LLVM 19.0.0git
Symbolize.cpp
Go to the documentation of this file.
1//===-- LLVMSymbolize.cpp -------------------------------------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// Implementation for LLVM symbolization library.
10//
11//===----------------------------------------------------------------------===//
12
14
15#include "llvm/ADT/STLExtras.h"
22#include "llvm/Object/BuildID.h"
23#include "llvm/Object/COFF.h"
25#include "llvm/Object/MachO.h"
27#include "llvm/Support/CRC.h"
30#include "llvm/Support/Errc.h"
33#include "llvm/Support/Path.h"
34#include <algorithm>
35#include <cassert>
36#include <cstring>
37
38namespace llvm {
39namespace codeview {
40union DebugInfo;
41}
42namespace symbolize {
43
45
47 : Opts(Opts),
48 BIDFetcher(std::make_unique<BuildIDFetcher>(Opts.DebugFileDirectory)) {}
49
51
52template <typename T>
54LLVMSymbolizer::symbolizeCodeCommon(const T &ModuleSpecifier,
55 object::SectionedAddress ModuleOffset) {
56
57 auto InfoOrErr = getOrCreateModuleInfo(ModuleSpecifier);
58 if (!InfoOrErr)
59 return InfoOrErr.takeError();
60
61 SymbolizableModule *Info = *InfoOrErr;
62
63 // A null module means an error has already been reported. Return an empty
64 // result.
65 if (!Info)
66 return DILineInfo();
67
68 // If the user is giving us relative addresses, add the preferred base of the
69 // object to the offset before we do the query. It's what DIContext expects.
70 if (Opts.RelativeAddresses)
71 ModuleOffset.Address += Info->getModulePreferredBase();
72
73 DILineInfo LineInfo = Info->symbolizeCode(
74 ModuleOffset, DILineInfoSpecifier(Opts.PathStyle, Opts.PrintFunctions),
75 Opts.UseSymbolTable);
76 if (Opts.Demangle)
77 LineInfo.FunctionName = DemangleName(LineInfo.FunctionName, Info);
78 return LineInfo;
79}
80
83 object::SectionedAddress ModuleOffset) {
84 return symbolizeCodeCommon(Obj, ModuleOffset);
85}
86
89 object::SectionedAddress ModuleOffset) {
90 return symbolizeCodeCommon(ModuleName, ModuleOffset);
91}
92
95 object::SectionedAddress ModuleOffset) {
96 return symbolizeCodeCommon(BuildID, ModuleOffset);
97}
98
99template <typename T>
100Expected<DIInliningInfo> LLVMSymbolizer::symbolizeInlinedCodeCommon(
101 const T &ModuleSpecifier, object::SectionedAddress ModuleOffset) {
102 auto InfoOrErr = getOrCreateModuleInfo(ModuleSpecifier);
103 if (!InfoOrErr)
104 return InfoOrErr.takeError();
105
106 SymbolizableModule *Info = *InfoOrErr;
107
108 // A null module means an error has already been reported. Return an empty
109 // result.
110 if (!Info)
111 return DIInliningInfo();
112
113 // If the user is giving us relative addresses, add the preferred base of the
114 // object to the offset before we do the query. It's what DIContext expects.
115 if (Opts.RelativeAddresses)
116 ModuleOffset.Address += Info->getModulePreferredBase();
117
118 DIInliningInfo InlinedContext = Info->symbolizeInlinedCode(
119 ModuleOffset, DILineInfoSpecifier(Opts.PathStyle, Opts.PrintFunctions),
120 Opts.UseSymbolTable);
121 if (Opts.Demangle) {
122 for (int i = 0, n = InlinedContext.getNumberOfFrames(); i < n; i++) {
123 auto *Frame = InlinedContext.getMutableFrame(i);
124 Frame->FunctionName = DemangleName(Frame->FunctionName, Info);
125 }
126 }
127 return InlinedContext;
128}
129
130Expected<DIInliningInfo>
132 object::SectionedAddress ModuleOffset) {
133 return symbolizeInlinedCodeCommon(Obj, ModuleOffset);
134}
135
138 object::SectionedAddress ModuleOffset) {
139 return symbolizeInlinedCodeCommon(ModuleName, ModuleOffset);
140}
141
144 object::SectionedAddress ModuleOffset) {
145 return symbolizeInlinedCodeCommon(BuildID, ModuleOffset);
146}
147
148template <typename T>
150LLVMSymbolizer::symbolizeDataCommon(const T &ModuleSpecifier,
151 object::SectionedAddress ModuleOffset) {
152
153 auto InfoOrErr = getOrCreateModuleInfo(ModuleSpecifier);
154 if (!InfoOrErr)
155 return InfoOrErr.takeError();
156
157 SymbolizableModule *Info = *InfoOrErr;
158 // A null module means an error has already been reported. Return an empty
159 // result.
