OLD | NEW |
---|---|
1 // Copyright (c) 2011 The Chromium Authors. All rights reserved. | 1 // Copyright (c) 2011 The Chromium Authors. All rights reserved. |
2 // Use of this source code is governed by a BSD-style license that can be | 2 // Use of this source code is governed by a BSD-style license that can be |
3 // found in the LICENSE file. | 3 // found in the LICENSE file. |
4 | 4 |
5 #ifndef COURGETTE_ASSEMBLY_PROGRAM_H_ | 5 #ifndef COURGETTE_ASSEMBLY_PROGRAM_H_ |
6 #define COURGETTE_ASSEMBLY_PROGRAM_H_ | 6 #define COURGETTE_ASSEMBLY_PROGRAM_H_ |
7 | 7 |
8 #include <stddef.h> | 8 #include <stddef.h> |
9 #include <stdint.h> | 9 #include <stdint.h> |
10 | 10 |
11 #include <map> | 11 #include <map> |
12 #include <memory> | 12 #include <memory> |
13 #include <set> | 13 #include <set> |
14 #include <vector> | 14 #include <vector> |
15 | 15 |
16 #include "base/bind.h" | 16 #include "base/bind.h" |
17 #include "base/macros.h" | 17 #include "base/macros.h" |
18 #include "base/memory/free_deleter.h" | 18 #include "base/memory/free_deleter.h" |
19 #include "courgette/courgette.h" | 19 #include "courgette/courgette.h" |
20 #include "courgette/image_utils.h" | 20 #include "courgette/image_utils.h" |
21 #include "courgette/label_manager.h" | 21 #include "courgette/label_manager.h" |
22 #include "courgette/memory_allocator.h" | 22 #include "courgette/memory_allocator.h" |
23 | 23 |
24 namespace courgette { | 24 namespace courgette { |
25 | 25 |
26 class AssemblyProgram; | |
26 class EncodedProgram; | 27 class EncodedProgram; |
27 | 28 |
28 // Opcodes of simple assembly language | 29 // Opcodes of simple assembly language |
29 enum OP { | 30 enum OP { |
30 ORIGIN, // ORIGIN <rva> - set current address for assembly. | 31 ORIGIN, // ORIGIN <rva> - set current address for assembly. |
31 MAKEPERELOCS, // Generates a base relocation table. | 32 MAKEPERELOCS, // Generates a base relocation table. |
32 MAKEELFRELOCS, // Generates a base relocation table. | 33 MAKEELFRELOCS, // Generates a base relocation table. |
33 DEFBYTE, // DEFBYTE <value> - emit a byte literal. | 34 DEFBYTE, // DEFBYTE <value> - emit a byte literal. |
34 REL32, // REL32 <label> - emit a rel32 encoded reference to 'label'. | 35 REL32, // REL32 <label> - emit a rel32 encoded reference to 'label'. |
35 ABS32, // ABS32 <label> - emit an abs32 encoded reference to 'label'. | 36 ABS32, // ABS32 <label> - emit an abs32 encoded reference to 'label'. |
(...skipping 14 matching lines...) Expand all Loading... | |
50 explicit Instruction(OP op) : op_(op), info_(0) {} | 51 explicit Instruction(OP op) : op_(op), info_(0) {} |
51 Instruction(OP op, unsigned int info) : op_(op), info_(info) {} | 52 Instruction(OP op, unsigned int info) : op_(op), info_(info) {} |
52 | 53 |
53 uint32_t op_ : 4; // A few bits to store the OP code. | 54 uint32_t op_ : 4; // A few bits to store the OP code. |
54 uint32_t info_ : 28; // Remaining bits in first word available to subclass. | 55 uint32_t info_ : 28; // Remaining bits in first word available to subclass. |
55 | 56 |
56 private: | 57 private: |
57 DISALLOW_COPY_AND_ASSIGN(Instruction); | 58 DISALLOW_COPY_AND_ASSIGN(Instruction); |
58 }; | 59 }; |
59 | 60 |
60 typedef NoThrowBuffer<Instruction*> InstructionVector; | 61 class InstructionReceptor { |
62 public: | |
63 InstructionReceptor() = default; | |
