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_endx: {
61     Optional<SectionedAddress> LowPC = LookupAddr(E.Value0);
62     if (!LowPC)
63       return createResolverError(E.Value0, E.Kind);
64     Optional<SectionedAddress> HighPC = LookupAddr(E.Value1);
65     if (!HighPC)
66       return createResolverError(E.Value1, E.Kind);
67     return DWARFLocationExpression{
68         DWARFAddressRange{LowPC->Address, HighPC->Address, LowPC->SectionIndex},
69         E.Loc};
70   }
71   case dwarf::DW_LLE_startx_length: {
72     Optional<SectionedAddress> LowPC = LookupAddr(E.Value0);
73     if (!LowPC)
74       return createResolverError(E.Value0, E.Kind);
75     return DWARFLocationExpression{DWARFAddressRange{LowPC->Address,
76                                                      LowPC->Address + E.Value1,
77                                                      LowPC->SectionIndex},
78                                    E.Loc};
79   }
80   case dwarf::DW_LLE_offset_pair: {
81     if (!Base) {
82       return createStringError(inconvertibleErrorCode(),
83                                "Unable to resolve location list offset pair: "
84                                "Base address not defined");
85     }
86     DWARFAddressRange Range{Base->Address + E.Value0, Base->Address + E.Value1,
87                             Base->SectionIndex};
88     if (Range.SectionIndex == SectionedAddress::UndefSection)
89       Range.SectionIndex = E.SectionIndex;
90     return DWARFLocationExpression{Range, E.Loc};
91   }
92   case dwarf::DW_LLE_default_location:
93     return DWARFLocationExpression{None, E.Loc};
94   case dwarf::DW_LLE_base_address:
95     Base = SectionedAddress{E.Value0, E.SectionIndex};
96     return None;
97   case dwarf::DW_LLE_start_end:
98     return DWARFLocationExpression{
99         DWARFAddressRange{E.Value0, E.Value1, E.SectionIndex}, E.Loc};
100   case dwarf::DW_LLE_start_length:
101     return DWARFLocationExpression{
102         DWARFAddressRange{E.Value0, E.Value0 + E.Value1, E.SectionIndex},
103         E.Loc};
104   default:
105     llvm_unreachable("unreachable locations list kind");
106   }
107 }
108 
109 // When directly dumping the .debug_loc without a compile unit, we have to guess
110 // at the DWARF version. This only affects DW_OP_call_ref, which is a rare
111 // expression that LLVM doesn't produce. Guessing the wrong version means we
112 // won't be able to pretty print expressions in DWARF2 binaries produced by
113 // non-LLVM tools.
114 static void dumpExpression(raw_ostream &OS, ArrayRef<uint8_t> Data,
115                            bool IsLittleEndian, unsigned AddressSize,
116                            const MCRegisterInfo *MRI, DWARFUnit *U) {
117   DWARFDataExtractor Extractor(toStringRef(Data), IsLittleEndian, AddressSize);
118   DWARFExpression(Extractor, dwarf::DWARF_VERSION, AddressSize).print(OS, MRI, U);
119 }
120 
121 bool DWARFLocationTable::dumpLocationList(uint64_t *Offset, raw_ostream &OS,
122                                           Optional<SectionedAddress> BaseAddr,
123                                           const MCRegisterInfo *MRI,
124                                           const DWARFObject &Obj, DWARFUnit *U,
125                                           DIDumpOptions DumpOpts,
126                                           unsigned Indent) const {
127   DWARFLocationInterpreter Interp(
128       BaseAddr, [U](uint32_t Index) -> Optional<SectionedAddress> {
129         if (U)
130           return U->getAddrOffsetSectionItem(Index);
