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/IntervalMap.h" |
10 | #include "llvm/ADT/STLExtras.h" |
11 | #include "llvm/ADT/SmallSet.h" |
12 | #include "llvm/BinaryFormat/Dwarf.h" |
13 | #include "llvm/DebugInfo/DWARF/DWARFAbbreviationDeclaration.h" |
14 | #include "llvm/DebugInfo/DWARF/DWARFAttribute.h" |
15 | #include "llvm/DebugInfo/DWARF/DWARFCompileUnit.h" |
16 | #include "llvm/DebugInfo/DWARF/DWARFContext.h" |
17 | #include "llvm/DebugInfo/DWARF/DWARFDataExtractor.h" |
18 | #include "llvm/DebugInfo/DWARF/DWARFDebugAbbrev.h" |
19 | #include "llvm/DebugInfo/DWARF/DWARFDebugLine.h" |
20 | #include "llvm/DebugInfo/DWARF/DWARFDebugLoc.h" |
21 | #include "llvm/DebugInfo/DWARF/DWARFDie.h" |
22 | #include "llvm/DebugInfo/DWARF/DWARFExpression.h" |
23 | #include "llvm/DebugInfo/DWARF/DWARFFormValue.h" |
24 | #include "llvm/DebugInfo/DWARF/DWARFLocationExpression.h" |
25 | #include "llvm/DebugInfo/DWARF/DWARFObject.h" |
26 | #include "llvm/DebugInfo/DWARF/DWARFSection.h" |
27 | #include "llvm/DebugInfo/DWARF/DWARFUnit.h" |
28 | #include "llvm/Object/Error.h" |
29 | #include "llvm/Support/DJB.h" |
30 | #include "llvm/Support/Error.h" |
31 | #include "llvm/Support/ErrorHandling.h" |
32 | #include "llvm/Support/FileSystem.h" |
33 | #include "llvm/Support/FormatVariadic.h" |
34 | #include "llvm/Support/JSON.h" |
35 | #include "llvm/Support/WithColor.h" |
36 | #include "llvm/Support/raw_ostream.h" |
37 | #include <map> |
38 | #include <set> |
39 | #include <vector> |
40 | |
41 | using namespace llvm; |
42 | using namespace dwarf; |
43 | using namespace object; |
44 | |
45 | namespace llvm { |
46 | class DWARFDebugInfoEntry; |
47 | } |
48 | |
49 | std::optional<DWARFAddressRange> |
50 | DWARFVerifier::DieRangeInfo::insert(const DWARFAddressRange &R) { |
51 | auto Begin = Ranges.begin(); |
52 | auto End = Ranges.end(); |
53 | auto Pos = std::lower_bound(Begin, End, R); |
54 | |
55 | if (Pos != End) { |
56 | DWARFAddressRange Range(*Pos); |
57 | if (Pos->merge(RHS: R)) |
58 | return Range; |
59 | } |
60 | if (Pos != Begin) { |
61 | auto Iter = Pos - 1; |
62 | DWARFAddressRange Range(*Iter); |
63 | if (Iter->merge(RHS: R)) |
64 | return Range; |
65 | } |
66 | |
67 | Ranges.insert(position: Pos, x: R); |
68 | return std::nullopt; |
69 | } |
70 | |
71 | DWARFVerifier::DieRangeInfo::die_range_info_iterator |
72 | DWARFVerifier::DieRangeInfo::insert(const DieRangeInfo &RI) { |
73 | if (RI.Ranges.empty()) |
74 | return Children.end(); |
75 | |
76 | auto End = Children.end(); |
77 | auto Iter = Children.begin(); |
78 | while (Iter != End) { |
79 | if (Iter->intersects(RHS: RI)) |
80 | return Iter; |
81 | ++Iter; |
82 | } |
83 | Children.insert(x: RI); |
84 | return Children.end(); |
85 | } |
86 | |
87 | bool DWARFVerifier::DieRangeInfo::contains(const DieRangeInfo &RHS) const { |
88 | auto I1 = Ranges.begin(), E1 = Ranges.end(); |
89 | auto I2 = RHS.Ranges.begin(), E2 = RHS.Ranges.end(); |
90 | if (I2 == E2) |
91 | return true; |
92 | |
93 | DWARFAddressRange R = *I2; |
94 | while (I1 != E1) { |
95 | bool Covered = I1->LowPC <= R.LowPC; |
96 | if (R.LowPC == R.HighPC || (Covered && R.HighPC <= I1->HighPC)) { |
97 | if (++I2 == E2) |
98 | return true; |
99 | R = *I2; |
100 | continue; |
101 | } |
102 | if (!Covered) |
103 | return false; |
104 | if (R.LowPC < I1->HighPC) |
105 | R.LowPC = I1->HighPC; |
106 | ++I1; |
107 | } |
108 | return false; |
109 | } |
110 | |
111 | bool DWARFVerifier::DieRangeInfo::intersects(const DieRangeInfo &RHS) const { |
112 | auto I1 = Ranges.begin(), E1 = Ranges.end(); |
113 | auto I2 = RHS.Ranges.begin(), E2 = RHS.Ranges.end(); |
114 | while (I1 != E1 && I2 != E2) { |
115 | if (I1->intersects(RHS: *I2)) |
116 | return true; |
117 | if (I1->LowPC < I2->LowPC) |
118 | ++I1; |
119 | else |
120 | ++I2; |
121 | } |
122 | return false; |
123 | } |
124 | |
125 | bool DWARFVerifier::(const DWARFDataExtractor DebugInfoData, |
126 | uint64_t *Offset, unsigned UnitIndex, |
127 | uint8_t &UnitType, bool &isUnitDWARF64) { |
128 | uint64_t AbbrOffset, Length; |
129 | uint8_t AddrSize = 0; |
130 | uint16_t Version; |
131 | bool Success = true; |
132 | |
133 | bool ValidLength = false; |
134 | bool ValidVersion = false; |
135 | bool ValidAddrSize = false; |
136 | bool ValidType = true; |
137 | bool ValidAbbrevOffset = true; |
138 | |
139 | uint64_t OffsetStart = *Offset; |
140 | DwarfFormat Format; |
141 | std::tie(args&: Length, args&: Format) = DebugInfoData.getInitialLength(Off: Offset); |
142 | isUnitDWARF64 = Format == DWARF64; |
143 | Version = DebugInfoData.getU16(offset_ptr: Offset); |
144 | |
145 | if (Version >= 5) { |
146 | UnitType = DebugInfoData.getU8(offset_ptr: Offset); |
147 | AddrSize = DebugInfoData.getU8(offset_ptr: Offset); |
148 | AbbrOffset = isUnitDWARF64 ? DebugInfoData.getU64(offset_ptr: Offset) : DebugInfoData.getU32(offset_ptr: Offset); |
149 | ValidType = dwarf::isUnitType(UnitType); |
150 | } else { |
151 | UnitType = 0; |
152 | AbbrOffset = isUnitDWARF64 ? DebugInfoData.getU64(offset_ptr: Offset) : DebugInfoData.getU32(offset_ptr: Offset); |
153 | AddrSize = DebugInfoData.getU8(offset_ptr: Offset); |
154 | } |
155 | |
156 | Expected<const DWARFAbbreviationDeclarationSet *> AbbrevSetOrErr = |
157 | DCtx.getDebugAbbrev()->getAbbreviationDeclarationSet(CUAbbrOffset: AbbrOffset); |
158 | if (!AbbrevSetOrErr) { |
159 | ValidAbbrevOffset = false; |
160 | // FIXME: A problematic debug_abbrev section is reported below in the form |
161 | // of a `note:`. We should propagate this error there (or elsewhere) to |
162 | // avoid losing the specific problem with the debug_abbrev section. |
163 | consumeError(Err: AbbrevSetOrErr.takeError()); |
164 | } |
165 | |
166 | ValidLength = DebugInfoData.isValidOffset(offset: OffsetStart + Length + 3); |
167 | ValidVersion = DWARFContext::isSupportedVersion(version: Version); |
168 | ValidAddrSize = DWARFContext::isAddressSizeSupported(AddressSize: AddrSize); |
169 | if (!ValidLength || !ValidVersion || !ValidAddrSize || !ValidAbbrevOffset || |
170 | !ValidType) { |
171 | Success = false; |
172 | bool = false; |
173 | auto = [&]() { |
174 | if (!HeaderShown) { |
175 | error() << format(Fmt: "Units[%d] - start offset: 0x%08" PRIx64 " \n" , |
176 | Vals: UnitIndex, Vals: OffsetStart); |
177 | HeaderShown = true; |
178 | } |
179 | }; |
180 | if (!ValidLength) |
181 | ErrorCategory.Report( |
182 | s: "Unit Header Length: Unit too large for .debug_info provided" , detailCallback: [&]() { |
183 | ShowHeaderOnce(); |
184 | note() << "The length for this unit is too " |
185 | "large for the .debug_info provided.\n" ; |
186 | }); |
187 | if (!ValidVersion) |
188 | ErrorCategory.Report( |
189 | s: "Unit Header Length: 16 bit unit header version is not valid" , detailCallback: [&]() { |
190 | ShowHeaderOnce(); |
191 | note() << "The 16 bit unit header version is not valid.\n" ; |
192 | }); |
193 | if (!ValidType) |
194 | ErrorCategory.Report( |
195 | s: "Unit Header Length: Unit type encoding is not valid" , detailCallback: [&]() { |
196 | ShowHeaderOnce(); |
197 | note() << "The unit type encoding is not valid.\n" ; |
198 | }); |
199 | if (!ValidAbbrevOffset) |
200 | ErrorCategory.Report( |
201 | s: "Unit Header Length: Offset into the .debug_abbrev section is not " |
202 | "valid" , |
203 | detailCallback: [&]() { |
204 | ShowHeaderOnce(); |
205 | note() << "The offset into the .debug_abbrev section is " |
206 | "not valid.\n" ; |
207 | }); |
208 | if (!ValidAddrSize) |
209 | ErrorCategory.Report(s: "Unit Header Length: Address size is unsupported" , |
210 | detailCallback: [&]() { |
211 | ShowHeaderOnce(); |
212 | note() << "The address size is unsupported.\n" ; |
213 | }); |
214 | } |
215 | *Offset = OffsetStart + Length + (isUnitDWARF64 ? 12 : 4); |
216 | return Success; |
217 | } |
218 | |
219 | bool DWARFVerifier::verifyName(const DWARFDie &Die) { |
220 | // FIXME Add some kind of record of which DIE names have already failed and |
221 | // don't bother checking a DIE that uses an already failed DIE. |
222 | |
223 | std::string ReconstructedName; |
224 | raw_string_ostream OS(ReconstructedName); |
225 | std::string OriginalFullName; |
226 | Die.getFullName(OS, OriginalFullName: &OriginalFullName); |
227 | OS.flush(); |
228 | if (OriginalFullName.empty() || OriginalFullName == ReconstructedName) |
229 | return false; |
230 | |
231 | ErrorCategory.Report( |
232 | s: "Simplified template DW_AT_name could not be reconstituted" , detailCallback: [&]() { |
233 | error() |
234 | << "Simplified template DW_AT_name could not be reconstituted:\n" |
235 | << formatv(Fmt: " original: {0}\n" |
236 | " reconstituted: {1}\n" , |
237 | Vals&: OriginalFullName, Vals&: ReconstructedName); |
238 | dump(Die) << '\n'; |
239 | dump(Die: Die.getDwarfUnit()->getUnitDIE()) << '\n'; |
240 | }); |
241 | return true; |
242 | } |
243 | |
244 | unsigned DWARFVerifier::verifyUnitContents(DWARFUnit &Unit, |
245 | ReferenceMap &UnitLocalReferences, |
246 | ReferenceMap &CrossUnitReferences) { |
247 | unsigned NumUnitErrors = 0; |
248 | unsigned NumDies = Unit.getNumDIEs(); |
249 | for (unsigned I = 0; I < NumDies; ++I) { |
250 | auto Die = Unit.getDIEAtIndex(Index: I); |
251 | |
252 | if (Die.getTag() == DW_TAG_null) |
253 | continue; |
254 | |
255 | for (auto AttrValue : Die.attributes()) { |
256 | NumUnitErrors += verifyDebugInfoAttribute(Die, AttrValue); |
257 | NumUnitErrors += verifyDebugInfoForm(Die, AttrValue, UnitLocalReferences, |
258 | CrossUnitReferences); |
259 | } |
260 | |
261 | NumUnitErrors += verifyName(Die); |
262 | |
263 | if (Die.hasChildren()) { |
264 | if (Die.getFirstChild().isValid() && |
265 | Die.getFirstChild().getTag() == DW_TAG_null) { |
266 | warn() << dwarf::TagString(Tag: Die.getTag()) |
267 | << " has DW_CHILDREN_yes but DIE has no children: " ; |
268 | Die.dump(OS); |
269 | } |
270 | } |
271 | |
272 | NumUnitErrors += verifyDebugInfoCallSite(Die); |
273 | } |
274 | |
275 | DWARFDie Die = Unit.getUnitDIE(/* ExtractUnitDIEOnly = */ false); |
276 | if (!Die) { |
277 | ErrorCategory.Report(s: "Compilation unit missing DIE" , detailCallback: [&]() { |
278 | error() << "Compilation unit without DIE.\n" ; |
279 | }); |
280 | NumUnitErrors++; |
281 | return NumUnitErrors; |
282 | } |
283 | |
284 | if (!dwarf::isUnitType(T: Die.getTag())) { |
285 | ErrorCategory.Report(s: "Compilation unit root DIE is not a unit DIE" , detailCallback: [&]() { |
286 | error() << "Compilation unit root DIE is not a unit DIE: " |
287 | << dwarf::TagString(Tag: Die.getTag()) << ".\n" ; |
288 | }); |
289 | NumUnitErrors++; |
290 | } |
291 | |
292 | uint8_t UnitType = Unit.getUnitType(); |
293 | if (!DWARFUnit::isMatchingUnitTypeAndTag(UnitType, Tag: Die.getTag())) { |
294 | ErrorCategory.Report(s: "Mismatched unit type" , detailCallback: [&]() { |