64 virtual ~InstructionReceptor() = default; | |
65 | |
66 // Generates an entire base relocation table. | |
67 virtual CheckBool EmitPeRelocs() WARN_UNUSED_RESULT = 0; | |
68 | |
69 // Generates an ELF style relocation table for X86. | |
70 virtual CheckBool EmitElfRelocation() WARN_UNUSED_RESULT = 0; | |
71 | |
72 // Generates an ELF style relocation table for ARM. | |
73 virtual CheckBool EmitElfARMRelocation() WARN_UNUSED_RESULT = 0; | |
74 | |
75 // Following instruction will be assembled at address 'rva'. | |
76 virtual CheckBool EmitOrigin(RVA rva) WARN_UNUSED_RESULT = 0; | |
77 | |
78 // Generates a single byte of data or machine instruction. | |
79 virtual CheckBool EmitSingleByte(uint8_t byte) WARN_UNUSED_RESULT = 0; | |
80 | |
81 // Generates multiple bytes of data or machine instructions. | |
82 virtual CheckBool EmitMultipleBytes(const uint8_t* bytes, | |
83 size_t len) WARN_UNUSED_RESULT = 0; | |
84 | |
85 // Generates a 4-byte relative reference to address of 'label'. | |
86 virtual CheckBool EmitRel32(Label* label) WARN_UNUSED_RESULT = 0; | |
87 | |
88 // Generates a 4-byte relative reference to address of 'label' for ARM. | |
89 virtual CheckBool EmitRel32ARM(uint16_t op, | |
90 Label* label, | |
91 const uint8_t* arm_op, | |
92 uint16_t op_size) WARN_UNUSED_RESULT = 0; | |
93 | |
94 // Generates a 4-byte absolute reference to address of 'label'. | |
95 virtual CheckBool EmitAbs32(Label* label) WARN_UNUSED_RESULT = 0; | |
96 | |
97 // Generates an 8-byte absolute reference to address of 'label'. | |
98 virtual CheckBool EmitAbs64(Label* label) WARN_UNUSED_RESULT = 0; | |
99 | |
100 private: | |
101 DISALLOW_COPY_AND_ASSIGN(InstructionReceptor); | |
102 }; | |
103 | |
104 class InstructionCountReceptor : public InstructionReceptor { | |
105 public: | |
106 InstructionCountReceptor() = default; | |
107 | |
108 size_t size() const { return size_; } | |
109 | |
110 // InstructionReceptor: | |
111 CheckBool EmitPeRelocs() override WARN_UNUSED_RESULT; | |
112 CheckBool EmitElfRelocation() override WARN_UNUSED_RESULT; | |
113 CheckBool EmitElfARMRelocation() override WARN_UNUSED_RESULT; | |
114 CheckBool EmitOrigin(RVA rva) override WARN_UNUSED_RESULT; | |
115 CheckBool EmitSingleByte(uint8_t byte) override WARN_UNUSED_RESULT; | |
116 CheckBool EmitMultipleBytes(const uint8_t* bytes, | |
117 size_t len) override WARN_UNUSED_RESULT; | |
118 CheckBool EmitRel32(Label* label) override WARN_UNUSED_RESULT; | |
119 CheckBool EmitRel32ARM(uint16_t op, | |
120 Label* label, | |
121 const uint8_t* arm_op, | |
122 uint16_t op_size) override WARN_UNUSED_RESULT; | |
123 CheckBool EmitAbs32(Label* label) override WARN_UNUSED_RESULT; | |
124 CheckBool EmitAbs64(Label* label) override WARN_UNUSED_RESULT; | |
125 | |
126 private: | |
127 size_t size_ = 0; | |
128 | |
129 DISALLOW_COPY_AND_ASSIGN(InstructionCountReceptor); | |
130 }; | |
131 | |
132 class InstructionStoreReceptor : public InstructionReceptor { | |
133 public: | |
134 explicit InstructionStoreReceptor(AssemblyProgram* program); | |
135 | |
136 // TODO(huangs): 2016/11: Add Reserve(). | |
137 | |
138 // InstructionReceptor: | |
139 CheckBool EmitPeRelocs() override WARN_UNUSED_RESULT; | |
140 CheckBool EmitElfRelocation() override WARN_UNUSED_RESULT; | |
141 CheckBool EmitElfARMRelocation() override WARN_UNUSED_RESULT; | |