131         return None;
132       });
133   OS << format("0x%8.8" PRIx64 ": ", *Offset);
134   Error E = visitLocationList(Offset, [&](const DWARFLocationEntry &E) {
135     Expected<Optional<DWARFLocationExpression>> Loc = Interp.Interpret(E);
136     if (!Loc || DumpOpts.DisplayRawContents)
137       dumpRawEntry(E, OS, Indent, DumpOpts, Obj);
138     if (Loc && *Loc) {
139       OS << "\n";
140       OS.indent(Indent);
141       if (DumpOpts.DisplayRawContents)
142         OS << "          => ";
143 
144       DIDumpOptions RangeDumpOpts(DumpOpts);
145       RangeDumpOpts.DisplayRawContents = false;
146       if (Loc.get()->Range)
147         Loc.get()->Range->dump(OS, Data.getAddressSize(), RangeDumpOpts, &Obj);
148       else
149         OS << "<default>";
150     }
151     if (!Loc)
152       consumeError(Loc.takeError());
153 
154     if (E.Kind != dwarf::DW_LLE_base_address &&
155         E.Kind != dwarf::DW_LLE_base_addressx &&
156         E.Kind != dwarf::DW_LLE_end_of_list) {
157       OS << ": ";
158       dumpExpression(OS, E.Loc, Data.isLittleEndian(), Data.getAddressSize(),
159                      MRI, U);
160     }
161     return true;
162   });
163   if (E) {
164     OS << "\n";
165     OS.indent(Indent);
166     OS << "error: " << toString(std::move(E));
167     return false;
168   }
169   return true;
170 }
171 
172 Error DWARFLocationTable::visitAbsoluteLocationList(
173     uint64_t Offset, Optional<SectionedAddress> BaseAddr,
174     std::function<Optional<SectionedAddress>(uint32_t)> LookupAddr,
175     function_ref<bool(Expected<DWARFLocationExpression>)> Callback) const {
176   DWARFLocationInterpreter Interp(BaseAddr, std::move(LookupAddr));
177   return visitLocationList(&Offset, [&](const DWARFLocationEntry &E) {
178     Expected<Optional<DWARFLocationExpression>> Loc = Interp.Interpret(E);
179     if (!Loc)
180       return Callback(Loc.takeError());
181     if (*Loc)
182       return Callback(**Loc);
183     return true;
184   });
185 }
186 
187 void DWARFDebugLoc::dump(raw_ostream &OS, const MCRegisterInfo *MRI,
188                          const DWARFObject &Obj, DIDumpOptions DumpOpts,
189                          Optional<uint64_t> DumpOffset) const {
190   auto BaseAddr = None;
191   unsigned Indent = 12;
192   if (DumpOffset) {
193     dumpLocationList(&*DumpOffset, OS, BaseAddr, MRI, Obj, nullptr, DumpOpts,
194                      Indent);
195   } else {
196     uint64_t Offset = 0;
197     StringRef Separator;
198     bool CanContinue = true;
199     while (CanContinue && Data.isValidOffset(Offset)) {
200       OS << Separator;
201       Separator = "\n";
202 
203       CanContinue = dumpLocationList(&Offset, OS, BaseAddr, MRI, Obj, nullptr,
204                                      DumpOpts, Indent);
205       OS << '\n';
206     }
207   }
208 }
209 
210 Error DWARFDebugLoc::visitLocationList(
211     uint64_t *Offset,
212     function_ref<bool(const DWARFLocationEntry &)> Callback) const {
213   DataExtractor::Cursor C(*Offset);
214   while (true) {
215     uint64_t SectionIndex;
216     uint64_t Value0 = Data.getRelocatedAddress(C);
217     uint64_t Value1 = Data.getRelocatedAddress(C, &SectionIndex);
218 
219     DWARFLocationEntry E;
220 
221     // The end of any given location list is marked by an end of list entry,
222     // which consists of a 0 for the beginning address offset and a 0 for the
223     // ending address offset. A beginning offset of 0xff...f marks the base
224     // address selection entry.
