xref: /freebsd/contrib/llvm-project/llvm/lib/DebugInfo/DWARF/DWARFVerifier.cpp (revision 77013d11e6483b970af25e13c9b892075742f7e5)
1 //===- DWARFVerifier.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 #include "llvm/DebugInfo/DWARF/DWARFVerifier.h"
9 #include "llvm/ADT/SmallSet.h"
10 #include "llvm/DebugInfo/DWARF/DWARFCompileUnit.h"
11 #include "llvm/DebugInfo/DWARF/DWARFContext.h"
12 #include "llvm/DebugInfo/DWARF/DWARFDebugLine.h"
13 #include "llvm/DebugInfo/DWARF/DWARFDie.h"
14 #include "llvm/DebugInfo/DWARF/DWARFExpression.h"
15 #include "llvm/DebugInfo/DWARF/DWARFFormValue.h"
16 #include "llvm/DebugInfo/DWARF/DWARFSection.h"
17 #include "llvm/Support/DJB.h"
18 #include "llvm/Support/FormatVariadic.h"
19 #include "llvm/Support/WithColor.h"
20 #include "llvm/Support/raw_ostream.h"
21 #include <map>
22 #include <set>
23 #include <vector>
24 
25 using namespace llvm;
26 using namespace dwarf;
27 using namespace object;
28 
29 Optional<DWARFAddressRange>
30 DWARFVerifier::DieRangeInfo::insert(const DWARFAddressRange &R) {
31   auto Begin = Ranges.begin();
32   auto End = Ranges.end();
33   auto Pos = std::lower_bound(Begin, End, R);
34 
35   if (Pos != End) {
36     DWARFAddressRange Range(*Pos);
37     if (Pos->merge(R))
38       return Range;
39   }
40   if (Pos != Begin) {
41     auto Iter = Pos - 1;
42     DWARFAddressRange Range(*Iter);
43     if (Iter->merge(R))
44       return Range;
45   }
46 
47   Ranges.insert(Pos, R);
48   return None;
49 }
50 
51 DWARFVerifier::DieRangeInfo::die_range_info_iterator
52 DWARFVerifier::DieRangeInfo::insert(const DieRangeInfo &RI) {
53   auto End = Children.end();
54   auto Iter = Children.begin();
55   while (Iter != End) {
56     if (Iter->intersects(RI))
57       return Iter;
58     ++Iter;
59   }
60   Children.insert(RI);
61   return Children.end();
62 }
63 
64 bool DWARFVerifier::DieRangeInfo::contains(const DieRangeInfo &RHS) const {
65   auto I1 = Ranges.begin(), E1 = Ranges.end();
66   auto I2 = RHS.Ranges.begin(), E2 = RHS.Ranges.end();
67   if (I2 == E2)
68     return true;
69 
70   DWARFAddressRange R = *I2;
71   while (I1 != E1) {
72     bool Covered = I1->LowPC <= R.LowPC;
73     if (R.LowPC == R.HighPC || (Covered && R.HighPC <= I1->HighPC)) {
74       if (++I2 == E2)
75         return true;
76       R = *I2;
77       continue;
78     }
79     if (!Covered)
80       return false;
81     if (R.LowPC < I1->HighPC)
82       R.LowPC = I1->HighPC;
83     ++I1;
84   }
85   return false;
86 }
87 
88 bool DWARFVerifier::DieRangeInfo::intersects(const DieRangeInfo &RHS) const {
89   auto I1 = Ranges.begin(), E1 = Ranges.end();
90   auto I2 = RHS.Ranges.begin(), E2 = RHS.Ranges.end();
91   while (I1 != E1 && I2 != E2) {
92     if (I1->intersects(*I2))
93       return true;
94     if (I1->LowPC < I2->LowPC)
95       ++I1;
96     else
97       ++I2;
98   }
99   return false;
100 }
101 
102 bool DWARFVerifier::verifyUnitHeader(const DWARFDataExtractor DebugInfoData,
103                                      uint64_t *Offset, unsigned UnitIndex,
104                                      uint8_t &UnitType, bool &isUnitDWARF64) {
105   uint64_t AbbrOffset, Length;
106   uint8_t AddrSize = 0;
107   uint16_t Version;
108   bool Success = true;
109 
110   bool ValidLength = false;
111   bool ValidVersion = false;
112   bool ValidAddrSize = false;
113   bool ValidType = true;
114   bool ValidAbbrevOffset = true;
115 
116   uint64_t OffsetStart = *Offset;
117   DwarfFormat Format;
118   std::tie(Length, Format) = DebugInfoData.getInitialLength(Offset);
119   isUnitDWARF64 = Format == DWARF64;
120   Version = DebugInfoData.getU16(Offset);
121 
122   if (Version >= 5) {
123     UnitType = DebugInfoData.getU8(Offset);
124     AddrSize = DebugInfoData.getU8(Offset);
125     AbbrOffset = isUnitDWARF64 ? DebugInfoData.getU64(Offset) : DebugInfoData.getU32(Offset);
126     ValidType = dwarf::isUnitType(UnitType);
127   } else {
128     UnitType = 0;
129     AbbrOffset = isUnitDWARF64 ? DebugInfoData.getU64(Offset) : DebugInfoData.getU32(Offset);
130     AddrSize = DebugInfoData.getU8(Offset);
131   }
132 
133   if (!DCtx.getDebugAbbrev()->getAbbreviationDeclarationSet(AbbrOffset))
134     ValidAbbrevOffset = false;
135 
136   ValidLength = DebugInfoData.isValidOffset(OffsetStart + Length + 3);
137   ValidVersion = DWARFContext::isSupportedVersion(Version);
138   ValidAddrSize = DWARFContext::isAddressSizeSupported(AddrSize);
139   if (!ValidLength || !ValidVersion || !ValidAddrSize || !ValidAbbrevOffset ||
140       !ValidType) {
141     Success = false;
142     error() << format("Units[%d] - start offset: 0x%08" PRIx64 " \n", UnitIndex,
143                       OffsetStart);
144     if (!ValidLength)
145       note() << "The length for this unit is too "
146                 "large for the .debug_info provided.\n";
147     if (!ValidVersion)
148       note() << "The 16 bit unit header version is not valid.\n";
149     if (!ValidType)
150       note() << "The unit type encoding is not valid.\n";
151     if (!ValidAbbrevOffset)
152       note() << "The offset into the .debug_abbrev section is "
153                 "not valid.\n";
154     if (!ValidAddrSize)
155       note() << "The address size is unsupported.\n";
156   }
157   *Offset = OffsetStart + Length + (isUnitDWARF64 ? 12 : 4);
158   return Success;
159 }
160 
161 unsigned DWARFVerifier::verifyUnitContents(DWARFUnit &Unit) {
162   unsigned NumUnitErrors = 0;
163   unsigned NumDies = Unit.getNumDIEs();
164   for (unsigned I = 0; I < NumDies; ++I) {
165     auto Die = Unit.getDIEAtIndex(I);
166 
167     if (Die.getTag() == DW_TAG_null)
168       continue;
169 
170     for (auto AttrValue : Die.attributes()) {
171       NumUnitErrors += verifyDebugInfoAttribute(Die, AttrValue);
172       NumUnitErrors += verifyDebugInfoForm(Die, AttrValue);
173     }
174 
175     if (Die.hasChildren()) {
176       if (Die.getFirstChild().isValid() &&
177           Die.getFirstChild().getTag() == DW_TAG_null) {
178         warn() << dwarf::TagString(Die.getTag())
179                << " has DW_CHILDREN_yes but DIE has no children: ";
180         Die.dump(OS);
181       }
182     }
183 
184     NumUnitErrors += verifyDebugInfoCallSite(Die);
185   }
186 
187   DWARFDie Die = Unit.getUnitDIE(/* ExtractUnitDIEOnly = */ false);
188   if (!Die) {
189     error() << "Compilation unit without DIE.\n";
190     NumUnitErrors++;
191     return NumUnitErrors;
192   }
193 
194   if (!dwarf::isUnitType(Die.getTag())) {
195     error() << "Compilation unit root DIE is not a unit DIE: "
196             << dwarf::TagString(Die.getTag()) << ".\n";
197     NumUnitErrors++;
198   }
199 
200   uint8_t UnitType = Unit.getUnitType();
201   if (!DWARFUnit::isMatchingUnitTypeAndTag(UnitType, Die.getTag())) {
202     error() << "Compilation unit type (" << dwarf::UnitTypeString(UnitType)
203             << ") and root DIE (" << dwarf::TagString(Die.getTag())
204             << ") do not match.\n";
205     NumUnitErrors++;
206   }
207 
208   //  According to DWARF Debugging Information Format Version 5,
209   //  3.1.2 Skeleton Compilation Unit Entries:
210   //  "A skeleton compilation unit has no children."
