1/*
2 * Copyright (c) 2012 Google
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are
7 * met: redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer;
9 * redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution;
12 * neither the name of the copyright holders nor the names of its
13 * contributors may be used to endorse or promote products derived from
14 * this software without specific prior written permission.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
17 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
18 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
19 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
20 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 *
28 * Authors: Gabe Black
29 */
30
31#ifndef __ARCH_X86_DECODER_HH__
32#define __ARCH_X86_DECODER_HH__
33
34#include <cassert>
35#include <vector>
36
37#include "arch/x86/regs/misc.hh"
38#include "arch/x86/types.hh"
39#include "base/bitfield.hh"
40#include "base/misc.hh"
41#include "base/trace.hh"
42#include "base/types.hh"
43#include "cpu/decode_cache.hh"
44#include "cpu/static_inst.hh"
45#include "debug/Decoder.hh"
46
47namespace X86ISA
48{
49
50class ISA;
51class Decoder
52{
53 private:
54 //These are defined and documented in decoder_tables.cc
55 static const uint8_t SizeTypeToSize[3][10];
56 typedef const uint8_t ByteTable[256];
57 static ByteTable Prefixes;
58
59 static ByteTable UsesModRMOneByte;
60 static ByteTable UsesModRMTwoByte;
61 static ByteTable UsesModRMThreeByte0F38;
62 static ByteTable UsesModRMThreeByte0F3A;
63
64 static ByteTable ImmediateTypeOneByte;
65 static ByteTable ImmediateTypeTwoByte;
66 static ByteTable ImmediateTypeThreeByte0F38;
67 static ByteTable ImmediateTypeThreeByte0F3A;
68 static ByteTable ImmediateTypeVex[10];
69
70 protected:
71 struct InstBytes
72 {
73 StaticInstPtr si;
74 std::vector<MachInst> chunks;
75 std::vector<MachInst> masks;
76 int lastOffset;
77
78 InstBytes() : lastOffset(0)
79 {}
80 };
81
82 static InstBytes dummy;
83
84 //The bytes to be predecoded
85 MachInst fetchChunk;
86 InstBytes *instBytes;
87 int chunkIdx;
88 //The pc of the start of fetchChunk
89 Addr basePC;
90 //The pc the current instruction started at
91 Addr origPC;
92 //The offset into fetchChunk of current processing
93 int offset;
94 //The extended machine instruction being generated
95 ExtMachInst emi;
96 //Predecoding state
97 X86Mode mode;
98 X86SubMode submode;
99 uint8_t altOp;
100 uint8_t defOp;
101 uint8_t altAddr;
102 uint8_t defAddr;
103 uint8_t stack;
104
105 uint8_t getNextByte()
106 {
107 return ((uint8_t *)&fetchChunk)[offset];
108 }
109
110 void getImmediate(int &collected, uint64_t &current, int size)
111 {
112 //Figure out how many bytes we still need to get for the
113 //immediate.
114 int toGet = size - collected;
115 //Figure out how many bytes are left in our "buffer"
116 int remaining = sizeof(MachInst) - offset;
117 //Get as much as we need, up to the amount available.
118 toGet = toGet > remaining ? remaining : toGet;
119
120 //Shift the bytes we want to be all the way to the right
121 uint64_t partialImm = fetchChunk >> (offset * 8);
122 //Mask off what we don't want
123 partialImm &= mask(toGet * 8);
124 //Shift it over to overlay with our displacement.
125 partialImm <<= (immediateCollected * 8);
126 //Put it into our displacement
127 current |= partialImm;
128 //Update how many bytes we've collected.
129 collected += toGet;
130 consumeBytes(toGet);
131 }
132
133 void updateOffsetState()
134 {
135 assert(offset <= sizeof(MachInst));
136 if (offset == sizeof(MachInst)) {
137 DPRINTF(Decoder, "At the end of a chunk, idx = %d, chunks = %d.\n",
138 chunkIdx, instBytes->chunks.size());
139 chunkIdx++;
140 if (chunkIdx == instBytes->chunks.size()) {
141 outOfBytes = true;
142 } else {
143 offset = 0;
144 fetchChunk = instBytes->chunks[chunkIdx];
145 basePC += sizeof(MachInst);
146 }
147 }
148 }
149
150 void consumeByte()
151 {
152 offset++;
153 updateOffsetState();
154 }
155
156 void consumeBytes(int numBytes)
157 {
158 offset += numBytes;
159 updateOffsetState();
160 }
161
162 //State machine state
163 protected:
164 //Whether or not we're out of bytes
165 bool outOfBytes;
166 //Whether we've completed generating an ExtMachInst
167 bool instDone;
168 //The size of the displacement value
169 int displacementSize;
170 //The size of the immediate value
171 int immediateSize;
172 //This is how much of any immediate value we've gotten. This is used
173 //for both the actual immediate and the displacement.
174 int immediateCollected;
175
176 enum State {
177 ResetState,
178 FromCacheState,
179 PrefixState,
180 TwoByteVexState,
181 ThreeByteVexFirstState,
182 ThreeByteVexSecondState,
183 OneByteOpcodeState,
184 TwoByteOpcodeState,
185 ThreeByte0F38OpcodeState,
186 ThreeByte0F3AOpcodeState,
187 ModRMState,
188 SIBState,
189 DisplacementState,
190 ImmediateState,
191 //We should never get to this state. Getting here is an error.
