1 //===- DWARFDebugLoc.cpp --------------------------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #include "llvm/DebugInfo/DWARF/DWARFDebugLoc.h"
10 #include "llvm/ADT/StringRef.h"
11 #include "llvm/BinaryFormat/Dwarf.h"
12 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
13 #include "llvm/DebugInfo/DWARF/DWARFExpression.h"
14 #include "llvm/DebugInfo/DWARF/DWARFRelocMap.h"
15 #include "llvm/DebugInfo/DWARF/DWARFUnit.h"
16 #include "llvm/Support/Compiler.h"
17 #include "llvm/Support/Format.h"
18 #include "llvm/Support/WithColor.h"
19 #include "llvm/Support/raw_ostream.h"
20 #include <algorithm>
21 #include <cinttypes>
22 #include <cstdint>
23 
24 using namespace llvm;
25 using object::SectionedAddress;
26 
27 namespace {
28 class DWARFLocationInterpreter {
29   Optional<object::SectionedAddress> Base;
30   std::function<Optional<object::SectionedAddress>(uint32_t)> LookupAddr;
31 
32 public:
33   DWARFLocationInterpreter(
34       Optional<object::SectionedAddress> Base,
35       std::function<Optional<object::SectionedAddress>(uint32_t)> LookupAddr)
36       : Base(Base), LookupAddr(std::move(LookupAddr)) {}
37 
38   Expected<Optional<DWARFLocationExpression>>
39   Interpret(const DWARFLocationEntry &E);
40 };
41 } // namespace
42 
43 static Error createResolverError(uint32_t Index, unsigned Kind) {
44   return createStringError(errc::invalid_argument,
45                            "Unable to resolve indirect address %u for: %s",
46                            Index, dwarf::LocListEncodingString(Kind).data());
47 }
48 
49 Expected<Optional<DWARFLocationExpression>>
50 DWARFLocationInterpreter::Interpret(const DWARFLocationEntry &E) {
51   switch (E.Kind) {
52   case dwarf::DW_LLE_end_of_list:
53     return None;
54   case dwarf::DW_LLE_base_addressx: {
55     Base = LookupAddr(E.Value0);
56     if (!Base)
57       return createResolverError(E.Value0, E.Kind);
58     return None;
59   }
60   case dwarf::DW_LLE_startx_length: {
61     Optional<SectionedAddress> LowPC = LookupAddr(E.Value0);
62     if (!LowPC)
63       return createResolverError(E.Value0, E.Kind);
64     return DWARFLocationExpression{DWARFAddressRange{LowPC->Address,
65                                                      LowPC->Address + E.Value1,
66                                                      LowPC->SectionIndex},
67                                    E.Loc};
68   }
69   case dwarf::DW_LLE_offset_pair:
70     if (!Base) {
71       return createStringError(
72           inconvertibleErrorCode(),
73           "Unable to resolve DW_LLE_offset_pair: base address unknown");
74     }
75     return DWARFLocationExpression{DWARFAddressRange{Base->Address + E.Value0,
76                                                      Base->Address + E.Value1,
77                                                      Base->SectionIndex},
78                                    E.Loc};
79   case dwarf::DW_LLE_base_address:
80     Base = SectionedAddress{E.Value0, SectionedAddress::UndefSection};
81     return None;
82   case dwarf::DW_LLE_start_length:
83     return DWARFLocationExpression{
84         DWARFAddressRange{E.Value0, E.Value0 + E.Value1,
85                           SectionedAddress::UndefSection},
86         E.Loc};
87   default:
88     llvm_unreachable("unreachable locations list kind");
89   }
90 }
91 
92 // When directly dumping the .debug_loc without a compile unit, we have to guess
93 // at the DWARF version. This only affects DW_OP_call_ref, which is a rare
94 // expression that LLVM doesn't produce. Guessing the wrong version means we
95 // won't be able to pretty print expressions in DWARF2 binaries produced by
96 // non-LLVM tools.
