1 /*
2  * Copyright (C) 2015 The Android Open Source Project
3  *
4  * Licensed under the Apache License, Version 2.0 (the "License");
5  * you may not use this file except in compliance with the License.
6  * You may obtain a copy of the License at
7  *
8  *      http://www.apache.org/licenses/LICENSE-2.0
9  *
10  * Unless required by applicable law or agreed to in writing, software
11  * distributed under the License is distributed on an "AS IS" BASIS,
12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13  * See the License for the specific language governing permissions and
14  * limitations under the License.
15  */
16 
17 #ifndef ART_LIBDEXFILE_DEX_TEST_DEX_FILE_BUILDER_H_
18 #define ART_LIBDEXFILE_DEX_TEST_DEX_FILE_BUILDER_H_
19 
20 #include <zlib.h>
21 
22 #include <cstring>
23 #include <map>
24 #include <set>
25 #include <vector>
26 
27 #include <android-base/logging.h>
28 
29 #include "dex/dex_file_loader.h"
30 #include "dex/standard_dex_file.h"
31 
32 namespace art {
33 
34 class TestDexFileBuilder {
35  public:
TestDexFileBuilder()36   TestDexFileBuilder()
37       : strings_(), types_(), fields_(), protos_(), dex_file_data_() {
38   }
39 
AddString(const std::string & str)40   void AddString(const std::string& str) {
41     CHECK(dex_file_data_.empty());
42     auto it = strings_.emplace(str, IdxAndDataOffset()).first;
43     CHECK_LT(it->first.length(), 128u);  // Don't allow multi-byte length in uleb128.
44   }
45 
AddType(const std::string & descriptor)46   void AddType(const std::string& descriptor) {
47     CHECK(dex_file_data_.empty());
48     AddString(descriptor);
49     types_.emplace(descriptor, 0u);
50   }
51 
AddField(const std::string & class_descriptor,const std::string & type,const std::string & name)52   void AddField(const std::string& class_descriptor, const std::string& type,
53                 const std::string& name) {
54     CHECK(dex_file_data_.empty());
55     AddType(class_descriptor);
56     AddType(type);
57     AddString(name);
58     FieldKey key = { class_descriptor, type, name };
59     fields_.emplace(key, 0u);
60   }
61 
AddMethod(const std::string & class_descriptor,const std::string & signature,const std::string & name)62   void AddMethod(const std::string& class_descriptor, const std::string& signature,
63                  const std::string& name) {
64     CHECK(dex_file_data_.empty());
65     AddType(class_descriptor);
66     AddString(name);
67 
68     ProtoKey proto_key = CreateProtoKey(signature);
69     AddString(proto_key.shorty);
70     AddType(proto_key.return_type);
71     for (const auto& arg_type : proto_key.args) {
72       AddType(arg_type);
73     }
74     auto it = protos_.emplace(proto_key, IdxAndDataOffset()).first;
75     const ProtoKey* proto = &it->first;  // Valid as long as the element remains in protos_.
76 
77     MethodKey method_key = {
78         class_descriptor, name, proto
79     };
80     methods_.emplace(method_key, 0u);
81   }
82 
83   // NOTE: The builder holds the actual data, so it must live as long as the dex file.
Build(const std::string & dex_location)84   std::unique_ptr<const DexFile> Build(const std::string& dex_location) {
85     CHECK(dex_file_data_.empty());
86     union {
87       uint8_t data[sizeof(DexFile::Header)];
88       uint64_t force_alignment;
89     } header_data;
90     std::memset(header_data.data, 0, sizeof(header_data.data));
91     DexFile::Header* header = reinterpret_cast<DexFile::Header*>(&header_data.data);
92     std::copy_n(StandardDexFile::kDexMagic, 4u, header->magic_);
93     std::copy_n(StandardDexFile::kDexMagicVersions[0], 4u, header->magic_ + 4u);
94     header->header_size_ = sizeof(DexFile::Header);
95     header->endian_tag_ = DexFile::kDexEndianConstant;
96     header->link_size_ = 0u;  // Unused.
97     header->link_off_ = 0u;  // Unused.
