Offset);
492 else
493 read<uint32_t>(Contents.data() + Offset);
494 }
495
496private:
497 template<typename T>
498 static uint64_t readNext(const char *&Buf) {
499 using llvm::support::little;
500 using llvm::support::unaligned;
501
502 uint64_t Val = support::endian::read<T, little, unaligned>(Buf);
503 Buf += sizeof(T);
504 return Val;
505 }
506
507 template<typename UIntPtr>
508 void read(const char *Buf) {
509 FunctionAddr = readNext<UIntPtr>(Buf);
510 Length = readNext<uint32_t>(Buf);
511 CompactEncoding = readNext<uint32_t>(Buf);
512 PersonalityAddr = readNext<UIntPtr>(Buf);
513 LSDAAddr = readNext<UIntPtr>(Buf);
514 }
515};
516}
517
518/// Given a relocation from __compact_unwind, consisting of the RelocationRef
519/// and data being relocated, determine the best base Name and Addend to use for
520/// display purposes.
521///
522/// 1. An Extern relocation will directly reference a symbol (and the data is
523/// then already an addend), so use that.
524/// 2. Otherwise the data is an offset in the object file's layout; try to find
525// a symbol before it in the same section, and use the offset from there.
526/// 3. Finally, if all that fails, fall back to an offset from the start of the
527/// referenced section.
528static void findUnwindRelocNameAddend(const MachOObjectFile *Obj,
529 std::map<uint64_t, SymbolRef> &Symbols,
530 const RelocationRef &Reloc,
531 uint64_t Addr,
532 StringRef &Name, uint64_t &Addend) {
533 if (Reloc.getSymbol() != Obj->symbol_end()) {
534 Reloc.getSymbol()->getName(Name);
535 Addend = Addr;
536 return;
537 }
538
539 auto RE = Obj->getRelocation(Reloc.getRawDataRefImpl());
540 SectionRef RelocSection = Obj->getRelocationSection(RE);
541
542 uint64_t SectionAddr;
543 RelocSection.getAddress(SectionAddr);
544
545 auto Sym = Symbols.upper_bound(Addr);
546 if (Sym == Symbols.begin()) {
547 // The first symbol in the object is after this reference, the best we can
548 // do is section-relative notation.
549 RelocSection.getName(Name);
550 Addend = Addr - SectionAddr;
551 return;
552 }
553
554 // Go back one so that SymbolAddress <= Addr.
555 --Sym;
556
557 section_iterator SymSection = Obj->section_end();
558 Sym->second.getSection(SymSection);
559 if (RelocSection == *SymSection) {
560 // There's a valid symbol in the same section before this reference.
561 Sym->second.getName(Name);
562 Addend = Addr - Sym->first;
563 return;
564 }
565
566 // There is a symbol before this reference, but it's in a different
567 // section. Probably not helpful to mention it, so use the section name.
568 RelocSection.getName(Name);
569 Addend = Addr - SectionAddr;
570}
571
572static void printUnwindRelocDest(const MachOObjectFile *Obj,
573 std::map<uint64_t, SymbolRef> &Symbols,
574 const RelocationRef &Reloc,
575 uint64_t Addr) {
576 StringRef Name;
577 uint64_t Addend;
578
579 findUnwindRelocNameAddend(Obj, Symbols, Reloc, Addr, Name, Addend);
580
581 outs() << Name;
582 if (Addend)
583 outs() << " + " << format("0x%x", Addend);
584}
585
586static void
587printMachOCompactUnwindSection(const MachOObjectFile *Obj,
588 std::map<uint64_t, SymbolRef> &Symbols,
589 const SectionRef &CompactUnwind) {
590
591 assert(Obj->isLittleEndian() &&
592 "There should not be a big-endian .o with __compact_unwind");
593
594 bool Is64 = Obj->is64Bit();
595 uint32_t PointerSize = Is64 ? sizeof(uint64_t) : sizeof(uint32_t);
596 uint32_t EntrySize = 3 * PointerSize + 2 * sizeof(uint32_t);
597
598 StringRef Contents;
599 CompactUnwind.getContents(Contents);
600
601 SmallVector<CompactUnwindEntry, 4> CompactUnwinds;
602
603 // First populate the initial raw offsets, encodings and so on from the entry.
604 for (unsigned Offset = 0; Offset < Contents.size(); Offset += EntrySize) {
605 CompactUnwindEntry Entry(Contents.data(), Offset, Is64);
606 CompactUnwinds.push_back(Entry);
607 }
608
609 // Next we need to look at the relocations to find out what objects are
610 // actually being referred to.