160 if (!Info)
161 return DIGlobal();
162
163 // If the user is giving us relative addresses, add the preferred base of
164 // the object to the offset before we do the query. It's what DIContext
165 // expects.
166 if (Opts.RelativeAddresses)
167 ModuleOffset.Address += Info->getModulePreferredBase();
168
169 DIGlobal Global = Info->symbolizeData(ModuleOffset);
170 if (Opts.Demangle)
171 Global.Name = DemangleName(Global.Name, Info);
172 return Global;
173}
174
177 object::SectionedAddress ModuleOffset) {
178 return symbolizeDataCommon(Obj, ModuleOffset);
179}
180
183 object::SectionedAddress ModuleOffset) {
184 return symbolizeDataCommon(ModuleName, ModuleOffset);
185}
186
189 object::SectionedAddress ModuleOffset) {
190 return symbolizeDataCommon(BuildID, ModuleOffset);
191}
192
193template <typename T>
195LLVMSymbolizer::symbolizeFrameCommon(const T &ModuleSpecifier,
196 object::SectionedAddress ModuleOffset) {
197 auto InfoOrErr = getOrCreateModuleInfo(ModuleSpecifier);
198 if (!InfoOrErr)
199 return InfoOrErr.takeError();
200
201 SymbolizableModule *Info = *InfoOrErr;
202 // A null module means an error has already been reported. Return an empty
203 // result.
204 if (!Info)
205 return std::vector<DILocal>();
206
207 // If the user is giving us relative addresses, add the preferred base of
208 // the object to the offset before we do the query. It's what DIContext
209 // expects.
210 if (Opts.RelativeAddresses)
211 ModuleOffset.Address += Info->getModulePreferredBase();
212
213 return Info->symbolizeFrame(ModuleOffset);
214}
215
218 object::SectionedAddress ModuleOffset) {
219 return symbolizeFrameCommon(Obj, ModuleOffset);
220}
221
224 object::SectionedAddress ModuleOffset) {
225 return symbolizeFrameCommon(ModuleName, ModuleOffset);
226}
227
230 object::SectionedAddress ModuleOffset) {
231 return symbolizeFrameCommon(BuildID, ModuleOffset);
232}
233
234template <typename T>
236LLVMSymbolizer::findSymbolCommon(const T &ModuleSpecifier, StringRef Symbol,
238 auto InfoOrErr = getOrCreateModuleInfo(ModuleSpecifier);
239 if (!InfoOrErr)
240 return InfoOrErr.takeError();
241
242 SymbolizableModule *Info = *InfoOrErr;
243 std::vector<DILineInfo> Result;
244
245 // A null module means an error has already been reported. Return an empty
246 // result.
247 if (!Info)
248 return Result;
249
250 for (object::SectionedAddress A : Info->findSymbol(Symbol, Offset)) {
251 DILineInfo LineInfo = Info->symbolizeCode(
253 Opts.UseSymbolTable);
254 if (LineInfo.FileName != DILineInfo::BadString) {
255 if (Opts.Demangle)
256 LineInfo.FunctionName = DemangleName(LineInfo.FunctionName, Info);
257 Result.push_back(LineInfo);
258 }
259 }
260
261 return Result;
262}
263
264Expected<std::vector<DILineInfo>>
267 return findSymbolCommon(Obj, Symbol, Offset);
268}
269
273 return findSymbolCommon(ModuleName, Symbol, Offset);
274}
275
279 return findSymbolCommon(BuildID, Symbol, Offset);
280}
281
283 ObjectForUBPathAndArch.clear();
284 LRUBinaries.clear();
285 CacheSize = 0;
286 BinaryForPath.clear();
287 ObjectPairForPathArch.clear();
288 Modules.clear();
289 BuildIDPaths.clear();
290}
291
292namespace {
293
294// For Path="/path/to/foo" and Basename="foo" assume that debug info is in
295// /path/to/foo.dSYM/Contents/Resources/DWARF/foo.
296// For Path="/path/to/bar.dSYM" and Basename="foo" assume that debug info is in
297// /path/to/bar.dSYM/Contents/Resources/DWARF/foo.