225     if (Value0 == 0 && Value1 == 0) {
226       E.Kind = dwarf::DW_LLE_end_of_list;
227     } else if (Value0 == (Data.getAddressSize() == 4 ? -1U : -1ULL)) {
228       E.Kind = dwarf::DW_LLE_base_address;
229       E.Value0 = Value1;
230       E.SectionIndex = SectionIndex;
231     } else {
232       E.Kind = dwarf::DW_LLE_offset_pair;
233       E.Value0 = Value0;
234       E.Value1 = Value1;
235       E.SectionIndex = SectionIndex;
236       unsigned Bytes = Data.getU16(C);
237       // A single location description describing the location of the object...
238       Data.getU8(C, E.Loc, Bytes);
239     }
240 
241     if (!C)
242       return C.takeError();
243     if (!Callback(E) || E.Kind == dwarf::DW_LLE_end_of_list)
244       break;
245   }
246   *Offset = C.tell();
247   return Error::success();
248 }
249 
250 void DWARFDebugLoc::dumpRawEntry(const DWARFLocationEntry &Entry,
251                                  raw_ostream &OS, unsigned Indent,
252                                  DIDumpOptions DumpOpts,
253                                  const DWARFObject &Obj) const {
254   uint64_t Value0, Value1;
255   switch (Entry.Kind) {
256   case dwarf::DW_LLE_base_address:
257     Value0 = Data.getAddressSize() == 4 ? -1U : -1ULL;
258     Value1 = Entry.Value0;
259     break;
260   case dwarf::DW_LLE_offset_pair:
261     Value0 = Entry.Value0;
262     Value1 = Entry.Value1;
263     break;
264   case dwarf::DW_LLE_end_of_list:
265     Value0 = Value1 = 0;
266     return;
267   default:
268     llvm_unreachable("Not possible in DWARF4!");
269   }
270   OS << '\n';
271   OS.indent(Indent);
272   OS << '(' << format_hex(Value0, 2 + Data.getAddressSize() * 2) << ", "
273      << format_hex(Value1, 2 + Data.getAddressSize() * 2) << ')';
274   DWARFFormValue::dumpAddressSection(Obj, OS, DumpOpts, Entry.SectionIndex);
275 }
276 
277 Error DWARFDebugLoclists::visitLocationList(
278     uint64_t *Offset, function_ref<bool(const DWARFLocationEntry &)> F) const {
279 
280   DataExtractor::Cursor C(*Offset);
281   bool Continue = true;
282   while (Continue) {
283     DWARFLocationEntry E;
284     E.Kind = Data.getU8(C);
285     switch (E.Kind) {
286     case dwarf::DW_LLE_end_of_list:
287       break;
288     case dwarf::DW_LLE_base_addressx:
289       E.Value0 = Data.getULEB128(C);
290       break;
291     case dwarf::DW_LLE_startx_endx:
292       E.Value0 = Data.getULEB128(C);
293       E.Value1 = Data.getULEB128(C);
294       break;
295     case dwarf::DW_LLE_startx_length:
296       E.Value0 = Data.getULEB128(C);
297       // Pre-DWARF 5 has different interpretation of the length field. We have
298       // to support both pre- and standartized styles for the compatibility.
299       if (Version < 5)
300         E.Value1 = Data.getU32(C);
301       else
302         E.Value1 = Data.getULEB128(C);
303       break;
304     case dwarf::DW_LLE_offset_pair:
305       E.Value0 = Data.getULEB128(C);
306       E.Value1 = Data.getULEB128(C);
307       E.SectionIndex = SectionedAddress::UndefSection;
308       break;
309     case dwarf::DW_LLE_default_location:
310       break;
311     case dwarf::DW_LLE_base_address:
312       E.Value0 = Data.getRelocatedAddress(C, &E.SectionIndex);
313       break;
314     case dwarf::DW_LLE_start_end:
315       E.Value0 = Data.getRelocatedAddress(C, &E.SectionIndex);
316       E.Value1 = Data.getRelocatedAddress(C);
317       break;
318     case dwarf::DW_LLE_start_length:
319       E.Value0 = Data.getRelocatedAddress(C, &E.SectionIndex);
320       E.Value1 = Data.getULEB128(C);
321       break;
322     default:
323       cantFail(C.takeError());
324       return createStringError(errc::illegal_byte_sequence,
325                                "LLE of kind %x not supported", (int)E.Kind);
326     }
327 
328     if (E.Kind != dwarf::DW_LLE_base_address &&
329         E.Kind != dwarf::DW_LLE_base_addressx &&
330         E.Kind != dwarf::DW_LLE_end_of_list) {
331       unsigned Bytes = Version >= 5 ? Data.getULEB128(C) : Data.getU16(C);
332       // A single location description describing the location of the object...
