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                                           uint64_t BaseAddr,
106                                           const MCRegisterInfo *MRI,
107                                           DWARFUnit *U, DIDumpOptions DumpOpts,
108                                           unsigned Indent) const {
109   DWARFLocationInterpreter Interp(
110       SectionedAddress{BaseAddr, SectionedAddress::UndefSection},
111       [U](uint32_t Index) -> Optional<SectionedAddress> {
112         if (U)
113           return U->getAddrOffsetSectionItem(Index);
114         return None;
115       });
116   OS << format("0x%8.8" PRIx64 ": ", *Offset);
117   Error E = visitLocationList(Offset, [&](const DWARFLocationEntry &E) {
118     Expected<Optional<DWARFLocationExpression>> Loc = Interp.Interpret(E);
119     if (!Loc || DumpOpts.Verbose)
120       dumpRawEntry(E, OS, Indent);
121     if (Loc && *Loc) {
122       OS << "\n";
123       OS.indent(Indent);
124       if (DumpOpts.Verbose)
125         OS << "          => ";
126       Loc.get()->Range->dump(OS, Data.getAddressSize(), DumpOpts);
127     }
128     if (!Loc)
129       consumeError(Loc.takeError());
130 
131     if (E.Kind != dwarf::DW_LLE_base_address &&
132         E.Kind != dwarf::DW_LLE_base_addressx &&
133         E.Kind != dwarf::DW_LLE_end_of_list) {
134       OS << ": ";
135       dumpExpression(OS, E.Loc, Data.isLittleEndian(), Data.getAddressSize(),
136                      MRI, U);
137     }
138     return true;
139   });
140   if (E) {
141     OS << "\n";
142     OS.indent(Indent);
143     OS << "error: " << toString(std::move(E));
144     return false;
145   }
146   return true;
147 }
148 
149 void DWARFDebugLoc::LocationList::dump(raw_ostream &OS, uint64_t BaseAddress,
150                                        bool IsLittleEndian,
151                                        unsigned AddressSize,
152                                        const MCRegisterInfo *MRI, DWARFUnit *U,
153                                        DIDumpOptions DumpOpts,
154                                        unsigned Indent) const {
155   for (const Entry &E : Entries) {
156     OS << '\n';
157     OS.indent(Indent);
158     OS << format("[0x%*.*" PRIx64 ", ", AddressSize * 2, AddressSize * 2,
159                  BaseAddress + E.Begin);
160     OS << format(" 0x%*.*" PRIx64 ")", AddressSize * 2, AddressSize * 2,
161                  BaseAddress + E.End);
162     OS << ": ";
163 
164     dumpExpression(OS, E.Loc, IsLittleEndian, AddressSize, MRI, U);
165   }
166 }
167 
168 DWARFDebugLoc::LocationList const *
169 DWARFDebugLoc::getLocationListAtOffset(uint64_t Offset) const {
170   auto It = partition_point(
171       Locations, [=](const LocationList &L) { return L.Offset < Offset; });
172   if (It != Locations.end() && It->Offset == Offset)
173     return &(*It);
174   return nullptr;
175 }
176 
177 void DWARFDebugLoc::dump(raw_ostream &OS, const MCRegisterInfo *MRI, DIDumpOptions DumpOpts,
178                          Optional<uint64_t> Offset) const {
179   auto DumpLocationList = [&](const LocationList &L) {
180     OS << format("0x%8.8" PRIx64 ": ", L.Offset);
181     L.dump(OS, 0, IsLittleEndian, AddressSize, MRI, nullptr, DumpOpts, 12);
182     OS << "\n";
183   };
184 
185   if (Offset) {
186     if (auto *L = getLocationListAtOffset(*Offset))
187       DumpLocationList(*L);
188     return;
189   }
190 
191   for (const LocationList &L : Locations) {
192     DumpLocationList(L);
193     if (&L != &Locations.back())
194       OS << '\n';
195   }
196 }
197 
198 Expected<DWARFDebugLoc::LocationList>
199 DWARFDebugLoc::parseOneLocationList(const DWARFDataExtractor &Data,
200                                     uint64_t *Offset) {
201   LocationList LL;
202   LL.Offset = *Offset;
203   AddressSize = Data.getAddressSize();
204   DataExtractor::Cursor C(*Offset);
205 
206   // 2.6.2 Location Lists
207   // A location list entry consists of:
208   while (true) {
209     Entry E;
210 
211     // 1. A beginning address offset. ...
212     E.Begin = Data.getRelocatedAddress(C);
213 
214     // 2. An ending address offset. ...
215     E.End = Data.getRelocatedAddress(C);
216 
217     if (Error Err = C.takeError())
218       return std::move(Err);
219 
220     // The end of any given location list is marked by an end of list entry,
221     // which consists of a 0 for the beginning address offset and a 0 for the
222     // ending address offset.
223     if (E.Begin == 0 && E.End == 0) {
224       *Offset = C.tell();
225       return LL;
226     }
227 
228     if (E.Begin != (AddressSize == 4 ? -1U : -1ULL)) {
229       unsigned Bytes = Data.getU16(C);
230       // A single location description describing the location of the object...
