1 //===-- SpecialCaseList.cpp - special case list for sanitizers ------------===//
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 // This is a utility class for instrumentation passes (like AddressSanitizer
10 // or ThreadSanitizer) to avoid instrumenting some functions or global
11 // variables, or to instrument some functions or global variables in a specific
12 // way, based on a user-supplied list.
13 //
14 //===----------------------------------------------------------------------===//
15 
16 #include "llvm/Support/SpecialCaseList.h"
17 #include "llvm/ADT/SmallVector.h"
18 #include "llvm/Support/MemoryBuffer.h"
19 #include "llvm/Support/Regex.h"
20 #include "llvm/Support/VirtualFileSystem.h"
21 #include <string>
22 #include <system_error>
23 #include <utility>
24 
25 #include <stdio.h>
26 namespace llvm {
27 
28 bool SpecialCaseList::Matcher::insert(std::string Regexp,
29                                       unsigned LineNumber,
30                                       std::string &REError) {
31   if (Regexp.empty()) {
32     REError = "Supplied regexp was blank";
33     return false;
34   }
35 
36   if (Regex::isLiteralERE(Regexp)) {
37     Strings[Regexp] = LineNumber;
38     return true;
39   }
40 
41   // Replace * with .*
42   for (size_t pos = 0; (pos = Regexp.find('*', pos)) != std::string::npos;
43        pos += strlen(".*")) {
44     Regexp.replace(pos, strlen("*"), ".*");
45   }
46 
47   Regexp = (Twine("^(") + StringRef(Regexp) + ")$").str();
48 
49   // Check that the regexp is valid.
50   Regex CheckRE(Regexp);
51   if (!CheckRE.isValid(REError))
52     return false;
53 
54   RegExes.emplace_back(
55       std::make_pair(std::make_unique<Regex>(std::move(CheckRE)), LineNumber));
56   return true;
57 }
58 
59 unsigned SpecialCaseList::Matcher::match(StringRef Query) const {
60   auto It = Strings.find(Query);
61   if (It != Strings.end())
62     return It->second;
63   for (const auto &RegExKV : RegExes)
64     if (RegExKV.first->match(Query))
65       return RegExKV.second;
66   return 0;
67 }
68 
69 std::unique_ptr<SpecialCaseList>
70 SpecialCaseList::create(const std::vector<std::string> &Paths,
71                         llvm::vfs::FileSystem &FS, std::string &Error) {
72   std::unique_ptr<SpecialCaseList> SCL(new SpecialCaseList());
73   if (SCL->createInternal(Paths, FS, Error))
74     return SCL;
75   return nullptr;
76 }
77 
78 std::unique_ptr<SpecialCaseList> SpecialCaseList::create(const MemoryBuffer *MB,
79                                                          std::string &Error) {
80   std::unique_ptr<SpecialCaseList> SCL(new SpecialCaseList());
81   if (SCL->createInternal(MB, Error))
82     return SCL;
83   return nullptr;
84 }
85 
86 std::unique_ptr<SpecialCaseList>
87 SpecialCaseList::createOrDie(const std::vector<std::string> &Paths,
88                              llvm::vfs::FileSystem &FS) {
89   std::string Error;
90   if (auto SCL = create(Paths, FS, Error))
91     return SCL;
92   report_fatal_error(Twine(Error));
93 }
94 
95 bool SpecialCaseList::createInternal(const std::vector<std::string> &Paths,
96                                      vfs::FileSystem &VFS, std::string &Error) {
97   StringMap<size_t> Sections;
98   for (const auto &Path : Paths) {
99     ErrorOr<std::unique_ptr<MemoryBuffer>> FileOrErr =
100         VFS.getBufferForFile(Path);
101     if (std::error_code EC = FileOrErr.getError()) {
102       Error = (Twine("can't open file '") + Path + "': " + EC.message()).str();
103       return false;
104     }
105     std::string ParseError;
106     if (!parse(FileOrErr.get().get(), Sections, ParseError)) {
107       Error = (Twine("error parsing file '") + Path + "': " + ParseError).str();
108       return false;
109     }
110   }
111   return true;
112 }
113 
114 bool SpecialCaseList::createInternal(const MemoryBuffer *MB,
115                                      std::string &Error) {
116   StringMap<size_t> Sections;
117   if (!parse(MB, Sections, Error))
118     return false;
119   return true;
120 }
121 
122 bool SpecialCaseList::parse(const MemoryBuffer *MB,
123                             StringMap<size_t> &SectionsMap,
124                             std::string &Error) {
125   // Iterate through each line in the exclusion list file.