295 | error() << "Compilation unit type (" << dwarf::UnitTypeString(UnitType) |
296 | << ") and root DIE (" << dwarf::TagString(Tag: Die.getTag()) |
297 | << ") do not match.\n" ; |
298 | }); |
299 | NumUnitErrors++; |
300 | } |
301 | |
302 | // According to DWARF Debugging Information Format Version 5, |
303 | // 3.1.2 Skeleton Compilation Unit Entries: |
304 | // "A skeleton compilation unit has no children." |
305 | if (Die.getTag() == dwarf::DW_TAG_skeleton_unit && Die.hasChildren()) { |
306 | ErrorCategory.Report(s: "Skeleton CU has children" , detailCallback: [&]() { |
307 | error() << "Skeleton compilation unit has children.\n" ; |
308 | }); |
309 | NumUnitErrors++; |
310 | } |
311 | |
312 | DieRangeInfo RI; |
313 | NumUnitErrors += verifyDieRanges(Die, ParentRI&: RI); |
314 | |
315 | return NumUnitErrors; |
316 | } |
317 | |
318 | unsigned DWARFVerifier::verifyDebugInfoCallSite(const DWARFDie &Die) { |
319 | if (Die.getTag() != DW_TAG_call_site && Die.getTag() != DW_TAG_GNU_call_site) |
320 | return 0; |
321 | |
322 | DWARFDie Curr = Die.getParent(); |
323 | for (; Curr.isValid() && !Curr.isSubprogramDIE(); Curr = Die.getParent()) { |
324 | if (Curr.getTag() == DW_TAG_inlined_subroutine) { |
325 | ErrorCategory.Report( |
326 | s: "Call site nested entry within inlined subroutine" , detailCallback: [&]() { |
327 | error() << "Call site entry nested within inlined subroutine:" ; |
328 | Curr.dump(OS); |
329 | }); |
330 | return 1; |
331 | } |
332 | } |
333 | |
334 | if (!Curr.isValid()) { |
335 | ErrorCategory.Report( |
336 | s: "Call site entry not nested within valid subprogram" , detailCallback: [&]() { |
337 | error() << "Call site entry not nested within a valid subprogram:" ; |
338 | Die.dump(OS); |
339 | }); |
340 | return 1; |
341 | } |
342 | |
343 | std::optional<DWARFFormValue> CallAttr = Curr.find( |
344 | Attrs: {DW_AT_call_all_calls, DW_AT_call_all_source_calls, |
345 | DW_AT_call_all_tail_calls, DW_AT_GNU_all_call_sites, |
346 | DW_AT_GNU_all_source_call_sites, DW_AT_GNU_all_tail_call_sites}); |
347 | if (!CallAttr) { |
348 | ErrorCategory.Report( |
349 | s: "Subprogram with call site entry has no DW_AT_call attribute" , detailCallback: [&]() { |
350 | error() |
351 | << "Subprogram with call site entry has no DW_AT_call attribute:" ; |
352 | Curr.dump(OS); |
353 | Die.dump(OS, /*indent*/ 1); |
354 | }); |
355 | return 1; |
356 | } |
357 | |
358 | return 0; |
359 | } |
360 | |
361 | unsigned DWARFVerifier::verifyAbbrevSection(const DWARFDebugAbbrev *Abbrev) { |
362 | if (!Abbrev) |
363 | return 0; |
364 | |
365 | Expected<const DWARFAbbreviationDeclarationSet *> AbbrDeclsOrErr = |
366 | Abbrev->getAbbreviationDeclarationSet(CUAbbrOffset: 0); |
367 | if (!AbbrDeclsOrErr) { |
368 | std::string ErrMsg = toString(E: AbbrDeclsOrErr.takeError()); |
369 | ErrorCategory.Report(s: "Abbreviation Declaration error" , |
370 | detailCallback: [&]() { error() << ErrMsg << "\n" ; }); |
371 | return 1; |
372 | } |
373 | |
374 | const auto *AbbrDecls = *AbbrDeclsOrErr; |
375 | unsigned NumErrors = 0; |
376 | for (auto AbbrDecl : *AbbrDecls) { |
377 | SmallDenseSet<uint16_t> AttributeSet; |
378 | for (auto Attribute : AbbrDecl.attributes()) { |
379 | auto Result = AttributeSet.insert(V: Attribute.Attr); |
380 | if (!Result.second) { |
381 | ErrorCategory.Report( |
382 | s: "Abbreviation declartion contains multiple attributes" , detailCallback: [&]() { |
383 | error() << "Abbreviation declaration contains multiple " |
384 | << AttributeString(Attribute: Attribute.Attr) << " attributes.\n" ; |
385 | AbbrDecl.dump(OS); |
386 | }); |
387 | ++NumErrors; |
388 | } |
389 | } |
390 | } |
391 | return NumErrors; |
392 | } |
393 | |
394 | bool DWARFVerifier::handleDebugAbbrev() { |
395 | OS << "Verifying .debug_abbrev...\n" ; |
396 | |
397 | const DWARFObject &DObj = DCtx.getDWARFObj(); |
398 | unsigned NumErrors = 0; |
399 | if (!DObj.getAbbrevSection().empty()) |
400 | NumErrors += verifyAbbrevSection(Abbrev: DCtx.getDebugAbbrev()); |
401 | if (!DObj.getAbbrevDWOSection().empty()) |
402 | NumErrors += verifyAbbrevSection(Abbrev: DCtx.getDebugAbbrevDWO()); |
403 | |
404 | return NumErrors == 0; |
405 | } |
406 | |
407 | unsigned DWARFVerifier::verifyUnits(const DWARFUnitVector &Units) { |
408 | unsigned NumDebugInfoErrors = 0; |
409 | ReferenceMap CrossUnitReferences; |
410 | |
411 | unsigned Index = 1; |
412 | for (const auto &Unit : Units) { |
413 | OS << "Verifying unit: " << Index << " / " << Units.getNumUnits(); |
414 | if (const char* Name = Unit->getUnitDIE(ExtractUnitDIEOnly: true).getShortName()) |
415 | OS << ", \"" << Name << '\"'; |
416 | OS << '\n'; |
417 | OS.flush(); |
418 | ReferenceMap UnitLocalReferences; |
419 | NumDebugInfoErrors += |
420 | verifyUnitContents(Unit&: *Unit, UnitLocalReferences, CrossUnitReferences); |
421 | NumDebugInfoErrors += verifyDebugInfoReferences( |
422 | UnitLocalReferences, GetUnitForDieOffset: [&](uint64_t Offset) { return Unit.get(); }); |
423 | ++Index; |
424 | } |
425 | |
426 | NumDebugInfoErrors += verifyDebugInfoReferences( |
427 | CrossUnitReferences, GetUnitForDieOffset: [&](uint64_t Offset) -> DWARFUnit * { |
428 | if (DWARFUnit *U = Units.getUnitForOffset(Offset)) |
429 | return U; |
430 | return nullptr; |
431 | }); |
432 | |
433 | return NumDebugInfoErrors; |
434 | } |
435 | |
436 | unsigned DWARFVerifier::verifyUnitSection(const DWARFSection &S) { |
437 | const DWARFObject &DObj = DCtx.getDWARFObj(); |
438 | DWARFDataExtractor DebugInfoData(DObj, S, DCtx.isLittleEndian(), 0); |
439 | unsigned NumDebugInfoErrors = 0; |
440 | uint64_t Offset = 0, UnitIdx = 0; |
441 | uint8_t UnitType = 0; |
442 | bool isUnitDWARF64 = false; |
443 | bool = true; |
444 | bool hasDIE = DebugInfoData.isValidOffset(offset: Offset); |
445 | DWARFUnitVector TypeUnitVector; |
446 | DWARFUnitVector CompileUnitVector; |
447 | /// A map that tracks all references (converted absolute references) so we |
448 | /// can verify each reference points to a valid DIE and not an offset that |
449 | /// lies between to valid DIEs. |
450 | ReferenceMap CrossUnitReferences; |
451 | while (hasDIE) { |
452 | if (!verifyUnitHeader(DebugInfoData, Offset: &Offset, UnitIndex: UnitIdx, UnitType, |
453 | isUnitDWARF64)) { |
454 | isHeaderChainValid = false; |
455 | if (isUnitDWARF64) |
456 | break; |
457 | } |
458 | hasDIE = DebugInfoData.isValidOffset(offset: Offset); |
459 | ++UnitIdx; |
460 | } |
461 | if (UnitIdx == 0 && !hasDIE) { |
462 | warn() << "Section is empty.\n" ; |
463 | isHeaderChainValid = true; |
464 | } |
465 | if (!isHeaderChainValid) |
466 | ++NumDebugInfoErrors; |
467 | return NumDebugInfoErrors; |
468 | } |
469 | |
470 | unsigned DWARFVerifier::verifyIndex(StringRef Name, |
471 | DWARFSectionKind InfoColumnKind, |
472 | StringRef IndexStr) { |
473 | if (IndexStr.empty()) |
474 | return 0; |
475 | OS << "Verifying " << Name << "...\n" ; |
476 | DWARFUnitIndex Index(InfoColumnKind); |
477 | DataExtractor D(IndexStr, DCtx.isLittleEndian(), 0); |
478 | if (!Index.parse(IndexData: D)) |
479 | return 1; |
480 | using MapType = IntervalMap<uint64_t, uint64_t>; |
481 | MapType::Allocator Alloc; |
482 | std::vector<std::unique_ptr<MapType>> Sections(Index.getColumnKinds().size()); |
483 | for (const DWARFUnitIndex::Entry &E : Index.getRows()) { |
484 | uint64_t Sig = E.getSignature(); |
485 | if (!E.getContributions()) |
486 | continue; |
487 | for (auto E : enumerate( |
488 | First: InfoColumnKind == DW_SECT_INFO |
489 | ? ArrayRef(E.getContributions(), Index.getColumnKinds().size()) |
490 | : ArrayRef(E.getContribution(), 1))) { |
491 | const DWARFUnitIndex::Entry::SectionContribution &SC = E.value(); |
492 | int Col = E.index(); |
493 | if (SC.getLength() == 0) |
494 | continue; |
495 | if (!Sections[Col]) |
496 | Sections[Col] = std::make_unique<MapType>(args&: Alloc); |
497 | auto &M = *Sections[Col]; |
498 | auto I = M.find(x: SC.getOffset()); |
499 | if (I != M.end() && I.start() < (SC.getOffset() + SC.getLength())) { |
500 | StringRef Category = InfoColumnKind == DWARFSectionKind::DW_SECT_INFO |
501 | ? "Overlapping CU index entries" |
502 | : "Overlapping TU index entries" ; |
503 | ErrorCategory.Report(s: Category, detailCallback: [&]() { |
504 | error() << llvm::formatv( |
505 | Fmt: "overlapping index entries for entries {0:x16} " |
506 | "and {1:x16} for column {2}\n" , |
507 | Vals: *I, Vals&: Sig, Vals: toString(Kind: Index.getColumnKinds()[Col])); |
508 | }); |
509 | return 1; |
510 | } |
511 | M.insert(a: SC.getOffset(), b: SC.getOffset() + SC.getLength() - 1, y: Sig); |
512 | } |
513 | } |
514 | |
515 | return 0; |
516 | } |
517 | |
518 | bool DWARFVerifier::handleDebugCUIndex() { |
519 | return verifyIndex(Name: ".debug_cu_index" , InfoColumnKind: DWARFSectionKind::DW_SECT_INFO, |
520 | IndexStr: DCtx.getDWARFObj().getCUIndexSection()) == 0; |
521 | } |
522 | |
523 | bool DWARFVerifier::handleDebugTUIndex() { |
524 | return verifyIndex(Name: ".debug_tu_index" , InfoColumnKind: DWARFSectionKind::DW_SECT_EXT_TYPES, |
525 | IndexStr: DCtx.getDWARFObj().getTUIndexSection()) == 0; |
526 | } |
527 | |
528 | bool DWARFVerifier::handleDebugInfo() { |
529 | const DWARFObject &DObj = DCtx.getDWARFObj(); |
530 | unsigned NumErrors = 0; |
531 | |
532 | OS << "Verifying .debug_info Unit Header Chain...\n" ; |
533 | DObj.forEachInfoSections(F: [&](const DWARFSection &S) { |
534 | NumErrors += verifyUnitSection(S); |
535 | }); |
536 | |
537 | OS << "Verifying .debug_types Unit Header Chain...\n" ; |
538 | DObj.forEachTypesSections(F: [&](const DWARFSection &S) { |
539 | NumErrors += verifyUnitSection(S); |
540 | }); |
541 | |
542 | OS << "Verifying non-dwo Units...\n" ; |
543 | NumErrors += verifyUnits(Units: DCtx.getNormalUnitsVector()); |
544 | |
545 | OS << "Verifying dwo Units...\n" ; |
546 | NumErrors += verifyUnits(Units: DCtx.getDWOUnitsVector()); |
547 | return NumErrors == 0; |
548 | } |
549 | |
550 | unsigned DWARFVerifier::verifyDieRanges(const DWARFDie &Die, |
551 | DieRangeInfo &ParentRI) { |
552 | unsigned NumErrors = 0; |
553 | |
554 | if (!Die.isValid()) |
555 | return NumErrors; |
556 | |
557 | DWARFUnit *Unit = Die.getDwarfUnit(); |
558 | |
559 | auto RangesOrError = Die.getAddressRanges(); |
560 | if (!RangesOrError) { |
561 | // FIXME: Report the error. |
562 | if (!Unit->isDWOUnit()) |
563 | ++NumErrors; |
564 | llvm::consumeError(Err: RangesOrError.takeError()); |
565 | return NumErrors; |
566 | } |
567 | |
568 | const DWARFAddressRangesVector &Ranges = RangesOrError.get(); |
569 | // Build RI for this DIE and check that ranges within this DIE do not |
570 | // overlap. |
571 | DieRangeInfo RI(Die); |
572 | |
573 | // TODO support object files better |
574 | // |
575 | // Some object file formats (i.e. non-MachO) support COMDAT. ELF in |