298std::string getDarwinDWARFResourceForPath(const std::string &Path,
299 const std::string &Basename) {
300 SmallString<16> ResourceName = StringRef(Path);
301 if (sys::path::extension(Path) != ".dSYM") {
302 ResourceName += ".dSYM";
303 }
304 sys::path::append(ResourceName, "Contents", "Resources", "DWARF");
305 sys::path::append(ResourceName, Basename);
306 return std::string(ResourceName);
307}
308
309bool checkFileCRC(StringRef Path, uint32_t CRCHash) {
310 ErrorOr<std::unique_ptr<MemoryBuffer>> MB =
312 if (!MB)
313 return false;
314 return CRCHash == llvm::crc32(arrayRefFromStringRef(MB.get()->getBuffer()));
315}
316
317bool getGNUDebuglinkContents(const ObjectFile *Obj, std::string &DebugName,
318 uint32_t &CRCHash) {
319 if (!Obj)
320 return false;
321 for (const SectionRef &Section : Obj->sections()) {
322 StringRef Name;
323 consumeError(Section.getName().moveInto(Name));
324
325 Name = Name.substr(Name.find_first_not_of("._"));
326 if (Name == "gnu_debuglink") {
327 Expected<StringRef> ContentsOrErr = Section.getContents();
328 if (!ContentsOrErr) {
329 consumeError(ContentsOrErr.takeError());
330 return false;
331 }
332 DataExtractor DE(*ContentsOrErr, Obj->isLittleEndian(), 0);
333 uint64_t Offset = 0;
334 if (const char *DebugNameStr = DE.getCStr(&Offset)) {
335 // 4-byte align the offset.
336 Offset = (Offset + 3) & ~0x3;
337 if (DE.isValidOffsetForDataOfSize(Offset, 4)) {
338 DebugName = DebugNameStr;
339 CRCHash = DE.getU32(&Offset);
340 return true;
341 }
342 }
343 break;
344 }
345 }
346 return false;
347}
348
349bool darwinDsymMatchesBinary(const MachOObjectFile *DbgObj,
350 const MachOObjectFile *Obj) {
351 ArrayRef<uint8_t> dbg_uuid = DbgObj->getUuid();
352 ArrayRef<uint8_t> bin_uuid = Obj->getUuid();
353 if (dbg_uuid.empty() || bin_uuid.empty())
354 return false;
355 return !memcmp(dbg_uuid.data(), bin_uuid.data(), dbg_uuid.size());
356}
357
358} // end anonymous namespace
359
360ObjectFile *LLVMSymbolizer::lookUpDsymFile(const std::string &ExePath,
361 const MachOObjectFile *MachExeObj,
362 const std::string &ArchName) {
363 // On Darwin we may find DWARF in separate object file in
364 // resource directory.
365 std::vector<std::string> DsymPaths;
366 StringRef Filename = sys::path::filename(ExePath);
367 DsymPaths.push_back(
368 getDarwinDWARFResourceForPath(ExePath, std::string(Filename)));
369 for (const auto &Path : Opts.DsymHints) {
370 DsymPaths.push_back(
371 getDarwinDWARFResourceForPath(Path, std::string(Filename)));
372 }
373 for (const auto &Path : DsymPaths) {
374 auto DbgObjOrErr = getOrCreateObject(Path, ArchName);
375 if (!DbgObjOrErr) {
376 // Ignore errors, the file might not exist.
377 consumeError(DbgObjOrErr.takeError());
378 continue;
379 }
380 ObjectFile *DbgObj = DbgObjOrErr.get();
381 if (!DbgObj)
382 continue;
383 const MachOObjectFile *MachDbgObj = dyn_cast<const MachOObjectFile>(DbgObj);
384 if (!MachDbgObj)
385 continue;
386 if (darwinDsymMatchesBinary(MachDbgObj, MachExeObj))
387 return DbgObj;
388 }
389 return nullptr;
390}
391
392ObjectFile *LLVMSymbolizer::lookUpDebuglinkObject(const std::string &Path,
393 const ObjectFile *Obj,
394 const std::string &ArchName) {
395 std::string DebuglinkName;
396 uint32_t CRCHash;
397 std::string DebugBinaryPath;
398 if (!getGNUDebuglinkContents(Obj, DebuglinkName, CRCHash))
399 return nullptr;
400 if (!findDebugBinary(Path, DebuglinkName, CRCHash, DebugBinaryPath))
401 return nullptr;
402 auto DbgObjOrErr = getOrCreateObject(DebugBinaryPath, ArchName);
403 if (!DbgObjOrErr) {
404 // Ignore errors, the file might not exist.