142 CheckBool EmitOrigin(RVA rva) override WARN_UNUSED_RESULT; | |
143 CheckBool EmitSingleByte(uint8_t byte) override WARN_UNUSED_RESULT; | |
144 CheckBool EmitMultipleBytes(const uint8_t* bytes, | |
145 size_t len) override WARN_UNUSED_RESULT; | |
146 CheckBool EmitRel32(Label* label) override WARN_UNUSED_RESULT; | |
147 CheckBool EmitRel32ARM(uint16_t op, | |
148 Label* label, | |
149 const uint8_t* arm_op, | |
150 uint16_t op_size) override WARN_UNUSED_RESULT; | |
151 CheckBool EmitAbs32(Label* label) override WARN_UNUSED_RESULT; | |
152 CheckBool EmitAbs64(Label* label) override WARN_UNUSED_RESULT; | |
153 | |
154 private: | |
155 AssemblyProgram* program_; | |
156 | |
157 DISALLOW_COPY_AND_ASSIGN(InstructionStoreReceptor); | |
158 }; | |
61 | 159 |
62 // An AssemblyProgram is the result of disassembling an executable file. | 160 // An AssemblyProgram is the result of disassembling an executable file. |
63 // | 161 // |
64 // * The disassembler creates labels in the AssemblyProgram and emits | 162 // * The disassembler creates labels in the AssemblyProgram and emits |
65 // 'Instructions'. | 163 // 'Instructions'. |
66 // * The disassembler then calls DefaultAssignIndexes to assign | 164 // * The disassembler then calls DefaultAssignIndexes to assign |
67 // addresses to positions in the address tables. | 165 // addresses to positions in the address tables. |
68 // * [Optional step] | 166 // * [Optional step] |
69 // * At this point the AssemblyProgram can be converted into an | 167 // * At this point the AssemblyProgram can be converted into an |
70 // EncodedProgram and serialized to an output stream. | 168 // EncodedProgram and serialized to an output stream. |
71 // * Later, the EncodedProgram can be deserialized and assembled into | 169 // * Later, the EncodedProgram can be deserialized and assembled into |
72 // the original file. | 170 // the original file. |
73 // | 171 // |
74 // The optional step is to modify the AssemblyProgram. One form of modification | 172 // The optional step is to modify the AssemblyProgram. One form of modification |
75 // is to assign indexes in such a way as to make the EncodedProgram for this | 173 // is to assign indexes in such a way as to make the EncodedProgram for this |
76 // AssemblyProgram look more like the EncodedProgram for some other | 174 // AssemblyProgram look more like the EncodedProgram for some other |
77 // AssemblyProgram. The modification process should call UnassignIndexes, do | 175 // AssemblyProgram. The modification process should call UnassignIndexes, do |
78 // its own assignment, and then call AssignRemainingIndexes to ensure all | 176 // its own assignment, and then call AssignRemainingIndexes to ensure all |
79 // indexes are assigned. | 177 // indexes are assigned. |
80 // | 178 |
81 class AssemblyProgram { | 179 class AssemblyProgram { |
82 public: | 180 public: |
83 using LabelHandler = base::Callback<void(Label*)>; | 181 using LabelHandler = base::Callback<void(Label*)>; |
84 using LabelHandlerMap = std::map<OP, LabelHandler>; | 182 using LabelHandlerMap = std::map<OP, LabelHandler>; |
85 | 183 |
86 explicit AssemblyProgram(ExecutableType kind); | 184 explicit AssemblyProgram(ExecutableType kind); |
87 ~AssemblyProgram(); | 185 ~AssemblyProgram(); |
88 | 186 |
89 ExecutableType kind() const { return kind_; } | 187 ExecutableType kind() const { return kind_; } |