576 | // particular does so by placing each function into a section. The DWARF data |
577 | // for the function at that point uses a section relative DW_FORM_addrp for |
578 | // the DW_AT_low_pc and a DW_FORM_data4 for the offset as the DW_AT_high_pc. |
579 | // In such a case, when the Die is the CU, the ranges will overlap, and we |
580 | // will flag valid conflicting ranges as invalid. |
581 | // |
582 | // For such targets, we should read the ranges from the CU and partition them |
583 | // by the section id. The ranges within a particular section should be |
584 | // disjoint, although the ranges across sections may overlap. We would map |
585 | // the child die to the entity that it references and the section with which |
586 | // it is associated. The child would then be checked against the range |
587 | // information for the associated section. |
588 | // |
589 | // For now, simply elide the range verification for the CU DIEs if we are |
590 | // processing an object file. |
591 | |
592 | if (!IsObjectFile || IsMachOObject || Die.getTag() != DW_TAG_compile_unit) { |
593 | bool DumpDieAfterError = false; |
594 | for (const auto &Range : Ranges) { |
595 | if (!Range.valid()) { |
596 | ++NumErrors; |
597 | ErrorCategory.Report(s: "Invalid address range" , detailCallback: [&]() { |
598 | error() << "Invalid address range " << Range << "\n" ; |
599 | DumpDieAfterError = true; |
600 | }); |
601 | continue; |
602 | } |
603 | |
604 | // Verify that ranges don't intersect and also build up the DieRangeInfo |
605 | // address ranges. Don't break out of the loop below early, or we will |
606 | // think this DIE doesn't have all of the address ranges it is supposed |
607 | // to have. Compile units often have DW_AT_ranges that can contain one or |
608 | // more dead stripped address ranges which tend to all be at the same |
609 | // address: 0 or -1. |
610 | if (auto PrevRange = RI.insert(R: Range)) { |
611 | ++NumErrors; |
612 | ErrorCategory.Report(s: "DIE has overlapping DW_AT_ranges" , detailCallback: [&]() { |
613 | error() << "DIE has overlapping ranges in DW_AT_ranges attribute: " |
614 | << *PrevRange << " and " << Range << '\n'; |
615 | DumpDieAfterError = true; |
616 | }); |
617 | } |
618 | } |
619 | if (DumpDieAfterError) |
620 | dump(Die, indent: 2) << '\n'; |
621 | } |
622 | |
623 | // Verify that children don't intersect. |
624 | const auto IntersectingChild = ParentRI.insert(RI); |
625 | if (IntersectingChild != ParentRI.Children.end()) { |
626 | ++NumErrors; |
627 | ErrorCategory.Report(s: "DIEs have overlapping address ranges" , detailCallback: [&]() { |
628 | error() << "DIEs have overlapping address ranges:" ; |
629 | dump(Die); |
630 | dump(Die: IntersectingChild->Die) << '\n'; |
631 | }); |
632 | } |
633 | |
634 | // Verify that ranges are contained within their parent. |
635 | bool ShouldBeContained = !RI.Ranges.empty() && !ParentRI.Ranges.empty() && |
636 | !(Die.getTag() == DW_TAG_subprogram && |
637 | ParentRI.Die.getTag() == DW_TAG_subprogram); |
638 | if (ShouldBeContained && !ParentRI.contains(RHS: RI)) { |
639 | ++NumErrors; |
640 | ErrorCategory.Report( |
641 | s: "DIE address ranges are not contained by parent ranges" , detailCallback: [&]() { |
642 | error() |
643 | << "DIE address ranges are not contained in its parent's ranges:" ; |
644 | dump(Die: ParentRI.Die); |
645 | dump(Die, indent: 2) << '\n'; |
646 | }); |
647 | } |
648 | |
649 | // Recursively check children. |
650 | for (DWARFDie Child : Die) |
651 | NumErrors += verifyDieRanges(Die: Child, ParentRI&: RI); |
652 | |
653 | return NumErrors; |
654 | } |
655 | |
656 | unsigned DWARFVerifier::verifyDebugInfoAttribute(const DWARFDie &Die, |
657 | DWARFAttribute &AttrValue) { |
658 | unsigned NumErrors = 0; |
659 | auto ReportError = [&](StringRef category, const Twine &TitleMsg) { |
660 | ++NumErrors; |
661 | ErrorCategory.Report(s: category, detailCallback: [&]() { |
662 | error() << TitleMsg << '\n'; |
663 | dump(Die) << '\n'; |
664 | }); |
665 | }; |
666 | |
667 | const DWARFObject &DObj = DCtx.getDWARFObj(); |
668 | DWARFUnit *U = Die.getDwarfUnit(); |
669 | const auto Attr = AttrValue.Attr; |
670 | switch (Attr) { |
671 | case DW_AT_ranges: |
672 | // Make sure the offset in the DW_AT_ranges attribute is valid. |
673 | if (auto SectionOffset = AttrValue.Value.getAsSectionOffset()) { |
674 | unsigned DwarfVersion = U->getVersion(); |
675 | const DWARFSection &RangeSection = DwarfVersion < 5 |
676 | ? DObj.getRangesSection() |
677 | : DObj.getRnglistsSection(); |
678 | if (U->isDWOUnit() && RangeSection.Data.empty()) |
679 | break; |
680 | if (*SectionOffset >= RangeSection.Data.size()) |
681 | ReportError("DW_AT_ranges offset out of bounds" , |
682 | "DW_AT_ranges offset is beyond " + |
683 | StringRef(DwarfVersion < 5 ? ".debug_ranges" |
684 | : ".debug_rnglists" ) + |
685 | " bounds: " + llvm::formatv(Fmt: "{0:x8}" , Vals&: *SectionOffset)); |
686 | break; |
687 | } |
688 | ReportError("Invalid DW_AT_ranges encoding" , |
689 | "DIE has invalid DW_AT_ranges encoding:" ); |
690 | break; |
691 | case DW_AT_stmt_list: |
692 | // Make sure the offset in the DW_AT_stmt_list attribute is valid. |
693 | if (auto SectionOffset = AttrValue.Value.getAsSectionOffset()) { |
694 | if (*SectionOffset >= U->getLineSection().Data.size()) |
695 | ReportError("DW_AT_stmt_list offset out of bounds" , |
696 | "DW_AT_stmt_list offset is beyond .debug_line bounds: " + |
697 | llvm::formatv(Fmt: "{0:x8}" , Vals&: *SectionOffset)); |
698 | break; |
699 | } |
700 | ReportError("Invalid DW_AT_stmt_list encoding" , |
701 | "DIE has invalid DW_AT_stmt_list encoding:" ); |
702 | break; |
703 | case DW_AT_location: { |
704 | // FIXME: It might be nice if there's a way to walk location expressions |
705 | // without trying to resolve the address ranges - it'd be a more efficient |
706 | // API (since the API is currently unnecessarily resolving addresses for |
707 | // this use case which only wants to validate the expressions themselves) & |
708 | // then the expressions could be validated even if the addresses can't be |
709 | // resolved. |
710 | // That sort of API would probably look like a callback "for each |
711 | // expression" with some way to lazily resolve the address ranges when |
712 | // needed (& then the existing API used here could be built on top of that - |
713 | // using the callback API to build the data structure and return it). |
714 | if (Expected<std::vector<DWARFLocationExpression>> Loc = |
715 | Die.getLocations(Attr: DW_AT_location)) { |
716 | for (const auto &Entry : *Loc) { |
717 | DataExtractor Data(toStringRef(Input: Entry.Expr), DCtx.isLittleEndian(), 0); |
718 | DWARFExpression Expression(Data, U->getAddressByteSize(), |
719 | U->getFormParams().Format); |
720 | bool Error = |
721 | any_of(Range&: Expression, P: [](const DWARFExpression::Operation &Op) { |
722 | return Op.isError(); |
723 | }); |
724 | if (Error || !Expression.verify(U)) |
725 | ReportError("Invalid DWARF expressions" , |
726 | "DIE contains invalid DWARF expression:" ); |
727 | } |
728 | } else if (Error Err = handleErrors( |
729 | E: Loc.takeError(), Hs: [&](std::unique_ptr<ResolverError> E) { |
730 | return U->isDWOUnit() ? Error::success() |
731 | : Error(std::move(E)); |
732 | })) |
733 | ReportError("Invalid DW_AT_location" , toString(E: std::move(Err))); |
734 | break; |
735 | } |
736 | case DW_AT_specification: |
737 | case DW_AT_abstract_origin: { |
738 | if (auto ReferencedDie = Die.getAttributeValueAsReferencedDie(Attr)) { |
739 | auto DieTag = Die.getTag(); |
740 | auto RefTag = ReferencedDie.getTag(); |
741 | if (DieTag == RefTag) |
742 | break; |
743 | if (DieTag == DW_TAG_inlined_subroutine && RefTag == DW_TAG_subprogram) |
744 | break; |
745 | if (DieTag == DW_TAG_variable && RefTag == DW_TAG_member) |
746 | break; |
747 | // This might be reference to a function declaration. |
748 | if (DieTag == DW_TAG_GNU_call_site && RefTag == DW_TAG_subprogram) |
749 | break; |
750 | ReportError("Incompatible DW_AT_abstract_origin tag reference" , |
751 | "DIE with tag " + TagString(Tag: DieTag) + " has " + |
752 | AttributeString(Attribute: Attr) + |
753 | " that points to DIE with " |
754 | "incompatible tag " + |
755 | TagString(Tag: RefTag)); |
756 | } |
757 | break; |
758 | } |
759 | case DW_AT_type: { |
760 | DWARFDie TypeDie = Die.getAttributeValueAsReferencedDie(Attr: DW_AT_type); |
761 | if (TypeDie && !isType(T: TypeDie.getTag())) { |
762 | ReportError("Incompatible DW_AT_type attribute tag" , |
763 | "DIE has " + AttributeString(Attribute: Attr) + |
764 | " with incompatible tag " + TagString(Tag: TypeDie.getTag())); |
765 | } |
766 | break; |
767 | } |
768 | case DW_AT_call_file: |
769 | case DW_AT_decl_file: { |
770 | if (auto FileIdx = AttrValue.Value.getAsUnsignedConstant()) { |
771 | if (U->isDWOUnit() && !U->isTypeUnit()) |
772 | break; |
773 | const auto *LT = U->getContext().getLineTableForUnit(U); |
774 | if (LT) { |
775 | if (!LT->hasFileAtIndex(FileIndex: *FileIdx)) { |
776 | bool IsZeroIndexed = LT->Prologue.getVersion() >= 5; |
777 | if (std::optional<uint64_t> LastFileIdx = |
778 | LT->getLastValidFileIndex()) { |
779 | ReportError("Invalid file index in DW_AT_decl_file" , |
780 | "DIE has " + AttributeString(Attribute: Attr) + |
781 | " with an invalid file index " + |
782 | llvm::formatv(Fmt: "{0}" , Vals&: *FileIdx) + |
783 | " (valid values are [" + |
784 | (IsZeroIndexed ? "0-" : "1-" ) + |
785 | llvm::formatv(Fmt: "{0}" , Vals&: *LastFileIdx) + "])" ); |
786 | } else { |
787 | ReportError("Invalid file index in DW_AT_decl_file" , |
788 | "DIE has " + AttributeString(Attribute: Attr) + |
789 | " with an invalid file index " + |
790 | llvm::formatv(Fmt: "{0}" , Vals&: *FileIdx) + |
791 | " (the file table in the prologue is empty)" ); |
792 | } |
793 | } |
794 | } else { |
795 | ReportError( |
796 | "File index in DW_AT_decl_file reference CU with no line table" , |
797 | "DIE has " + AttributeString(Attribute: Attr) + |
798 | " that references a file with index " + |
799 | llvm::formatv(Fmt: "{0}" , Vals&: *FileIdx) + |
800 | " and the compile unit has no line table" ); |
801 | } |
802 | } else { |
803 | ReportError("Invalid encoding in DW_AT_decl_file" , |
804 | "DIE has " + AttributeString(Attribute: Attr) + |
805 | " with invalid encoding" ); |
806 | } |
807 | break; |
808 | } |
809 | case DW_AT_call_line: |
810 | case DW_AT_decl_line: { |
811 | if (!AttrValue.Value.getAsUnsignedConstant()) { |
812 | ReportError( |
813 | Attr == DW_AT_call_line ? "Invalid file index in DW_AT_decl_line" |
814 | : "Invalid file index in DW_AT_call_line" , |
815 | "DIE has " + AttributeString(Attribute: Attr) + " with invalid encoding" ); |
816 | } |
817 | break; |
818 | } |
819 | default: |
820 | break; |
821 | } |