405 consumeError(DbgObjOrErr.takeError());
406 return nullptr;
407 }
408 return DbgObjOrErr.get();
409}
410
411ObjectFile *LLVMSymbolizer::lookUpBuildIDObject(const std::string &Path,
412 const ELFObjectFileBase *Obj,
413 const std::string &ArchName) {
414 auto BuildID = getBuildID(Obj);
415 if (BuildID.size() < 2)
416 return nullptr;
417 std::string DebugBinaryPath;
418 if (!getOrFindDebugBinary(BuildID, DebugBinaryPath))
419 return nullptr;
420 auto DbgObjOrErr = getOrCreateObject(DebugBinaryPath, ArchName);
421 if (!DbgObjOrErr) {
422 consumeError(DbgObjOrErr.takeError());
423 return nullptr;
424 }
425 return DbgObjOrErr.get();
426}
427
428bool LLVMSymbolizer::findDebugBinary(const std::string &OrigPath,
429 const std::string &DebuglinkName,
430 uint32_t CRCHash, std::string &Result) {
431 SmallString<16> OrigDir(OrigPath);
433 SmallString<16> DebugPath = OrigDir;
434 // Try relative/path/to/original_binary/debuglink_name
435 llvm::sys::path::append(DebugPath, DebuglinkName);
436 if (checkFileCRC(DebugPath, CRCHash)) {
437 Result = std::string(DebugPath);
438 return true;
439 }
440 // Try relative/path/to/original_binary/.debug/debuglink_name
441 DebugPath = OrigDir;
442 llvm::sys::path::append(DebugPath, ".debug", DebuglinkName);
443 if (checkFileCRC(DebugPath, CRCHash)) {
444 Result = std::string(DebugPath);
445 return true;
446 }
447 // Make the path absolute so that lookups will go to
448 // "/usr/lib/debug/full/path/to/debug", not
449 // "/usr/lib/debug/to/debug"
451 if (!Opts.FallbackDebugPath.empty()) {
452 // Try <FallbackDebugPath>/absolute/path/to/original_binary/debuglink_name
453 DebugPath = Opts.FallbackDebugPath;
454 } else {
455#if defined(__NetBSD__)
456 // Try /usr/libdata/debug/absolute/path/to/original_binary/debuglink_name
457 DebugPath = "/usr/libdata/debug";
458#else
459 // Try /usr/lib/debug/absolute/path/to/original_binary/debuglink_name
460 DebugPath = "/usr/lib/debug";
461#endif
462 }
464 DebuglinkName);
465 if (checkFileCRC(DebugPath, CRCHash)) {
466 Result = std::string(DebugPath);
467 return true;
468 }
469 return false;
470}
471
473 return StringRef(reinterpret_cast<const char *>(BuildID.data()),
474 BuildID.size());
475}
476
477bool LLVMSymbolizer::getOrFindDebugBinary(const ArrayRef<uint8_t> BuildID,
478 std::string &Result) {
479 StringRef BuildIDStr = getBuildIDStr(BuildID);
480 auto I = BuildIDPaths.find(BuildIDStr);
481 if (I != BuildIDPaths.end()) {
482 Result = I->second;
483 return true;
484 }
485 if (!BIDFetcher)
486 return false;
487 if (std::optional<std::string> Path = BIDFetcher->fetch(BuildID)) {
488 Result = *Path;
489 auto InsertResult = BuildIDPaths.insert({BuildIDStr, Result});
490 assert(InsertResult.second);
491 (void)InsertResult;
492 return true;
493 }
494
495 return false;
496}
497
498Expected<LLVMSymbolizer::ObjectPair>
499LLVMSymbolizer::getOrCreateObjectPair(const std::string &Path,
500 const std::string &ArchName) {
501 auto I = ObjectPairForPathArch.find(std::make_pair(Path, ArchName));
502 if (I != ObjectPairForPathArch.end()) {
503 recordAccess(BinaryForPath.find(Path)->second);
504 return I->second;
505 }
506
507 auto ObjOrErr = getOrCreateObject(Path, ArchName);
508 if (!ObjOrErr) {
509 ObjectPairForPathArch.emplace(std::make_pair(Path, ArchName),
510 ObjectPair(nullptr, nullptr));
511 return ObjOrErr.takeError();
512 }
513
514 ObjectFile *Obj = ObjOrErr.get();
515 assert(Obj != nullptr);
516 ObjectFile *DbgObj = nullptr;
517
518 if (auto MachObj = dyn_cast<const MachOObjectFile>(Obj))
519 DbgObj = lookUpDsymFile(Path, MachObj, ArchName);
520 else if (auto ELFObj = dyn_cast<const ELFObjectFileBase>(Obj))
521 DbgObj = lookUpBuildIDObject(Path, ELFObj, ArchName);
522 if (!DbgObj)
523 DbgObj = lookUpDebuglinkObject(Path, Obj, ArchName);
524 if (!DbgObj)
525 DbgObj = Obj;
526 ObjectPair Res = std::make_pair(Obj, DbgObj);
527 std::string DbgObjPath = DbgObj->getFileName().str();