90 | 188 |
91 void set_image_base(uint64_t image_base) { image_base_ = image_base; } | 189 void set_image_base(uint64_t image_base) { image_base_ = image_base; } |
92 | 190 |
93 // Instructions will be assembled in the order they are emitted. | |
94 | |
95 // Generates an entire base relocation table. | |
96 CheckBool EmitPeRelocsInstruction() WARN_UNUSED_RESULT; | |
97 | |
98 // Generates an ELF style relocation table for X86. | |
99 CheckBool EmitElfRelocationInstruction() WARN_UNUSED_RESULT; | |
100 | |
101 // Generates an ELF style relocation table for ARM. | |
102 CheckBool EmitElfARMRelocationInstruction() WARN_UNUSED_RESULT; | |
103 | |
104 // Following instruction will be assembled at address 'rva'. | |
105 CheckBool EmitOriginInstruction(RVA rva) WARN_UNUSED_RESULT; | |
106 | |
107 // Generates a single byte of data or machine instruction. | |
108 CheckBool EmitByteInstruction(uint8_t byte) WARN_UNUSED_RESULT; | |
109 | |
110 // Generates multiple bytes of data or machine instructions. | |
111 CheckBool EmitBytesInstruction(const uint8_t* value, | |
112 size_t len) WARN_UNUSED_RESULT; | |
113 | |
114 // Generates 4-byte relative reference to address of 'label'. | |
115 CheckBool EmitRel32(Label* label) WARN_UNUSED_RESULT; | |
116 | |
117 // Generates 4-byte relative reference to address of 'label' for | |
118 // ARM. | |
119 CheckBool EmitRel32ARM(uint16_t op, | |
120 Label* label, | |
121 const uint8_t* arm_op, | |
122 uint16_t op_size) WARN_UNUSED_RESULT; | |
123 | |
124 // Generates 4-byte absolute reference to address of 'label'. | |
125 CheckBool EmitAbs32(Label* label) WARN_UNUSED_RESULT; | |
126 | |
127 // Generates 8-byte absolute reference to address of 'label'. | |
128 CheckBool EmitAbs64(Label* label) WARN_UNUSED_RESULT; | |
129 | |
130 // Traverses RVAs in |abs32_visitor| and |rel32_visitor| to precompute Labels. | 191 // Traverses RVAs in |abs32_visitor| and |rel32_visitor| to precompute Labels. |
131 void PrecomputeLabels(RvaVisitor* abs32_visitor, RvaVisitor* rel32_visitor); | 192 void PrecomputeLabels(RvaVisitor* abs32_visitor, RvaVisitor* rel32_visitor); |
132 | 193 |
133 // Removes underused Labels. Thresholds used (0 = no trimming) is | 194 // Removes underused Labels. Thresholds used (0 = no trimming) is |
134 // architecture-dependent. | 195 // architecture-dependent. |
135 void TrimLabels(); | 196 void TrimLabels(); |
136 | 197 |
137 void UnassignIndexes(); | 198 void UnassignIndexes(); |
138 void DefaultAssignIndexes(); | 199 void DefaultAssignIndexes(); |
139 void AssignRemainingIndexes(); | 200 void AssignRemainingIndexes(); |
140 | 201 |
141 // Looks up abs32 label. Returns null if none found. | 202 // Looks up abs32 label. Returns null if none found. |
142 Label* FindAbs32Label(RVA rva); | 203 Label* FindAbs32Label(RVA rva); |
143 | 204 |
144 // Looks up rel32 label. Returns null if none found. | 205 // Looks up rel32 label. Returns null if none found. |
145 Label* FindRel32Label(RVA rva); | 206 Label* FindRel32Label(RVA rva); |
146 | 207 |
147 std::unique_ptr<EncodedProgram> Encode() const; | 208 std::unique_ptr<EncodedProgram> Encode() const; |
148 | 209 |
149 // For each |instruction| in |instructions_|, looks up its opcode from | 210 // For each |instruction| in |instructions_|, looks up its opcode from |