97 static void dumpExpression(raw_ostream &OS, ArrayRef<uint8_t> Data,
98                            bool IsLittleEndian, unsigned AddressSize,
99                            const MCRegisterInfo *MRI, DWARFUnit *U) {
100   DWARFDataExtractor Extractor(toStringRef(Data), IsLittleEndian, AddressSize);
101   DWARFExpression(Extractor, dwarf::DWARF_VERSION, AddressSize).print(OS, MRI, U);
102 }
103 
104 bool DWARFLocationTable::dumpLocationList(uint64_t *Offset, raw_ostream &OS,
105                                           Optional<SectionedAddress> BaseAddr,
106                                           const MCRegisterInfo *MRI,
107                                           DWARFUnit *U, DIDumpOptions DumpOpts,
108                                           unsigned Indent) const {
109   DWARFLocationInterpreter Interp(
110       BaseAddr, [U](uint32_t Index) -> Optional<SectionedAddress> {
111         if (U)
112           return U->getAddrOffsetSectionItem(Index);
113         return None;
114       });
115   OS << format("0x%8.8" PRIx64 ": ", *Offset);
116   Error E = visitLocationList(Offset, [&](const DWARFLocationEntry &E) {
117     Expected<Optional<DWARFLocationExpression>> Loc = Interp.Interpret(E);
118     if (!Loc || DumpOpts.Verbose)
119       dumpRawEntry(E, OS, Indent);
120     if (Loc && *Loc) {
121       OS << "\n";
122       OS.indent(Indent);
123       if (DumpOpts.Verbose)
124         OS << "          => ";
125       Loc.get()->Range->dump(OS, Data.getAddressSize(), DumpOpts);
126     }
127     if (!Loc)
128       consumeError(Loc.takeError());
129 
130     if (E.Kind != dwarf::DW_LLE_base_address &&
131         E.Kind != dwarf::DW_LLE_base_addressx &&
132         E.Kind != dwarf::DW_LLE_end_of_list) {
133       OS << ": ";
134       dumpExpression(OS, E.Loc, Data.isLittleEndian(), Data.getAddressSize(),
135                      MRI, U);
136     }
137     return true;
138   });
139   if (E) {
140     OS << "\n";
141     OS.indent(Indent);
142     OS << "error: " << toString(std::move(E));
143     return false;
144   }
145   return true;
146 }
147 
148 DWARFDebugLoc::LocationList const *
149 DWARFDebugLoc::getLocationListAtOffset(uint64_t Offset) const {
150   auto It = partition_point(
151       Locations, [=](const LocationList &L) { return L.Offset < Offset; });
152   if (It != Locations.end() && It->Offset == Offset)
153     return &(*It);
154   return nullptr;
155 }
156 
157 void DWARFDebugLoc::dump(raw_ostream &OS, const MCRegisterInfo *MRI,
158                          DIDumpOptions DumpOpts,
159                          Optional<uint64_t> DumpOffset) const {
160   auto BaseAddr = None;
161   unsigned Indent = 12;
162   if (DumpOffset) {
163     dumpLocationList(&*DumpOffset, OS, BaseAddr, MRI, nullptr, DumpOpts,
164                      Indent);
165   } else {
166     uint64_t Offset = 0;
167     StringRef Separator;
168     bool CanContinue = true;
169     while (CanContinue && Data.isValidOffset(Offset)) {
170       OS << Separator;
171       Separator = "\n";
172 
173       CanContinue = dumpLocationList(&Offset, OS, BaseAddr, MRI, nullptr,
174                                      DumpOpts, Indent);
175       OS << '\n';
176     }
177   }
178 }
179 
180 Error DWARFDebugLoc::visitLocationList(
181     uint64_t *Offset,
182     function_ref<bool(const DWARFLocationEntry &)> Callback) const {
183   DataExtractor::Cursor C(*Offset);
184   while (true) {
185     uint64_t Value0 = Data.getRelocatedAddress(C);
186     uint64_t Value1 = Data.getRelocatedAddress(C);
187 
188     DWARFLocationEntry E;
189 
190     // The end of any given location list is marked by an end of list entry,
191     // which consists of a 0 for the beginning address offset and a 0 for the
192     // ending address offset. A beginning offset of 0xff...f marks the base
193     // address selection entry.