528 auto Pair =
529 ObjectPairForPathArch.emplace(std::make_pair(Path, ArchName), Res);
530 BinaryForPath.find(DbgObjPath)->second.pushEvictor([this, I = Pair.first]() {
531 ObjectPairForPathArch.erase(I);
532 });
533 return Res;
534}
535
536Expected<ObjectFile *>
537LLVMSymbolizer::getOrCreateObject(const std::string &Path,
538 const std::string &ArchName) {
539 Binary *Bin;
540 auto Pair = BinaryForPath.emplace(Path, OwningBinary<Binary>());
541 if (!Pair.second) {
542 Bin = Pair.first->second->getBinary();
543 recordAccess(Pair.first->second);
544 } else {
545 Expected<OwningBinary<Binary>> BinOrErr = createBinary(Path);
546 if (!BinOrErr)
547 return BinOrErr.takeError();
548
549 CachedBinary &CachedBin = Pair.first->second;
550 CachedBin = std::move(BinOrErr.get());
551 CachedBin.pushEvictor([this, I = Pair.first]() { BinaryForPath.erase(I); });
552 LRUBinaries.push_back(CachedBin);
553 CacheSize += CachedBin.size();
554 Bin = CachedBin->getBinary();
555 }
556
557 if (!Bin)
558 return static_cast<ObjectFile *>(nullptr);
559
560 if (MachOUniversalBinary *UB = dyn_cast_or_null<MachOUniversalBinary>(Bin)) {
561 auto I = ObjectForUBPathAndArch.find(std::make_pair(Path, ArchName));
562 if (I != ObjectForUBPathAndArch.end())
563 return I->second.get();
564
565 Expected<std::unique_ptr<ObjectFile>> ObjOrErr =
566 UB->getMachOObjectForArch(ArchName);
567 if (!ObjOrErr) {
568 ObjectForUBPathAndArch.emplace(std::make_pair(Path, ArchName),
569 std::unique_ptr<ObjectFile>());
570 return ObjOrErr.takeError();
571 }
572 ObjectFile *Res = ObjOrErr->get();
573 auto Pair = ObjectForUBPathAndArch.emplace(std::make_pair(Path, ArchName),
574 std::move(ObjOrErr.get()));
575 BinaryForPath.find(Path)->second.pushEvictor(
576 [this, Iter = Pair.first]() { ObjectForUBPathAndArch.erase(Iter); });
577 return Res;
578 }
579 if (Bin->isObject()) {
580 return cast<ObjectFile>(Bin);
581 }
582 return errorCodeToError(object_error::arch_not_found);
583}
584
585Expected<SymbolizableModule *>
586LLVMSymbolizer::createModuleInfo(const ObjectFile *Obj,
587 std::unique_ptr<DIContext> Context,
588 StringRef ModuleName) {
589 auto InfoOrErr = SymbolizableObjectFile::create(Obj, std::move(Context),
590 Opts.UntagAddresses);
591 std::unique_ptr<SymbolizableModule> SymMod;
592 if (InfoOrErr)
593 SymMod = std::move(*InfoOrErr);
594 auto InsertResult = Modules.insert(
595 std::make_pair(std::string(ModuleName), std::move(SymMod)));
596 assert(InsertResult.second);
597 if (!InfoOrErr)
598 return InfoOrErr.takeError();
599 return InsertResult.first->second.get();
600}
601
602Expected<SymbolizableModule *>
604 std::string BinaryName = ModuleName;
605 std::string ArchName = Opts.DefaultArch;
606 size_t ColonPos = ModuleName.find_last_of(':');
607 // Verify that substring after colon form a valid arch name.
608 if (ColonPos != std::string::npos) {
609 std::string ArchStr = ModuleName.substr(ColonPos + 1);
610 if (Triple(ArchStr).getArch() != Triple::UnknownArch) {
611 BinaryName = ModuleName.substr(0, ColonPos);
612 ArchName = ArchStr;
613 }
614 }
615
616 auto I = Modules.find(ModuleName);
617 if (I != Modules.end()) {
618 recordAccess(BinaryForPath.find(BinaryName)->second);
619 return I->second.get();
620 }
621
622 auto ObjectsOrErr = getOrCreateObjectPair(BinaryName, ArchName);
623 if (!ObjectsOrErr) {
624 // Failed to find valid object file.
625 Modules.emplace(ModuleName, std::unique_ptr<SymbolizableModule>());
626 return ObjectsOrErr.takeError();
627 }
628 ObjectPair Objects = ObjectsOrErr.get();
629
630 std::unique_ptr<DIContext> Context;
631 // If this is a COFF object containing PDB info and not containing DWARF
632 // section, use a PDBContext to symbolize. Otherwise, use DWARF.
633 if (auto CoffObject = dyn_cast<COFFObjectFile>(Objects.first)) {
635 StringRef PDBFileName;
636 auto EC = CoffObject->getDebugPDBInfo(DebugInfo, PDBFileName);
637 // Use DWARF if there're DWARF sections.