150 // |handler_map| for a handler. If a handler exists, invoke it by passing the | 211 // |handler_map| for a handler. If a handler exists, invoke it by passing the |
151 // |instruction|'s label. We assume that |handler_map| has correct keys, i.e., | 212 // |instruction|'s label. We assume that |handler_map| has correct keys, i.e., |
152 // opcodes for an instruction that have label. | 213 // opcodes for an instruction that have label. |
153 void HandleInstructionLabels(const LabelHandlerMap& handler_map) const; | 214 void HandleInstructionLabels(const LabelHandlerMap& handler_map) const; |
154 | 215 |
216 // Returns via |ret| the InstructionCountReceptor instance owned by |this|. | |
217 // Must only be called once. Returns true on success. | |
218 CheckBool CreateInstructionCountReceptor(InstructionCountReceptor** ret); | |
grt (UTC plus 2)
2016/11/07 09:36:17
consumers of these methods don't need to know the
huangs
2016/11/07 19:46:52
Move implementations entirely into .cc file.
| |
219 | |
220 // Returns via |ret| the InstructionStoreReceptor instance owned by |this|. | |
221 // Must only be called once, and only after CreateInstructionCountReceptor() | |
222 // has been called. Returns true on success. | |
223 CheckBool CreateInstructionStoreReceptor(InstructionStoreReceptor** ret); | |
224 | |
155 private: | 225 private: |
226 using InstructionVector = NoThrowBuffer<Instruction*>; | |
227 | |
156 using ScopedInstruction = | 228 using ScopedInstruction = |
157 std::unique_ptr<Instruction, UncheckedDeleter<Instruction>>; | 229 std::unique_ptr<Instruction, UncheckedDeleter<Instruction>>; |
158 | 230 |
231 friend InstructionStoreReceptor; | |
232 | |
233 // Instructions will be assembled in the order they are emitted. | |
234 | |
235 // TODO(huangs): Implement these in InstructionStoreReceptor. | |
236 | |
237 // Generates an entire base relocation table. | |
238 CheckBool EmitPeRelocs() WARN_UNUSED_RESULT; | |
239 | |
240 // Generates an ELF style relocation table for X86. | |
241 CheckBool EmitElfRelocation() WARN_UNUSED_RESULT; | |
242 | |
243 // Generates an ELF style relocation table for ARM. | |
244 CheckBool EmitElfARMRelocation() WARN_UNUSED_RESULT; | |
245 | |
246 // Following instruction will be assembled at address 'rva'. | |
247 CheckBool EmitOrigin(RVA rva) WARN_UNUSED_RESULT; | |
248 | |
249 // Generates a single byte of data or machine instruction. | |
250 CheckBool EmitSingleByte(uint8_t byte) WARN_UNUSED_RESULT; | |
251 | |
252 // Generates multiple bytes of data or machine instructions. | |
253 CheckBool EmitMultipleBytes(const uint8_t* bytes, | |
254 size_t len) WARN_UNUSED_RESULT; | |
255 | |
256 // Generates a 4-byte relative reference to address of 'label'. | |
257 CheckBool EmitRel32(Label* label) WARN_UNUSED_RESULT; | |
258 | |
259 // Generates a 4-byte relative reference to address of 'label' for ARM. | |
260 CheckBool EmitRel32ARM(uint16_t op, | |
261 Label* label, | |
262 const uint8_t* arm_op, | |
263 uint16_t op_size) WARN_UNUSED_RESULT; | |
264 | |
265 // Generates a 4-byte absolute reference to address of 'label'. | |
266 CheckBool EmitAbs32(Label* label) WARN_UNUSED_RESULT; | |
267 | |
268 // Generates an 8-byte absolute reference to address of 'label'. | |
269 CheckBool EmitAbs64(Label* label) WARN_UNUSED_RESULT; | |
270 | |
159 ExecutableType kind_; | 271 ExecutableType kind_; |