211   if (Die.getTag() == dwarf::DW_TAG_skeleton_unit && Die.hasChildren()) {
212     error() << "Skeleton compilation unit has children.\n";
213     NumUnitErrors++;
214   }
215 
216   DieRangeInfo RI;
217   NumUnitErrors += verifyDieRanges(Die, RI);
218 
219   return NumUnitErrors;
220 }
221 
222 unsigned DWARFVerifier::verifyDebugInfoCallSite(const DWARFDie &Die) {
223   if (Die.getTag() != DW_TAG_call_site && Die.getTag() != DW_TAG_GNU_call_site)
224     return 0;
225 
226   DWARFDie Curr = Die.getParent();
227   for (; Curr.isValid() && !Curr.isSubprogramDIE(); Curr = Die.getParent()) {
228     if (Curr.getTag() == DW_TAG_inlined_subroutine) {
229       error() << "Call site entry nested within inlined subroutine:";
230       Curr.dump(OS);
231       return 1;
232     }
233   }
234 
235   if (!Curr.isValid()) {
236     error() << "Call site entry not nested within a valid subprogram:";
237     Die.dump(OS);
238     return 1;
239   }
240 
241   Optional<DWARFFormValue> CallAttr =
242       Curr.find({DW_AT_call_all_calls, DW_AT_call_all_source_calls,
243                  DW_AT_call_all_tail_calls, DW_AT_GNU_all_call_sites,
244                  DW_AT_GNU_all_source_call_sites,
245                  DW_AT_GNU_all_tail_call_sites});
246   if (!CallAttr) {
247     error() << "Subprogram with call site entry has no DW_AT_call attribute:";
248     Curr.dump(OS);
249     Die.dump(OS, /*indent*/ 1);
250     return 1;
251   }
252 
253   return 0;
254 }
255 
256 unsigned DWARFVerifier::verifyAbbrevSection(const DWARFDebugAbbrev *Abbrev) {
257   unsigned NumErrors = 0;
258   if (Abbrev) {
259     const DWARFAbbreviationDeclarationSet *AbbrDecls =
260         Abbrev->getAbbreviationDeclarationSet(0);
261     for (auto AbbrDecl : *AbbrDecls) {
262       SmallDenseSet<uint16_t> AttributeSet;
263       for (auto Attribute : AbbrDecl.attributes()) {
264         auto Result = AttributeSet.insert(Attribute.Attr);
265         if (!Result.second) {
266           error() << "Abbreviation declaration contains multiple "
267                   << AttributeString(Attribute.Attr) << " attributes.\n";
268           AbbrDecl.dump(OS);
269           ++NumErrors;
270         }
271       }
272     }
273   }
274   return NumErrors;
275 }
276 
277 bool DWARFVerifier::handleDebugAbbrev() {
278   OS << "Verifying .debug_abbrev...\n";
279 
280   const DWARFObject &DObj = DCtx.getDWARFObj();
281   unsigned NumErrors = 0;
282   if (!DObj.getAbbrevSection().empty())
283     NumErrors += verifyAbbrevSection(DCtx.getDebugAbbrev());
284   if (!DObj.getAbbrevDWOSection().empty())
285     NumErrors += verifyAbbrevSection(DCtx.getDebugAbbrevDWO());
286 
287   return NumErrors == 0;
288 }
289 
290 unsigned DWARFVerifier::verifyUnitSection(const DWARFSection &S,
291                                           DWARFSectionKind SectionKind) {
292   const DWARFObject &DObj = DCtx.getDWARFObj();
293   DWARFDataExtractor DebugInfoData(DObj, S, DCtx.isLittleEndian(), 0);
294   unsigned NumDebugInfoErrors = 0;
295   uint64_t OffsetStart = 0, Offset = 0, UnitIdx = 0;
296   uint8_t UnitType = 0;
297   bool isUnitDWARF64 = false;
298   bool isHeaderChainValid = true;
299   bool hasDIE = DebugInfoData.isValidOffset(Offset);
300   DWARFUnitVector TypeUnitVector;
301   DWARFUnitVector CompileUnitVector;
302   while (hasDIE) {
303     OffsetStart = Offset;
304     if (!verifyUnitHeader(DebugInfoData, &Offset, UnitIdx, UnitType,
305                           isUnitDWARF64)) {
306       isHeaderChainValid = false;
307       if (isUnitDWARF64)
308         break;
309     } else {
310       DWARFUnitHeader Header;
311       Header.extract(DCtx, DebugInfoData, &OffsetStart, SectionKind);
312       DWARFUnit *Unit;
313       switch (UnitType) {
314       case dwarf::DW_UT_type:
315       case dwarf::DW_UT_split_type: {
316         Unit = TypeUnitVector.addUnit(std::make_unique<DWARFTypeUnit>(
317             DCtx, S, Header, DCtx.getDebugAbbrev(), &DObj.getRangesSection(),
318             &DObj.getLocSection(), DObj.getStrSection(),
319             DObj.getStrOffsetsSection(), &DObj.getAddrSection(),
320             DObj.getLineSection(), DCtx.isLittleEndian(), false,
321             TypeUnitVector));
322         break;
323       }
324       case dwarf::DW_UT_skeleton:
325       case dwarf::DW_UT_split_compile:
326       case dwarf::DW_UT_compile:
327       case dwarf::DW_UT_partial:
328       // UnitType = 0 means that we are verifying a compile unit in DWARF v4.
329       case 0: {
330         Unit = CompileUnitVector.addUnit(std::make_unique<DWARFCompileUnit>(
331             DCtx, S, Header, DCtx.getDebugAbbrev(), &DObj.getRangesSection(),
332             &DObj.getLocSection(), DObj.getStrSection(),
333             DObj.getStrOffsetsSection(), &DObj.getAddrSection(),
334             DObj.getLineSection(), DCtx.isLittleEndian(), false,
335             CompileUnitVector));
336         break;
337       }
338       default: { llvm_unreachable("Invalid UnitType."); }
339       }
340       NumDebugInfoErrors += verifyUnitContents(*Unit);
341     }
342     hasDIE = DebugInfoData.isValidOffset(Offset);
343     ++UnitIdx;
344   }
345   if (UnitIdx == 0 && !hasDIE) {
346     warn() << "Section is empty.\n";
347     isHeaderChainValid = true;
348   }
349   if (!isHeaderChainValid)
350     ++NumDebugInfoErrors;
351   NumDebugInfoErrors += verifyDebugInfoReferences();
352   return NumDebugInfoErrors;
353 }
354 
355 bool DWARFVerifier::handleDebugInfo() {
356   const DWARFObject &DObj = DCtx.getDWARFObj();
357   unsigned NumErrors = 0;
358 
359   OS << "Verifying .debug_info Unit Header Chain...\n";
360   DObj.forEachInfoSections([&](const DWARFSection &S) {
361     NumErrors += verifyUnitSection(S, DW_SECT_INFO);
362   });
363 
364   OS << "Verifying .debug_types Unit Header Chain...\n";
365   DObj.forEachTypesSections([&](const DWARFSection &S) {
366     NumErrors += verifyUnitSection(S, DW_SECT_EXT_TYPES);
367   });
368   return NumErrors == 0;
369 }
370 
371 unsigned DWARFVerifier::verifyDieRanges(const DWARFDie &Die,
372                                         DieRangeInfo &ParentRI) {
373   unsigned NumErrors = 0;
374 
375   if (!Die.isValid())
376     return NumErrors;
377 
378   auto RangesOrError = Die.getAddressRanges();
379   if (!RangesOrError) {
380     // FIXME: Report the error.
381     ++NumErrors;
382     llvm::consumeError(RangesOrError.takeError());
383     return NumErrors;
384   }
385 
386   DWARFAddressRangesVector Ranges = RangesOrError.get();
387   // Build RI for this DIE and check that ranges within this DIE do not
388   // overlap.
389   DieRangeInfo RI(Die);
390 
391   // TODO support object files better
392   //
393   // Some object file formats (i.e. non-MachO) support COMDAT.  ELF in
394   // particular does so by placing each function into a section.  The DWARF data
395   // for the function at that point uses a section relative DW_FORM_addrp for
396   // the DW_AT_low_pc and a DW_FORM_data4 for the offset as the DW_AT_high_pc.
397   // In such a case, when the Die is the CU, the ranges will overlap, and we
398   // will flag valid conflicting ranges as invalid.
399   //
400   // For such targets, we should read the ranges from the CU and partition them
401   // by the section id.  The ranges within a particular section should be
402   // disjoint, although the ranges across sections may overlap.  We would map
403   // the child die to the entity that it references and the section with which
404   // it is associated.  The child would then be checked against the range
405   // information for the associated section.