98     header->map_off_ = 0u;  // Unused. TODO: This is wrong. Dex files created by this builder
99                             //               cannot be verified. b/26808512
100 
101     uint32_t data_section_size = 0u;
102 
103     uint32_t string_ids_offset = sizeof(DexFile::Header);
104     uint32_t string_idx = 0u;
105     for (auto& entry : strings_) {
106       entry.second.idx = string_idx;
107       string_idx += 1u;
108       entry.second.data_offset = data_section_size;
109       data_section_size += entry.first.length() + 1u /* length */ + 1u /* null-terminator */;
110     }
111     header->string_ids_size_ = strings_.size();
112     header->string_ids_off_ = strings_.empty() ? 0u : string_ids_offset;
113 
114     uint32_t type_ids_offset = string_ids_offset + strings_.size() * sizeof(DexFile::StringId);
115     uint32_t type_idx = 0u;
116     for (auto& entry : types_) {
117       entry.second = type_idx;
118       type_idx += 1u;
119     }
120     header->type_ids_size_ = types_.size();
121     header->type_ids_off_ = types_.empty() ? 0u : type_ids_offset;
122 
123     uint32_t proto_ids_offset = type_ids_offset + types_.size() * sizeof(DexFile::TypeId);
124     uint32_t proto_idx = 0u;
125     for (auto& entry : protos_) {
126       entry.second.idx = proto_idx;
127       proto_idx += 1u;
128       size_t num_args = entry.first.args.size();
129       if (num_args != 0u) {
130         entry.second.data_offset = RoundUp(data_section_size, 4u);
131         data_section_size = entry.second.data_offset + 4u + num_args * sizeof(DexFile::TypeItem);
132       } else {
133         entry.second.data_offset = 0u;
134       }
135     }
136     header->proto_ids_size_ = protos_.size();
137     header->proto_ids_off_ = protos_.empty() ? 0u : proto_ids_offset;
138 
139     uint32_t field_ids_offset = proto_ids_offset + protos_.size() * sizeof(DexFile::ProtoId);
140     uint32_t field_idx = 0u;
141     for (auto& entry : fields_) {
142       entry.second = field_idx;
143       field_idx += 1u;
144     }
145     header->field_ids_size_ = fields_.size();
146     header->field_ids_off_ = fields_.empty() ? 0u : field_ids_offset;
147 
148     uint32_t method_ids_offset = field_ids_offset + fields_.size() * sizeof(DexFile::FieldId);
149     uint32_t method_idx = 0u;
150     for (auto& entry : methods_) {
151       entry.second = method_idx;
152       method_idx += 1u;
153     }
154     header->method_ids_size_ = methods_.size();
155     header->method_ids_off_ = methods_.empty() ? 0u : method_ids_offset;
156 
157     // No class defs.
158     header->class_defs_size_ = 0u;
159     header->class_defs_off_ = 0u;
160 
161     uint32_t data_section_offset = method_ids_offset + methods_.size() * sizeof(DexFile::MethodId);
162     header->data_size_ = data_section_size;
163     header->data_off_ = (data_section_size != 0u) ? data_section_offset : 0u;
164 
165     uint32_t total_size = data_section_offset + data_section_size;
166 
167     dex_file_data_.resize(total_size);
168 
169     for (const auto& entry : strings_) {
170       CHECK_LT(entry.first.size(), 128u);
171       uint32_t raw_offset = data_section_offset + entry.second.data_offset;
172       dex_file_data_[raw_offset] = static_cast<uint8_t>(entry.first.size());
173       std::memcpy(&dex_file_data_[raw_offset + 1], entry.first.c_str(), entry.first.size() + 1);
174       Write32(string_ids_offset + entry.second.idx * sizeof(DexFile::StringId), raw_offset);
175     }
176 
177     for (const auto& entry : types_) {
178       Write32(type_ids_offset + entry.second * sizeof(DexFile::TypeId), GetStringIdx(entry.first));
179       ++type_idx;
180     }
181 
182     for (const auto& entry : protos_) {
183       size_t num_args = entry.first.args.size();
184       uint32_t type_list_offset =
185           (num_args != 0u) ? data_section_offset + entry.second.data_offset : 0u;
186       uint32_t raw_offset = proto_ids_offset + entry.second.idx * sizeof(DexFile::ProtoId);
187       Write32(raw_offset + 0u, GetStringIdx(entry.first.shorty));
188       Write16(raw_offset + 4u, GetTypeIdx(entry.first.return_type));
189       Write32(raw_offset + 8u, type_list_offset);
190       if (num_args != 0u) {
191         CHECK_NE(entry.second.data_offset, 0u);
192         Write32(type_list_offset, num_args);
193         for (size_t i = 0; i != num_args; ++i) {
194           Write16(type_list_offset + 4u + i * sizeof(DexFile::TypeItem),
195                   GetTypeIdx(entry.first.args[i]));
196         }
197       }
198     }
199 
200     for (const auto& entry : fields_) {
201       uint32_t raw_offset = field_ids_offset + entry.second * sizeof(DexFile::FieldId);
202       Write16(raw_offset + 0u, GetTypeIdx(entry.first.class_descriptor));
203       Write16(raw_offset + 2u, GetTypeIdx(entry.first.type));
204       Write32(raw_offset + 4u, GetStringIdx(entry.first.name));
205     }
206 
207     for (const auto& entry : methods_) {
208       uint32_t raw_offset = method_ids_offset + entry.second * sizeof(DexFile::MethodId);
209       Write16(raw_offset + 0u, GetTypeIdx(entry.first.class_descriptor));
210       auto it = protos_.find(*entry.first.proto);
211       CHECK(it != protos_.end());
212       Write16(raw_offset + 2u, it->second.idx);
213       Write32(raw_offset + 4u, GetStringIdx(entry.first.name));
214     }
215 
216     // Leave signature as zeros.
