1//===-- SpecialCaseList.cpp - special case list for sanitizers ------------===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// This is a utility class for instrumentation passes (like AddressSanitizer
11// or ThreadSanitizer) to avoid instrumenting some functions or global
12// variables, or to instrument some functions or global variables in a specific
13// way, based on a user-supplied list.
14//
15//===----------------------------------------------------------------------===//
16
17#include "llvm/Support/SpecialCaseList.h"
18#include "llvm/ADT/SmallVector.h"
19#include "llvm/ADT/StringExtras.h"
20#include "llvm/ADT/StringSet.h"
21#include "llvm/Support/MemoryBuffer.h"
22#include "llvm/Support/Regex.h"
23#include <string>
24#include <system_error>
25#include <utility>
26
27namespace llvm {
28
29/// Represents a set of regular expressions.  Regular expressions which are
30/// "literal" (i.e. no regex metacharacters) are stored in Strings, while all
31/// others are represented as a single pipe-separated regex in RegEx.  The
32/// reason for doing so is efficiency; StringSet is much faster at matching
33/// literal strings than Regex.
34struct SpecialCaseList::Entry {
35  Entry() {}
36  Entry(Entry &&Other)
37      : Strings(std::move(Other.Strings)), RegEx(std::move(Other.RegEx)) {}
38
39  StringSet<> Strings;
40  std::unique_ptr<Regex> RegEx;
41
42  bool match(StringRef Query) const {
43    return Strings.count(Query) || (RegEx && RegEx->match(Query));
44  }
45};
46
47SpecialCaseList::SpecialCaseList() : Entries(), Regexps(), IsCompiled(false) {}
48
49std::unique_ptr<SpecialCaseList>
50SpecialCaseList::create(const std::vector<std::string> &Paths,
51                        std::string &Error) {
52  std::unique_ptr<SpecialCaseList> SCL(new SpecialCaseList());
53  for (const auto &Path : Paths) {
54    ErrorOr<std::unique_ptr<MemoryBuffer>> FileOrErr =
55        MemoryBuffer::getFile(Path);
56    if (std::error_code EC = FileOrErr.getError()) {
57      Error = (Twine("can't open file '") + Path + "': " + EC.message()).str();
58      return nullptr;
59    }
60    std::string ParseError;
61    if (!SCL->parse(FileOrErr.get().get(), ParseError)) {
62      Error = (Twine("error parsing file '") + Path + "': " + ParseError).str();
63      return nullptr;
64    }
65  }
66  SCL->compile();
67  return SCL;
68}
69
70std::unique_ptr<SpecialCaseList> SpecialCaseList::create(const MemoryBuffer *MB,
71                                                         std::string &Error) {
72  std::unique_ptr<SpecialCaseList> SCL(new SpecialCaseList());
73  if (!SCL->parse(MB, Error))
74    return nullptr;
75  SCL->compile();
76  return SCL;
77}
78
79std::unique_ptr<SpecialCaseList>
80SpecialCaseList::createOrDie(const std::vector<std::string> &Paths) {
81  std::string Error;
82  if (auto SCL = create(Paths, Error))
83    return SCL;
84  report_fatal_error(Error);
85}
86
87bool SpecialCaseList::parse(const MemoryBuffer *MB, std::string &Error) {
88  // Iterate through each line in the blacklist file.
89  SmallVector<StringRef, 16> Lines;
90  SplitString(MB->getBuffer(), Lines, "\n\r");
91  int LineNo = 1;
92  for (auto I = Lines.begin(), E = Lines.end(); I != E; ++I, ++LineNo) {
93    // Ignore empty lines and lines starting with "#"
94    if (I->empty() || I->startswith("#"))
95      continue;
96    // Get our prefix and unparsed regexp.
97    std::pair<StringRef, StringRef> SplitLine = I->split(":");
98    StringRef Prefix = SplitLine.first;
99    if (SplitLine.second.empty()) {
100      // Missing ':' in the line.
101      Error = (Twine("malformed line ") + Twine(LineNo) + ": '" +
102               SplitLine.first + "'").str();
103      return false;
104    }
105
106    std::pair<StringRef, StringRef> SplitRegexp = SplitLine.second.split("=");
107    std::string Regexp = SplitRegexp.first;
108    StringRef Category = SplitRegexp.second;
109
110    // See if we can store Regexp in Strings.
111    if (Regex::isLiteralERE(Regexp)) {
112      Entries[Prefix][Category].Strings.insert(Regexp);
113      continue;
114    }
115
116    // Replace * with .*
117    for (size_t pos = 0; (pos = Regexp.find("*", pos)) != std::string::npos;
118         pos += strlen(".*")) {
119      Regexp.replace(pos, strlen("*"), ".*");
120    }
121
122    // Check that the regexp is valid.
123    Regex CheckRE(Regexp);
124    std::string REError;
125    if (!CheckRE.isValid(REError)) {
126      Error = (Twine("malformed regex in line ") + Twine(LineNo) + ": '" +
127               SplitLine.second + "': " + REError).str();
128      return false;
129    }
130
131    // Add this regexp into the proper group by its prefix.
132    if (!Regexps[Prefix][Category].empty())
133      Regexps[Prefix][Category] += "|";
134    Regexps[Prefix][Category] += "^" + Regexp + "$";
135  }
136  return true;
137}
138
139void SpecialCaseList::compile() {
140  assert(!IsCompiled && "compile() should only be called once");
141  // Iterate through each of the prefixes, and create Regexs for them.
142  for (StringMap<StringMap<std::string>>::const_iterator I = Regexps.begin(),
143                                                         E = Regexps.end();
144       I != E; ++I) {
145    for (StringMap<std::string>::const_iterator II = I->second.begin(),
146                                                IE = I->second.end();
147         II != IE; ++II) {
148      Entries[I->getKey()][II->getKey()].RegEx.reset(new Regex(II->getValue()));
149    }
150  }
151  Regexps.clear();
152  IsCompiled = true;
153}
154
155SpecialCaseList::~SpecialCaseList() {}
156
157bool SpecialCaseList::inSection(StringRef Section, StringRef Query,
158                                StringRef Category) const {
159  assert(IsCompiled && "SpecialCaseList::compile() was not called!");
160  StringMap<StringMap<Entry> >::const_iterator I = Entries.find(Section);
161  if (I == Entries.end()) return false;
162  StringMap<Entry>::const_iterator II = I->second.find(Category);
163  if (II == I->second.end()) return false;
164
165  return II->getValue().match(Query);
166}
167
168}  // namespace llvm
169