406   //
407   // For now, simply elide the range verification for the CU DIEs if we are
408   // processing an object file.
409 
410   if (!IsObjectFile || IsMachOObject || Die.getTag() != DW_TAG_compile_unit) {
411     bool DumpDieAfterError = false;
412     for (auto Range : Ranges) {
413       if (!Range.valid()) {
414         ++NumErrors;
415         error() << "Invalid address range " << Range << "\n";
416         DumpDieAfterError = true;
417         continue;
418       }
419 
420       // Verify that ranges don't intersect and also build up the DieRangeInfo
421       // address ranges. Don't break out of the loop below early, or we will
422       // think this DIE doesn't have all of the address ranges it is supposed
423       // to have. Compile units often have DW_AT_ranges that can contain one or
424       // more dead stripped address ranges which tend to all be at the same
425       // address: 0 or -1.
426       if (auto PrevRange = RI.insert(Range)) {
427         ++NumErrors;
428         error() << "DIE has overlapping ranges in DW_AT_ranges attribute: "
429                 << *PrevRange << " and " << Range << '\n';
430         DumpDieAfterError = true;
431       }
432     }
433     if (DumpDieAfterError)
434       dump(Die, 2) << '\n';
435   }
436 
437   // Verify that children don't intersect.
438   const auto IntersectingChild = ParentRI.insert(RI);
439   if (IntersectingChild != ParentRI.Children.end()) {
440     ++NumErrors;
441     error() << "DIEs have overlapping address ranges:";
442     dump(Die);
443     dump(IntersectingChild->Die) << '\n';
444   }
445 
446   // Verify that ranges are contained within their parent.
447   bool ShouldBeContained = !Ranges.empty() && !ParentRI.Ranges.empty() &&
448                            !(Die.getTag() == DW_TAG_subprogram &&
449                              ParentRI.Die.getTag() == DW_TAG_subprogram);
450   if (ShouldBeContained && !ParentRI.contains(RI)) {
451     ++NumErrors;
452     error() << "DIE address ranges are not contained in its parent's ranges:";
453     dump(ParentRI.Die);
454     dump(Die, 2) << '\n';
455   }
456 
457   // Recursively check children.
458   for (DWARFDie Child : Die)
459     NumErrors += verifyDieRanges(Child, RI);
460 
461   return NumErrors;
462 }
463 
464 unsigned DWARFVerifier::verifyDebugInfoAttribute(const DWARFDie &Die,
465                                                  DWARFAttribute &AttrValue) {
466   unsigned NumErrors = 0;
467   auto ReportError = [&](const Twine &TitleMsg) {
468     ++NumErrors;
469     error() << TitleMsg << '\n';
470     dump(Die) << '\n';
471   };
472 
473   const DWARFObject &DObj = DCtx.getDWARFObj();
474   const auto Attr = AttrValue.Attr;
475   switch (Attr) {
476   case DW_AT_ranges:
477     // Make sure the offset in the DW_AT_ranges attribute is valid.
478     if (auto SectionOffset = AttrValue.Value.getAsSectionOffset()) {
479       unsigned DwarfVersion = Die.getDwarfUnit()->getVersion();
480       const DWARFSection &RangeSection = DwarfVersion < 5
481                                              ? DObj.getRangesSection()
482                                              : DObj.getRnglistsSection();
483       if (*SectionOffset >= RangeSection.Data.size())
484         ReportError(
485             "DW_AT_ranges offset is beyond " +
486             StringRef(DwarfVersion < 5 ? ".debug_ranges" : ".debug_rnglists") +
487             " bounds: " + llvm::formatv("{0:x8}", *SectionOffset));
488       break;
489     }
490     ReportError("DIE has invalid DW_AT_ranges encoding:");
491     break;
492   case DW_AT_stmt_list:
493     // Make sure the offset in the DW_AT_stmt_list attribute is valid.
494     if (auto SectionOffset = AttrValue.Value.getAsSectionOffset()) {
495       if (*SectionOffset >= DObj.getLineSection().Data.size())
496         ReportError("DW_AT_stmt_list offset is beyond .debug_line bounds: " +
497                     llvm::formatv("{0:x8}", *SectionOffset));
498       break;
499     }
500     ReportError("DIE has invalid DW_AT_stmt_list encoding:");
501     break;
502   case DW_AT_location: {
503     if (Expected<std::vector<DWARFLocationExpression>> Loc =
504             Die.getLocations(DW_AT_location)) {
505       DWARFUnit *U = Die.getDwarfUnit();
506       for (const auto &Entry : *Loc) {
507         DataExtractor Data(toStringRef(Entry.Expr), DCtx.isLittleEndian(), 0);
508         DWARFExpression Expression(Data, U->getAddressByteSize(),
509                                    U->getFormParams().Format);
510         bool Error = any_of(Expression, [](DWARFExpression::Operation &Op) {
511           return Op.isError();
512         });
513         if (Error || !Expression.verify(U))
514           ReportError("DIE contains invalid DWARF expression:");
515       }
516     } else
517       ReportError(toString(Loc.takeError()));
518     break;
519   }
520   case DW_AT_specification:
521   case DW_AT_abstract_origin: {
522     if (auto ReferencedDie = Die.getAttributeValueAsReferencedDie(Attr)) {
523       auto DieTag = Die.getTag();
524       auto RefTag = ReferencedDie.getTag();
525       if (DieTag == RefTag)
526         break;
527       if (DieTag == DW_TAG_inlined_subroutine && RefTag == DW_TAG_subprogram)
528         break;
529       if (DieTag == DW_TAG_variable && RefTag == DW_TAG_member)
530         break;
531       // This might be reference to a function declaration.
532       if (DieTag == DW_TAG_GNU_call_site && RefTag == DW_TAG_subprogram)
533         break;
534       ReportError("DIE with tag " + TagString(DieTag) + " has " +
535                   AttributeString(Attr) +
536                   " that points to DIE with "
537                   "incompatible tag " +
538                   TagString(RefTag));
539     }
540     break;
541   }
542   case DW_AT_type: {
543     DWARFDie TypeDie = Die.getAttributeValueAsReferencedDie(DW_AT_type);
544     if (TypeDie && !isType(TypeDie.getTag())) {
545       ReportError("DIE has " + AttributeString(Attr) +
546                   " with incompatible tag " + TagString(TypeDie.getTag()));
547     }
548     break;
549   }
550   case DW_AT_call_file:
551   case DW_AT_decl_file: {
552     if (auto FileIdx = AttrValue.Value.getAsUnsignedConstant()) {
553       DWARFUnit *U = Die.getDwarfUnit();
554       const auto *LT = U->getContext().getLineTableForUnit(U);
555       if (LT) {
556         if (!LT->hasFileAtIndex(*FileIdx)) {
557           bool IsZeroIndexed = LT->Prologue.getVersion() >= 5;
558           if (Optional<uint64_t> LastFileIdx = LT->getLastValidFileIndex()) {
559             ReportError("DIE has " + AttributeString(Attr) +
560                         " with an invalid file index " +
561                         llvm::formatv("{0}", *FileIdx) +
562                         " (valid values are [" + (IsZeroIndexed ? "0-" : "1-") +
563                         llvm::formatv("{0}", *LastFileIdx) + "])");
564           } else {
565             ReportError("DIE has " + AttributeString(Attr) +
566                         " with an invalid file index " +
567                         llvm::formatv("{0}", *FileIdx) +
568                         " (the file table in the prologue is empty)");
569           }
570         }
571       } else {
572         ReportError("DIE has " + AttributeString(Attr) +
573                     " that references a file with index " +
574                     llvm::formatv("{0}", *FileIdx) +
575                     " and the compile unit has no line table");
576       }
577     } else {
578       ReportError("DIE has " + AttributeString(Attr) +
579                   " with invalid encoding");
580     }
581     break;
582   }
583   default:
584     break;
585   }
586   return NumErrors;
587 }
588 
589 unsigned DWARFVerifier::verifyDebugInfoForm(const DWARFDie &Die,
590                                             DWARFAttribute &AttrValue) {
591   const DWARFObject &DObj = DCtx.getDWARFObj();
592   auto DieCU = Die.getDwarfUnit();
593   unsigned NumErrors = 0;
594   const auto Form = AttrValue.Value.getForm();
595   switch (Form) {
596   case DW_FORM_ref1:
597   case DW_FORM_ref2:
598   case DW_FORM_ref4:
599   case DW_FORM_ref8:
600   case DW_FORM_ref_udata: {
601     // Verify all CU relative references are valid CU offsets.
