OLD | NEW |
---|---|
(Empty) | |
1 // Copyright 2016 The Chromium Authors. All rights reserved. | |
2 // Use of this source code is governed by a BSD-style license that can be | |
3 // found in the LICENSE file. | |
4 | |
5 #include <memory> | |
6 #include <stdio.h> | |
7 | |
8 #include "clang/ASTMatchers/ASTMatchFinder.h" | |
9 #include "clang/ASTMatchers/ASTMatchers.h" | |
10 #include "clang/Basic/SourceManager.h" | |
11 #include "clang/Frontend/FrontendActions.h" | |
12 #include "clang/Lex/Lexer.h" | |
13 #include "clang/Tooling/CommonOptionsParser.h" | |
14 #include "clang/Tooling/Refactoring.h" | |
15 #include "clang/Tooling/Tooling.h" | |
16 #include "llvm/Support/CommandLine.h" | |
17 //--> TODO: #include "../../traffic_annotation/traffic_annotation.pb.h" | |
battre
2016/10/26 14:29:18
remove?
Ramin Halavati
2016/10/27 09:40:08
I think if we want to add protobuf to clang tool i
| |
18 | |
19 using namespace clang::ast_matchers; | |
20 using clang::tooling::CommonOptionsParser; | |
21 | |
22 namespace { | |
23 | |
24 // Structure to collect instances. | |
battre
2016/10/26 14:29:19
instances of what? Can you add more comments to th
Ramin Halavati
2016/10/27 09:40:08
Done.
| |
25 typedef struct Collector { | |
battre
2016/10/26 14:29:18
I think that in C++ we don't need to "typedef stru
Ramin Halavati
2016/10/27 09:40:10
Done.
| |
26 typedef struct Instance { | |
27 Instance() { | |
battre
2016/10/26 14:29:18
Can you use an initializer list?
Instance() : is_
Ramin Halavati
2016/10/27 09:40:08
Done.
| |
28 is_direct_call = false; | |
29 transitive_parameter = false; | |
30 variable_reference = NULL; | |
battre
2016/10/26 14:29:19
nullptr
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
31 } | |
32 | |
33 // Information about where this annotation or call has happened. | |
34 struct Location { | |
35 std::string file_name; | |
36 int line_number; | |
battre
2016/10/26 14:29:19
line number is not default initialized.
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
37 // Name of the function including this instance. | |
38 std::string function_name; | |
39 // Name of the variable that contains annotation or the function called | |
40 // with annotation. | |
41 std::string object_name; | |
42 }; | |
43 | |
44 // Annotation content | |
battre
2016/10/26 14:29:18
Comments should end with period. (please also chec
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
45 struct Annotation { | |
46 std::string unique_id; | |
47 std::string text; | |
48 }; | |
49 | |
50 Location location; | |
51 Annotation annotation; | |
52 | |
53 // Possible error (empty if no error) | |
54 std::string error; | |
55 // A reference to the variable containing annotation. Null if not available. | |
56 const clang::NamedDecl* variable_reference; | |
57 // Flag stating that parameter is directly passed to annotate function here | |
58 // or it's through a variable. | |
59 bool is_direct_call; | |
60 // Flag stating that a variable is a parameter received by upper level | |
61 // function. | |
62 bool transitive_parameter; | |
63 } Instance; | |
64 | |
65 std::vector<Instance> variable_definitions, calls; | |
battre
2016/10/26 14:29:19
one variable definition per line.