126   SmallVector<StringRef, 16> Lines;
127   MB->getBuffer().split(Lines, '\n');
128 
129   unsigned LineNo = 1;
130   StringRef Section = "*";
131 
132   for (auto I = Lines.begin(), E = Lines.end(); I != E; ++I, ++LineNo) {
133     *I = I->trim();
134     // Ignore empty lines and lines starting with "#"
135     if (I->empty() || I->startswith("#"))
136       continue;
137 
138     // Save section names
139     if (I->startswith("[")) {
140       if (!I->endswith("]")) {
141         Error = (Twine("malformed section header on line ") + Twine(LineNo) +
142                  ": " + *I).str();
143         return false;
144       }
145 
146       Section = I->slice(1, I->size() - 1);
147 
148       std::string REError;
149       Regex CheckRE(Section);
150       if (!CheckRE.isValid(REError)) {
151         Error =
152             (Twine("malformed regex for section ") + Section + ": '" + REError)
153                 .str();
154         return false;
155       }
156 
157       continue;
158     }
159 
160     // Get our prefix and unparsed regexp.
161     std::pair<StringRef, StringRef> SplitLine = I->split(":");
162     StringRef Prefix = SplitLine.first;
163     if (SplitLine.second.empty()) {
164       // Missing ':' in the line.
165       Error = (Twine("malformed line ") + Twine(LineNo) + ": '" +
166                SplitLine.first + "'").str();
167       return false;
168     }
169 
170     std::pair<StringRef, StringRef> SplitRegexp = SplitLine.second.split("=");
171     std::string Regexp = std::string(SplitRegexp.first);
172     StringRef Category = SplitRegexp.second;
173 
174     // Create this section if it has not been seen before.
175     if (!SectionsMap.contains(Section)) {
176       std::unique_ptr<Matcher> M = std::make_unique<Matcher>();
177       std::string REError;
178       if (!M->insert(std::string(Section), LineNo, REError)) {
179         Error = (Twine("malformed section ") + Section + ": '" + REError).str();
180         return false;
181       }
182 
183       SectionsMap[Section] = Sections.size();
184       Sections.emplace_back(std::move(M));
185     }
186 
187     auto &Entry = Sections[SectionsMap[Section]].Entries[Prefix][Category];
188     std::string REError;
189     if (!Entry.insert(std::move(Regexp), LineNo, REError)) {
190       Error = (Twine("malformed regex in line ") + Twine(LineNo) + ": '" +
191                SplitLine.second + "': " + REError).str();
192       return false;
193     }
194   }
195   return true;
196 }
197 
198 SpecialCaseList::~SpecialCaseList() = default;
199 
200 bool SpecialCaseList::inSection(StringRef Section, StringRef Prefix,
201                                 StringRef Query, StringRef Category) const {
202   return inSectionBlame(Section, Prefix, Query, Category);
203 }
204 
205 unsigned SpecialCaseList::inSectionBlame(StringRef Section, StringRef Prefix,
206                                          StringRef Query,
207                                          StringRef Category) const {
208   for (const auto &SectionIter : Sections)
209     if (SectionIter.SectionMatcher->match(Section)) {
210       unsigned Blame =
211           inSectionBlame(SectionIter.Entries, Prefix, Query, Category);
212       if (Blame)
213         return Blame;
214     }
215   return 0;
216 }
217 
218 unsigned SpecialCaseList::inSectionBlame(const SectionEntries &Entries,
219                                          StringRef Prefix, StringRef Query,
220                                          StringRef Category) const {
221   SectionEntries::const_iterator I = Entries.find(Prefix);
222   if (I == Entries.end()) return 0;
223   StringMap<Matcher>::const_iterator II = I->second.find(Category);
224   if (II == I->second.end()) return 0;
225 
226   return II->getValue().match(Query);
227 }
228 
229 }  // namespace llvm
230