602     Optional<uint64_t> RefVal = AttrValue.Value.getAsReference();
603     assert(RefVal);
604     if (RefVal) {
605       auto CUSize = DieCU->getNextUnitOffset() - DieCU->getOffset();
606       auto CUOffset = AttrValue.Value.getRawUValue();
607       if (CUOffset >= CUSize) {
608         ++NumErrors;
609         error() << FormEncodingString(Form) << " CU offset "
610                 << format("0x%08" PRIx64, CUOffset)
611                 << " is invalid (must be less than CU size of "
612                 << format("0x%08" PRIx64, CUSize) << "):\n";
613         Die.dump(OS, 0, DumpOpts);
614         dump(Die) << '\n';
615       } else {
616         // Valid reference, but we will verify it points to an actual
617         // DIE later.
618         ReferenceToDIEOffsets[*RefVal].insert(Die.getOffset());
619       }
620     }
621     break;
622   }
623   case DW_FORM_ref_addr: {
624     // Verify all absolute DIE references have valid offsets in the
625     // .debug_info section.
626     Optional<uint64_t> RefVal = AttrValue.Value.getAsReference();
627     assert(RefVal);
628     if (RefVal) {
629       if (*RefVal >= DieCU->getInfoSection().Data.size()) {
630         ++NumErrors;
631         error() << "DW_FORM_ref_addr offset beyond .debug_info "
632                    "bounds:\n";
633         dump(Die) << '\n';
634       } else {
635         // Valid reference, but we will verify it points to an actual
636         // DIE later.
637         ReferenceToDIEOffsets[*RefVal].insert(Die.getOffset());
638       }
639     }
640     break;
641   }
642   case DW_FORM_strp: {
643     auto SecOffset = AttrValue.Value.getAsSectionOffset();
644     assert(SecOffset); // DW_FORM_strp is a section offset.
645     if (SecOffset && *SecOffset >= DObj.getStrSection().size()) {
646       ++NumErrors;
647       error() << "DW_FORM_strp offset beyond .debug_str bounds:\n";
648       dump(Die) << '\n';
649     }
650     break;
651   }
652   case DW_FORM_strx:
653   case DW_FORM_strx1:
654   case DW_FORM_strx2:
655   case DW_FORM_strx3:
656   case DW_FORM_strx4: {
657     auto Index = AttrValue.Value.getRawUValue();
658     auto DieCU = Die.getDwarfUnit();
659     // Check that we have a valid DWARF v5 string offsets table.
660     if (!DieCU->getStringOffsetsTableContribution()) {
661       ++NumErrors;
662       error() << FormEncodingString(Form)
663               << " used without a valid string offsets table:\n";
664       dump(Die) << '\n';
665       break;
666     }
667     // Check that the index is within the bounds of the section.
668     unsigned ItemSize = DieCU->getDwarfStringOffsetsByteSize();
669     // Use a 64-bit type to calculate the offset to guard against overflow.
670     uint64_t Offset =
671         (uint64_t)DieCU->getStringOffsetsBase() + Index * ItemSize;
672     if (DObj.getStrOffsetsSection().Data.size() < Offset + ItemSize) {
673       ++NumErrors;
674       error() << FormEncodingString(Form) << " uses index "
675               << format("%" PRIu64, Index) << ", which is too large:\n";
676       dump(Die) << '\n';
677       break;
678     }
679     // Check that the string offset is valid.
680     uint64_t StringOffset = *DieCU->getStringOffsetSectionItem(Index);
681     if (StringOffset >= DObj.getStrSection().size()) {
682       ++NumErrors;
683       error() << FormEncodingString(Form) << " uses index "
684               << format("%" PRIu64, Index)
685               << ", but the referenced string"
686                  " offset is beyond .debug_str bounds:\n";
687       dump(Die) << '\n';
688     }
689     break;
690   }
691   default:
692     break;
693   }
694   return NumErrors;
695 }
696 
697 unsigned DWARFVerifier::verifyDebugInfoReferences() {
698   // Take all references and make sure they point to an actual DIE by
699   // getting the DIE by offset and emitting an error
700   OS << "Verifying .debug_info references...\n";
701   unsigned NumErrors = 0;
702   for (const std::pair<const uint64_t, std::set<uint64_t>> &Pair :
703        ReferenceToDIEOffsets) {
704     if (DCtx.getDIEForOffset(Pair.first))
705       continue;
706     ++NumErrors;
707     error() << "invalid DIE reference " << format("0x%08" PRIx64, Pair.first)
708             << ". Offset is in between DIEs:\n";
709     for (auto Offset : Pair.second)
710       dump(DCtx.getDIEForOffset(Offset)) << '\n';
711     OS << "\n";
712   }
713   return NumErrors;
714 }
715 
716 void DWARFVerifier::verifyDebugLineStmtOffsets() {
717   std::map<uint64_t, DWARFDie> StmtListToDie;
718   for (const auto &CU : DCtx.compile_units()) {
719     auto Die = CU->getUnitDIE();
720     // Get the attribute value as a section offset. No need to produce an
721     // error here if the encoding isn't correct because we validate this in
722     // the .debug_info verifier.
723     auto StmtSectionOffset = toSectionOffset(Die.find(DW_AT_stmt_list));
724     if (!StmtSectionOffset)
725       continue;
726     const uint64_t LineTableOffset = *StmtSectionOffset;
727     auto LineTable = DCtx.getLineTableForUnit(CU.get());
728     if (LineTableOffset < DCtx.getDWARFObj().getLineSection().Data.size()) {
729       if (!LineTable) {
730         ++NumDebugLineErrors;
731         error() << ".debug_line[" << format("0x%08" PRIx64, LineTableOffset)
732                 << "] was not able to be parsed for CU:\n";
733         dump(Die) << '\n';
734         continue;
735       }
736     } else {
737       // Make sure we don't get a valid line table back if the offset is wrong.
738       assert(LineTable == nullptr);
739       // Skip this line table as it isn't valid. No need to create an error
740       // here because we validate this in the .debug_info verifier.
741       continue;
742     }
743     auto Iter = StmtListToDie.find(LineTableOffset);
744     if (Iter != StmtListToDie.end()) {
745       ++NumDebugLineErrors;
746       error() << "two compile unit DIEs, "
747               << format("0x%08" PRIx64, Iter->second.getOffset()) << " and "
748               << format("0x%08" PRIx64, Die.getOffset())
749               << ", have the same DW_AT_stmt_list section offset:\n";
750       dump(Iter->second);
751       dump(Die) << '\n';
752       // Already verified this line table before, no need to do it again.
753       continue;
754     }
755     StmtListToDie[LineTableOffset] = Die;
756   }
757 }
758 
759 void DWARFVerifier::verifyDebugLineRows() {
760   for (const auto &CU : DCtx.compile_units()) {
761     auto Die = CU->getUnitDIE();
762     auto LineTable = DCtx.getLineTableForUnit(CU.get());
763     // If there is no line table we will have created an error in the
764     // .debug_info verifier or in verifyDebugLineStmtOffsets().
765     if (!LineTable)
766       continue;
767 
768     // Verify prologue.
769     uint32_t MaxDirIndex = LineTable->Prologue.IncludeDirectories.size();
770     uint32_t FileIndex = 1;
771     StringMap<uint16_t> FullPathMap;
772     for (const auto &FileName : LineTable->Prologue.FileNames) {
773       // Verify directory index.
774       if (FileName.DirIdx > MaxDirIndex) {
775         ++NumDebugLineErrors;
776         error() << ".debug_line["
777                 << format("0x%08" PRIx64,
778                           *toSectionOffset(Die.find(DW_AT_stmt_list)))
779                 << "].prologue.file_names[" << FileIndex
780                 << "].dir_idx contains an invalid index: " << FileName.DirIdx
781                 << "\n";
782       }
783 
784       // Check file paths for duplicates.
785       std::string FullPath;
786       const bool HasFullPath = LineTable->getFileNameByIndex(
787           FileIndex, CU->getCompilationDir(),
788           DILineInfoSpecifier::FileLineInfoKind::AbsoluteFilePath, FullPath);
789       assert(HasFullPath && "Invalid index?");
790       (void)HasFullPath;
791       auto It = FullPathMap.find(FullPath);
792       if (It == FullPathMap.end())
793         FullPathMap[FullPath] = FileIndex;
794       else if (It->second != FileIndex) {
795         warn() << ".debug_line["
796                << format("0x%08" PRIx64,
797                          *toSectionOffset(Die.find(DW_AT_stmt_list)))
798                << "].prologue.file_names[" << FileIndex
799                << "] is a duplicate of file_names[" << It->second << "]\n";
800       }
801 
802       FileIndex++;
803     }
804 
805     // Verify rows.