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
66 } Collector; | |
67 | |
68 // Returns the function that includes the given clang::Decl. | |
69 std::string GetCoveringFunction(const clang::Decl* token, | |
70 const MatchFinder::MatchResult& result); | |
71 | |
72 // Checks if a token matches a name, with or without net:: namespace | |
73 bool net_match(const std::string& token, const std::string& name) { | |
74 return token == name || token == (std::string("net::") + name); | |
75 } | |
76 | |
77 // Returns the text of a given statement or subclass. | |
78 std::string GetStmtText(const clang::Stmt* token, | |
79 const clang::SourceManager& source_manager) { | |
80 clang::LangOptions lopt; | |
81 // Get text range | |
82 clang::SourceLocation start = token->getLocStart(); | |
83 clang::SourceLocation end = token->getLocEnd(); | |
84 | |
85 // If it's a macro, go to definition. | |
86 if (start.isMacroID()) | |
87 start = source_manager.getSpellingLoc(start); | |
88 if (end.isMacroID()) | |
89 end = source_manager.getSpellingLoc(end); | |
90 | |
91 // Get the real end of the token. | |
92 end = clang::Lexer::getLocForEndOfToken(end, 0, source_manager, lopt); | |
93 | |
94 // Extract text. | |
95 std::string output = std::string(source_manager.getCharacterData(start), | |
96 source_manager.getCharacterData(end) - | |
97 source_manager.getCharacterData(start)); | |
battre
2016/10/26 14:29:19
can you also write
std::string output(source_manag
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
98 | |
99 // Raw string? | |
100 if (output == "R") | |
battre
2016/10/26 14:29:19
{} around multiline blocks. Same in the next line.
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
101 if (auto* c1 = clang::dyn_cast<clang::ImplicitCastExpr>(token)) | |
102 if (const clang::StringLiteral* c2 = | |
103 clang::dyn_cast<clang::StringLiteral>( | |
104 c1->getSubExprAsWritten())) { | |
105 output = c2->getString(); | |
106 } | |
107 | |
108 return output; | |
109 } | |
110 | |
111 // Returns annotation text of a call to "DefineNetworkTrafficAnnotation" | |
112 // function. | |
113 void GetAnnotationText(const clang::CallExpr* call_expr, | |
114 const clang::SourceManager& source_manager, | |
115 Collector::Instance* instance) { | |
116 if (net_match(GetStmtText(call_expr->getCallee(), source_manager), | |
117 "DefineNetworkTrafficAnnotation") && | |
118 call_expr->getNumArgs() == 2) { | |
119 instance->annotation.unique_id = | |
120 GetStmtText(call_expr->getArgs()[0], source_manager); | |
121 instance->annotation.text = | |
122 GetStmtText(call_expr->getArgs()[1], source_manager); | |
123 instance->error = ""; | |
124 } else { | |
125 instance->annotation.unique_id = ""; | |
126 instance->annotation.text = ""; | |
127 instance->error = "Unexpected function."; | |
128 } | |
129 } | |
130 | |
131 // Returns the function that includes the given clang::Stmt. | |
132 std::string GetCoveringFunction(const clang::Stmt* token, | |
133 const MatchFinder::MatchResult& result) { | |
134 auto parents = result.Context->getParents(*token); | |
135 if (parents.size() == 1) { | |
136 if (const clang::Stmt* s = parents[0].get<clang::Stmt>()) | |
137 return GetCoveringFunction(s, result); | |
138 else if (const clang::Decl* d = parents[0].get<clang::Decl>()) | |
139 return GetCoveringFunction(d, result); | |
140 } | |
141 return "Unknown"; | |
142 } | |
143 | |
144 // Returns the function that includes the given clang::Decl. | |
145 std::string GetCoveringFunction(const clang::Decl* token, | |
146 const MatchFinder::MatchResult& result) { | |
147 if (auto f = clang::dyn_cast<clang::FunctionDecl>(token)) | |
148 return f->getQualifiedNameAsString(); | |
149 | |
150 auto parents = result.Context->getParents(*token); | |
151 if (parents.size() == 1) { | |
battre
2016/10/26 14:29:18
Add a comment what having a single parent means?
