1/* 2 * New-style decoder for i386 instructions 3 * 4 * Copyright (c) 2022 Red Hat, Inc. 5 * 6 * Author: Paolo Bonzini <pbonzini@redhat.com> 7 * 8 * This library is free software; you can redistribute it and/or 9 * modify it under the terms of the GNU Lesser General Public 10 * License as published by the Free Software Foundation; either 11 * version 2.1 of the License, or (at your option) any later version. 12 * 13 * This library is distributed in the hope that it will be useful, 14 * but WITHOUT ANY WARRANTY; without even the implied warranty of 15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 16 * Lesser General Public License for more details. 17 * 18 * You should have received a copy of the GNU Lesser General Public 19 * License along with this library; if not, see <http://www.gnu.org/licenses/>. 20 */ 21 22/* 23 * The decoder is mostly based on tables copied from the Intel SDM. As 24 * a result, most operand load and writeback is done entirely in common 25 * table-driven code using the same operand type (X86_TYPE_*) and 26 * size (X86_SIZE_*) codes used in the manual. There are a few differences 27 * though. 28 * 29 * Operand sizes 30 * ------------- 31 * 32 * The manual lists d64 ("cannot encode 32-bit size in 64-bit mode") and f64 33 * ("cannot encode 16-bit or 32-bit size in 64-bit mode") as modifiers of the 34 * "v" or "z" sizes. The decoder simply makes them separate operand sizes. 35 * 36 * The manual lists immediate far destinations as Ap (technically an implicit 37 * argument). The decoder splits them into two immediates, using "Ip" for 38 * the offset part (that comes first in the instruction stream) and "Iw" for 39 * the segment/selector part. The size of the offset is given by s->dflag 40 * and the instructions are illegal in 64-bit mode, so the choice of "Ip" 41 * is somewhat arbitrary; "Iv" or "Iz" would work just as well. 42 * 43 * Operand types 44 * ------------- 45 * 46 * For memory-only operands, if the emitter functions wants to rely on 47 * generic load and writeback, the decoder needs to know the type of the 48 * operand. Therefore, M is often replaced by the more specific EM and WM 49 * (respectively selecting an ALU operand, like the operand type E, or a 50 * vector operand like the operand type W). 51 * 52 * Immediates are almost always signed or masked away in helpers. Two 53 * common exceptions are IN/OUT and absolute jumps. For these, there is 54 * an additional custom operand type "I_unsigned". Alternatively, the 55 * mask could be applied (and the original sign-extended value would be 56 * optimized away by TCG) in the emitter function. 57 * 58 * Finally, a "nop" operand type is used for multi-byte NOPs. It accepts 59 * any value of mod including 11b (unlike M) but it does not try to 60 * interpret the operand (like M). 