806     uint64_t PrevAddress = 0;
807     uint32_t RowIndex = 0;
808     for (const auto &Row : LineTable->Rows) {
809       // Verify row address.
810       if (Row.Address.Address < PrevAddress) {
811         ++NumDebugLineErrors;
812         error() << ".debug_line["
813                 << format("0x%08" PRIx64,
814                           *toSectionOffset(Die.find(DW_AT_stmt_list)))
815                 << "] row[" << RowIndex
816                 << "] decreases in address from previous row:\n";
817 
818         DWARFDebugLine::Row::dumpTableHeader(OS, 0);
819         if (RowIndex > 0)
820           LineTable->Rows[RowIndex - 1].dump(OS);
821         Row.dump(OS);
822         OS << '\n';
823       }
824 
825       // Verify file index.
826       if (!LineTable->hasFileAtIndex(Row.File)) {
827         ++NumDebugLineErrors;
828         bool isDWARF5 = LineTable->Prologue.getVersion() >= 5;
829         error() << ".debug_line["
830                 << format("0x%08" PRIx64,
831                           *toSectionOffset(Die.find(DW_AT_stmt_list)))
832                 << "][" << RowIndex << "] has invalid file index " << Row.File
833                 << " (valid values are [" << (isDWARF5 ? "0," : "1,")
834                 << LineTable->Prologue.FileNames.size()
835                 << (isDWARF5 ? ")" : "]") << "):\n";
836         DWARFDebugLine::Row::dumpTableHeader(OS, 0);
837         Row.dump(OS);
838         OS << '\n';
839       }
840       if (Row.EndSequence)
841         PrevAddress = 0;
842       else
843         PrevAddress = Row.Address.Address;
844       ++RowIndex;
845     }
846   }
847 }
848 
849 DWARFVerifier::DWARFVerifier(raw_ostream &S, DWARFContext &D,
850                              DIDumpOptions DumpOpts)
851     : OS(S), DCtx(D), DumpOpts(std::move(DumpOpts)), IsObjectFile(false),
852       IsMachOObject(false) {
853   if (const auto *F = DCtx.getDWARFObj().getFile()) {
854     IsObjectFile = F->isRelocatableObject();
855     IsMachOObject = F->isMachO();
856   }
857 }
858 
859 bool DWARFVerifier::handleDebugLine() {
860   NumDebugLineErrors = 0;
861   OS << "Verifying .debug_line...\n";
862   verifyDebugLineStmtOffsets();
863   verifyDebugLineRows();
864   return NumDebugLineErrors == 0;
865 }
866 
867 unsigned DWARFVerifier::verifyAppleAccelTable(const DWARFSection *AccelSection,
868                                               DataExtractor *StrData,
869                                               const char *SectionName) {
870   unsigned NumErrors = 0;
871   DWARFDataExtractor AccelSectionData(DCtx.getDWARFObj(), *AccelSection,
872                                       DCtx.isLittleEndian(), 0);
873   AppleAcceleratorTable AccelTable(AccelSectionData, *StrData);
874 
875   OS << "Verifying " << SectionName << "...\n";
876 
877   // Verify that the fixed part of the header is not too short.
878   if (!AccelSectionData.isValidOffset(AccelTable.getSizeHdr())) {
879     error() << "Section is too small to fit a section header.\n";
880     return 1;
881   }
882 
883   // Verify that the section is not too short.
884   if (Error E = AccelTable.extract()) {
885     error() << toString(std::move(E)) << '\n';
886     return 1;
887   }
888 
889   // Verify that all buckets have a valid hash index or are empty.
890   uint32_t NumBuckets = AccelTable.getNumBuckets();
891   uint32_t NumHashes = AccelTable.getNumHashes();
892 
893   uint64_t BucketsOffset =
894       AccelTable.getSizeHdr() + AccelTable.getHeaderDataLength();
895   uint64_t HashesBase = BucketsOffset + NumBuckets * 4;
896   uint64_t OffsetsBase = HashesBase + NumHashes * 4;
897   for (uint32_t BucketIdx = 0; BucketIdx < NumBuckets; ++BucketIdx) {
898     uint32_t HashIdx = AccelSectionData.getU32(&BucketsOffset);
899     if (HashIdx >= NumHashes && HashIdx != UINT32_MAX) {
900       error() << format("Bucket[%d] has invalid hash index: %u.\n", BucketIdx,
901                         HashIdx);
902       ++NumErrors;
903     }
904   }
905   uint32_t NumAtoms = AccelTable.getAtomsDesc().size();
906   if (NumAtoms == 0) {
907     error() << "No atoms: failed to read HashData.\n";
908     return 1;
909   }
910   if (!AccelTable.validateForms()) {
911     error() << "Unsupported form: failed to read HashData.\n";
912     return 1;
913   }
914 
915   for (uint32_t HashIdx = 0; HashIdx < NumHashes; ++HashIdx) {
916     uint64_t HashOffset = HashesBase + 4 * HashIdx;
917     uint64_t DataOffset = OffsetsBase + 4 * HashIdx;
918     uint32_t Hash = AccelSectionData.getU32(&HashOffset);
919     uint64_t HashDataOffset = AccelSectionData.getU32(&DataOffset);
920     if (!AccelSectionData.isValidOffsetForDataOfSize(HashDataOffset,
921                                                      sizeof(uint64_t))) {
922       error() << format("Hash[%d] has invalid HashData offset: "
923                         "0x%08" PRIx64 ".\n",
924                         HashIdx, HashDataOffset);
925       ++NumErrors;
926     }
927 
928     uint64_t StrpOffset;
929     uint64_t StringOffset;
930     uint32_t StringCount = 0;
931     uint64_t Offset;
932     unsigned Tag;
933     while ((StrpOffset = AccelSectionData.getU32(&HashDataOffset)) != 0) {
934       const uint32_t NumHashDataObjects =
935           AccelSectionData.getU32(&HashDataOffset);
936       for (uint32_t HashDataIdx = 0; HashDataIdx < NumHashDataObjects;
937            ++HashDataIdx) {
938         std::tie(Offset, Tag) = AccelTable.readAtoms(&HashDataOffset);
939         auto Die = DCtx.getDIEForOffset(Offset);
940         if (!Die) {
941           const uint32_t BucketIdx =
942               NumBuckets ? (Hash % NumBuckets) : UINT32_MAX;
943           StringOffset = StrpOffset;
944           const char *Name = StrData->getCStr(&StringOffset);
945           if (!Name)
946             Name = "<NULL>";
947 
948           error() << format(
949               "%s Bucket[%d] Hash[%d] = 0x%08x "
950               "Str[%u] = 0x%08" PRIx64 " DIE[%d] = 0x%08" PRIx64 " "
951               "is not a valid DIE offset for \"%s\".\n",
952               SectionName, BucketIdx, HashIdx, Hash, StringCount, StrpOffset,
953               HashDataIdx, Offset, Name);
954 
955           ++NumErrors;
956           continue;
957         }
958         if ((Tag != dwarf::DW_TAG_null) && (Die.getTag() != Tag)) {
959           error() << "Tag " << dwarf::TagString(Tag)
960                   << " in accelerator table does not match Tag "
961                   << dwarf::TagString(Die.getTag()) << " of DIE[" << HashDataIdx
962                   << "].\n";
963           ++NumErrors;
964         }
965       }
966       ++StringCount;
967     }
968   }
969   return NumErrors;
970 }
971 
972 unsigned
973 DWARFVerifier::verifyDebugNamesCULists(const DWARFDebugNames &AccelTable) {
974   // A map from CU offset to the (first) Name Index offset which claims to index
975   // this CU.
