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.DisplayRawContents)
119       dumpRawEntry(E, OS, Indent);
120     if (Loc && *Loc) {
121       OS << "\n";
122       OS.indent(Indent);
123       if (DumpOpts.DisplayRawContents)
124         OS << "          => ";
125 
126       DIDumpOptions RangeDumpOpts(DumpOpts);
127       RangeDumpOpts.DisplayRawContents = false;
128       const DWARFObject *Obj = nullptr;
129       if (U)
130         Obj = &U->getContext().getDWARFObj();
131       Loc.get()->Range->dump(OS, Data.getAddressSize(), RangeDumpOpts, Obj);
132     }
133     if (!Loc)
134       consumeError(Loc.takeError());
135 
136     if (E.Kind != dwarf::DW_LLE_base_address &&
137         E.Kind != dwarf::DW_LLE_base_addressx &&
138         E.Kind != dwarf::DW_LLE_end_of_list) {
139       OS << ": ";
140       dumpExpression(OS, E.Loc, Data.isLittleEndian(), Data.getAddressSize(),
141                      MRI, U);
142     }
143     return true;
144   });
145   if (E) {
146     OS << "\n";
147     OS.indent(Indent);
148     OS << "error: " << toString(std::move(E));
149     return false;
150   }
151   return true;
152 }
153 
154 DWARFDebugLoc::LocationList const *
155 DWARFDebugLoc::getLocationListAtOffset(uint64_t Offset) const {
156   auto It = partition_point(
157       Locations, [=](const LocationList &L) { return L.Offset < Offset; });
158   if (It != Locations.end() && It->Offset == Offset)
159     return &(*It);
160   return nullptr;
161 }
162 
163 void DWARFDebugLoc::dump(raw_ostream &OS, const MCRegisterInfo *MRI,
164                          DIDumpOptions DumpOpts,
165                          Optional<uint64_t> DumpOffset) const {
166   auto BaseAddr = None;
167   unsigned Indent = 12;
168   if (DumpOffset) {
169     dumpLocationList(&*DumpOffset, OS, BaseAddr, MRI, nullptr, DumpOpts,
170                      Indent);
171   } else {
172     uint64_t Offset = 0;
173     StringRef Separator;
174     bool CanContinue = true;
175     while (CanContinue && Data.isValidOffset(Offset)) {
176       OS << Separator;
177       Separator = "\n";
178 
179       CanContinue = dumpLocationList(&Offset, OS, BaseAddr, MRI, nullptr,
180                                      DumpOpts, Indent);
181       OS << '\n';
182     }
183   }
184 }
185 
186 Error DWARFDebugLoc::visitLocationList(
187     uint64_t *Offset,
188     function_ref<bool(const DWARFLocationEntry &)> Callback) const {
189   DataExtractor::Cursor C(*Offset);
190   while (true) {
191     uint64_t Value0 = Data.getRelocatedAddress(C);
192     uint64_t Value1 = Data.getRelocatedAddress(C);
193 
194     DWARFLocationEntry E;
195 
196     // The end of any given location list is marked by an end of list entry,
197     // which consists of a 0 for the beginning address offset and a 0 for the
198     // ending address offset. A beginning offset of 0xff...f marks the base
199     // address selection entry.
200     if (Value0 == 0 && Value1 == 0) {
201       E.Kind = dwarf::DW_LLE_end_of_list;
202     } else if (Value0 == (Data.getAddressSize() == 4 ? -1U : -1ULL)) {
203       E.Kind = dwarf::DW_LLE_base_address;
204       E.Value0 = Value1;
205     } else {
206       E.Kind = dwarf::DW_LLE_offset_pair;
207       E.Value0 = Value0;
208       E.Value1 = Value1;
209       unsigned Bytes = Data.getU16(C);
210       // A single location description describing the location of the object...
211       Data.getU8(C, E.Loc, Bytes);
212     }
213 
214     if (!C)
215       return C.takeError();
216     if (!Callback(E) || E.Kind == dwarf::DW_LLE_end_of_list)
217       break;
218   }
219   *Offset = C.tell();
220   return Error::success();
221 }
222 
223 Expected<DWARFDebugLoc::LocationList>
224 DWARFDebugLoc::parseOneLocationList(uint64_t *Offset) {
225   LocationList LL;
226   LL.Offset = *Offset;
227 
228   Error E = visitLocationList(Offset, [&](const DWARFLocationEntry &E) {
229     LL.Entries.push_back(E);
230     return true;
231   });
232   if (E)
233     return std::move(E);
234   return std::move(LL);
235 }
236 
237 void DWARFDebugLoc::parse() {
238   uint64_t Offset = 0;
239   while (Offset < Data.getData().size()) {
240     if (auto LL = parseOneLocationList(&Offset))
241       Locations.push_back(std::move(*LL));
242     else {
243       logAllUnhandledErrors(LL.takeError(), WithColor::error());
244       break;
245     }
246   }
247 }
248 
249 void DWARFDebugLoc::dumpRawEntry(const DWARFLocationEntry &Entry,
250                                  raw_ostream &OS, unsigned Indent) const {
251   uint64_t Value0, Value1;
252   switch (Entry.Kind) {
253   case dwarf::DW_LLE_base_address:
254     Value0 = Data.getAddressSize() == 4 ? -1U : -1ULL;
255     Value1 = Entry.Value0;
256     break;
257   case dwarf::DW_LLE_offset_pair:
258     Value0 = Entry.Value0;
259     Value1 = Entry.Value1;
260     break;
261   case dwarf::DW_LLE_end_of_list:
262     Value0 = Value1 = 0;
263     return;
264   default:
265     llvm_unreachable("Not possible in DWARF4!");
266   }
267   OS << '\n';
268   OS.indent(Indent);
269   OS << '(' << format_hex(Value0, 2 + Data.getAddressSize() * 2) << ", "
270      << format_hex(Value1, 2 + Data.getAddressSize() * 2) << ')';
271 }
272 
273 Error DWARFDebugLoclists::visitLocationList(
274     uint64_t *Offset, function_ref<bool(const DWARFLocationEntry &)> F) const {
275 
276   DataExtractor::Cursor C(*Offset);
277   bool Continue = true;
278   while (Continue) {
279     DWARFLocationEntry E;
280     E.Kind = Data.getU8(C);
281     switch (E.Kind) {
282     case dwarf::DW_LLE_end_of_list:
283       break;
284     case dwarf::DW_LLE_base_addressx:
285       E.Value0 = Data.getULEB128(C);
286       break;
287     case dwarf::DW_LLE_startx_length:
288       E.Value0 = Data.getULEB128(C);
289       // Pre-DWARF 5 has different interpretation of the length field. We have
290       // to support both pre- and standartized styles for the compatibility.