61 * 62 * Vector operands 63 * --------------- 64 * 65 * The main difference is that the V, U and W types are extended to 66 * cover MMX as well; if an instruction is like 67 * 68 * por Pq, Qq 69 * 66 por Vx, Hx, Wx 70 * 71 * only the second row is included and the instruction is marked as a 72 * valid MMX instruction. The MMX flag directs the decoder to rewrite 73 * the V/U/H/W types to P/N/P/Q if there is no prefix, as well as changing 74 * "x" to "q" if there is no prefix. 75 * 76 * In addition, the ss/ps/sd/pd types are sometimes mushed together as "x" 77 * if the difference is expressed via prefixes. Individual instructions 78 * are separated by prefix in the generator functions. 79 * 80 * There is a custom size "xh" used to address half of a SSE/AVX operand. 81 * This points to a 64-bit operand for SSE operations, 128-bit operand 82 * for 256-bit AVX operands, etc. It is used for conversion operations 83 * such as VCVTPH2PS or VCVTSS2SD. 84 * 85 * There are a couple cases in which instructions (e.g. MOVD) write the 86 * whole XMM or MM register but are established incorrectly in the manual 87 * as "d" or "q". These have to be fixed for the decoder to work correctly. 88 * 89 * VEX exception classes 90 * --------------------- 91 * 92 * Speaking about imprecisions in the manual, the decoder treats all 93 * exception-class 4 instructions as having an optional VEX prefix, and 94 * all exception-class 6 instructions as having a mandatory VEX prefix. 95 * This is true except for a dozen instructions; these are in exception 96 * class 4 but do not ignore the VEX.W bit (which does not even exist 97 * without a VEX prefix). These instructions are mostly listed in Intel's 98 * table 2-16, but with a few exceptions. 99 * 100 * The AMD manual has more precise subclasses for exceptions, and unlike Intel 101 * they list the VEX.W requirements in the exception classes as well (except 102 * when they don't). AMD describes class 6 as "AVX Mixed Memory Argument" 103 * without defining what a mixed memory argument is, but still use 4 as the 104 * primary exception class... except when they don't. 105 * 106 * The summary is: 107 * Intel AMD VEX.W note 108 * ------------------------------------------------------------------- 109 * vpblendd 4 4J 0 110 * vpblendvb 4 4E-X 0 (*) 111 * vpbroadcastq 6 6D 0 (+) 112 * vpermd/vpermps 4 4H 0 (§) 113 * vpermq/vpermpd 4 4H-1 1 (§) 114 * vpermilpd/vpermilps 4 6E 0 (^) 115 * vpmaskmovd 6 4K significant (^) 116 * vpsllv 4 4K significant 117 * vpsrav 4 4J 0 118 * vpsrlv 4 4K significant 119 * vtestps/vtestpd 4 4G 0 120 * 121 * (*) AMD lists VPBLENDVB as related to SSE4.1 PBLENDVB, which may 122 * explain why it is considered exception class 4. However, 123 * Intel says that VEX-only instructions should be in class 6... 124 * 125 * (+) Not found in Intel's table 2-16 126 * 127 * (§) 4H and 4H-1 do not mention VEX.W requirements, which are 128 * however present in the description of the instruction 129 * 130 * (^) these are the two cases in which Intel and AMD disagree on the 131 * primary exception class 132 */ 133 134#define X86_OP_NONE { 0 }, 135 136#define X86_OP_GROUP3(op, op0_, s0_, op1_, s1_, op2_, s2_, ...) { \ 137 .decode = glue(decode_, op), \ 138 .op0 = glue(X86_TYPE_, op0_), \ 