638 bool HasDwarf =
639 llvm::any_of(Objects.first->sections(), [](SectionRef Section) -> bool {
640 if (Expected<StringRef> SectionName = Section.getName())
641 return SectionName.get() == ".debug_info";
642 return false;
643 });
644 if (!EC && !HasDwarf && DebugInfo != nullptr && !PDBFileName.empty()) {
645 using namespace pdb;
646 std::unique_ptr<IPDBSession> Session;
647
648 PDB_ReaderType ReaderType =
649 Opts.UseDIA ? PDB_ReaderType::DIA : PDB_ReaderType::Native;
650 if (auto Err = loadDataForEXE(ReaderType, Objects.first->getFileName(),
651 Session)) {
652 Modules.emplace(ModuleName, std::unique_ptr<SymbolizableModule>());
653 // Return along the PDB filename to provide more context
654 return createFileError(PDBFileName, std::move(Err));
655 }
656 Context.reset(new PDBContext(*CoffObject, std::move(Session)));
657 }
658 }
659 if (!Context)
662 nullptr, Opts.DWPName);
663 auto ModuleOrErr =
664 createModuleInfo(Objects.first, std::move(Context), ModuleName);
665 if (ModuleOrErr) {
666 auto I = Modules.find(ModuleName);
667 BinaryForPath.find(BinaryName)->second.pushEvictor([this, I]() {
668 Modules.erase(I);
669 });
670 }
671 return ModuleOrErr;
672}
673
674// For BPF programs .BTF.ext section contains line numbers information,
675// use it if regular DWARF is not available (e.g. for stripped binary).
676static bool useBTFContext(const ObjectFile &Obj) {
677 return Obj.makeTriple().isBPF() && !Obj.hasDebugInfo() &&
679}
680
683 StringRef ObjName = Obj.getFileName();
684 auto I = Modules.find(ObjName);
685 if (I != Modules.end())
686 return I->second.get();
687
688 std::unique_ptr<DIContext> Context;
689 if (useBTFContext(Obj))
691 else
693 // FIXME: handle COFF object with PDB info to use PDBContext
694 return createModuleInfo(&Obj, std::move(Context), ObjName);
695}
696
697Expected<SymbolizableModule *>
699 std::string Path;
700 if (!getOrFindDebugBinary(BuildID, Path)) {
702 "could not find build ID");
703 }
704 return getOrCreateModuleInfo(Path);
705}
706
707namespace {
708
709// Undo these various manglings for Win32 extern "C" functions:
710// cdecl - _foo
711// stdcall - _foo@12
712// fastcall - @foo@12
713// vectorcall - foo@@12
714// These are all different linkage names for 'foo'.
715StringRef demanglePE32ExternCFunc(StringRef SymbolName) {
716 char Front = SymbolName.empty() ? '\0' : SymbolName[0];
717
718 // Remove any '@[0-9]+' suffix.
719 bool HasAtNumSuffix = false;
720 if (Front != '?') {
721 size_t AtPos = SymbolName.rfind('@');
722 if (AtPos != StringRef::npos &&
723 all_of(drop_begin(SymbolName, AtPos + 1), isDigit)) {
724 SymbolName = SymbolName.substr(0, AtPos);
725 HasAtNumSuffix = true;
726 }
727 }
728
729 // Remove any ending '@' for vectorcall.
730 bool IsVectorCall = false;
731 if (HasAtNumSuffix && SymbolName.ends_with("@")) {
732 SymbolName = SymbolName.drop_back();
733 IsVectorCall = true;
734 }
735
736 // If not vectorcall, remove any '_' or '@' prefix.
737 if (!IsVectorCall && (Front == '_' || Front == '@'))
738 SymbolName = SymbolName.drop_front();
739
740 return SymbolName;
741}
742
743} // end anonymous namespace
744
745std::string
747 const SymbolizableModule *DbiModuleDescriptor) {
748 std::string Result;
749 if (nonMicrosoftDemangle(Name, Result))
750 return Result;
751
752 if (!Name.empty() && Name.front() == '?') {
753 // Only do MSVC C++ demangling on symbols starting with '?'.
754 int status = 0;
755 char *DemangledName = microsoftDemangle(
756 Name, nullptr, &status,
759 if (status != 0)
760 return Name;
761 Result = DemangledName;
762 free(DemangledName);
763 return Result;
764 }
765
766 if (DbiModuleDescriptor && DbiModuleDescriptor->isWin32Module()) {
767 std::string DemangledCName(demanglePE32ExternCFunc(Name));
768 // On i386 Windows, the C name mangling for different calling conventions
769 // may also be applied on top of the Itanium or Rust name mangling.
770 if (nonMicrosoftDemangle(DemangledCName, Result))
771 return Result;
772 return DemangledCName;
773 }
774 return Name;
775}
776
777void LLVMSymbolizer::recordAccess(CachedBinary &Bin) {
778 if (Bin->getBinary())
779 LRUBinaries.splice(LRUBinaries.end(), LRUBinaries, Bin.getIterator());
780}
781
783 // Evict the LRU binary until the max cache size is reached or there's <= 1
784 // item in the cache. The MRU binary is always kept to avoid thrashing if it's
785 // larger than the cache size.