976   DenseMap<uint64_t, uint64_t> CUMap;
977   const uint64_t NotIndexed = std::numeric_limits<uint64_t>::max();
978 
979   CUMap.reserve(DCtx.getNumCompileUnits());
980   for (const auto &CU : DCtx.compile_units())
981     CUMap[CU->getOffset()] = NotIndexed;
982 
983   unsigned NumErrors = 0;
984   for (const DWARFDebugNames::NameIndex &NI : AccelTable) {
985     if (NI.getCUCount() == 0) {
986       error() << formatv("Name Index @ {0:x} does not index any CU\n",
987                          NI.getUnitOffset());
988       ++NumErrors;
989       continue;
990     }
991     for (uint32_t CU = 0, End = NI.getCUCount(); CU < End; ++CU) {
992       uint64_t Offset = NI.getCUOffset(CU);
993       auto Iter = CUMap.find(Offset);
994 
995       if (Iter == CUMap.end()) {
996         error() << formatv(
997             "Name Index @ {0:x} references a non-existing CU @ {1:x}\n",
998             NI.getUnitOffset(), Offset);
999         ++NumErrors;
1000         continue;
1001       }
1002 
1003       if (Iter->second != NotIndexed) {
1004         error() << formatv("Name Index @ {0:x} references a CU @ {1:x}, but "
1005                            "this CU is already indexed by Name Index @ {2:x}\n",
1006                            NI.getUnitOffset(), Offset, Iter->second);
1007         continue;
1008       }
1009       Iter->second = NI.getUnitOffset();
1010     }
1011   }
1012 
1013   for (const auto &KV : CUMap) {
1014     if (KV.second == NotIndexed)
1015       warn() << formatv("CU @ {0:x} not covered by any Name Index\n", KV.first);
1016   }
1017 
1018   return NumErrors;
1019 }
1020 
1021 unsigned
1022 DWARFVerifier::verifyNameIndexBuckets(const DWARFDebugNames::NameIndex &NI,
1023                                       const DataExtractor &StrData) {
1024   struct BucketInfo {
1025     uint32_t Bucket;
1026     uint32_t Index;
1027 
1028     constexpr BucketInfo(uint32_t Bucket, uint32_t Index)
1029         : Bucket(Bucket), Index(Index) {}
1030     bool operator<(const BucketInfo &RHS) const { return Index < RHS.Index; }
1031   };
1032 
1033   uint32_t NumErrors = 0;
1034   if (NI.getBucketCount() == 0) {
1035     warn() << formatv("Name Index @ {0:x} does not contain a hash table.\n",
1036                       NI.getUnitOffset());
1037     return NumErrors;
1038   }
1039 
1040   // Build up a list of (Bucket, Index) pairs. We use this later to verify that
1041   // each Name is reachable from the appropriate bucket.
1042   std::vector<BucketInfo> BucketStarts;
1043   BucketStarts.reserve(NI.getBucketCount() + 1);
1044   for (uint32_t Bucket = 0, End = NI.getBucketCount(); Bucket < End; ++Bucket) {
1045     uint32_t Index = NI.getBucketArrayEntry(Bucket);
1046     if (Index > NI.getNameCount()) {
1047       error() << formatv("Bucket {0} of Name Index @ {1:x} contains invalid "
1048                          "value {2}. Valid range is [0, {3}].\n",
1049                          Bucket, NI.getUnitOffset(), Index, NI.getNameCount());
1050       ++NumErrors;
1051       continue;
1052     }
1053     if (Index > 0)
1054       BucketStarts.emplace_back(Bucket, Index);
1055   }
1056 
1057   // If there were any buckets with invalid values, skip further checks as they
1058   // will likely produce many errors which will only confuse the actual root
1059   // problem.
1060   if (NumErrors > 0)
1061     return NumErrors;
1062 
1063   // Sort the list in the order of increasing "Index" entries.
1064   array_pod_sort(BucketStarts.begin(), BucketStarts.end());
1065 
1066   // Insert a sentinel entry at the end, so we can check that the end of the
1067   // table is covered in the loop below.
1068   BucketStarts.emplace_back(NI.getBucketCount(), NI.getNameCount() + 1);
1069 
1070   // Loop invariant: NextUncovered is the (1-based) index of the first Name
1071   // which is not reachable by any of the buckets we processed so far (and
1072   // hasn't been reported as uncovered).
1073   uint32_t NextUncovered = 1;
1074   for (const BucketInfo &B : BucketStarts) {
1075     // Under normal circumstances B.Index be equal to NextUncovered, but it can
1076     // be less if a bucket points to names which are already known to be in some
1077     // bucket we processed earlier. In that case, we won't trigger this error,
1078     // but report the mismatched hash value error instead. (We know the hash
1079     // will not match because we have already verified that the name's hash
1080     // puts it into the previous bucket.)
1081     if (B.Index > NextUncovered) {
1082       error() << formatv("Name Index @ {0:x}: Name table entries [{1}, {2}] "
1083                          "are not covered by the hash table.\n",
1084                          NI.getUnitOffset(), NextUncovered, B.Index - 1);
1085       ++NumErrors;
1086     }
1087     uint32_t Idx = B.Index;
1088 
1089     // The rest of the checks apply only to non-sentinel entries.
1090     if (B.Bucket == NI.getBucketCount())
1091       break;
1092 
1093     // This triggers if a non-empty bucket points to a name with a mismatched
1094     // hash. Clients are likely to interpret this as an empty bucket, because a
1095     // mismatched hash signals the end of a bucket, but if this is indeed an
1096     // empty bucket, the producer should have signalled this by marking the
1097     // bucket as empty.
1098     uint32_t FirstHash = NI.getHashArrayEntry(Idx);
1099     if (FirstHash % NI.getBucketCount() != B.Bucket) {
1100       error() << formatv(
1101           "Name Index @ {0:x}: Bucket {1} is not empty but points to a "
1102           "mismatched hash value {2:x} (belonging to bucket {3}).\n",
1103           NI.getUnitOffset(), B.Bucket, FirstHash,
1104           FirstHash % NI.getBucketCount());
1105       ++NumErrors;
1106     }
1107 
1108     // This find the end of this bucket and also verifies that all the hashes in
1109     // this bucket are correct by comparing the stored hashes to the ones we
1110     // compute ourselves.
1111     while (Idx <= NI.getNameCount()) {
1112       uint32_t Hash = NI.getHashArrayEntry(Idx);
1113       if (Hash % NI.getBucketCount() != B.Bucket)
1114         break;
1115 
1116       const char *Str = NI.getNameTableEntry(Idx).getString();
1117       if (caseFoldingDjbHash(Str) != Hash) {
1118         error() << formatv("Name Index @ {0:x}: String ({1}) at index {2} "
1119                            "hashes to {3:x}, but "
1120                            "the Name Index hash is {4:x}\n",
1121                            NI.getUnitOffset(), Str, Idx,
1122                            caseFoldingDjbHash(Str), Hash);
1123         ++NumErrors;
1124       }
1125 
1126       ++Idx;
1127     }
1128     NextUncovered = std::max(NextUncovered, Idx);
1129   }
1130   return NumErrors;
1131 }
1132 
1133 unsigned DWARFVerifier::verifyNameIndexAttribute(
1134     const DWARFDebugNames::NameIndex &NI, const DWARFDebugNames::Abbrev &Abbr,
1135     DWARFDebugNames::AttributeEncoding AttrEnc) {
1136   StringRef FormName = dwarf::FormEncodingString(AttrEnc.Form);
1137   if (FormName.empty()) {
1138     error() << formatv("NameIndex @ {0:x}: Abbreviation {1:x}: {2} uses an "
1139                        "unknown form: {3}.\n",
1140                        NI.getUnitOffset(), Abbr.Code, AttrEnc.Index,
1141                        AttrEnc.Form);
1142     return 1;
1143   }
1144 
1145   if (AttrEnc.Index == DW_IDX_type_hash) {
1146     if (AttrEnc.Form != dwarf::DW_FORM_data8) {
1147       error() << formatv(
1148           "NameIndex @ {0:x}: Abbreviation {1:x}: DW_IDX_type_hash "
1149           "uses an unexpected form {2} (should be {3}).\n",
1150           NI.getUnitOffset(), Abbr.Code, AttrEnc.Form, dwarf::DW_FORM_data8);
1151       return 1;
1152     }
1153   }
1154 
1155   // A list of known index attributes and their expected form classes.
1156   // DW_IDX_type_hash is handled specially in the check above, as it has a
1157   // specific form (not just a form class) we should expect.