333       Data.getU8(C, E.Loc, Bytes);
334     }
335 
336     if (!C)
337       return C.takeError();
338     Continue = F(E) && E.Kind != dwarf::DW_LLE_end_of_list;
339   }
340   *Offset = C.tell();
341   return Error::success();
342 }
343 
344 void DWARFDebugLoclists::dumpRawEntry(const DWARFLocationEntry &Entry,
345                                       raw_ostream &OS, unsigned Indent,
346                                       DIDumpOptions DumpOpts,
347                                       const DWARFObject &Obj) const {
348   size_t MaxEncodingStringLength = 0;
349 #define HANDLE_DW_LLE(ID, NAME)                                                \
350   MaxEncodingStringLength = std::max(MaxEncodingStringLength,                  \
351                                      dwarf::LocListEncodingString(ID).size());
352 #include "llvm/BinaryFormat/Dwarf.def"
353 
354   OS << "\n";
355   OS.indent(Indent);
356   StringRef EncodingString = dwarf::LocListEncodingString(Entry.Kind);
357   // Unsupported encodings should have been reported during parsing.
358   assert(!EncodingString.empty() && "Unknown loclist entry encoding");
359   OS << format("%-*s(", MaxEncodingStringLength, EncodingString.data());
360   unsigned FieldSize = 2 + 2 * Data.getAddressSize();
361   switch (Entry.Kind) {
362   case dwarf::DW_LLE_end_of_list:
363   case dwarf::DW_LLE_default_location:
364     break;
365   case dwarf::DW_LLE_startx_endx:
366   case dwarf::DW_LLE_startx_length:
367   case dwarf::DW_LLE_offset_pair:
368   case dwarf::DW_LLE_start_end:
369   case dwarf::DW_LLE_start_length:
370     OS << format_hex(Entry.Value0, FieldSize) << ", "
371        << format_hex(Entry.Value1, FieldSize);
372     break;
373   case dwarf::DW_LLE_base_addressx:
374   case dwarf::DW_LLE_base_address:
375     OS << format_hex(Entry.Value0, FieldSize);
376     break;
377   }
378   OS << ')';
379   switch (Entry.Kind) {
380   case dwarf::DW_LLE_base_address:
381   case dwarf::DW_LLE_start_end:
382   case dwarf::DW_LLE_start_length:
383     DWARFFormValue::dumpAddressSection(Obj, OS, DumpOpts, Entry.SectionIndex);
384     break;
385   default:
386     break;
387   }
388 }
389 
390 void DWARFDebugLoclists::dumpRange(uint64_t StartOffset, uint64_t Size,
391                                    raw_ostream &OS, const MCRegisterInfo *MRI,
392                                    const DWARFObject &Obj,
393                                    DIDumpOptions DumpOpts) {
394   if (!Data.isValidOffsetForDataOfSize(StartOffset, Size))  {
395     OS << "Invalid dump range\n";
396     return;
397   }
398   uint64_t Offset = StartOffset;
399   StringRef Separator;
400   bool CanContinue = true;
401   while (CanContinue && Offset < StartOffset + Size) {
402     OS << Separator;
403     Separator = "\n";
404 
405     CanContinue = dumpLocationList(&Offset, OS, /*BaseAddr=*/None, MRI, Obj,
406                                    nullptr, DumpOpts, /*Indent=*/12);
407     OS << '\n';
408   }
409 }
410