217 
218     header->file_size_ = dex_file_data_.size();
219 
220     // Write the complete header early, as part of it needs to be checksummed.
221     std::memcpy(&dex_file_data_[0], header_data.data, sizeof(DexFile::Header));
222 
223     // Checksum starts after the checksum field.
224     size_t skip = sizeof(header->magic_) + sizeof(header->checksum_);
225     header->checksum_ = adler32(adler32(0L, Z_NULL, 0),
226                                 dex_file_data_.data() + skip,
227                                 dex_file_data_.size() - skip);
228 
229     // Write the complete header again, just simpler that way.
230     std::memcpy(&dex_file_data_[0], header_data.data, sizeof(DexFile::Header));
231 
232     static constexpr bool kVerify = false;
233     static constexpr bool kVerifyChecksum = false;
234     std::string error_msg;
235     const DexFileLoader dex_file_loader;
236     std::unique_ptr<const DexFile> dex_file(dex_file_loader.Open(
237         &dex_file_data_[0],
238         dex_file_data_.size(),
239         dex_location,
240         0u,
241         nullptr,
242         kVerify,
243         kVerifyChecksum,
244         &error_msg));
245     CHECK(dex_file != nullptr) << error_msg;
246     return dex_file;
247   }
248 
GetStringIdx(const std::string & type)249   uint32_t GetStringIdx(const std::string& type) {
250     auto it = strings_.find(type);
251     CHECK(it != strings_.end());
252     return it->second.idx;
253   }
254 
GetTypeIdx(const std::string & type)255   uint32_t GetTypeIdx(const std::string& type) {
256     auto it = types_.find(type);
257     CHECK(it != types_.end());
258     return it->second;
259   }
260 
GetFieldIdx(const std::string & class_descriptor,const std::string & type,const std::string & name)261   uint32_t GetFieldIdx(const std::string& class_descriptor, const std::string& type,
262                        const std::string& name) {
263     FieldKey key = { class_descriptor, type, name };
264     auto it = fields_.find(key);
265     CHECK(it != fields_.end());
266     return it->second;
267   }
268 
GetMethodIdx(const std::string & class_descriptor,const std::string & signature,const std::string & name)269   uint32_t GetMethodIdx(const std::string& class_descriptor, const std::string& signature,
270                         const std::string& name) {
271     ProtoKey proto_key = CreateProtoKey(signature);
272     MethodKey method_key = { class_descriptor, name, &proto_key };
273     auto it = methods_.find(method_key);
274     CHECK(it != methods_.end());
275     return it->second;
276   }
277 
278  private:
279   struct IdxAndDataOffset {
280     uint32_t idx;
281     uint32_t data_offset;
282   };
283 
284   struct FieldKey {
285     const std::string class_descriptor;
286     const std::string type;
287     const std::string name;
288   };
289   struct FieldKeyComparator {
operatorFieldKeyComparator290     bool operator()(const FieldKey& lhs, const FieldKey& rhs) const {
291       if (lhs.class_descriptor != rhs.class_descriptor) {
292         return lhs.class_descriptor < rhs.class_descriptor;
293       }
294       if (lhs.name != rhs.name) {
295         return lhs.name < rhs.name;
296       }
297       return lhs.type < rhs.type;
298     }
299   };
300 
301   struct ProtoKey {
302     std::string shorty;
303     std::string return_type;
304     std::vector<std::string> args;
305   };
306   struct ProtoKeyComparator {
operatorProtoKeyComparator307     bool operator()(const ProtoKey& lhs, const ProtoKey& rhs) const {