822 | return NumErrors; |
823 | } |
824 | |
825 | unsigned DWARFVerifier::verifyDebugInfoForm(const DWARFDie &Die, |
826 | DWARFAttribute &AttrValue, |
827 | ReferenceMap &LocalReferences, |
828 | ReferenceMap &CrossUnitReferences) { |
829 | auto DieCU = Die.getDwarfUnit(); |
830 | unsigned NumErrors = 0; |
831 | const auto Form = AttrValue.Value.getForm(); |
832 | switch (Form) { |
833 | case DW_FORM_ref1: |
834 | case DW_FORM_ref2: |
835 | case DW_FORM_ref4: |
836 | case DW_FORM_ref8: |
837 | case DW_FORM_ref_udata: { |
838 | // Verify all CU relative references are valid CU offsets. |
839 | std::optional<uint64_t> RefVal = AttrValue.Value.getAsRelativeReference(); |
840 | assert(RefVal); |
841 | if (RefVal) { |
842 | auto CUSize = DieCU->getNextUnitOffset() - DieCU->getOffset(); |
843 | auto CUOffset = AttrValue.Value.getRawUValue(); |
844 | if (CUOffset >= CUSize) { |
845 | ++NumErrors; |
846 | ErrorCategory.Report(s: "Invalid CU offset" , detailCallback: [&]() { |
847 | error() << FormEncodingString(Encoding: Form) << " CU offset " |
848 | << format(Fmt: "0x%08" PRIx64, Vals: CUOffset) |
849 | << " is invalid (must be less than CU size of " |
850 | << format(Fmt: "0x%08" PRIx64, Vals: CUSize) << "):\n" ; |
851 | Die.dump(OS, indent: 0, DumpOpts); |
852 | dump(Die) << '\n'; |
853 | }); |
854 | } else { |
855 | // Valid reference, but we will verify it points to an actual |
856 | // DIE later. |
857 | LocalReferences[AttrValue.Value.getUnit()->getOffset() + *RefVal] |
858 | .insert(x: Die.getOffset()); |
859 | } |
860 | } |
861 | break; |
862 | } |
863 | case DW_FORM_ref_addr: { |
864 | // Verify all absolute DIE references have valid offsets in the |
865 | // .debug_info section. |
866 | std::optional<uint64_t> RefVal = AttrValue.Value.getAsDebugInfoReference(); |
867 | assert(RefVal); |
868 | if (RefVal) { |
869 | if (*RefVal >= DieCU->getInfoSection().Data.size()) { |
870 | ++NumErrors; |
871 | ErrorCategory.Report(s: "DW_FORM_ref_addr offset out of bounds" , detailCallback: [&]() { |
872 | error() << "DW_FORM_ref_addr offset beyond .debug_info " |
873 | "bounds:\n" ; |
874 | dump(Die) << '\n'; |
875 | }); |
876 | } else { |
877 | // Valid reference, but we will verify it points to an actual |
878 | // DIE later. |
879 | CrossUnitReferences[*RefVal].insert(x: Die.getOffset()); |
880 | } |
881 | } |
882 | break; |
883 | } |
884 | case DW_FORM_strp: |
885 | case DW_FORM_strx: |
886 | case DW_FORM_strx1: |
887 | case DW_FORM_strx2: |
888 | case DW_FORM_strx3: |
889 | case DW_FORM_strx4: |
890 | case DW_FORM_line_strp: { |
891 | if (Error E = AttrValue.Value.getAsCString().takeError()) { |
892 | ++NumErrors; |
893 | std::string ErrMsg = toString(E: std::move(E)); |
894 | ErrorCategory.Report(s: "Invalid DW_FORM attribute" , detailCallback: [&]() { |
895 | error() << ErrMsg << ":\n" ; |
896 | dump(Die) << '\n'; |
897 | }); |
898 | } |
899 | break; |
900 | } |
901 | default: |
902 | break; |
903 | } |
904 | return NumErrors; |
905 | } |
906 | |
907 | unsigned DWARFVerifier::verifyDebugInfoReferences( |
908 | const ReferenceMap &References, |
909 | llvm::function_ref<DWARFUnit *(uint64_t)> GetUnitForOffset) { |
910 | auto GetDIEForOffset = [&](uint64_t Offset) { |
911 | if (DWARFUnit *U = GetUnitForOffset(Offset)) |
912 | return U->getDIEForOffset(Offset); |
913 | return DWARFDie(); |
914 | }; |
915 | unsigned NumErrors = 0; |
916 | for (const std::pair<const uint64_t, std::set<uint64_t>> &Pair : |
917 | References) { |
918 | if (GetDIEForOffset(Pair.first)) |
919 | continue; |
920 | ++NumErrors; |
921 | ErrorCategory.Report(s: "Invalid DIE reference" , detailCallback: [&]() { |
922 | error() << "invalid DIE reference " << format(Fmt: "0x%08" PRIx64, Vals: Pair.first) |
923 | << ". Offset is in between DIEs:\n" ; |
924 | for (auto Offset : Pair.second) |
925 | dump(Die: GetDIEForOffset(Offset)) << '\n'; |
926 | OS << "\n" ; |
927 | }); |
928 | } |
929 | return NumErrors; |
930 | } |
931 | |
932 | void DWARFVerifier::verifyDebugLineStmtOffsets() { |
933 | std::map<uint64_t, DWARFDie> StmtListToDie; |
934 | for (const auto &CU : DCtx.compile_units()) { |
935 | auto Die = CU->getUnitDIE(); |
936 | // Get the attribute value as a section offset. No need to produce an |
937 | // error here if the encoding isn't correct because we validate this in |
938 | // the .debug_info verifier. |
939 | auto StmtSectionOffset = toSectionOffset(V: Die.find(Attr: DW_AT_stmt_list)); |
940 | if (!StmtSectionOffset) |
941 | continue; |
942 | const uint64_t LineTableOffset = *StmtSectionOffset; |
943 | auto LineTable = DCtx.getLineTableForUnit(U: CU.get()); |
944 | if (LineTableOffset < DCtx.getDWARFObj().getLineSection().Data.size()) { |
945 | if (!LineTable) { |
946 | ++NumDebugLineErrors; |
947 | ErrorCategory.Report(s: "Unparsable .debug_line entry" , detailCallback: [&]() { |
948 | error() << ".debug_line[" << format(Fmt: "0x%08" PRIx64, Vals: LineTableOffset) |
949 | << "] was not able to be parsed for CU:\n" ; |
950 | dump(Die) << '\n'; |
951 | }); |
952 | continue; |
953 | } |
954 | } else { |
955 | // Make sure we don't get a valid line table back if the offset is wrong. |
956 | assert(LineTable == nullptr); |
957 | // Skip this line table as it isn't valid. No need to create an error |
958 | // here because we validate this in the .debug_info verifier. |
959 | continue; |
960 | } |
961 | auto Iter = StmtListToDie.find(x: LineTableOffset); |
962 | if (Iter != StmtListToDie.end()) { |
963 | ++NumDebugLineErrors; |
964 | ErrorCategory.Report(s: "Identical DW_AT_stmt_list section offset" , detailCallback: [&]() { |
965 | error() << "two compile unit DIEs, " |
966 | << format(Fmt: "0x%08" PRIx64, Vals: Iter->second.getOffset()) << " and " |
967 | << format(Fmt: "0x%08" PRIx64, Vals: Die.getOffset()) |
968 | << ", have the same DW_AT_stmt_list section offset:\n" ; |
969 | dump(Die: Iter->second); |
970 | dump(Die) << '\n'; |
971 | }); |
972 | // Already verified this line table before, no need to do it again. |
973 | continue; |
974 | } |
975 | StmtListToDie[LineTableOffset] = Die; |
976 | } |
977 | } |
978 | |
979 | void DWARFVerifier::verifyDebugLineRows() { |
980 | for (const auto &CU : DCtx.compile_units()) { |
981 | auto Die = CU->getUnitDIE(); |
982 | auto LineTable = DCtx.getLineTableForUnit(U: CU.get()); |
983 | // If there is no line table we will have created an error in the |
984 | // .debug_info verifier or in verifyDebugLineStmtOffsets(). |
985 | if (!LineTable) |
986 | continue; |
987 | |
988 | // Verify prologue. |
989 | bool isDWARF5 = LineTable->Prologue.getVersion() >= 5; |
990 | uint32_t MaxDirIndex = LineTable->Prologue.IncludeDirectories.size(); |
991 | uint32_t MinFileIndex = isDWARF5 ? 0 : 1; |
992 | uint32_t FileIndex = MinFileIndex; |
993 | StringMap<uint16_t> FullPathMap; |
994 | for (const auto &FileName : LineTable->Prologue.FileNames) { |
995 | // Verify directory index. |
996 | if (FileName.DirIdx > MaxDirIndex) { |
997 | ++NumDebugLineErrors; |
998 | ErrorCategory.Report( |
999 | s: "Invalid index in .debug_line->prologue.file_names->dir_idx" , |
1000 | detailCallback: [&]() { |
1001 | error() << ".debug_line[" |
1002 | << format(Fmt: "0x%08" PRIx64, |
1003 | Vals: *toSectionOffset(V: Die.find(Attr: DW_AT_stmt_list))) |
1004 | << "].prologue.file_names[" << FileIndex |
1005 | << "].dir_idx contains an invalid index: " |
1006 | << FileName.DirIdx << "\n" ; |
1007 | }); |
1008 | } |
1009 | |
1010 | // Check file paths for duplicates. |
1011 | std::string FullPath; |
1012 | const bool HasFullPath = LineTable->getFileNameByIndex( |
1013 | FileIndex, CompDir: CU->getCompilationDir(), |
1014 | Kind: DILineInfoSpecifier::FileLineInfoKind::AbsoluteFilePath, Result&: FullPath); |
1015 | assert(HasFullPath && "Invalid index?" ); |
1016 | (void)HasFullPath; |
1017 | auto It = FullPathMap.find(Key: FullPath); |
1018 | if (It == FullPathMap.end()) |
1019 | FullPathMap[FullPath] = FileIndex; |
1020 | else if (It->second != FileIndex && DumpOpts.Verbose) { |
1021 | warn() << ".debug_line[" |
1022 | << format(Fmt: "0x%08" PRIx64, |
1023 | Vals: *toSectionOffset(V: Die.find(Attr: DW_AT_stmt_list))) |
1024 | << "].prologue.file_names[" << FileIndex |
1025 | << "] is a duplicate of file_names[" << It->second << "]\n" ; |
1026 | } |
1027 | |
1028 | FileIndex++; |
1029 | } |
1030 | |
1031 | // Nothing to verify in a line table with a single row containing the end |
1032 | // sequence. |
1033 | if (LineTable->Rows.size() == 1 && LineTable->Rows.front().EndSequence) |
1034 | continue; |
1035 | |
1036 | // Verify rows. |
1037 | uint64_t PrevAddress = 0; |
1038 | uint32_t RowIndex = 0; |
1039 | for (const auto &Row : LineTable->Rows) { |
1040 | // Verify row address. |
1041 | if (Row.Address.Address < PrevAddress) { |
1042 | ++NumDebugLineErrors; |
1043 | ErrorCategory.Report( |
1044 | s: "decreasing address between debug_line rows" , detailCallback: [&]() { |
1045 | error() << ".debug_line[" |
1046 | << format(Fmt: "0x%08" PRIx64, |
1047 | Vals: *toSectionOffset(V: Die.find(Attr: DW_AT_stmt_list))) |
1048 | << "] row[" << RowIndex |
1049 | << "] decreases in address from previous row:\n" ; |
1050 | |
1051 | DWARFDebugLine::Row::dumpTableHeader(OS, Indent: 0); |
1052 | if (RowIndex > 0) |
1053 | LineTable->Rows[RowIndex - 1].dump(OS); |
1054 | Row.dump(OS); |
1055 | OS << '\n'; |
1056 | }); |
1057 | } |
1058 | |
1059 | if (!LineTable->hasFileAtIndex(FileIndex: Row.File)) { |
1060 | ++NumDebugLineErrors; |
1061 | ErrorCategory.Report(s: "Invalid file index in debug_line" , detailCallback: [&]() { |
1062 | error() << ".debug_line[" |
1063 | << format(Fmt: "0x%08" PRIx64, |
1064 | Vals: *toSectionOffset(V: Die.find(Attr: DW_AT_stmt_list))) |
1065 | << "][" << RowIndex << "] has invalid file index " << Row.File |
1066 | << " (valid values are [" << MinFileIndex << ',' |
1067 | << LineTable->Prologue.FileNames.size() |
1068 | << (isDWARF5 ? ")" : "]" ) << "):\n" ; |
1069 | DWARFDebugLine::Row::dumpTableHeader(OS, Indent: 0); |
1070 | Row.dump(OS); |
1071 | OS << '\n'; |
1072 | }); |
1073 | } |
1074 | if (Row.EndSequence) |
1075 | PrevAddress = 0; |
1076 | else |
1077 | PrevAddress = Row.Address.Address; |
1078 | ++RowIndex; |
1079 | } |
1080 | } |
1081 | } |
1082 | |
1083 | DWARFVerifier::DWARFVerifier(raw_ostream &S, DWARFContext &D, |
1084 | DIDumpOptions DumpOpts) |
1085 | : OS(S), DCtx(D), DumpOpts(std::move(DumpOpts)), IsObjectFile(false), |
1086 | IsMachOObject(false) { |
1087 | ErrorCategory.ShowDetail(showDetail: this->DumpOpts.Verbose || |
1088 | !this->DumpOpts.ShowAggregateErrors); |
1089 | if (const auto *F = DCtx.getDWARFObj().getFile()) { |
1090 | IsObjectFile = F->isRelocatableObject(); |
1091 | IsMachOObject = F->isMachO(); |
1092 | } |
1093 | } |
1094 | |
1095 | bool DWARFVerifier::handleDebugLine() { |