1158   struct FormClassTable {
1159     dwarf::Index Index;
1160     DWARFFormValue::FormClass Class;
1161     StringLiteral ClassName;
1162   };
1163   static constexpr FormClassTable Table[] = {
1164       {dwarf::DW_IDX_compile_unit, DWARFFormValue::FC_Constant, {"constant"}},
1165       {dwarf::DW_IDX_type_unit, DWARFFormValue::FC_Constant, {"constant"}},
1166       {dwarf::DW_IDX_die_offset, DWARFFormValue::FC_Reference, {"reference"}},
1167       {dwarf::DW_IDX_parent, DWARFFormValue::FC_Constant, {"constant"}},
1168   };
1169 
1170   ArrayRef<FormClassTable> TableRef(Table);
1171   auto Iter = find_if(TableRef, [AttrEnc](const FormClassTable &T) {
1172     return T.Index == AttrEnc.Index;
1173   });
1174   if (Iter == TableRef.end()) {
1175     warn() << formatv("NameIndex @ {0:x}: Abbreviation {1:x} contains an "
1176                       "unknown index attribute: {2}.\n",
1177                       NI.getUnitOffset(), Abbr.Code, AttrEnc.Index);
1178     return 0;
1179   }
1180 
1181   if (!DWARFFormValue(AttrEnc.Form).isFormClass(Iter->Class)) {
1182     error() << formatv("NameIndex @ {0:x}: Abbreviation {1:x}: {2} uses an "
1183                        "unexpected form {3} (expected form class {4}).\n",
1184                        NI.getUnitOffset(), Abbr.Code, AttrEnc.Index,
1185                        AttrEnc.Form, Iter->ClassName);
1186     return 1;
1187   }
1188   return 0;
1189 }
1190 
1191 unsigned
1192 DWARFVerifier::verifyNameIndexAbbrevs(const DWARFDebugNames::NameIndex &NI) {
1193   if (NI.getLocalTUCount() + NI.getForeignTUCount() > 0) {
1194     warn() << formatv("Name Index @ {0:x}: Verifying indexes of type units is "
1195                       "not currently supported.\n",
1196                       NI.getUnitOffset());
1197     return 0;
1198   }
1199 
1200   unsigned NumErrors = 0;
1201   for (const auto &Abbrev : NI.getAbbrevs()) {
1202     StringRef TagName = dwarf::TagString(Abbrev.Tag);
1203     if (TagName.empty()) {
1204       warn() << formatv("NameIndex @ {0:x}: Abbreviation {1:x} references an "
1205                         "unknown tag: {2}.\n",
1206                         NI.getUnitOffset(), Abbrev.Code, Abbrev.Tag);
1207     }
1208     SmallSet<unsigned, 5> Attributes;
1209     for (const auto &AttrEnc : Abbrev.Attributes) {
1210       if (!Attributes.insert(AttrEnc.Index).second) {
1211         error() << formatv("NameIndex @ {0:x}: Abbreviation {1:x} contains "
1212                            "multiple {2} attributes.\n",
1213                            NI.getUnitOffset(), Abbrev.Code, AttrEnc.Index);
1214         ++NumErrors;
1215         continue;
1216       }
1217       NumErrors += verifyNameIndexAttribute(NI, Abbrev, AttrEnc);
1218     }
1219 
1220     if (NI.getCUCount() > 1 && !Attributes.count(dwarf::DW_IDX_compile_unit)) {
1221       error() << formatv("NameIndex @ {0:x}: Indexing multiple compile units "
1222                          "and abbreviation {1:x} has no {2} attribute.\n",
1223                          NI.getUnitOffset(), Abbrev.Code,
1224                          dwarf::DW_IDX_compile_unit);
1225       ++NumErrors;
1226     }
1227     if (!Attributes.count(dwarf::DW_IDX_die_offset)) {
1228       error() << formatv(
1229           "NameIndex @ {0:x}: Abbreviation {1:x} has no {2} attribute.\n",
1230           NI.getUnitOffset(), Abbrev.Code, dwarf::DW_IDX_die_offset);
1231       ++NumErrors;
1232     }
1233   }
1234   return NumErrors;
1235 }
1236 
1237 static SmallVector<StringRef, 2> getNames(const DWARFDie &DIE,
1238                                           bool IncludeLinkageName = true) {
1239   SmallVector<StringRef, 2> Result;
1240   if (const char *Str = DIE.getName(DINameKind::ShortName))
1241     Result.emplace_back(Str);
1242   else if (DIE.getTag() == dwarf::DW_TAG_namespace)
1243     Result.emplace_back("(anonymous namespace)");
1244 
1245   if (IncludeLinkageName) {
1246     if (const char *Str = DIE.getName(DINameKind::LinkageName)) {
1247       if (Result.empty() || Result[0] != Str)
1248         Result.emplace_back(Str);
1249     }
1250   }
1251 
1252   return Result;
1253 }
1254 
1255 unsigned DWARFVerifier::verifyNameIndexEntries(
1256     const DWARFDebugNames::NameIndex &NI,
1257     const DWARFDebugNames::NameTableEntry &NTE) {
1258   // Verifying type unit indexes not supported.
1259   if (NI.getLocalTUCount() + NI.getForeignTUCount() > 0)
1260     return 0;
1261 
1262   const char *CStr = NTE.getString();
1263   if (!CStr) {
1264     error() << formatv(
1265         "Name Index @ {0:x}: Unable to get string associated with name {1}.\n",
1266         NI.getUnitOffset(), NTE.getIndex());
1267     return 1;
1268   }
1269   StringRef Str(CStr);
1270 
1271   unsigned NumErrors = 0;
1272   unsigned NumEntries = 0;
1273   uint64_t EntryID = NTE.getEntryOffset();
1274   uint64_t NextEntryID = EntryID;
1275   Expected<DWARFDebugNames::Entry> EntryOr = NI.getEntry(&NextEntryID);
1276   for (; EntryOr; ++NumEntries, EntryID = NextEntryID,
1277                                 EntryOr = NI.getEntry(&NextEntryID)) {
1278     uint32_t CUIndex = *EntryOr->getCUIndex();
1279     if (CUIndex > NI.getCUCount()) {
1280       error() << formatv("Name Index @ {0:x}: Entry @ {1:x} contains an "
1281                          "invalid CU index ({2}).\n",
1282                          NI.getUnitOffset(), EntryID, CUIndex);
1283       ++NumErrors;
1284       continue;
1285     }
1286     uint64_t CUOffset = NI.getCUOffset(CUIndex);
1287     uint64_t DIEOffset = CUOffset + *EntryOr->getDIEUnitOffset();
1288     DWARFDie DIE = DCtx.getDIEForOffset(DIEOffset);
1289     if (!DIE) {
1290       error() << formatv("Name Index @ {0:x}: Entry @ {1:x} references a "
1291                          "non-existing DIE @ {2:x}.\n",
1292                          NI.getUnitOffset(), EntryID, DIEOffset);
1293       ++NumErrors;
1294       continue;
1295     }
1296     if (DIE.getDwarfUnit()->getOffset() != CUOffset) {
1297       error() << formatv("Name Index @ {0:x}: Entry @ {1:x}: mismatched CU of "
1298                          "DIE @ {2:x}: index - {3:x}; debug_info - {4:x}.\n",
1299                          NI.getUnitOffset(), EntryID, DIEOffset, CUOffset,
1300                          DIE.getDwarfUnit()->getOffset());
1301       ++NumErrors;
1302     }
1303     if (DIE.getTag() != EntryOr->tag()) {
1304       error() << formatv("Name Index @ {0:x}: Entry @ {1:x}: mismatched Tag of "
1305                          "DIE @ {2:x}: index - {3}; debug_info - {4}.\n",
1306                          NI.getUnitOffset(), EntryID, DIEOffset, EntryOr->tag(),
1307                          DIE.getTag());
1308       ++NumErrors;
1309     }
1310 
1311     auto EntryNames = getNames(DIE);
1312     if (!is_contained(EntryNames, Str)) {
1313       error() << formatv("Name Index @ {0:x}: Entry @ {1:x}: mismatched Name "
1314                          "of DIE @ {2:x}: index - {3}; debug_info - {4}.\n",
1315                          NI.getUnitOffset(), EntryID, DIEOffset, Str,
1316                          make_range(EntryNames.begin(), EntryNames.end()));
1317       ++NumErrors;
1318     }
1319   }
1320   handleAllErrors(EntryOr.takeError(),
1321                   [&](const DWARFDebugNames::SentinelError &) {
1322                     if (NumEntries > 0)
1323                       return;
1324                     error() << formatv("Name Index @ {0:x}: Name {1} ({2}) is "
1325                                        "not associated with any entries.\n",
1326                                        NI.getUnitOffset(), NTE.getIndex(), Str);
1327                     ++NumErrors;
1328                   },
1329                   [&](const ErrorInfoBase &Info) {
1330                     error()
1331                         << formatv("Name Index @ {0:x}: Name {1} ({2}): {3}\n",
1332                                    NI.getUnitOffset(), NTE.getIndex(), Str,
1333                                    Info.message());
1334                     ++NumErrors;
1335                   });
1336   return NumErrors;
1337 }
1338 
1339 static bool isVariableIndexable(const DWARFDie &Die, DWARFContext &DCtx) {
1340   Expected<std::vector<DWARFLocationExpression>> Loc =
1341       Die.getLocations(DW_AT_location);
1342   if (!Loc) {
1343     consumeError(Loc.takeError());
1344     return false;
1345   }
1346   DWARFUnit *U = Die.getDwarfUnit();
1347   for (const auto &Entry : *Loc) {
1348     DataExtractor Data(toStringRef(Entry.Expr), DCtx.isLittleEndian(),
1349                        U->getAddressByteSize());
1350     DWARFExpression Expression(Data, U->getAddressByteSize(),
1351                                U->getFormParams().Format);
1352     bool IsInteresting = any_of(Expression, [](DWARFExpression::Operation &Op) {
1353       return !Op.isError() && (Op.getCode() == DW_OP_addr ||
1354                                Op.getCode() == DW_OP_form_tls_address ||
1355                                Op.getCode() == DW_OP_GNU_push_tls_address);
1356     });
1357     if (IsInteresting)
1358       return true;
1359   }
1360   return false;
1361 }
1362 
1363 unsigned DWARFVerifier::verifyNameIndexCompleteness(
1364     const DWARFDie &Die, const DWARFDebugNames::NameIndex &NI) {
1365 
1366   // First check, if the Die should be indexed. The code follows the DWARF v5
1367   // wording as closely as possible.