194     if (Value0 == 0 && Value1 == 0) {
195       E.Kind = dwarf::DW_LLE_end_of_list;
196     } else if (Value0 == (Data.getAddressSize() == 4 ? -1U : -1ULL)) {
197       E.Kind = dwarf::DW_LLE_base_address;
198       E.Value0 = Value1;
199     } else {
200       E.Kind = dwarf::DW_LLE_offset_pair;
201       E.Value0 = Value0;
202       E.Value1 = Value1;
203       unsigned Bytes = Data.getU16(C);
204       // A single location description describing the location of the object...
205       Data.getU8(C, E.Loc, Bytes);
206     }
207 
208     if (!C)
209       return C.takeError();
210     if (!Callback(E) || E.Kind == dwarf::DW_LLE_end_of_list)
211       break;
212   }
213   *Offset = C.tell();
214   return Error::success();
215 }
216 
217 Expected<DWARFDebugLoc::LocationList>
218 DWARFDebugLoc::parseOneLocationList(uint64_t *Offset) {
219   LocationList LL;
220   LL.Offset = *Offset;
221 
222   Error E = visitLocationList(Offset, [&](const DWARFLocationEntry &E) {
223     LL.Entries.push_back(E);
224     return true;
225   });
226   if (E)
227     return std::move(E);
228   return std::move(LL);
229 }
230 
231 void DWARFDebugLoc::parse() {
232   uint64_t Offset = 0;
233   while (Offset < Data.getData().size()) {
234     if (auto LL = parseOneLocationList(&Offset))
235       Locations.push_back(std::move(*LL));
236     else {
237       logAllUnhandledErrors(LL.takeError(), WithColor::error());
238       break;
239     }
240   }
241 }
242 
243 void DWARFDebugLoc::dumpRawEntry(const DWARFLocationEntry &Entry,
244                                  raw_ostream &OS, unsigned Indent) const {
245   uint64_t Value0, Value1;
246   switch (Entry.Kind) {
247   case dwarf::DW_LLE_base_address:
248     Value0 = Data.getAddressSize() == 4 ? -1U : -1ULL;
249     Value1 = Entry.Value0;
250     break;
251   case dwarf::DW_LLE_offset_pair:
252     Value0 = Entry.Value0;
253     Value1 = Entry.Value1;
254     break;
255   case dwarf::DW_LLE_end_of_list:
256     Value0 = Value1 = 0;
257     return;
258   default:
259     llvm_unreachable("Not possible in DWARF4!");
260   }
261   OS << '\n';
262   OS.indent(Indent);
263   OS << '(' << format_hex(Value0, 2 + Data.getAddressSize() * 2) << ", "
264      << format_hex(Value1, 2 + Data.getAddressSize() * 2) << ')';
265 }
266 
267 Error DWARFDebugLoclists::visitLocationList(
268     uint64_t *Offset, function_ref<bool(const DWARFLocationEntry &)> F) const {
269 
270   DataExtractor::Cursor C(*Offset);
271   bool Continue = true;
272   while (Continue) {
273     DWARFLocationEntry E;
274     E.Kind = Data.getU8(C);
275     switch (E.Kind) {
276     case dwarf::DW_LLE_end_of_list:
277       break;
278     case dwarf::DW_LLE_base_addressx:
279       E.Value0 = Data.getULEB128(C);
280       break;
281     case dwarf::DW_LLE_startx_length:
282       E.Value0 = Data.getULEB128(C);
283       // Pre-DWARF 5 has different interpretation of the length field. We have
284       // to support both pre- and standartized styles for the compatibility.