Ramin Halavati
2016/10/27 09:40:09
I did not find a conclusive description on this an
| |
152 if (const clang::Stmt* s = parents[0].get<clang::Stmt>()) | |
153 return GetCoveringFunction(s, result); | |
154 else if (const clang::Decl* d = parents[0].get<clang::Decl>()) | |
155 return GetCoveringFunction(d, result); | |
156 } | |
157 return "Unknown"; | |
158 } | |
159 | |
160 // Returns the file name and line number of the given token. | |
161 template <class T> | |
162 void GetOccuranceLocation(const T* token, | |
battre
2016/10/26 14:29:19
typo: occurr*e*nce
What's the difference between
Ramin Halavati
2016/10/27 09:40:09
Renamed it to GetLocation as it can be called gene
| |
163 const MatchFinder::MatchResult& result, | |
164 Collector::Instance::Location* location) { | |
165 clang::SourceLocation source_location = token->getLocStart(); | |
166 location->file_name = result.SourceManager->getFilename(source_location); | |
167 location->line_number = | |
168 result.SourceManager->getSpellingLineNumber(source_location); | |
169 } | |
170 | |
171 class TheCallback : public MatchFinder::MatchCallback { | |
battre
2016/10/26 14:29:19
Description of the responsibility of this class?
Ramin Halavati
2016/10/27 09:40:08
Done.
| |
172 public: | |
173 explicit TheCallback(Collector* collector) : collector_(collector) {} | |
battre
2016/10/26 14:29:18
~TheCallback() override = default;
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
174 | |
175 virtual void run(const MatchFinder::MatchResult& result) override { | |
176 if (const clang::VarDecl* var_decl = | |
177 result.Nodes.getNodeAs<clang::VarDecl>("annotation_variable")) | |
battre
2016/10/26 14:29:19
also {} if the is statement spans more than one li
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
178 AddVariable(var_decl, result); | |
179 else if (const clang::CallExpr* call_expr = | |
180 result.Nodes.getNodeAs<clang::CallExpr>("user_function")) | |
181 AddFunction(call_expr, result); | |
182 } | |
183 | |
184 // Stores an annotation variable defintion in the Collector. | |
185 void AddVariable(const clang::VarDecl* var_decl, | |
186 const MatchFinder::MatchResult& result) { | |
187 Collector::Instance instance; | |
188 | |
189 GetOccuranceLocation(var_decl, result, &instance.location); | |
190 instance.location.object_name = var_decl->getQualifiedNameAsString(); | |
191 instance.variable_reference = clang::dyn_cast<clang::NamedDecl>(var_decl); | |
192 | |
193 // Get annotation text. | |
194 // If it doesn't have initialization, but it's a parameter, store it. | |
195 if (!var_decl->hasInit() && var_decl->isLocalVarDeclOrParm() && | |
196 !var_decl->isLocalVarDecl()) { | |
197 instance.transitive_parameter = true; | |
198 } else if (auto* init_expr = var_decl->getInit()) { | |
199 if (auto* call_expr = clang::dyn_cast<clang::CallExpr>(init_expr)) | |
200 GetAnnotationText(call_expr, *result.SourceManager, &instance); | |
201 } | |
202 // If nothing is set, issue an error. | |
203 if (!instance.transitive_parameter && | |
204 instance.annotation.unique_id.empty() && instance.error.empty()) | |
battre
2016/10/26 14:29:19
{}
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
205 instance.error = "Could not resolve variable initialization."; | |
206 | |
207 collector_->variable_definitions.push_back(instance); | |
208 } | |
209 | |
210 // Stores a function call that should be monitored. | |
211 void AddFunction(const clang::CallExpr* call_expr, | |
212 const MatchFinder::MatchResult& result) { | |
213 Collector::Instance instance; | |
214 | |
215 GetOccuranceLocation(call_expr, result, &instance.location); | |