1096 | NumDebugLineErrors = 0; |
1097 | OS << "Verifying .debug_line...\n" ; |
1098 | verifyDebugLineStmtOffsets(); |
1099 | verifyDebugLineRows(); |
1100 | return NumDebugLineErrors == 0; |
1101 | } |
1102 | |
1103 | unsigned DWARFVerifier::(const DWARFSection *AccelSection, |
1104 | DataExtractor *StrData, |
1105 | const char *SectionName) { |
1106 | unsigned NumErrors = 0; |
1107 | DWARFDataExtractor AccelSectionData(DCtx.getDWARFObj(), *AccelSection, |
1108 | DCtx.isLittleEndian(), 0); |
1109 | AppleAcceleratorTable AccelTable(AccelSectionData, *StrData); |
1110 | |
1111 | OS << "Verifying " << SectionName << "...\n" ; |
1112 | |
1113 | // Verify that the fixed part of the header is not too short. |
1114 | if (!AccelSectionData.isValidOffset(offset: AccelTable.getSizeHdr())) { |
1115 | ErrorCategory.Report(s: "Section is too small to fit a section header" , detailCallback: [&]() { |
1116 | error() << "Section is too small to fit a section header.\n" ; |
1117 | }); |
1118 | return 1; |
1119 | } |
1120 | |
1121 | // Verify that the section is not too short. |
1122 | if (Error E = AccelTable.extract()) { |
1123 | std::string Msg = toString(E: std::move(E)); |
1124 | ErrorCategory.Report(s: "Section is too small to fit a section header" , |
1125 | detailCallback: [&]() { error() << Msg << '\n'; }); |
1126 | return 1; |
1127 | } |
1128 | |
1129 | // Verify that all buckets have a valid hash index or are empty. |
1130 | uint32_t NumBuckets = AccelTable.getNumBuckets(); |
1131 | uint32_t NumHashes = AccelTable.getNumHashes(); |
1132 | |
1133 | uint64_t BucketsOffset = |
1134 | AccelTable.getSizeHdr() + AccelTable.getHeaderDataLength(); |
1135 | uint64_t HashesBase = BucketsOffset + NumBuckets * 4; |
1136 | uint64_t OffsetsBase = HashesBase + NumHashes * 4; |
1137 | for (uint32_t BucketIdx = 0; BucketIdx < NumBuckets; ++BucketIdx) { |
1138 | uint32_t HashIdx = AccelSectionData.getU32(offset_ptr: &BucketsOffset); |
1139 | if (HashIdx >= NumHashes && HashIdx != UINT32_MAX) { |
1140 | ErrorCategory.Report(s: "Invalid hash index" , detailCallback: [&]() { |
1141 | error() << format(Fmt: "Bucket[%d] has invalid hash index: %u.\n" , Vals: BucketIdx, |
1142 | Vals: HashIdx); |
1143 | }); |
1144 | ++NumErrors; |
1145 | } |
1146 | } |
1147 | uint32_t NumAtoms = AccelTable.getAtomsDesc().size(); |
1148 | if (NumAtoms == 0) { |
1149 | ErrorCategory.Report(s: "No atoms" , detailCallback: [&]() { |
1150 | error() << "No atoms: failed to read HashData.\n" ; |
1151 | }); |
1152 | return 1; |
1153 | } |
1154 | if (!AccelTable.validateForms()) { |
1155 | ErrorCategory.Report(s: "Unsupported form" , detailCallback: [&]() { |
1156 | error() << "Unsupported form: failed to read HashData.\n" ; |
1157 | }); |
1158 | return 1; |
1159 | } |
1160 | |
1161 | for (uint32_t HashIdx = 0; HashIdx < NumHashes; ++HashIdx) { |
1162 | uint64_t HashOffset = HashesBase + 4 * HashIdx; |
1163 | uint64_t DataOffset = OffsetsBase + 4 * HashIdx; |
1164 | uint32_t Hash = AccelSectionData.getU32(offset_ptr: &HashOffset); |
1165 | uint64_t HashDataOffset = AccelSectionData.getU32(offset_ptr: &DataOffset); |
1166 | if (!AccelSectionData.isValidOffsetForDataOfSize(offset: HashDataOffset, |
1167 | length: sizeof(uint64_t))) { |
1168 | ErrorCategory.Report(s: "Invalid HashData offset" , detailCallback: [&]() { |
1169 | error() << format(Fmt: "Hash[%d] has invalid HashData offset: " |
1170 | "0x%08" PRIx64 ".\n" , |
1171 | Vals: HashIdx, Vals: HashDataOffset); |
1172 | }); |
1173 | ++NumErrors; |
1174 | } |
1175 | |
1176 | uint64_t StrpOffset; |
1177 | uint64_t StringOffset; |
1178 | uint32_t StringCount = 0; |
1179 | uint64_t Offset; |
1180 | unsigned Tag; |
1181 | while ((StrpOffset = AccelSectionData.getU32(offset_ptr: &HashDataOffset)) != 0) { |
1182 | const uint32_t NumHashDataObjects = |
1183 | AccelSectionData.getU32(offset_ptr: &HashDataOffset); |
1184 | for (uint32_t HashDataIdx = 0; HashDataIdx < NumHashDataObjects; |
1185 | ++HashDataIdx) { |
1186 | std::tie(args&: Offset, args&: Tag) = AccelTable.readAtoms(HashDataOffset: &HashDataOffset); |
1187 | auto Die = DCtx.getDIEForOffset(Offset); |
1188 | if (!Die) { |
1189 | const uint32_t BucketIdx = |
1190 | NumBuckets ? (Hash % NumBuckets) : UINT32_MAX; |
1191 | StringOffset = StrpOffset; |
1192 | const char *Name = StrData->getCStr(OffsetPtr: &StringOffset); |
1193 | if (!Name) |
1194 | Name = "<NULL>" ; |
1195 | |
1196 | ErrorCategory.Report(s: "Invalid DIE offset" , detailCallback: [&]() { |
1197 | error() << format( |
1198 | Fmt: "%s Bucket[%d] Hash[%d] = 0x%08x " |
1199 | "Str[%u] = 0x%08" PRIx64 " DIE[%d] = 0x%08" PRIx64 " " |
1200 | "is not a valid DIE offset for \"%s\".\n" , |
1201 | Vals: SectionName, Vals: BucketIdx, Vals: HashIdx, Vals: Hash, Vals: StringCount, Vals: StrpOffset, |
1202 | Vals: HashDataIdx, Vals: Offset, Vals: Name); |
1203 | }); |
1204 | |
1205 | ++NumErrors; |
1206 | continue; |
1207 | } |
1208 | if ((Tag != dwarf::DW_TAG_null) && (Die.getTag() != Tag)) { |
1209 | ErrorCategory.Report(s: "Mismatched Tag in accellerator table" , detailCallback: [&]() { |
1210 | error() << "Tag " << dwarf::TagString(Tag) |
1211 | << " in accelerator table does not match Tag " |
1212 | << dwarf::TagString(Tag: Die.getTag()) << " of DIE[" |
1213 | << HashDataIdx << "].\n" ; |
1214 | }); |
1215 | ++NumErrors; |
1216 | } |
1217 | } |
1218 | ++StringCount; |
1219 | } |
1220 | } |
1221 | return NumErrors; |
1222 | } |
1223 | |
1224 | unsigned |
1225 | DWARFVerifier::verifyDebugNamesCULists(const DWARFDebugNames &AccelTable) { |
1226 | // A map from CU offset to the (first) Name Index offset which claims to index |
1227 | // this CU. |
1228 | DenseMap<uint64_t, uint64_t> CUMap; |
1229 | const uint64_t NotIndexed = std::numeric_limits<uint64_t>::max(); |
1230 | |
1231 | CUMap.reserve(NumEntries: DCtx.getNumCompileUnits()); |
1232 | for (const auto &CU : DCtx.compile_units()) |
1233 | CUMap[CU->getOffset()] = NotIndexed; |
1234 | |
1235 | unsigned NumErrors = 0; |
1236 | for (const DWARFDebugNames::NameIndex &NI : AccelTable) { |
1237 | if (NI.getCUCount() == 0) { |
1238 | ErrorCategory.Report(s: "Name Index doesn't index any CU" , detailCallback: [&]() { |
1239 | error() << formatv(Fmt: "Name Index @ {0:x} does not index any CU\n" , |
1240 | Vals: NI.getUnitOffset()); |
1241 | }); |
1242 | ++NumErrors; |
1243 | continue; |
1244 | } |
1245 | for (uint32_t CU = 0, End = NI.getCUCount(); CU < End; ++CU) { |
1246 | uint64_t Offset = NI.getCUOffset(CU); |
1247 | auto Iter = CUMap.find(Val: Offset); |
1248 | |
1249 | if (Iter == CUMap.end()) { |
1250 | ErrorCategory.Report(s: "Name Index references non-existing CU" , detailCallback: [&]() { |
1251 | error() << formatv( |
1252 | Fmt: "Name Index @ {0:x} references a non-existing CU @ {1:x}\n" , |
1253 | Vals: NI.getUnitOffset(), Vals&: Offset); |
1254 | }); |
1255 | ++NumErrors; |
1256 | continue; |
1257 | } |
1258 | |
1259 | if (Iter->second != NotIndexed) { |
1260 | ErrorCategory.Report(s: "Duplicate Name Index" , detailCallback: [&]() { |
1261 | error() << formatv( |
1262 | Fmt: "Name Index @ {0:x} references a CU @ {1:x}, but " |
1263 | "this CU is already indexed by Name Index @ {2:x}\n" , |
1264 | Vals: NI.getUnitOffset(), Vals&: Offset, Vals&: Iter->second); |
1265 | }); |
1266 | continue; |
1267 | } |
1268 | Iter->second = NI.getUnitOffset(); |
1269 | } |
1270 | } |
1271 | |
1272 | for (const auto &KV : CUMap) { |
1273 | if (KV.second == NotIndexed) |
1274 | warn() << formatv(Fmt: "CU @ {0:x} not covered by any Name Index\n" , Vals: KV.first); |
1275 | } |
1276 | |
1277 | return NumErrors; |
1278 | } |
1279 | |
1280 | unsigned |
1281 | DWARFVerifier::(const DWARFDebugNames::NameIndex &NI, |
1282 | const DataExtractor &StrData) { |
1283 | struct BucketInfo { |
1284 | uint32_t Bucket; |
1285 | uint32_t Index; |
1286 | |
1287 | constexpr BucketInfo(uint32_t Bucket, uint32_t Index) |
1288 | : Bucket(Bucket), Index(Index) {} |
1289 | bool operator<(const BucketInfo &RHS) const { return Index < RHS.Index; } |
1290 | }; |
1291 | |
1292 | uint32_t NumErrors = 0; |
1293 | if (NI.getBucketCount() == 0) { |
1294 | warn() << formatv(Fmt: "Name Index @ {0:x} does not contain a hash table.\n" , |
1295 | Vals: NI.getUnitOffset()); |
1296 | return NumErrors; |
1297 | } |
1298 | |
1299 | // Build up a list of (Bucket, Index) pairs. We use this later to verify that |
1300 | // each Name is reachable from the appropriate bucket. |
1301 | std::vector<BucketInfo> BucketStarts; |
1302 | BucketStarts.reserve(n: NI.getBucketCount() + 1); |
1303 | for (uint32_t Bucket = 0, End = NI.getBucketCount(); Bucket < End; ++Bucket) { |
1304 | uint32_t Index = NI.getBucketArrayEntry(Bucket); |
1305 | if (Index > NI.getNameCount()) { |
1306 | ErrorCategory.Report(s: "Name Index Bucket contains invalid value" , detailCallback: [&]() { |
1307 | error() << formatv(Fmt: "Bucket {0} of Name Index @ {1:x} contains invalid " |
1308 | "value {2}. Valid range is [0, {3}].\n" , |
1309 | Vals&: Bucket, Vals: NI.getUnitOffset(), Vals&: Index, |
1310 | Vals: NI.getNameCount()); |
1311 | }); |
1312 | ++NumErrors; |
1313 | continue; |
1314 | } |
1315 | if (Index > 0) |
1316 | BucketStarts.emplace_back(args&: Bucket, args&: Index); |
1317 | } |
1318 | |
1319 | // If there were any buckets with invalid values, skip further checks as they |
1320 | // will likely produce many errors which will only confuse the actual root |
1321 | // problem. |
1322 | if (NumErrors > 0) |
1323 | return NumErrors; |
1324 | |
1325 | // Sort the list in the order of increasing "Index" entries. |
1326 | array_pod_sort(Start: BucketStarts.begin(), End: BucketStarts.end()); |
1327 | |
1328 | // Insert a sentinel entry at the end, so we can check that the end of the |
1329 | // table is covered in the loop below. |
1330 | BucketStarts.emplace_back(args: NI.getBucketCount(), args: NI.getNameCount() + 1); |
1331 | |
1332 | // Loop invariant: NextUncovered is the (1-based) index of the first Name |
1333 | // which is not reachable by any of the buckets we processed so far (and |
1334 | // hasn't been reported as uncovered). |
1335 | uint32_t NextUncovered = 1; |
1336 | for (const BucketInfo &B : BucketStarts) { |
1337 | // Under normal circumstances B.Index be equal to NextUncovered, but it can |
1338 | // be less if a bucket points to names which are already known to be in some |
1339 | // bucket we processed earlier. In that case, we won't trigger this error, |
1340 | // but report the mismatched hash value error instead. (We know the hash |
1341 | // will not match because we have already verified that the name's hash |
1342 | // puts it into the previous bucket.) |
1343 | if (B.Index > NextUncovered) { |
1344 | ErrorCategory.Report(s: "Name table entries uncovered by hash table" , detailCallback: [&]() { |