1368 
1369   // "All non-defining declarations (that is, debugging information entries
1370   // with a DW_AT_declaration attribute) are excluded."
1371   if (Die.find(DW_AT_declaration))
1372     return 0;
1373 
1374   // "DW_TAG_namespace debugging information entries without a DW_AT_name
1375   // attribute are included with the name “(anonymous namespace)”.
1376   // All other debugging information entries without a DW_AT_name attribute
1377   // are excluded."
1378   // "If a subprogram or inlined subroutine is included, and has a
1379   // DW_AT_linkage_name attribute, there will be an additional index entry for
1380   // the linkage name."
1381   auto IncludeLinkageName = Die.getTag() == DW_TAG_subprogram ||
1382                             Die.getTag() == DW_TAG_inlined_subroutine;
1383   auto EntryNames = getNames(Die, IncludeLinkageName);
1384   if (EntryNames.empty())
1385     return 0;
1386 
1387   // We deviate from the specification here, which says:
1388   // "The name index must contain an entry for each debugging information entry
1389   // that defines a named subprogram, label, variable, type, or namespace,
1390   // subject to ..."
1391   // Explicitly exclude all TAGs that we know shouldn't be indexed.
1392   switch (Die.getTag()) {
1393   // Compile units and modules have names but shouldn't be indexed.
1394   case DW_TAG_compile_unit:
1395   case DW_TAG_module:
1396     return 0;
1397 
1398   // Function and template parameters are not globally visible, so we shouldn't
1399   // index them.
1400   case DW_TAG_formal_parameter:
1401   case DW_TAG_template_value_parameter:
1402   case DW_TAG_template_type_parameter:
1403   case DW_TAG_GNU_template_parameter_pack:
1404   case DW_TAG_GNU_template_template_param:
1405     return 0;
1406 
1407   // Object members aren't globally visible.
1408   case DW_TAG_member:
1409     return 0;
1410 
1411   // According to a strict reading of the specification, enumerators should not
1412   // be indexed (and LLVM currently does not do that). However, this causes
1413   // problems for the debuggers, so we may need to reconsider this.
1414   case DW_TAG_enumerator:
1415     return 0;
1416 
1417   // Imported declarations should not be indexed according to the specification
1418   // and LLVM currently does not do that.
1419   case DW_TAG_imported_declaration:
1420     return 0;
1421 
1422   // "DW_TAG_subprogram, DW_TAG_inlined_subroutine, and DW_TAG_label debugging
1423   // information entries without an address attribute (DW_AT_low_pc,
1424   // DW_AT_high_pc, DW_AT_ranges, or DW_AT_entry_pc) are excluded."
1425   case DW_TAG_subprogram:
1426   case DW_TAG_inlined_subroutine:
1427   case DW_TAG_label:
1428     if (Die.findRecursively(
1429             {DW_AT_low_pc, DW_AT_high_pc, DW_AT_ranges, DW_AT_entry_pc}))
1430       break;
1431     return 0;
1432 
1433   // "DW_TAG_variable debugging information entries with a DW_AT_location
1434   // attribute that includes a DW_OP_addr or DW_OP_form_tls_address operator are
1435   // included; otherwise, they are excluded."
1436   //
1437   // LLVM extension: We also add DW_OP_GNU_push_tls_address to this list.
1438   case DW_TAG_variable:
1439     if (isVariableIndexable(Die, DCtx))
1440       break;
1441     return 0;
1442 
1443   default:
1444     break;
1445   }
1446 
1447   // Now we know that our Die should be present in the Index. Let's check if
1448   // that's the case.
1449   unsigned NumErrors = 0;
1450   uint64_t DieUnitOffset = Die.getOffset() - Die.getDwarfUnit()->getOffset();
1451   for (StringRef Name : EntryNames) {
1452     if (none_of(NI.equal_range(Name), [&](const DWARFDebugNames::Entry &E) {
1453           return E.getDIEUnitOffset() == DieUnitOffset;
1454         })) {
1455       error() << formatv("Name Index @ {0:x}: Entry for DIE @ {1:x} ({2}) with "
1456                          "name {3} missing.\n",
1457                          NI.getUnitOffset(), Die.getOffset(), Die.getTag(),
1458                          Name);
1459       ++NumErrors;
1460     }
1461   }
1462   return NumErrors;
1463 }
1464 
1465 unsigned DWARFVerifier::verifyDebugNames(const DWARFSection &AccelSection,
1466                                          const DataExtractor &StrData) {
1467   unsigned NumErrors = 0;
1468   DWARFDataExtractor AccelSectionData(DCtx.getDWARFObj(), AccelSection,
1469                                       DCtx.isLittleEndian(), 0);
1470   DWARFDebugNames AccelTable(AccelSectionData, StrData);
1471 
1472   OS << "Verifying .debug_names...\n";
1473 
1474   // This verifies that we can read individual name indices and their
1475   // abbreviation tables.
1476   if (Error E = AccelTable.extract()) {
1477     error() << toString(std::move(E)) << '\n';
1478     return 1;
1479   }
1480 
1481   NumErrors += verifyDebugNamesCULists(AccelTable);
1482   for (const auto &NI : AccelTable)
1483     NumErrors += verifyNameIndexBuckets(NI, StrData);
1484   for (const auto &NI : AccelTable)
1485     NumErrors += verifyNameIndexAbbrevs(NI);
1486 
1487   // Don't attempt Entry validation if any of the previous checks found errors
1488   if (NumErrors > 0)
1489     return NumErrors;
1490   for (const auto &NI : AccelTable)
1491     for (DWARFDebugNames::NameTableEntry NTE : NI)
1492       NumErrors += verifyNameIndexEntries(NI, NTE);
1493 
1494   if (NumErrors > 0)
1495     return NumErrors;
1496 
1497   for (const std::unique_ptr<DWARFUnit> &U : DCtx.compile_units()) {
1498     if (const DWARFDebugNames::NameIndex *NI =
1499             AccelTable.getCUNameIndex(U->getOffset())) {
1500       auto *CU = cast<DWARFCompileUnit>(U.get());
1501       for (const DWARFDebugInfoEntry &Die : CU->dies())
1502         NumErrors += verifyNameIndexCompleteness(DWARFDie(CU, &Die), *NI);
1503     }
1504   }
1505   return NumErrors;
1506 }
1507 
1508 bool DWARFVerifier::handleAccelTables() {
1509   const DWARFObject &D = DCtx.getDWARFObj();
1510   DataExtractor StrData(D.getStrSection(), DCtx.isLittleEndian(), 0);
1511   unsigned NumErrors = 0;
1512   if (!D.getAppleNamesSection().Data.empty())
1513     NumErrors += verifyAppleAccelTable(&D.getAppleNamesSection(), &StrData,
1514                                        ".apple_names");
1515   if (!D.getAppleTypesSection().Data.empty())
1516     NumErrors += verifyAppleAccelTable(&D.getAppleTypesSection(), &StrData,
1517                                        ".apple_types");
1518   if (!D.getAppleNamespacesSection().Data.empty())
1519     NumErrors += verifyAppleAccelTable(&D.getAppleNamespacesSection(), &StrData,
1520                                        ".apple_namespaces");
1521   if (!D.getAppleObjCSection().Data.empty())
1522     NumErrors += verifyAppleAccelTable(&D.getAppleObjCSection(), &StrData,
1523                                        ".apple_objc");
1524 
1525   if (!D.getNamesSection().Data.empty())
1526     NumErrors += verifyDebugNames(D.getNamesSection(), StrData);
1527   return NumErrors == 0;
1528 }
1529 
1530 raw_ostream &DWARFVerifier::error() const { return WithColor::error(OS); }
1531 
1532 raw_ostream &DWARFVerifier::warn() const { return WithColor::warning(OS); }
1533 
1534 raw_ostream &DWARFVerifier::note() const { return WithColor::note(OS); }
1535 
1536 raw_ostream &DWARFVerifier::dump(const DWARFDie &Die, unsigned indent) const {
1537   Die.dump(OS, indent, DumpOpts);
1538   return OS;
1539 }
1540