611 for (const RelocationRef &Reloc : CompactUnwind.relocations()) {
612 uint64_t RelocAddress;
613 Reloc.getOffset(RelocAddress);
614
615 uint32_t EntryIdx = RelocAddress / EntrySize;
616 uint32_t OffsetInEntry = RelocAddress - EntryIdx * EntrySize;
617 CompactUnwindEntry &Entry = CompactUnwinds[EntryIdx];
618
619 if (OffsetInEntry == 0)
620 Entry.FunctionReloc = Reloc;
621 else if (OffsetInEntry == PointerSize + 2 * sizeof(uint32_t))
622 Entry.PersonalityReloc = Reloc;
623 else if (OffsetInEntry == 2 * PointerSize + 2 * sizeof(uint32_t))
624 Entry.LSDAReloc = Reloc;
625 else
626 llvm_unreachable("Unexpected relocation in __compact_unwind section");
627 }
628
629 // Finally, we're ready to print the data we've gathered.
630 outs() << "Contents of __compact_unwind section:\n";
631 for (auto &Entry : CompactUnwinds) {
Tim Northover06af2602014-08-08 12:08:51 +0000632 outs() << " Entry at offset "
633 << format("0x%" PRIx32, Entry.OffsetInSection) << ":\n";
Tim Northover4bd286a2014-08-01 13:07:19 +0000634
635 // 1. Start of the region this entry applies to.
636 outs() << " start: "
Tim Northoverb911bf82014-08-08 12:00:09 +0000637 << format("0x%" PRIx64, Entry.FunctionAddr) << ' ';
Tim Northover4bd286a2014-08-01 13:07:19 +0000638 printUnwindRelocDest(Obj, Symbols, Entry.FunctionReloc,
639 Entry.FunctionAddr);
640 outs() << '\n';
641
642 // 2. Length of the region this entry applies to.
643 outs() << " length: "
Tim Northoverb911bf82014-08-08 12:00:09 +0000644 << format("0x%" PRIx32, Entry.Length) << '\n';
Tim Northover4bd286a2014-08-01 13:07:19 +0000645 // 3. The 32-bit compact encoding.
646 outs() << " compact encoding: "
Tim Northoverb911bf82014-08-08 12:00:09 +0000647 << format("0x%08" PRIx32, Entry.CompactEncoding) << '\n';
Tim Northover4bd286a2014-08-01 13:07:19 +0000648
649 // 4. The personality function, if present.
650 if (Entry.PersonalityReloc.getObjectFile()) {
651 outs() << " personality function: "
Tim Northoverb911bf82014-08-08 12:00:09 +0000652 << format("0x%" PRIx64, Entry.PersonalityAddr) << ' ';
Tim Northover4bd286a2014-08-01 13:07:19 +0000653 printUnwindRelocDest(Obj, Symbols, Entry.PersonalityReloc,
654 Entry.PersonalityAddr);
655 outs() << '\n';
656 }
657
658 // 5. This entry's language-specific data area.
659 if (Entry.LSDAReloc.getObjectFile()) {
660 outs() << " LSDA: "
Tim Northoverb911bf82014-08-08 12:00:09 +0000661 << format("0x%" PRIx64, Entry.LSDAAddr) << ' ';
Tim Northover4bd286a2014-08-01 13:07:19 +0000662 printUnwindRelocDest(Obj, Symbols, Entry.LSDAReloc, Entry.LSDAAddr);
663 outs() << '\n';
664 }
665 }
666}
667
668void llvm::printMachOUnwindInfo(const MachOObjectFile *Obj) {
669 std::map<uint64_t, SymbolRef> Symbols;
670 for (const SymbolRef &SymRef : Obj->symbols()) {
671 // Discard any undefined or absolute symbols. They're not going to take part
672 // in the convenience lookup for unwind info and just take up resources.
673 section_iterator Section = Obj->section_end();
674 SymRef.getSection(Section);
675 if (Section == Obj->section_end())
676 continue;
677
678 uint64_t Addr;
679 SymRef.getAddress(Addr);
680 Symbols.insert(std::make_pair(Addr, SymRef));
681 }
682
683 for (const SectionRef &Section : Obj->sections()) {
684 StringRef SectName;
685 Section.getName(SectName);
686 if (SectName == "__compact_unwind")
687 printMachOCompactUnwindSection(Obj, Symbols, Section);
688 else if (SectName == "__unwind_info")
689 outs() << "llvm-objdump: warning: unhandled __unwind_info section\n";
690 else if (SectName == "__eh_frame")
691 outs() << "llvm-objdump: warning: unhandled __eh_frame section\n";
692
693 }
694}