1345 | error() << formatv(Fmt: "Name Index @ {0:x}: Name table entries [{1}, {2}] " |
1346 | "are not covered by the hash table.\n" , |
1347 | Vals: NI.getUnitOffset(), Vals&: NextUncovered, Vals: B.Index - 1); |
1348 | }); |
1349 | ++NumErrors; |
1350 | } |
1351 | uint32_t Idx = B.Index; |
1352 | |
1353 | // The rest of the checks apply only to non-sentinel entries. |
1354 | if (B.Bucket == NI.getBucketCount()) |
1355 | break; |
1356 | |
1357 | // This triggers if a non-empty bucket points to a name with a mismatched |
1358 | // hash. Clients are likely to interpret this as an empty bucket, because a |
1359 | // mismatched hash signals the end of a bucket, but if this is indeed an |
1360 | // empty bucket, the producer should have signalled this by marking the |
1361 | // bucket as empty. |
1362 | uint32_t FirstHash = NI.getHashArrayEntry(Index: Idx); |
1363 | if (FirstHash % NI.getBucketCount() != B.Bucket) { |
1364 | ErrorCategory.Report(s: "Name Index point to mismatched hash value" , detailCallback: [&]() { |
1365 | error() << formatv( |
1366 | Fmt: "Name Index @ {0:x}: Bucket {1} is not empty but points to a " |
1367 | "mismatched hash value {2:x} (belonging to bucket {3}).\n" , |
1368 | Vals: NI.getUnitOffset(), Vals: B.Bucket, Vals&: FirstHash, |
1369 | Vals: FirstHash % NI.getBucketCount()); |
1370 | }); |
1371 | ++NumErrors; |
1372 | } |
1373 | |
1374 | // This find the end of this bucket and also verifies that all the hashes in |
1375 | // this bucket are correct by comparing the stored hashes to the ones we |
1376 | // compute ourselves. |
1377 | while (Idx <= NI.getNameCount()) { |
1378 | uint32_t Hash = NI.getHashArrayEntry(Index: Idx); |
1379 | if (Hash % NI.getBucketCount() != B.Bucket) |
1380 | break; |
1381 | |
1382 | const char *Str = NI.getNameTableEntry(Index: Idx).getString(); |
1383 | if (caseFoldingDjbHash(Buffer: Str) != Hash) { |
1384 | ErrorCategory.Report( |
1385 | s: "String hash doesn't match Name Index hash" , detailCallback: [&]() { |
1386 | error() << formatv( |
1387 | Fmt: "Name Index @ {0:x}: String ({1}) at index {2} " |
1388 | "hashes to {3:x}, but " |
1389 | "the Name Index hash is {4:x}\n" , |
1390 | Vals: NI.getUnitOffset(), Vals&: Str, Vals&: Idx, Vals: caseFoldingDjbHash(Buffer: Str), Vals&: Hash); |
1391 | }); |
1392 | ++NumErrors; |
1393 | } |
1394 | |
1395 | ++Idx; |
1396 | } |
1397 | NextUncovered = std::max(a: NextUncovered, b: Idx); |
1398 | } |
1399 | return NumErrors; |
1400 | } |
1401 | |
1402 | unsigned DWARFVerifier::verifyNameIndexAttribute( |
1403 | const DWARFDebugNames::NameIndex &NI, const DWARFDebugNames::Abbrev &Abbr, |
1404 | DWARFDebugNames::AttributeEncoding AttrEnc) { |
1405 | StringRef FormName = dwarf::FormEncodingString(Encoding: AttrEnc.Form); |
1406 | if (FormName.empty()) { |
1407 | ErrorCategory.Report(s: "Unknown NameIndex Abbreviation" , detailCallback: [&]() { |
1408 | error() << formatv(Fmt: "NameIndex @ {0:x}: Abbreviation {1:x}: {2} uses an " |
1409 | "unknown form: {3}.\n" , |
1410 | Vals: NI.getUnitOffset(), Vals: Abbr.Code, Vals&: AttrEnc.Index, |
1411 | Vals&: AttrEnc.Form); |
1412 | }); |
1413 | return 1; |
1414 | } |
1415 | |
1416 | if (AttrEnc.Index == DW_IDX_type_hash) { |
1417 | if (AttrEnc.Form != dwarf::DW_FORM_data8) { |
1418 | ErrorCategory.Report(s: "Unexpected NameIndex Abbreviation" , detailCallback: [&]() { |
1419 | error() << formatv( |
1420 | Fmt: "NameIndex @ {0:x}: Abbreviation {1:x}: DW_IDX_type_hash " |
1421 | "uses an unexpected form {2} (should be {3}).\n" , |
1422 | Vals: NI.getUnitOffset(), Vals: Abbr.Code, Vals&: AttrEnc.Form, Vals: dwarf::DW_FORM_data8); |
1423 | }); |
1424 | return 1; |
1425 | } |
1426 | return 0; |
1427 | } |
1428 | |
1429 | if (AttrEnc.Index == dwarf::DW_IDX_parent) { |
1430 | constexpr static auto AllowedForms = {dwarf::Form::DW_FORM_flag_present, |
1431 | dwarf::Form::DW_FORM_ref4}; |
1432 | if (!is_contained(Set: AllowedForms, Element: AttrEnc.Form)) { |
1433 | ErrorCategory.Report(s: "Unexpected NameIndex Abbreviation" , detailCallback: [&]() { |
1434 | error() << formatv( |
1435 | Fmt: "NameIndex @ {0:x}: Abbreviation {1:x}: DW_IDX_parent " |
1436 | "uses an unexpected form {2} (should be " |
1437 | "DW_FORM_ref4 or DW_FORM_flag_present).\n" , |
1438 | Vals: NI.getUnitOffset(), Vals: Abbr.Code, Vals&: AttrEnc.Form); |
1439 | }); |
1440 | return 1; |
1441 | } |
1442 | return 0; |
1443 | } |
1444 | |
1445 | // A list of known index attributes and their expected form classes. |
1446 | // DW_IDX_type_hash is handled specially in the check above, as it has a |
1447 | // specific form (not just a form class) we should expect. |
1448 | struct FormClassTable { |
1449 | dwarf::Index Index; |
1450 | DWARFFormValue::FormClass Class; |
1451 | StringLiteral ClassName; |
1452 | }; |
1453 | static constexpr FormClassTable Table[] = { |
1454 | {.Index: dwarf::DW_IDX_compile_unit, .Class: DWARFFormValue::FC_Constant, .ClassName: {"constant" }}, |
1455 | {.Index: dwarf::DW_IDX_type_unit, .Class: DWARFFormValue::FC_Constant, .ClassName: {"constant" }}, |
1456 | {.Index: dwarf::DW_IDX_die_offset, .Class: DWARFFormValue::FC_Reference, .ClassName: {"reference" }}, |
1457 | }; |
1458 | |
1459 | ArrayRef<FormClassTable> TableRef(Table); |
1460 | auto Iter = find_if(Range&: TableRef, P: [AttrEnc](const FormClassTable &T) { |
1461 | return T.Index == AttrEnc.Index; |
1462 | }); |
1463 | if (Iter == TableRef.end()) { |
1464 | warn() << formatv(Fmt: "NameIndex @ {0:x}: Abbreviation {1:x} contains an " |
1465 | "unknown index attribute: {2}.\n" , |
1466 | Vals: NI.getUnitOffset(), Vals: Abbr.Code, Vals&: AttrEnc.Index); |
1467 | return 0; |
1468 | } |
1469 | |
1470 | if (!DWARFFormValue(AttrEnc.Form).isFormClass(FC: Iter->Class)) { |
1471 | ErrorCategory.Report(s: "Unexpected NameIndex Abbreviation" , detailCallback: [&]() { |
1472 | error() << formatv(Fmt: "NameIndex @ {0:x}: Abbreviation {1:x}: {2} uses an " |
1473 | "unexpected form {3} (expected form class {4}).\n" , |
1474 | Vals: NI.getUnitOffset(), Vals: Abbr.Code, Vals&: AttrEnc.Index, |
1475 | Vals&: AttrEnc.Form, Vals: Iter->ClassName); |
1476 | }); |
1477 | return 1; |
1478 | } |
1479 | return 0; |
1480 | } |
1481 | |
1482 | unsigned |
1483 | DWARFVerifier::verifyNameIndexAbbrevs(const DWARFDebugNames::NameIndex &NI) { |
1484 | if (NI.getLocalTUCount() + NI.getForeignTUCount() > 0) { |
1485 | warn() << formatv(Fmt: "Name Index @ {0:x}: Verifying indexes of type units is " |
1486 | "not currently supported.\n" , |
1487 | Vals: NI.getUnitOffset()); |
1488 | return 0; |
1489 | } |
1490 | |
1491 | unsigned NumErrors = 0; |
1492 | for (const auto &Abbrev : NI.getAbbrevs()) { |
1493 | StringRef TagName = dwarf::TagString(Tag: Abbrev.Tag); |
1494 | if (TagName.empty()) { |
1495 | warn() << formatv(Fmt: "NameIndex @ {0:x}: Abbreviation {1:x} references an " |
1496 | "unknown tag: {2}.\n" , |
1497 | Vals: NI.getUnitOffset(), Vals: Abbrev.Code, Vals: Abbrev.Tag); |
1498 | } |
1499 | SmallSet<unsigned, 5> Attributes; |
1500 | for (const auto &AttrEnc : Abbrev.Attributes) { |
1501 | if (!Attributes.insert(V: AttrEnc.Index).second) { |
1502 | ErrorCategory.Report( |
1503 | s: "NameIndex Abbreviateion contains multiple attributes" , detailCallback: [&]() { |
1504 | error() << formatv( |
1505 | Fmt: "NameIndex @ {0:x}: Abbreviation {1:x} contains " |
1506 | "multiple {2} attributes.\n" , |
1507 | Vals: NI.getUnitOffset(), Vals: Abbrev.Code, Vals: AttrEnc.Index); |
1508 | }); |
1509 | ++NumErrors; |
1510 | continue; |
1511 | } |
1512 | NumErrors += verifyNameIndexAttribute(NI, Abbr: Abbrev, AttrEnc); |
1513 | } |
1514 | |
1515 | if (NI.getCUCount() > 1 && !Attributes.count(V: dwarf::DW_IDX_compile_unit)) { |
1516 | ErrorCategory.Report(s: "Abbreviation contains no attribute" , detailCallback: [&]() { |
1517 | error() << formatv(Fmt: "NameIndex @ {0:x}: Indexing multiple compile units " |
1518 | "and abbreviation {1:x} has no {2} attribute.\n" , |
1519 | Vals: NI.getUnitOffset(), Vals: Abbrev.Code, |
1520 | Vals: dwarf::DW_IDX_compile_unit); |
1521 | }); |
1522 | ++NumErrors; |
1523 | } |
1524 | if (!Attributes.count(V: dwarf::DW_IDX_die_offset)) { |
1525 | ErrorCategory.Report(s: "Abbreviate in NameIndex missing attribute" , detailCallback: [&]() { |
1526 | error() << formatv( |
1527 | Fmt: "NameIndex @ {0:x}: Abbreviation {1:x} has no {2} attribute.\n" , |
1528 | Vals: NI.getUnitOffset(), Vals: Abbrev.Code, Vals: dwarf::DW_IDX_die_offset); |
1529 | }); |
1530 | ++NumErrors; |
1531 | } |
1532 | } |
1533 | return NumErrors; |
1534 | } |
1535 | |
1536 | static SmallVector<std::string, 3> getNames(const DWARFDie &DIE, |
1537 | bool IncludeStrippedTemplateNames, |
1538 | bool IncludeObjCNames = true, |
1539 | bool IncludeLinkageName = true) { |
1540 | SmallVector<std::string, 3> Result; |
1541 | if (const char *Str = DIE.getShortName()) { |
1542 | StringRef Name(Str); |
1543 | Result.emplace_back(Args&: Name); |
1544 | if (IncludeStrippedTemplateNames) { |
1545 | if (std::optional<StringRef> StrippedName = |
1546 | StripTemplateParameters(Name: Result.back())) |
1547 | // Convert to std::string and push; emplacing the StringRef may trigger |
1548 | // a vector resize which may destroy the StringRef memory. |
1549 | Result.push_back(Elt: StrippedName->str()); |
1550 | } |
1551 | |
1552 | if (IncludeObjCNames) { |
1553 | if (std::optional<ObjCSelectorNames> ObjCNames = |
1554 | getObjCNamesIfSelector(Name)) { |
1555 | Result.emplace_back(Args&: ObjCNames->ClassName); |
1556 | Result.emplace_back(Args&: ObjCNames->Selector); |
1557 | if (ObjCNames->ClassNameNoCategory) |
1558 | Result.emplace_back(Args&: *ObjCNames->ClassNameNoCategory); |
1559 | if (ObjCNames->MethodNameNoCategory) |
1560 | Result.push_back(Elt: std::move(*ObjCNames->MethodNameNoCategory)); |
1561 | } |
1562 | } |
1563 | } else if (DIE.getTag() == dwarf::DW_TAG_namespace) |
1564 | Result.emplace_back(Args: "(anonymous namespace)" ); |
1565 | |
1566 | if (IncludeLinkageName) { |
1567 | if (const char *Str = DIE.getLinkageName()) |
1568 | Result.emplace_back(Args&: Str); |
1569 | } |
1570 | |
1571 | return Result; |
1572 | } |
1573 | |
1574 | unsigned DWARFVerifier::verifyNameIndexEntries( |
1575 | const DWARFDebugNames::NameIndex &NI, |
1576 | const DWARFDebugNames::NameTableEntry &NTE) { |
1577 | // Verifying type unit indexes not supported. |
1578 | if (NI.getLocalTUCount() + NI.getForeignTUCount() > 0) |
1579 | return 0; |
1580 | |
1581 | const char *CStr = NTE.getString(); |
1582 | if (!CStr) { |
1583 | ErrorCategory.Report(s: "Unable to get string associated with name" , detailCallback: [&]() { |
1584 | error() << formatv(Fmt: "Name Index @ {0:x}: Unable to get string associated " |
1585 | "with name {1}.\n" , |
1586 | Vals: NI.getUnitOffset(), Vals: NTE.getIndex()); |