192 ErrorState
193 };
194
195 State state;
196
197 //Functions to handle each of the states
198 State doResetState();
199 State doFromCacheState();
200 State doPrefixState(uint8_t);
201 State doTwoByteVexState(uint8_t);
202 State doThreeByteVexFirstState(uint8_t);
203 State doThreeByteVexSecondState(uint8_t);
204 State doOneByteOpcodeState(uint8_t);
205 State doTwoByteOpcodeState(uint8_t);
206 State doThreeByte0F38OpcodeState(uint8_t);
207 State doThreeByte0F3AOpcodeState(uint8_t);
208 State doModRMState(uint8_t);
209 State doSIBState(uint8_t);
210 State doDisplacementState();
211 State doImmediateState();
212
213 //Process the actual opcode found earlier, using the supplied tables.
214 State processOpcode(ByteTable &immTable, ByteTable &modrmTable,
215 bool addrSizedImm = false);
216 // Process the opcode found with VEX / XOP prefix.
217 State processExtendedOpcode(ByteTable &immTable);
218
219 protected:
220 /// Caching for decoded instruction objects.
221
222 typedef MiscReg CacheKey;
223
224 typedef DecodeCache::AddrMap<Decoder::InstBytes> DecodePages;
225 DecodePages *decodePages;
226 typedef m5::hash_map<CacheKey, DecodePages *> AddrCacheMap;
227 AddrCacheMap addrCacheMap;
228
229 DecodeCache::InstMap *instMap;
230 typedef m5::hash_map<CacheKey, DecodeCache::InstMap *> InstCacheMap;
231 static InstCacheMap instCacheMap;
232
233 public:
234 Decoder(ISA* isa = nullptr) : basePC(0), origPC(0), offset(0),
235 outOfBytes(true), instDone(false),
236 state(ResetState)
237 {
238 memset(&emi, 0, sizeof(emi));
239 mode = LongMode;
240 submode = SixtyFourBitMode;
241 emi.mode.mode = mode;
242 emi.mode.submode = submode;
243 altOp = 0;
244 defOp = 0;
245 altAddr = 0;
246 defAddr = 0;
247 stack = 0;
248 instBytes = &dummy;
249 decodePages = NULL;
250 instMap = NULL;
251 }
252
253 void setM5Reg(HandyM5Reg m5Reg)
254 {
255 mode = (X86Mode)(uint64_t)m5Reg.mode;
256 submode = (X86SubMode)(uint64_t)m5Reg.submode;
257 emi.mode.mode = mode;
258 emi.mode.submode = submode;
259 altOp = m5Reg.altOp;
260 defOp = m5Reg.defOp;
261 altAddr = m5Reg.altAddr;
262 defAddr = m5Reg.defAddr;
263 stack = m5Reg.stack;
264
265 AddrCacheMap::iterator amIter = addrCacheMap.find(m5Reg);
266 if (amIter != addrCacheMap.end()) {
267 decodePages = amIter->second;
268 } else {
269 decodePages = new DecodePages;
270 addrCacheMap[m5Reg] = decodePages;
271 }
272
273 InstCacheMap::iterator imIter = instCacheMap.find(m5Reg);
274 if (imIter != instCacheMap.end()) {
275 instMap = imIter->second;
276 } else {
277 instMap = new DecodeCache::InstMap;
278 instCacheMap[m5Reg] = instMap;
279 }
280 }
281
282 void takeOverFrom(Decoder *old)
283 {
284 mode = old->mode;
285 submode = old->submode;
286 emi.mode.mode = mode;
287 emi.mode.submode = submode;
288 altOp = old->altOp;
289 defOp = old->defOp;
290 altAddr = old->altAddr;
291 defAddr = old->defAddr;
292 stack = old->stack;
293 }
294
295 void reset()
296 {
297 state = ResetState;
298 }
299
300 void process();
301
302 //Use this to give data to the decoder. This should be used
303 //when there is control flow.
304 void moreBytes(const PCState &pc, Addr fetchPC, MachInst data)
305 {
306 DPRINTF(Decoder, "Getting more bytes.\n");
307 basePC = fetchPC;
308 offset = (fetchPC >= pc.instAddr()) ? 0 : pc.instAddr() - fetchPC;
309 fetchChunk = data;
310 outOfBytes = false;
311 process();
312 }
313
314 bool needMoreBytes()
315 {
316 return outOfBytes;
317 }
318
319 bool instReady()
320 {
321 return instDone;
322 }
323
324 void
325 updateNPC(X86ISA::PCState &nextPC)
326 {
327 if (!nextPC.size()) {
328 int size = basePC + offset - origPC;
329 DPRINTF(Decoder,
330 "Calculating the instruction size: "
331 "basePC: %#x offset: %#x origPC: %#x size: %d\n",
332 basePC, offset, origPC, size);
333 nextPC.size(size);
334 nextPC.npc(nextPC.pc() + size);
335 }
336 }
337
338 public:
339 StaticInstPtr decodeInst(ExtMachInst mach_inst);
340
341 /// Decode a machine instruction.
342 /// @param mach_inst The binary instruction to decode.
343 /// @retval A pointer to the corresponding StaticInst object.
344 StaticInstPtr decode(ExtMachInst mach_inst, Addr addr);
345 StaticInstPtr decode(X86ISA::PCState &nextPC);
346};
347
348} // namespace X86ISA
349
350#endif // __ARCH_X86_DECODER_HH__