786 while (CacheSize > Opts.MaxCacheSize && !LRUBinaries.empty() &&
787 std::next(LRUBinaries.begin()) != LRUBinaries.end()) {
788 CachedBinary &Bin = LRUBinaries.front();
789 CacheSize -= Bin.size();
790 LRUBinaries.pop_front();
791 Bin.evict();
792 }
793}
794
795void CachedBinary::pushEvictor(std::function<void()> NewEvictor) {
796 if (Evictor) {
797 this->Evictor = [OldEvictor = std::move(this->Evictor),
798 NewEvictor = std::move(NewEvictor)]() {
799 NewEvictor();
800 OldEvictor();
801 };
802 } else {
803 this->Evictor = std::move(NewEvictor);
804 }
805}
806
807} // namespace symbolize
808} // namespace llvm
This file declares a library for handling Build IDs and using them to find debug info.
static GCRegistry::Add< ErlangGC > A("erlang", "erlang-compatible garbage collector")
Analysis containing CSE Info
Definition: CSEInfo.cpp:27
std::string Name
#define I(x, y, z)
Definition: MD5.cpp:58
Merge contiguous icmps into a memcmp
Definition: MergeICmps.cpp:911
LLVMContext & Context
static bool isDigit(const char C)
assert(ImpDefSCC.getReg()==AMDGPU::SCC &&ImpDefSCC.isDef())
This file contains some templates that are useful if you are working with the STL at all.
ArrayRef - Represent a constant reference to an array (0 or more elements consecutively in memory),...
Definition: ArrayRef.h:41
static std::unique_ptr< BTFContext > create(const object::ObjectFile &Obj, std::function< void(Error)> ErrorHandler=WithColor::defaultErrorHandler)
Definition: BTFContext.cpp:63
static bool hasBTFSections(const ObjectFile &Obj)
Definition: BTFParser.cpp:410
A format-neutral container for inlined code description.
Definition: DIContext.h:92
static std::unique_ptr< DWARFContext > create(const object::ObjectFile &Obj, ProcessDebugRelocations RelocAction=ProcessDebugRelocations::Process, const LoadedObjectInfo *L=nullptr, std::string DWPName="", std::function< void(Error)> RecoverableErrorHandler=WithColor::defaultErrorHandler, std::function< void(Error)> WarningHandler=WithColor::defaultWarningHandler, bool ThreadSafe=false)
Tagged union holding either a T or a Error.
Definition: Error.h:474
static ErrorOr< std::unique_ptr< MemoryBuffer > > getFileOrSTDIN(const Twine &Filename, bool IsText=false, bool RequiresNullTerminator=true, std::optional< Align > Alignment=std::nullopt)
Open the specified file as a MemoryBuffer, or open stdin if the Filename is "-".
SmallString - A SmallString is just a SmallVector with methods and accessors that make it work better...
Definition: SmallString.h:26
size_t size() const
Definition: SmallVector.h:91
pointer data()
Return a pointer to the vector's buffer, even if empty().
Definition: SmallVector.h:299
This is a 'vector' (really, a variable-sized array), optimized for the case when the array is small.
Definition: SmallVector.h:1209
iterator end()
Definition: StringMap.h:221
iterator find(StringRef Key)
Definition: StringMap.h:234
bool insert(MapEntryTy *KeyValue)
insert - Insert the specified key/value pair into the map.
Definition: StringMap.h:307
StringRef - Represent a constant reference to a string, i.e.
Definition: StringRef.h:50
constexpr bool empty() const
empty - Check if the string is empty.
Definition: StringRef.h:134
static constexpr size_t npos
Definition: StringRef.h:52
Triple - Helper class for working with autoconf configuration names.
Definition: Triple.h:44
bool isBPF() const
Tests whether the target is eBPF.
Definition: Triple.h:1027
@ UnknownArch
Definition: Triple.h:47
StringRef getFileName() const
Definition: Binary.cpp:41
BuildIDFetcher searches local cache directories for debug info.
Definition: BuildID.h:39
This class is the base class for all object file types.
Definition: ObjectFile.h:229
Triple makeTriple() const
Create a triple from the data in this object file.
Definition: ObjectFile.cpp:109
virtual bool hasDebugInfo() const
Definition: ObjectFile.cpp:99
This is a value type class that represents a single section in the list of sections in the object fil...
Definition: ObjectFile.h:81
void pushEvictor(std::function< void()> Evictor)
Definition: Symbolize.cpp:795
Expected< std::vector< DILineInfo > > findSymbol(const ObjectFile &Obj, StringRef Symbol, uint64_t Offset)
Definition: Symbolize.cpp:265
Expected< DIInliningInfo > symbolizeInlinedCode(const ObjectFile &Obj, object::SectionedAddress ModuleOffset)
Definition: Symbolize.cpp:131
Expected< SymbolizableModule * > getOrCreateModuleInfo(const std::string &ModuleName)
Returns a SymbolizableModule or an error if loading debug info failed.