1587 | }); |
1588 | return 1; |
1589 | } |
1590 | StringRef Str(CStr); |
1591 | |
1592 | unsigned NumErrors = 0; |
1593 | unsigned NumEntries = 0; |
1594 | uint64_t EntryID = NTE.getEntryOffset(); |
1595 | uint64_t NextEntryID = EntryID; |
1596 | Expected<DWARFDebugNames::Entry> EntryOr = NI.getEntry(Offset: &NextEntryID); |
1597 | for (; EntryOr; ++NumEntries, EntryID = NextEntryID, |
1598 | EntryOr = NI.getEntry(Offset: &NextEntryID)) { |
1599 | uint32_t CUIndex = *EntryOr->getCUIndex(); |
1600 | if (CUIndex > NI.getCUCount()) { |
1601 | ErrorCategory.Report(s: "Name Index entry contains invalid CU index" , detailCallback: [&]() { |
1602 | error() << formatv(Fmt: "Name Index @ {0:x}: Entry @ {1:x} contains an " |
1603 | "invalid CU index ({2}).\n" , |
1604 | Vals: NI.getUnitOffset(), Vals&: EntryID, Vals&: CUIndex); |
1605 | }); |
1606 | ++NumErrors; |
1607 | continue; |
1608 | } |
1609 | uint64_t CUOffset = NI.getCUOffset(CU: CUIndex); |
1610 | uint64_t DIEOffset = CUOffset + *EntryOr->getDIEUnitOffset(); |
1611 | DWARFDie DIE = DCtx.getDIEForOffset(Offset: DIEOffset); |
1612 | if (!DIE) { |
1613 | ErrorCategory.Report(s: "NameIndex references nonexistent DIE" , detailCallback: [&]() { |
1614 | error() << formatv(Fmt: "Name Index @ {0:x}: Entry @ {1:x} references a " |
1615 | "non-existing DIE @ {2:x}.\n" , |
1616 | Vals: NI.getUnitOffset(), Vals&: EntryID, Vals&: DIEOffset); |
1617 | }); |
1618 | ++NumErrors; |
1619 | continue; |
1620 | } |
1621 | if (DIE.getDwarfUnit()->getOffset() != CUOffset) { |
1622 | ErrorCategory.Report(s: "Name index contains mismatched CU of DIE" , detailCallback: [&]() { |
1623 | error() << formatv( |
1624 | Fmt: "Name Index @ {0:x}: Entry @ {1:x}: mismatched CU of " |
1625 | "DIE @ {2:x}: index - {3:x}; debug_info - {4:x}.\n" , |
1626 | Vals: NI.getUnitOffset(), Vals&: EntryID, Vals&: DIEOffset, Vals&: CUOffset, |
1627 | Vals: DIE.getDwarfUnit()->getOffset()); |
1628 | }); |
1629 | ++NumErrors; |
1630 | } |
1631 | if (DIE.getTag() != EntryOr->tag()) { |
1632 | ErrorCategory.Report(s: "Name Index contains mismatched Tag of DIE" , detailCallback: [&]() { |
1633 | error() << formatv( |
1634 | Fmt: "Name Index @ {0:x}: Entry @ {1:x}: mismatched Tag of " |
1635 | "DIE @ {2:x}: index - {3}; debug_info - {4}.\n" , |
1636 | Vals: NI.getUnitOffset(), Vals&: EntryID, Vals&: DIEOffset, Vals: EntryOr->tag(), |
1637 | Vals: DIE.getTag()); |
1638 | }); |
1639 | ++NumErrors; |
1640 | } |
1641 | |
1642 | // We allow an extra name for functions: their name without any template |
1643 | // parameters. |
1644 | auto IncludeStrippedTemplateNames = |
1645 | DIE.getTag() == DW_TAG_subprogram || |
1646 | DIE.getTag() == DW_TAG_inlined_subroutine; |
1647 | auto EntryNames = getNames(DIE, IncludeStrippedTemplateNames); |
1648 | if (!is_contained(Range&: EntryNames, Element: Str)) { |
1649 | ErrorCategory.Report(s: "Name Index contains mismatched name of DIE" , detailCallback: [&]() { |
1650 | error() << formatv(Fmt: "Name Index @ {0:x}: Entry @ {1:x}: mismatched Name " |
1651 | "of DIE @ {2:x}: index - {3}; debug_info - {4}.\n" , |
1652 | Vals: NI.getUnitOffset(), Vals&: EntryID, Vals&: DIEOffset, Vals&: Str, |
1653 | Vals: make_range(x: EntryNames.begin(), y: EntryNames.end())); |
1654 | }); |
1655 | ++NumErrors; |
1656 | } |
1657 | } |
1658 | handleAllErrors( |
1659 | E: EntryOr.takeError(), |
1660 | Handlers: [&](const DWARFDebugNames::SentinelError &) { |
1661 | if (NumEntries > 0) |
1662 | return; |
1663 | ErrorCategory.Report( |
1664 | s: "NameIndex Name is not associated with any entries" , detailCallback: [&]() { |
1665 | error() << formatv(Fmt: "Name Index @ {0:x}: Name {1} ({2}) is " |
1666 | "not associated with any entries.\n" , |
1667 | Vals: NI.getUnitOffset(), Vals: NTE.getIndex(), Vals&: Str); |
1668 | }); |
1669 | ++NumErrors; |
1670 | }, |
1671 | Handlers: [&](const ErrorInfoBase &Info) { |
1672 | ErrorCategory.Report(s: "Uncategorized NameIndex error" , detailCallback: [&]() { |
1673 | error() << formatv(Fmt: "Name Index @ {0:x}: Name {1} ({2}): {3}\n" , |
1674 | Vals: NI.getUnitOffset(), Vals: NTE.getIndex(), Vals&: Str, |
1675 | Vals: Info.message()); |
1676 | }); |
1677 | ++NumErrors; |
1678 | }); |
1679 | return NumErrors; |
1680 | } |
1681 | |
1682 | static bool isVariableIndexable(const DWARFDie &Die, DWARFContext &DCtx) { |
1683 | Expected<std::vector<DWARFLocationExpression>> Loc = |
1684 | Die.getLocations(Attr: DW_AT_location); |
1685 | if (!Loc) { |
1686 | consumeError(Err: Loc.takeError()); |
1687 | return false; |
1688 | } |
1689 | DWARFUnit *U = Die.getDwarfUnit(); |
1690 | for (const auto &Entry : *Loc) { |
1691 | DataExtractor Data(toStringRef(Input: Entry.Expr), DCtx.isLittleEndian(), |
1692 | U->getAddressByteSize()); |
1693 | DWARFExpression Expression(Data, U->getAddressByteSize(), |
1694 | U->getFormParams().Format); |
1695 | bool IsInteresting = |
1696 | any_of(Range&: Expression, P: [](const DWARFExpression::Operation &Op) { |
1697 | return !Op.isError() && (Op.getCode() == DW_OP_addr || |
1698 | Op.getCode() == DW_OP_form_tls_address || |
1699 | Op.getCode() == DW_OP_GNU_push_tls_address); |
1700 | }); |
1701 | if (IsInteresting) |
1702 | return true; |
1703 | } |
1704 | return false; |
1705 | } |
1706 | |
1707 | unsigned DWARFVerifier::verifyNameIndexCompleteness( |
1708 | const DWARFDie &Die, const DWARFDebugNames::NameIndex &NI) { |
1709 | |
1710 | // First check, if the Die should be indexed. The code follows the DWARF v5 |
1711 | // wording as closely as possible. |
1712 | |
1713 | // "All non-defining declarations (that is, debugging information entries |
1714 | // with a DW_AT_declaration attribute) are excluded." |
1715 | if (Die.find(Attr: DW_AT_declaration)) |
1716 | return 0; |
1717 | |
1718 | // "DW_TAG_namespace debugging information entries without a DW_AT_name |
1719 | // attribute are included with the name “(anonymous namespace)”. |
1720 | // All other debugging information entries without a DW_AT_name attribute |
1721 | // are excluded." |
1722 | // "If a subprogram or inlined subroutine is included, and has a |
1723 | // DW_AT_linkage_name attribute, there will be an additional index entry for |
1724 | // the linkage name." |
1725 | auto IncludeLinkageName = Die.getTag() == DW_TAG_subprogram || |
1726 | Die.getTag() == DW_TAG_inlined_subroutine; |
1727 | // We *allow* stripped template names / ObjectiveC names as extra entries into |
1728 | // the table, but we don't *require* them to pass the completeness test. |
1729 | auto IncludeStrippedTemplateNames = false; |
1730 | auto IncludeObjCNames = false; |
1731 | auto EntryNames = getNames(DIE: Die, IncludeStrippedTemplateNames, |
1732 | IncludeObjCNames, IncludeLinkageName); |
1733 | if (EntryNames.empty()) |
1734 | return 0; |
1735 | |
1736 | // We deviate from the specification here, which says: |
1737 | // "The name index must contain an entry for each debugging information entry |
1738 | // that defines a named subprogram, label, variable, type, or namespace, |
1739 | // subject to ..." |
1740 | // Explicitly exclude all TAGs that we know shouldn't be indexed. |
1741 | switch (Die.getTag()) { |
1742 | // Compile units and modules have names but shouldn't be indexed. |
1743 | case DW_TAG_compile_unit: |
1744 | case DW_TAG_module: |
1745 | return 0; |
1746 | |
1747 | // Function and template parameters are not globally visible, so we shouldn't |
1748 | // index them. |
1749 | case DW_TAG_formal_parameter: |
1750 | case DW_TAG_template_value_parameter: |
1751 | case DW_TAG_template_type_parameter: |
1752 | case DW_TAG_GNU_template_parameter_pack: |
1753 | case DW_TAG_GNU_template_template_param: |
1754 | return 0; |
1755 | |
1756 | // Object members aren't globally visible. |
1757 | case DW_TAG_member: |
1758 | return 0; |
1759 | |
1760 | // According to a strict reading of the specification, enumerators should not |
1761 | // be indexed (and LLVM currently does not do that). However, this causes |
1762 | // problems for the debuggers, so we may need to reconsider this. |
1763 | case DW_TAG_enumerator: |
1764 | return 0; |
1765 | |
1766 | // Imported declarations should not be indexed according to the specification |
1767 | // and LLVM currently does not do that. |
1768 | case DW_TAG_imported_declaration: |
1769 | return 0; |
1770 | |
1771 | // "DW_TAG_subprogram, DW_TAG_inlined_subroutine, and DW_TAG_label debugging |
1772 | // information entries without an address attribute (DW_AT_low_pc, |
1773 | // DW_AT_high_pc, DW_AT_ranges, or DW_AT_entry_pc) are excluded." |
1774 | case DW_TAG_subprogram: |
1775 | case DW_TAG_inlined_subroutine: |
1776 | case DW_TAG_label: |
1777 | if (Die.findRecursively( |
1778 | Attrs: {DW_AT_low_pc, DW_AT_high_pc, DW_AT_ranges, DW_AT_entry_pc})) |
1779 | break; |
1780 | return 0; |
1781 | |
1782 | // "DW_TAG_variable debugging information entries with a DW_AT_location |
1783 | // attribute that includes a DW_OP_addr or DW_OP_form_tls_address operator are |
1784 | // included; otherwise, they are excluded." |
1785 | // |
1786 | // LLVM extension: We also add DW_OP_GNU_push_tls_address to this list. |
1787 | case DW_TAG_variable: |
1788 | if (isVariableIndexable(Die, DCtx)) |
1789 | break; |
1790 | return 0; |
1791 | |
1792 | default: |
1793 | break; |
1794 | } |
1795 | |
1796 | // Now we know that our Die should be present in the Index. Let's check if |
1797 | // that's the case. |
1798 | unsigned NumErrors = 0; |
1799 | uint64_t DieUnitOffset = Die.getOffset() - Die.getDwarfUnit()->getOffset(); |
1800 | for (StringRef Name : EntryNames) { |
1801 | if (none_of(Range: NI.equal_range(Key: Name), P: [&](const DWARFDebugNames::Entry &E) { |
1802 | return E.getDIEUnitOffset() == DieUnitOffset; |
1803 | })) { |
1804 | ErrorCategory.Report(s: "Name Index DIE entry missing name" , detailCallback: [&]() { |
1805 | error() << formatv( |
1806 | Fmt: "Name Index @ {0:x}: Entry for DIE @ {1:x} ({2}) with " |
1807 | "name {3} missing.\n" , |
1808 | Vals: NI.getUnitOffset(), Vals: Die.getOffset(), Vals: Die.getTag(), Vals&: Name); |
1809 | }); |
1810 | ++NumErrors; |
1811 | } |
1812 | } |
1813 | return NumErrors; |
1814 | } |
1815 | |
1816 | unsigned DWARFVerifier::(const DWARFSection &AccelSection, |
1817 | const DataExtractor &StrData) { |
1818 | unsigned NumErrors = 0; |
1819 | DWARFDataExtractor AccelSectionData(DCtx.getDWARFObj(), AccelSection, |
1820 | DCtx.isLittleEndian(), 0); |
1821 | DWARFDebugNames AccelTable(AccelSectionData, StrData); |
1822 | |
1823 | OS << "Verifying .debug_names...\n" ; |
1824 | |
1825 | // This verifies that we can read individual name indices and their |
1826 | // abbreviation tables. |
1827 | if (Error E = AccelTable.extract()) { |
1828 | std::string Msg = toString(E: std::move(E)); |
1829 | ErrorCategory.Report(s: "Accelerator Table Error" , |