231       Data.getU8(C, E.Loc, Bytes);
232     }
233 
234     LL.Entries.push_back(std::move(E));
235   }
236 }
237 
238 void DWARFDebugLoc::parse(const DWARFDataExtractor &data) {
239   IsLittleEndian = data.isLittleEndian();
240   AddressSize = data.getAddressSize();
241 
242   uint64_t Offset = 0;
243   while (Offset < data.getData().size()) {
244     if (auto LL = parseOneLocationList(data, &Offset))
245       Locations.push_back(std::move(*LL));
246     else {
247       logAllUnhandledErrors(LL.takeError(), WithColor::error());
248       break;
249     }
250   }
251 }
252 
253 Error DWARFDebugLoclists::visitLocationList(
254     uint64_t *Offset, function_ref<bool(const DWARFLocationEntry &)> F) const {
255 
256   DataExtractor::Cursor C(*Offset);
257   bool Continue = true;
258   while (Continue) {
259     DWARFLocationEntry E;
260     E.Kind = Data.getU8(C);
261     switch (E.Kind) {
262     case dwarf::DW_LLE_end_of_list:
263       break;
264     case dwarf::DW_LLE_base_addressx:
265       E.Value0 = Data.getULEB128(C);
266       break;
267     case dwarf::DW_LLE_startx_length:
268       E.Value0 = Data.getULEB128(C);
269       // Pre-DWARF 5 has different interpretation of the length field. We have
270       // to support both pre- and standartized styles for the compatibility.
271       if (Version < 5)
272         E.Value1 = Data.getU32(C);
273       else
274         E.Value1 = Data.getULEB128(C);
275       break;
276     case dwarf::DW_LLE_offset_pair:
277       E.Value0 = Data.getULEB128(C);
278       E.Value1 = Data.getULEB128(C);
279       break;
280     case dwarf::DW_LLE_base_address:
281       E.Value0 = Data.getRelocatedAddress(C);
282       break;
283     case dwarf::DW_LLE_start_length:
284       E.Value0 = Data.getRelocatedAddress(C);
285       E.Value1 = Data.getULEB128(C);
286       break;
287     case dwarf::DW_LLE_startx_endx:
288     case dwarf::DW_LLE_default_location:
289     case dwarf::DW_LLE_start_end:
290     default:
291       cantFail(C.takeError());
292       return createStringError(errc::illegal_byte_sequence,
293                                "LLE of kind %x not supported", (int)E.Kind);
294     }
295 
296     if (E.Kind != dwarf::DW_LLE_base_address &&
297         E.Kind != dwarf::DW_LLE_base_addressx &&
298         E.Kind != dwarf::DW_LLE_end_of_list) {
299       unsigned Bytes = Version >= 5 ? Data.getULEB128(C) : Data.getU16(C);
300       // A single location description describing the location of the object...
301       Data.getU8(C, E.Loc, Bytes);
302     }
303 
304     if (!C)
305       return C.takeError();
306     Continue = F(E) && E.Kind != dwarf::DW_LLE_end_of_list;
307   }
308   *Offset = C.tell();
309   return Error::success();
310 }
311 
312 void DWARFDebugLoclists::dumpRawEntry(const DWARFLocationEntry &Entry,
313                                       raw_ostream &OS, unsigned Indent) const {
314   size_t MaxEncodingStringLength = 0;
315 #define HANDLE_DW_LLE(ID, NAME)                                                \
316   MaxEncodingStringLength = std::max(MaxEncodingStringLength,                  \
317                                      dwarf::LocListEncodingString(ID).size());
318 #include "llvm/BinaryFormat/Dwarf.def"
319 
320   OS << "\n";
321   OS.indent(Indent);
322   StringRef EncodingString = dwarf::LocListEncodingString(Entry.Kind);
323   // Unsupported encodings should have been reported during parsing.
324   assert(!EncodingString.empty() && "Unknown loclist entry encoding");
325   OS << format("%-*s(", MaxEncodingStringLength, EncodingString.data());
326   unsigned FieldSize = 2 + 2 * Data.getAddressSize();
327   switch (Entry.Kind) {
328   case dwarf::DW_LLE_startx_length:
329   case dwarf::DW_LLE_start_length:
330   case dwarf::DW_LLE_offset_pair:
331     OS << format_hex(Entry.Value0, FieldSize) << ", "
332        << format_hex(Entry.Value1, FieldSize);
333     break;
334   case dwarf::DW_LLE_base_addressx:
335   case dwarf::DW_LLE_base_address:
336     OS << format_hex(Entry.Value0, FieldSize);
337     break;
338   case dwarf::DW_LLE_end_of_list:
339     break;
340   }
341   OS << ')';
342 }
343 
344 void DWARFDebugLoclists::dumpRange(uint64_t StartOffset, uint64_t Size,
345                                    raw_ostream &OS, uint64_t BaseAddr,
346                                    const MCRegisterInfo *MRI,
347                                    DIDumpOptions DumpOpts) {
348   if (!Data.isValidOffsetForDataOfSize(StartOffset, Size))  {
349     OS << "Invalid dump range\n";
350     return;
351   }
352   uint64_t Offset = StartOffset;
353   StringRef Separator;
354   bool CanContinue = true;
355   while (CanContinue && Offset < StartOffset + Size) {
356     OS << Separator;
357     Separator = "\n";
358 
359     CanContinue = dumpLocationList(&Offset, OS, BaseAddr, MRI, nullptr,
360                                    DumpOpts, /*Indent=*/12);
361     OS << '\n';
362   }
363 }
364