160 | 272 |
161 CheckBool Emit(ScopedInstruction instruction) WARN_UNUSED_RESULT; | 273 CheckBool Emit(ScopedInstruction instruction) WARN_UNUSED_RESULT; |
162 CheckBool EmitShared(Instruction* instruction) WARN_UNUSED_RESULT; | 274 CheckBool EmitShared(Instruction* instruction) WARN_UNUSED_RESULT; |
163 | 275 |
164 static const int kLabelLowerLimit; | 276 static const int kLabelLowerLimit; |
165 | 277 |
166 // Looks up a label or creates a new one. Might return NULL. | 278 // Looks up a label or creates a new one. Might return NULL. |
167 Label* FindLabel(RVA rva, RVAToLabel* labels); | 279 Label* FindLabel(RVA rva, RVAToLabel* labels); |
168 | 280 |
169 // Helper methods for the public versions. | 281 // Helper methods for the public versions. |
170 static void UnassignIndexes(RVAToLabel* labels); | 282 static void UnassignIndexes(RVAToLabel* labels); |
171 static void DefaultAssignIndexes(RVAToLabel* labels); | 283 static void DefaultAssignIndexes(RVAToLabel* labels); |
172 static void AssignRemainingIndexes(RVAToLabel* labels); | 284 static void AssignRemainingIndexes(RVAToLabel* labels); |
173 | 285 |
174 // Sharing instructions that emit a single byte saves a lot of space. | 286 // Sharing instructions that emit a single byte saves a lot of space. |
175 Instruction* GetByteInstruction(uint8_t byte); | 287 Instruction* GetByteInstruction(uint8_t byte); |
176 std::unique_ptr<Instruction* [], base::FreeDeleter> byte_instruction_cache_; | 288 std::unique_ptr<Instruction* [], base::FreeDeleter> byte_instruction_cache_; |
177 | 289 |
178 uint64_t image_base_; // Desired or mandated base address of image. | 290 uint64_t image_base_; // Desired or mandated base address of image. |
179 | 291 |
180 InstructionVector instructions_; // All the instructions in program. | 292 InstructionVector instructions_; // All the instructions in program. |
181 | 293 |
182 // Storage and lookup of Labels associated with target addresses. We use | 294 // Storage and lookup of Labels associated with target addresses. We use |
183 // separate abs32 and rel32 labels. | 295 // separate abs32 and rel32 labels. |
184 LabelManager abs32_label_manager_; | 296 LabelManager abs32_label_manager_; |
185 LabelManager rel32_label_manager_; | 297 LabelManager rel32_label_manager_; |
186 | 298 |
299 std::unique_ptr<InstructionCountReceptor> count_receptor_; | |
300 std::unique_ptr<InstructionStoreReceptor> store_receptor_; | |
301 | |
187 DISALLOW_COPY_AND_ASSIGN(AssemblyProgram); | 302 DISALLOW_COPY_AND_ASSIGN(AssemblyProgram); |
188 }; | 303 }; |
189 | 304 |
190 // Converts |program| into encoded form, returning it as |*output|. | 305 // Converts |program| into encoded form, returning it as |*output|. |
191 // Returns C_OK if succeeded, otherwise returns an error status and sets | 306 // Returns C_OK if succeeded, otherwise returns an error status and sets |
192 // |*output| to null. | 307 // |*output| to null. |
193 Status Encode(const AssemblyProgram& program, | 308 Status Encode(const AssemblyProgram& program, |
194 std::unique_ptr<EncodedProgram>* output); | 309 std::unique_ptr<EncodedProgram>* output); |
195 | 310 |
196 } // namespace courgette | 311 } // namespace courgette |
197 | 312 |
198 #endif // COURGETTE_ASSEMBLY_PROGRAM_H_ | 313 #endif // COURGETTE_ASSEMBLY_PROGRAM_H_ |
OLD | NEW |