1830 | detailCallback: [&]() { error() << Msg << '\n'; }); |
1831 | return 1; |
1832 | } |
1833 | |
1834 | NumErrors += verifyDebugNamesCULists(AccelTable); |
1835 | for (const auto &NI : AccelTable) |
1836 | NumErrors += verifyNameIndexBuckets(NI, StrData); |
1837 | for (const auto &NI : AccelTable) |
1838 | NumErrors += verifyNameIndexAbbrevs(NI); |
1839 | |
1840 | // Don't attempt Entry validation if any of the previous checks found errors |
1841 | if (NumErrors > 0) |
1842 | return NumErrors; |
1843 | for (const auto &NI : AccelTable) |
1844 | for (const DWARFDebugNames::NameTableEntry &NTE : NI) |
1845 | NumErrors += verifyNameIndexEntries(NI, NTE); |
1846 | |
1847 | if (NumErrors > 0) |
1848 | return NumErrors; |
1849 | |
1850 | for (const std::unique_ptr<DWARFUnit> &U : DCtx.compile_units()) { |
1851 | if (const DWARFDebugNames::NameIndex *NI = |
1852 | AccelTable.getCUNameIndex(CUOffset: U->getOffset())) { |
1853 | auto *CU = cast<DWARFCompileUnit>(Val: U.get()); |
1854 | for (const DWARFDebugInfoEntry &Die : CU->dies()) |
1855 | NumErrors += verifyNameIndexCompleteness(Die: DWARFDie(CU, &Die), NI: *NI); |
1856 | } |
1857 | } |
1858 | return NumErrors; |
1859 | } |
1860 | |
1861 | bool DWARFVerifier::handleAccelTables() { |
1862 | const DWARFObject &D = DCtx.getDWARFObj(); |
1863 | DataExtractor StrData(D.getStrSection(), DCtx.isLittleEndian(), 0); |
1864 | unsigned NumErrors = 0; |
1865 | if (!D.getAppleNamesSection().Data.empty()) |
1866 | NumErrors += verifyAppleAccelTable(AccelSection: &D.getAppleNamesSection(), StrData: &StrData, |
1867 | SectionName: ".apple_names" ); |
1868 | if (!D.getAppleTypesSection().Data.empty()) |
1869 | NumErrors += verifyAppleAccelTable(AccelSection: &D.getAppleTypesSection(), StrData: &StrData, |
1870 | SectionName: ".apple_types" ); |
1871 | if (!D.getAppleNamespacesSection().Data.empty()) |
1872 | NumErrors += verifyAppleAccelTable(AccelSection: &D.getAppleNamespacesSection(), StrData: &StrData, |
1873 | SectionName: ".apple_namespaces" ); |
1874 | if (!D.getAppleObjCSection().Data.empty()) |
1875 | NumErrors += verifyAppleAccelTable(AccelSection: &D.getAppleObjCSection(), StrData: &StrData, |
1876 | SectionName: ".apple_objc" ); |
1877 | |
1878 | if (!D.getNamesSection().Data.empty()) |
1879 | NumErrors += verifyDebugNames(AccelSection: D.getNamesSection(), StrData); |
1880 | return NumErrors == 0; |
1881 | } |
1882 | |
1883 | bool DWARFVerifier::handleDebugStrOffsets() { |
1884 | OS << "Verifying .debug_str_offsets...\n" ; |
1885 | const DWARFObject &DObj = DCtx.getDWARFObj(); |
1886 | bool Success = true; |
1887 | |
1888 | // dwo sections may contain the legacy debug_str_offsets format (and they |
1889 | // can't be mixed with dwarf 5's format). This section format contains no |
1890 | // header. |
1891 | // As such, check the version from debug_info and, if we are in the legacy |
1892 | // mode (Dwarf <= 4), extract Dwarf32/Dwarf64. |
1893 | std::optional<DwarfFormat> DwoLegacyDwarf4Format; |
1894 | DObj.forEachInfoDWOSections(F: [&](const DWARFSection &S) { |
1895 | if (DwoLegacyDwarf4Format) |
1896 | return; |
1897 | DWARFDataExtractor DebugInfoData(DObj, S, DCtx.isLittleEndian(), 0); |
1898 | uint64_t Offset = 0; |
1899 | DwarfFormat InfoFormat = DebugInfoData.getInitialLength(Off: &Offset).second; |
1900 | if (uint16_t InfoVersion = DebugInfoData.getU16(offset_ptr: &Offset); InfoVersion <= 4) |
1901 | DwoLegacyDwarf4Format = InfoFormat; |
1902 | }); |
1903 | |
1904 | Success &= verifyDebugStrOffsets( |
1905 | LegacyFormat: DwoLegacyDwarf4Format, SectionName: ".debug_str_offsets.dwo" , |
1906 | Section: DObj.getStrOffsetsDWOSection(), StrData: DObj.getStrDWOSection()); |
1907 | Success &= verifyDebugStrOffsets( |
1908 | /*LegacyFormat=*/std::nullopt, SectionName: ".debug_str_offsets" , |
1909 | Section: DObj.getStrOffsetsSection(), StrData: DObj.getStrSection()); |
1910 | return Success; |
1911 | } |
1912 | |
1913 | bool DWARFVerifier::verifyDebugStrOffsets( |
1914 | std::optional<DwarfFormat> LegacyFormat, StringRef SectionName, |
1915 | const DWARFSection &Section, StringRef StrData) { |
1916 | const DWARFObject &DObj = DCtx.getDWARFObj(); |
1917 | |
1918 | DWARFDataExtractor DA(DObj, Section, DCtx.isLittleEndian(), 0); |
1919 | DataExtractor::Cursor C(0); |
1920 | uint64_t NextUnit = 0; |
1921 | bool Success = true; |
1922 | while (C.seek(NewOffSet: NextUnit), C.tell() < DA.getData().size()) { |
1923 | DwarfFormat Format; |
1924 | uint64_t Length; |
1925 | uint64_t StartOffset = C.tell(); |
1926 | if (LegacyFormat) { |
1927 | Format = *LegacyFormat; |
1928 | Length = DA.getData().size(); |
1929 | NextUnit = C.tell() + Length; |
1930 | } else { |
1931 | std::tie(args&: Length, args&: Format) = DA.getInitialLength(C); |
1932 | if (!C) |
1933 | break; |
1934 | if (C.tell() + Length > DA.getData().size()) { |
1935 | ErrorCategory.Report( |
1936 | s: "Section contribution length exceeds available space" , detailCallback: [&]() { |
1937 | error() << formatv( |
1938 | Fmt: "{0}: contribution {1:X}: length exceeds available space " |
1939 | "(contribution " |
1940 | "offset ({1:X}) + length field space ({2:X}) + length " |
1941 | "({3:X}) == " |
1942 | "{4:X} > section size {5:X})\n" , |
1943 | Vals&: SectionName, Vals&: StartOffset, Vals: C.tell() - StartOffset, Vals&: Length, |
1944 | Vals: C.tell() + Length, Vals: DA.getData().size()); |
1945 | }); |
1946 | Success = false; |
1947 | // Nothing more to do - no other contributions to try. |
1948 | break; |
1949 | } |
1950 | NextUnit = C.tell() + Length; |
1951 | uint8_t Version = DA.getU16(C); |
1952 | if (C && Version != 5) { |
1953 | ErrorCategory.Report(s: "Invalid Section version" , detailCallback: [&]() { |
1954 | error() << formatv(Fmt: "{0}: contribution {1:X}: invalid version {2}\n" , |
1955 | Vals&: SectionName, Vals&: StartOffset, Vals&: Version); |
1956 | }); |
1957 | Success = false; |
1958 | // Can't parse the rest of this contribution, since we don't know the |
1959 | // version, but we can pick up with the next contribution. |
1960 | continue; |
1961 | } |
1962 | (void)DA.getU16(C); // padding |
1963 | } |
1964 | uint64_t OffsetByteSize = getDwarfOffsetByteSize(Format); |
1965 | DA.setAddressSize(OffsetByteSize); |
1966 | uint64_t Remainder = (Length - 4) % OffsetByteSize; |
1967 | if (Remainder != 0) { |
1968 | ErrorCategory.Report(s: "Invalid section contribution length" , detailCallback: [&]() { |
1969 | error() << formatv( |
1970 | Fmt: "{0}: contribution {1:X}: invalid length ((length ({2:X}) " |
1971 | "- header (0x4)) % offset size {3:X} == {4:X} != 0)\n" , |
1972 | Vals&: SectionName, Vals&: StartOffset, Vals&: Length, Vals&: OffsetByteSize, Vals&: Remainder); |
1973 | }); |
1974 | Success = false; |
1975 | } |
1976 | for (uint64_t Index = 0; C && C.tell() + OffsetByteSize <= NextUnit; ++Index) { |
1977 | uint64_t OffOff = C.tell(); |
1978 | uint64_t StrOff = DA.getAddress(C); |
1979 | // check StrOff refers to the start of a string |
1980 | if (StrOff == 0) |
1981 | continue; |
1982 | if (StrData.size() <= StrOff) { |
1983 | ErrorCategory.Report( |
1984 | s: "String offset out of bounds of string section" , detailCallback: [&]() { |
1985 | error() << formatv( |
1986 | Fmt: "{0}: contribution {1:X}: index {2:X}: invalid string " |
1987 | "offset *{3:X} == {4:X}, is beyond the bounds of the string " |
1988 | "section of length {5:X}\n" , |
1989 | Vals&: SectionName, Vals&: StartOffset, Vals&: Index, Vals&: OffOff, Vals&: StrOff, |
1990 | Vals: StrData.size()); |
1991 | }); |
1992 | continue; |
1993 | } |
1994 | if (StrData[StrOff - 1] == '\0') |
1995 | continue; |
1996 | ErrorCategory.Report( |
1997 | s: "Section contribution contains invalid string offset" , detailCallback: [&]() { |
1998 | error() << formatv( |
1999 | Fmt: "{0}: contribution {1:X}: index {2:X}: invalid string " |
2000 | "offset *{3:X} == {4:X}, is neither zero nor " |
2001 | "immediately following a null character\n" , |
2002 | Vals&: SectionName, Vals&: StartOffset, Vals&: Index, Vals&: OffOff, Vals&: StrOff); |
2003 | }); |
2004 | Success = false; |
2005 | } |
2006 | } |
2007 | |
2008 | if (Error E = C.takeError()) { |
2009 | std::string Msg = toString(E: std::move(E)); |
2010 | ErrorCategory.Report(s: "String offset error" , detailCallback: [&]() { |
2011 | error() << SectionName << ": " << Msg << '\n'; |
2012 | return false; |
2013 | }); |
2014 | } |
2015 | return Success; |
2016 | } |
2017 | |
2018 | void OutputCategoryAggregator::Report( |
2019 | StringRef s, std::function<void(void)> detailCallback) { |
2020 | Aggregation[std::string(s)]++; |
2021 | if (IncludeDetail) |
2022 | detailCallback(); |
2023 | } |
2024 | |
2025 | void OutputCategoryAggregator::EnumerateResults( |
2026 | std::function<void(StringRef, unsigned)> handleCounts) { |
2027 | for (auto &&[name, count] : Aggregation) { |
2028 | handleCounts(name, count); |
2029 | } |
2030 | } |
2031 | |
2032 | void DWARFVerifier::summarize() { |
2033 | if (DumpOpts.ShowAggregateErrors && ErrorCategory.GetNumCategories()) { |
2034 | error() << "Aggregated error counts:\n" ; |
2035 | ErrorCategory.EnumerateResults(handleCounts: [&](StringRef s, unsigned count) { |
2036 | error() << s << " occurred " << count << " time(s).\n" ; |
2037 | }); |
2038 | } |
2039 | if (!DumpOpts.JsonErrSummaryFile.empty()) { |
2040 | std::error_code EC; |
2041 | raw_fd_ostream JsonStream(DumpOpts.JsonErrSummaryFile, EC, |
2042 | sys::fs::OF_Text); |
2043 | if (EC) { |
2044 | error() << "unable to open json summary file '" |
2045 | << DumpOpts.JsonErrSummaryFile |
2046 | << "' for writing: " << EC.message() << '\n'; |
2047 | return; |
2048 | } |
2049 | |
2050 | llvm::json::Object Categories; |
2051 | uint64_t ErrorCount = 0; |
2052 | ErrorCategory.EnumerateResults(handleCounts: [&](StringRef Category, unsigned Count) { |
2053 | llvm::json::Object Val; |
2054 | Val.try_emplace(K: "count" , Args&: Count); |
2055 | Categories.try_emplace(K: Category, Args: std::move(Val)); |
2056 | ErrorCount += Count; |
2057 | }); |
2058 | llvm::json::Object RootNode; |
2059 | RootNode.try_emplace(K: "error-categories" , Args: std::move(Categories)); |
2060 | RootNode.try_emplace(K: "error-count" , Args&: ErrorCount); |
2061 | |
2062 | JsonStream << llvm::json::Value(std::move(RootNode)); |
2063 | } |
2064 | } |
2065 | |
2066 | raw_ostream &DWARFVerifier::error() const { return WithColor::error(OS); } |
2067 | |
2068 | raw_ostream &DWARFVerifier::warn() const { return WithColor::warning(OS); } |
2069 | |
2070 | raw_ostream &DWARFVerifier::note() const { return WithColor::note(OS); } |
2071 | |
2072 | raw_ostream &DWARFVerifier::dump(const DWARFDie &Die, unsigned indent) const { |
2073 | Die.dump(OS, indent, DumpOpts); |
2074 | return OS; |
2075 | } |
2076 | |