285       if (Version < 5)
286         E.Value1 = Data.getU32(C);
287       else
288         E.Value1 = Data.getULEB128(C);
289       break;
290     case dwarf::DW_LLE_offset_pair:
291       E.Value0 = Data.getULEB128(C);
292       E.Value1 = Data.getULEB128(C);
293       break;
294     case dwarf::DW_LLE_base_address:
295       E.Value0 = Data.getRelocatedAddress(C);
296       break;
297     case dwarf::DW_LLE_start_length:
298       E.Value0 = Data.getRelocatedAddress(C);
299       E.Value1 = Data.getULEB128(C);
300       break;
301     case dwarf::DW_LLE_startx_endx:
302     case dwarf::DW_LLE_default_location:
303     case dwarf::DW_LLE_start_end:
304     default:
305       cantFail(C.takeError());
306       return createStringError(errc::illegal_byte_sequence,
307                                "LLE of kind %x not supported", (int)E.Kind);
308     }
309 
310     if (E.Kind != dwarf::DW_LLE_base_address &&
311         E.Kind != dwarf::DW_LLE_base_addressx &&
312         E.Kind != dwarf::DW_LLE_end_of_list) {
313       unsigned Bytes = Version >= 5 ? Data.getULEB128(C) : Data.getU16(C);
314       // A single location description describing the location of the object...
315       Data.getU8(C, E.Loc, Bytes);
316     }
317 
318     if (!C)
319       return C.takeError();
320     Continue = F(E) && E.Kind != dwarf::DW_LLE_end_of_list;
321   }
322   *Offset = C.tell();
323   return Error::success();
324 }
325 
326 void DWARFDebugLoclists::dumpRawEntry(const DWARFLocationEntry &Entry,
327                                       raw_ostream &OS, unsigned Indent) const {
328   size_t MaxEncodingStringLength = 0;
329 #define HANDLE_DW_LLE(ID, NAME)                                                \
330   MaxEncodingStringLength = std::max(MaxEncodingStringLength,                  \
331                                      dwarf::LocListEncodingString(ID).size());
332 #include "llvm/BinaryFormat/Dwarf.def"
333 
334   OS << "\n";
335   OS.indent(Indent);
336   StringRef EncodingString = dwarf::LocListEncodingString(Entry.Kind);
337   // Unsupported encodings should have been reported during parsing.
338   assert(!EncodingString.empty() && "Unknown loclist entry encoding");
339   OS << format("%-*s(", MaxEncodingStringLength, EncodingString.data());
340   unsigned FieldSize = 2 + 2 * Data.getAddressSize();
341   switch (Entry.Kind) {
342   case dwarf::DW_LLE_startx_length:
343   case dwarf::DW_LLE_start_length:
344   case dwarf::DW_LLE_offset_pair:
345     OS << format_hex(Entry.Value0, FieldSize) << ", "
346        << format_hex(Entry.Value1, FieldSize);
347     break;
348   case dwarf::DW_LLE_base_addressx:
349   case dwarf::DW_LLE_base_address:
350     OS << format_hex(Entry.Value0, FieldSize);
351     break;
352   case dwarf::DW_LLE_end_of_list:
353     break;
354   }
355   OS << ')';
356 }
357 
358 void DWARFDebugLoclists::dumpRange(uint64_t StartOffset, uint64_t Size,
359                                    raw_ostream &OS, const MCRegisterInfo *MRI,
360                                    DIDumpOptions DumpOpts) {
361   if (!Data.isValidOffsetForDataOfSize(StartOffset, Size))  {
362     OS << "Invalid dump range\n";
363     return;
364   }
365   uint64_t Offset = StartOffset;
366   StringRef Separator;
367   bool CanContinue = true;
368   while (CanContinue && Offset < StartOffset + Size) {
369     OS << Separator;
370     Separator = "\n";
371 
372     CanContinue = dumpLocationList(&Offset, OS, /*BaseAddr=*/None, MRI, nullptr,
373                                    DumpOpts, /*Indent=*/12);
374     OS << '\n';
375   }
376 }
377