216 instance.location.function_name = | |
217 GetCoveringFunction(clang::dyn_cast<clang::Stmt>(call_expr), result); | |
218 instance.location.object_name = | |
219 call_expr->getDirectCallee()->getQualifiedNameAsString(); | |
220 | |
221 // Get annotation text. | |
222 const clang::FunctionDecl* function_decl = call_expr->getDirectCallee(); | |
223 unsigned params_count = function_decl->getNumParams(); | |
224 unsigned args_count = call_expr->getNumArgs(); | |
225 | |
226 for (unsigned i = 0; i < params_count; i++) { | |
227 if (net_match(clang::QualType::getAsString( | |
228 function_decl->getParamDecl(i)->getType().split()), | |
229 "NetworkTrafficAnnotationTag")) { | |
230 if (i >= args_count) { | |
231 instance.error = "Function missing annotation argument."; | |
232 } else { | |
233 // Get the argument. | |
234 const clang::Expr* arg = call_expr->getArgs()[i]; | |
235 | |
236 // Is it a call to annotate function? | |
237 if (auto* inner_call_expr = clang::dyn_cast<clang::CallExpr>(arg)) { | |
238 instance.is_direct_call = true; | |
239 GetAnnotationText(inner_call_expr, *result.SourceManager, | |
240 &instance); | |
241 instance.error = ""; | |
242 } else { | |
243 // Then it's a variable. | |
244 instance.is_direct_call = false; | |
245 if (auto* pure_arg = | |
246 clang::dyn_cast<clang::DeclRefExpr>(arg->IgnoreCasts())) { | |
247 instance.variable_reference = pure_arg->getFoundDecl(); | |
248 instance.error = ""; | |
249 } else { | |
250 instance.error = "Unknwon parameter type."; | |
251 } | |
252 } | |
253 } | |
254 collector_->calls.push_back(instance); | |
255 } | |
256 } | |
257 } | |
258 | |
259 private: | |
260 // TODO store pointer | |
battre
2016/10/26 14:29:19
Delete the todo?
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
261 Collector* collector_; | |
262 }; | |
263 | |
264 // NRA class keeps the call back function and sets the matchers. | |
battre
2016/10/26 14:29:18
Please don't use abbreviations.
Ramin Halavati
2016/10/27 09:40:08
Done. The class is totally removed and replaced by
| |
265 class NetworkRequestAuditor { | |
battre
2016/10/26 14:29:19
Rename this to NetworkAnnotationTagExtractor? Then
Ramin Halavati
2016/10/27 09:40:09
The class is totally removed and replaced by a sin
| |
266 public: | |
267 explicit NetworkRequestAuditor(Collector* collector) | |
268 : the_callback_(collector) {} | |
269 | |
270 void SetupMatchers(MatchFinder* match_finder) { | |
271 // Find variables defined as Annotation | |
battre
2016/10/26 14:29:19
.
What does "Find variables defined as Annotation
Ramin Halavati
2016/10/27 09:40:08
Comments changed.
| |
272 match_finder->addMatcher( | |
273 varDecl(anyOf(hasType(asString("NetworkTrafficAnnotationTag")), | |
274 hasType(asString("net::NetworkTrafficAnnotationTag")))) | |
275 .bind("annotation_variable"), | |
276 &the_callback_); | |
277 | |
278 // Find instances of functions that have a parameter of type Annotate. | |
279 match_finder->addMatcher( | |
280 callExpr(hasDeclaration(functionDecl(hasAnyParameter(anyOf( | |
281 hasType(asString("NetworkTrafficAnnotationTag")), | |
282 hasType(asString("net::NetworkTrafficAnnotationTag"))))))) | |
283 .bind("user_function"), | |
284 &the_callback_); | |
285 } | |
286 | |
287 private: | |
288 TheCallback the_callback_; | |
289 }; | |
290 | |
291 } // namespace | |
292 | |
293 int main(int argc, const char* argv[]) { | |
294 llvm::cl::OptionCategory category("Network Request Audit Extractor Tool"); | |
295 CommonOptionsParser options(argc, argv, category); | |
296 clang::tooling::ClangTool tool(options.getCompilations(), | |