291       if (Version < 5)
292         E.Value1 = Data.getU32(C);
293       else
294         E.Value1 = Data.getULEB128(C);
295       break;
296     case dwarf::DW_LLE_offset_pair:
297       E.Value0 = Data.getULEB128(C);
298       E.Value1 = Data.getULEB128(C);
299       break;
300     case dwarf::DW_LLE_base_address:
301       E.Value0 = Data.getRelocatedAddress(C);
302       break;
303     case dwarf::DW_LLE_start_length:
304       E.Value0 = Data.getRelocatedAddress(C);
305       E.Value1 = Data.getULEB128(C);
306       break;
307     case dwarf::DW_LLE_startx_endx:
308     case dwarf::DW_LLE_default_location:
309     case dwarf::DW_LLE_start_end:
310     default:
311       cantFail(C.takeError());
312       return createStringError(errc::illegal_byte_sequence,
313                                "LLE of kind %x not supported", (int)E.Kind);
314     }
315 
316     if (E.Kind != dwarf::DW_LLE_base_address &&
317         E.Kind != dwarf::DW_LLE_base_addressx &&
318         E.Kind != dwarf::DW_LLE_end_of_list) {
319       unsigned Bytes = Version >= 5 ? Data.getULEB128(C) : Data.getU16(C);
320       // A single location description describing the location of the object...
321       Data.getU8(C, E.Loc, Bytes);
322     }
323 
324     if (!C)
325       return C.takeError();
326     Continue = F(E) && E.Kind != dwarf::DW_LLE_end_of_list;
327   }
328   *Offset = C.tell();
329   return Error::success();
330 }
331 
332 void DWARFDebugLoclists::dumpRawEntry(const DWARFLocationEntry &Entry,
333                                       raw_ostream &OS, unsigned Indent) const {
334   size_t MaxEncodingStringLength = 0;
335 #define HANDLE_DW_LLE(ID, NAME)                                                \
336   MaxEncodingStringLength = std::max(MaxEncodingStringLength,                  \
337                                      dwarf::LocListEncodingString(ID).size());
338 #include "llvm/BinaryFormat/Dwarf.def"
339 
340   OS << "\n";
341   OS.indent(Indent);
342   StringRef EncodingString = dwarf::LocListEncodingString(Entry.Kind);
343   // Unsupported encodings should have been reported during parsing.
344   assert(!EncodingString.empty() && "Unknown loclist entry encoding");
345   OS << format("%-*s(", MaxEncodingStringLength, EncodingString.data());
346   unsigned FieldSize = 2 + 2 * Data.getAddressSize();
347   switch (Entry.Kind) {
348   case dwarf::DW_LLE_startx_length:
349   case dwarf::DW_LLE_start_length:
350   case dwarf::DW_LLE_offset_pair:
351     OS << format_hex(Entry.Value0, FieldSize) << ", "
352        << format_hex(Entry.Value1, FieldSize);
353     break;
354   case dwarf::DW_LLE_base_addressx:
355   case dwarf::DW_LLE_base_address:
356     OS << format_hex(Entry.Value0, FieldSize);
357     break;
358   case dwarf::DW_LLE_end_of_list:
359     break;
360   }
361   OS << ')';
362 }
363 
364 void DWARFDebugLoclists::dumpRange(uint64_t StartOffset, uint64_t Size,
365                                    raw_ostream &OS, const MCRegisterInfo *MRI,
366                                    DIDumpOptions DumpOpts) {
367   if (!Data.isValidOffsetForDataOfSize(StartOffset, Size))  {
368     OS << "Invalid dump range\n";
369     return;
370   }
371   uint64_t Offset = StartOffset;
372   StringRef Separator;
373   bool CanContinue = true;
374   while (CanContinue && Offset < StartOffset + Size) {
375     OS << Separator;
376     Separator = "\n";
377 
378     CanContinue = dumpLocationList(&Offset, OS, /*BaseAddr=*/None, MRI, nullptr,
379                                    DumpOpts, /*Indent=*/12);
380     OS << '\n';
381   }
382 }
383