308       if (lhs.return_type != rhs.return_type) {
309         return lhs.return_type < rhs.return_type;
310       }
311       size_t min_args = std::min(lhs.args.size(), rhs.args.size());
312       for (size_t i = 0; i != min_args; ++i) {
313         if (lhs.args[i] != rhs.args[i]) {
314           return lhs.args[i] < rhs.args[i];
315         }
316       }
317       return lhs.args.size() < rhs.args.size();
318     }
319   };
320 
321   struct MethodKey {
322     std::string class_descriptor;
323     std::string name;
324     const ProtoKey* proto;
325   };
326   struct MethodKeyComparator {
operatorMethodKeyComparator327     bool operator()(const MethodKey& lhs, const MethodKey& rhs) const {
328       if (lhs.class_descriptor != rhs.class_descriptor) {
329         return lhs.class_descriptor < rhs.class_descriptor;
330       }
331       if (lhs.name != rhs.name) {
332         return lhs.name < rhs.name;
333       }
334       return ProtoKeyComparator()(*lhs.proto, *rhs.proto);
335     }
336   };
337 
CreateProtoKey(const std::string & signature)338   ProtoKey CreateProtoKey(const std::string& signature) {
339     CHECK_EQ(signature[0], '(');
340     const char* args = signature.c_str() + 1;
341     const char* args_end = std::strchr(args, ')');
342     CHECK(args_end != nullptr);
343     const char* return_type = args_end + 1;
344 
345     ProtoKey key = {
346         std::string() + ((*return_type == '[') ? 'L' : *return_type),
347         return_type,
348         std::vector<std::string>()
349     };
350     while (args != args_end) {
351       key.shorty += (*args == '[') ? 'L' : *args;
352       const char* arg_start = args;
353       while (*args == '[') {
354         ++args;
355       }
356       if (*args == 'L') {
357         do {
358           ++args;
359           CHECK_NE(args, args_end);
360         } while (*args != ';');
361       }
362       ++args;
363       key.args.emplace_back(arg_start, args);
364     }
365     return key;
366   }
367 
Write32(size_t offset,uint32_t value)368   void Write32(size_t offset, uint32_t value) {
369     CHECK_LE(offset + 4u, dex_file_data_.size());
370     CHECK_EQ(dex_file_data_[offset + 0], 0u);
371     CHECK_EQ(dex_file_data_[offset + 1], 0u);
372     CHECK_EQ(dex_file_data_[offset + 2], 0u);
373     CHECK_EQ(dex_file_data_[offset + 3], 0u);
374     dex_file_data_[offset + 0] = static_cast<uint8_t>(value >> 0);
375     dex_file_data_[offset + 1] = static_cast<uint8_t>(value >> 8);
376     dex_file_data_[offset + 2] = static_cast<uint8_t>(value >> 16);
377     dex_file_data_[offset + 3] = static_cast<uint8_t>(value >> 24);
378   }
379 
Write16(size_t offset,uint32_t value)380   void Write16(size_t offset, uint32_t value) {
381     CHECK_LE(value, 0xffffu);
382     CHECK_LE(offset + 2u, dex_file_data_.size());
383     CHECK_EQ(dex_file_data_[offset + 0], 0u);
384     CHECK_EQ(dex_file_data_[offset + 1], 0u);
385     dex_file_data_[offset + 0] = static_cast<uint8_t>(value >> 0);
386     dex_file_data_[offset + 1] = static_cast<uint8_t>(value >> 8);
387   }
388 
389   std::map<std::string, IdxAndDataOffset> strings_;
390   std::map<std::string, uint32_t> types_;
391   std::map<FieldKey, uint32_t, FieldKeyComparator> fields_;
392   std::map<ProtoKey, IdxAndDataOffset, ProtoKeyComparator> protos_;
393   std::map<MethodKey, uint32_t, MethodKeyComparator> methods_;
394 
395   std::vector<uint8_t> dex_file_data_;
396 };
397 
398 }  // namespace art
399 
400 #endif  // ART_LIBDEXFILE_DEX_TEST_DEX_FILE_BUILDER_H_
401