297 options.getSourcePathList()); | |
298 | |
299 Collector collector; | |
300 NetworkRequestAuditor auditor(&collector); | |
301 MatchFinder match_finder; | |
302 auditor.SetupMatchers(&match_finder); | |
303 | |
304 // Find output folder | |
305 const std::string kOutputSpecifier("output_dir="); | |
306 std::string output_dir; | |
307 for (int i = 0; i < argc; i++) { | |
308 if (!strncmp(argv[i], kOutputSpecifier.c_str(), | |
309 kOutputSpecifier.length())) { | |
battre
2016/10/26 14:29:19
if (argv[i] == kOutputSpecifier) does the same thi
Ramin Halavati
2016/10/27 09:40:08
argv[i] parameter includes the specifier and the p
| |
310 output_dir = argv[i] + kOutputSpecifier.length(); | |
311 break; | |
312 } | |
313 } | |
314 | |
315 if (output_dir == "") { | |
316 llvm::errs() << "Temporary files directory is not specified."; | |
317 return 1; | |
318 } | |
319 | |
320 std::unique_ptr<clang::tooling::FrontendActionFactory> frontend_factory = | |
321 clang::tooling::newFrontendActionFactory(&match_finder); | |
322 int result = tool.run(frontend_factory.get()); | |
323 | |
324 if (result != 0) | |
325 return result; | |
326 | |
327 llvm::outs() << "==== BEGIN EDITS ====\n"; | |
328 llvm::outs() << "==== END EDITS ====\n"; | |
329 | |
330 // For each call, if the parameter is not generated by a direct call to | |
331 // "DefineNetworkTrafficAnnotation", find the variable that holds the value. | |
332 for (auto& c : collector.calls) { | |
battre
2016/10/26 14:29:19
const auto& c?
Ramin Halavati
2016/10/27 09:40:09
the annotation field of iterators may be changed 5
| |
333 if (!c.is_direct_call) { | |
334 // Find the variable. | |
335 for (auto& v : collector.variable_definitions) | |
battre
2016/10/26 14:29:19
{}
const auto& ?
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
336 if (v.variable_reference == c.variable_reference) { | |
337 c.annotation = v.annotation; | |
338 c.transitive_parameter = v.transitive_parameter; | |
339 c.error = c.error + (c.error.length() ? "\n+" : "") + v.error; | |
340 break; | |
341 } | |
342 if (!c.annotation.unique_id.length()) | |
343 c.error = "Variable not found."; | |
344 } | |
345 | |
346 // If the function just receives the variable and passes it to another | |
347 // function, ignore it, otherwise write it to file. | |
348 if (!c.transitive_parameter) { | |
349 std::string s = c.location.file_name; | |
350 std::replace(s.begin(), s.end(), '/', '_'); | |
351 std::replace(s.begin(), s.end(), '.', '_'); | |
352 char file_name[1000]; | |
353 | |
354 snprintf(file_name, sizeof(file_name), "%s/%s(%i).txt", | |
355 output_dir.c_str(), s.c_str(), c.location.line_number); | |
battre
2016/10/26 14:29:18
std::string filename = output_dir + "/" + s + "("
Ramin Halavati
2016/10/27 09:40:09
Done.
| |
356 | |
357 FILE* file = fopen(file_name, "wt"); | |
battre
2016/10/26 14:29:18
how about staying C++ here and using std::ofstream
Ramin Halavati
2016/10/27 09:40:08
Done.
| |
358 if (file) { | |
359 fprintf(file, "%s\n", c.location.file_name.c_str()); | |
360 fprintf(file, "%s\n", c.location.function_name.c_str()); | |
361 fprintf(file, "%i\n", c.location.line_number); | |
362 fprintf(file, "%s\n", c.location.object_name.c_str()); | |
363 fprintf(file, "%s\n", c.error.c_str()); | |
364 fprintf(file, "%s\n", c.annotation.unique_id.c_str()); | |
365 fprintf(file, "%s", c.annotation.text.c_str()); | |
366 fclose(file); | |
367 } else { | |
368 llvm::errs() << "Could not write to file: " << file_name << " because " | |
369 << strerror(errno) << "\n"; | |
370 return 1; | |
371 } | |
372 } | |
373 } | |
374 | |
375 return 0; | |
376 } | |
OLD | NEW |