Definition: Symbolize.cpp:603
Expected< DILineInfo > symbolizeCode(const ObjectFile &Obj, object::SectionedAddress ModuleOffset)
Definition: Symbolize.cpp:82
static std::string DemangleName(const std::string &Name, const SymbolizableModule *DbiModuleDescriptor)
Definition: Symbolize.cpp:746
Expected< DIGlobal > symbolizeData(const ObjectFile &Obj, object::SectionedAddress ModuleOffset)
Definition: Symbolize.cpp:176
Expected< std::vector< DILocal > > symbolizeFrame(const ObjectFile &Obj, object::SectionedAddress ModuleOffset)
Definition: Symbolize.cpp:217
virtual bool isWin32Module() const =0
static Expected< std::unique_ptr< SymbolizableObjectFile > > create(const object::ObjectFile *Obj, std::unique_ptr< DIContext > DICtx, bool UntagAddresses)
constexpr char SymbolName[]
Key for Kernel::Metadata::mSymbolName.
SmallVector< uint8_t, 10 > BuildID
A build ID in binary form.
Definition: BuildID.h:25
BuildIDRef getBuildID(const ObjectFile *Obj)
Returns the build ID, if any, contained in the given object file.
Definition: BuildID.cpp:56
Expected< std::unique_ptr< Binary > > createBinary(MemoryBufferRef Source, LLVMContext *Context=nullptr, bool InitContent=true)
Create a Binary from Source, autodetecting the file type.
Definition: Binary.cpp:45
static bool useBTFContext(const ObjectFile &Obj)
Definition: Symbolize.cpp:676
static StringRef getBuildIDStr(ArrayRef< uint8_t > BuildID)
Definition: Symbolize.cpp:472
void make_absolute(const Twine &current_directory, SmallVectorImpl< char > &path)
Make path an absolute path.
Definition: Path.cpp:907
void remove_filename(SmallVectorImpl< char > &path, Style style=Style::native)
Remove the last component from path unless it is the root dir.
Definition: Path.cpp:475
StringRef filename(StringRef path, Style style=Style::native)
Get filename.
Definition: Path.cpp:578
void append(SmallVectorImpl< char > &path, const Twine &a, const Twine &b="", const Twine &c="", const Twine &d="")
Append to path.
Definition: Path.cpp:457
StringRef extension(StringRef path, Style style=Style::native)
Get extension.
Definition: Path.cpp:591
StringRef relative_path(StringRef path, Style style=Style::native)
Get relative path.
Definition: Path.cpp:414
This is an optimization pass for GlobalISel generic memory operations.
Definition: AddressRanges.h:18
auto drop_begin(T &&RangeOrContainer, size_t N=1)
Return a range covering RangeOrContainer with the first N elements excluded.
Definition: STLExtras.h:329
@ Offset
Definition: DWP.cpp:456
bool nonMicrosoftDemangle(std::string_view MangledName, std::string &Result, bool CanHaveLeadingDot=true, bool ParseParams=true)
Definition: Demangle.cpp:49
bool all_of(R &&range, UnaryPredicate P)
Provide wrappers to std::all_of which take ranges instead of having to pass begin/end explicitly.
Definition: STLExtras.h:1731
Error createFileError(const Twine &F, Error E)
Concatenate a source file path and/or name with an Error.
Definition: Error.h:1339
Error createStringError(std::error_code EC, char const *Fmt, const Ts &... Vals)
Create formatted StringError object.
Definition: Error.h:1258
@ no_such_file_or_directory
bool any_of(R &&range, UnaryPredicate P)
Provide wrappers to std::any_of which take ranges instead of having to pass begin/end explicitly.
Definition: STLExtras.h:1738
@ Global
Append to llvm.global_dtors.
uint32_t crc32(ArrayRef< uint8_t > Data)
Definition: CRC.cpp:101
char * microsoftDemangle(std::string_view mangled_name, size_t *n_read, int *status, MSDemangleFlags Flags=MSDF_None)
Demangles the Microsoft symbol pointed at by mangled_name and returns it.
MSDemangleFlags
Definition: Demangle.h:37
@ MSDF_NoReturnType
Definition: Demangle.h:42
@ MSDF_NoMemberType
Definition: Demangle.h:43
@ MSDF_NoCallingConvention
Definition: Demangle.h:41
@ MSDF_NoAccessSpecifier
Definition: Demangle.h:40
Error errorCodeToError(std::error_code EC)
Helper for converting an std::error_code to a Error.
Definition: Error.cpp:103
void consumeError(Error Err)
Consume a Error without doing anything.
Definition: Error.h:1041
Implement std::hash so that hash_code can be used in STL containers.
Definition: BitVector.h:858
Container for description of a global variable.
Definition: DIContext.h:118
Controls which fields of DILineInfo container should be filled with data.
Definition: DIContext.h:144
A format-neutral container for source line information.
Definition: DIContext.h:32
static constexpr const char *const BadString
Definition: DIContext.h:34
std::string FileName
Definition: DIContext.h:37
std::string FunctionName
Definition: DIContext.h:38
std::vector< std::string > DsymHints
Definition: Symbolize.h:61