139 .s0 = glue(X86_SIZE_, s0_), \ 140 .op1 = glue(X86_TYPE_, op1_), \ 141 .s1 = glue(X86_SIZE_, s1_), \ 142 .op2 = glue(X86_TYPE_, op2_), \ 143 .s2 = glue(X86_SIZE_, s2_), \ 144 .is_decode = true, \ 145 ## __VA_ARGS__ \ 146} 147 148#define X86_OP_GROUP1(op, op0, s0, ...) \ 149 X86_OP_GROUP3(op, op0, s0, 2op, s0, None, None, ## __VA_ARGS__) 150#define X86_OP_GROUP2(op, op0, s0, op1, s1, ...) \ 151 X86_OP_GROUP3(op, op0, s0, 2op, s0, op1, s1, ## __VA_ARGS__) 152#define X86_OP_GROUPw(op, op0, s0, ...) \ 153 X86_OP_GROUP3(op, op0, s0, None, None, None, None, ## __VA_ARGS__) 154#define X86_OP_GROUPwr(op, op0, s0, op1, s1, ...) \ 155 X86_OP_GROUP3(op, op0, s0, op1, s1, None, None, ## __VA_ARGS__) 156#define X86_OP_GROUP0(op, ...) \ 157 X86_OP_GROUP3(op, None, None, None, None, None, None, ## __VA_ARGS__) 158 159#define X86_OP_ENTRY3(op, op0_, s0_, op1_, s1_, op2_, s2_, ...) { \ 160 .gen = glue(gen_, op), \ 161 .op0 = glue(X86_TYPE_, op0_), \ 162 .s0 = glue(X86_SIZE_, s0_), \ 163 .op1 = glue(X86_TYPE_, op1_), \ 164 .s1 = glue(X86_SIZE_, s1_), \ 165 .op2 = glue(X86_TYPE_, op2_), \ 166 .s2 = glue(X86_SIZE_, s2_), \ 167 ## __VA_ARGS__ \ 168} 169 170#define X86_OP_ENTRY4(op, op0_, s0_, op1_, s1_, op2_, s2_, ...) \ 171 X86_OP_ENTRY3(op, op0_, s0_, op1_, s1_, op2_, s2_, \ 172 .op3 = X86_TYPE_I, .s3 = X86_SIZE_b, \ 173 ## __VA_ARGS__) 174 175/* 176 * Short forms that are mostly useful for ALU opcodes and other 177 * one-byte opcodes. For vector instructions it is usually 178 * clearer to write all three operands explicitly, because the 179 * corresponding gen_* function will use OP_PTRn rather than s->T0 180 * and s->T1. 181 */ 182#define X86_OP_ENTRYrr(op, op0, s0, op1, s1, ...) \ 183 X86_OP_ENTRY3(op, None, None, op0, s0, op1, s1, ## __VA_ARGS__) 184#define X86_OP_ENTRYwr(op, op0, s0, op1, s1, ...) \ 185 X86_OP_ENTRY3(op, op0, s0, op1, s1, None, None, ## __VA_ARGS__) 186#define X86_OP_ENTRY2(op, op0, s0, op1, s1, ...) \ 187 X86_OP_ENTRY3(op, op0, s0, 2op, s0, op1, s1, ## __VA_ARGS__) 188#define X86_OP_ENTRYw(op, op0, s0, ...) \ 189 X86_OP_ENTRY3(op, op0, s0, None, None, None, None, ## __VA_ARGS__) 190#define X86_OP_ENTRYr(op, op0, s0, ...) \ 191 X86_OP_ENTRY3(op, None, None, op0, s0, None, None, ## __VA_ARGS__) 192#define X86_OP_ENTRY1(op, op0, s0, ...) \ 193 X86_OP_ENTRY3(op, op0, s0, 2op, s0, None, None, ## __VA_ARGS__) 194#define X86_OP_ENTRY0(op, ...) \ 195 X86_OP_ENTRY3(op, None, None, None, None, None, None, ## __VA_ARGS__) 196 197#define cpuid(feat) .cpuid = X86_FEAT_##feat, 198#define nolea .special = X86_SPECIAL_NoLoadEA, 199#define xchg .special = X86_SPECIAL_Locked, 200#define lock .special = X86_SPECIAL_HasLock, 201#define mmx .special = X86_SPECIAL_MMX, 202#define op0_Rd .special = X86_SPECIAL_Op0_Rd, 203#define op2_Ry .special = X86_SPECIAL_Op2_Ry, 204#define avx_movx .special = X86_SPECIAL_AVXExtMov, 205#define sextT0 .special = X86_SPECIAL_SExtT0, 206#define zextT0 .special = X86_SPECIAL_ZExtT0, 207#define op0_Mw .special = X86_SPECIAL_Op0_Mw, 208 209#define vex1 .vex_class = 1, 210#define vex1_rep3 .vex_class = 1, .vex_special = X86_VEX_REPScalar, 211#define vex2 .vex_class = 2, 212#define vex2_rep3 .vex_class = 2, .vex_special = X86_VEX_REPScalar, 213#define vex3 .vex_class = 3, 214#define vex4 .vex_class = 4, 215#define vex4_unal .vex_class = 4, .vex_special = X86_VEX_SSEUnaligned, 216#define vex4_rep5 .vex_class = 4, .vex_special = X86_VEX_REPScalar, 217#define vex5 .vex_class = 5, 218#define vex6 .vex_class = 6, 219#define vex7 .vex_class = 7, 220#define vex8 .vex_class = 8, 221#define vex11 .vex_class = 11, 222#define vex12 .vex_class = 12, 223#define vex13 .vex_class = 13, 224 225#define chk(a) .check = X86_CHECK_##a, 226#define chk2(a, b) .check = X86_CHECK_##a | X86_CHECK_##b, 227#define chk3(a, b, c) .check = X86_CHECK_##a | X86_CHECK_##b | X86_CHECK_##c, 228#define svm(a) .intercept = SVM_EXIT_##a, .has_intercept = true, 229 230#define avx2_256 .vex_special = X86_VEX_AVX2_256, 231 232#define P_00 1 233#define P_66 (1 << PREFIX_DATA) 234#define P_F3 (1 << PREFIX_REPZ) 235#define P_F2 (1 << PREFIX_REPNZ) 236 237#define p_00 .valid_prefix = P_00, 238#define p_66 .valid_prefix = P_66, 239#define p_f3 .valid_prefix = P_F3, 240#define p_f2 .valid_prefix = P_F2, 241#define p_00_66 .valid_prefix = P_00 | P_66, 242#define p_00_f3 .valid_prefix = P_00 | P_F3, 243#define p_66_f2 .valid_prefix = P_66 | P_F2, 244#define p_00_66_f3 .valid_prefix = P_00 | P_66 | P_F3, 245#define p_66_f3_f2 .valid_prefix = P_66 | P_F3 | P_F2, 246#define p_00_66_f3_f2 .valid_prefix = P_00 | P_66 | P_F3 | P_F2, 247 248#define UNKNOWN_OPCODE ((X86OpEntry) {}) 249 250static uint8_t get_modrm(DisasContext *s, CPUX86State *env) 251{ 252 if (!s->has_modrm) { 253 s->modrm = x86_ldub_code(env, s); 254 s->has_modrm = true; 255 } 256 return s->modrm; 257} 258 259static inline const X86OpEntry *decode_by_prefix(DisasContext *s, const X86OpEntry entries[4]) 260{ 261 if (s->prefix & PREFIX_REPNZ) { 262 return &entries[3]; 263 } else if (s->prefix & PREFIX_REPZ) { 264 return &entries[2]; 265 } else if (s->prefix & PREFIX_DATA) { 266 return &entries[1]; 267 } else { 268 return &entries[0]; 269 } 270} 271 272static void decode_group15(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 273{ 274 static const X86OpEntry group15_reg[8] = { 275 [0] = X86_OP_ENTRYw(RDxxBASE, R,y, cpuid(FSGSBASE) chk(o64) p_f3), 276 [1] = X86_OP_ENTRYw(RDxxBASE, R,y, cpuid(FSGSBASE) chk(o64) p_f3), 277 [2] = X86_OP_ENTRYr(WRxxBASE, R,y, cpuid(FSGSBASE) chk(o64) p_f3 zextT0), 278 [3] = X86_OP_ENTRYr(WRxxBASE, R,y, cpuid(FSGSBASE) chk(o64) p_f3 zextT0), 279 [5] = X86_OP_ENTRY0(LFENCE, cpuid(SSE2) p_00), 280 [6] = X86_OP_ENTRY0(MFENCE, cpuid(SSE2) p_00), 281 [7] = X86_OP_ENTRY0(SFENCE, cpuid(SSE2) p_00), 282 }; 283 284 static const X86OpEntry group15_mem[8] = { 285 [0] = X86_OP_ENTRYw(FXSAVE, M,y, cpuid(FXSR) p_00), 286 [1] = X86_OP_ENTRYr(FXRSTOR, M,y, cpuid(FXSR) p_00), 287 [2] = X86_OP_ENTRYr(LDMXCSR, E,d, vex5 chk(VEX128) p_00), 288 [3] = X86_OP_ENTRYw(STMXCSR, E,d, vex5 chk(VEX128) p_00), 289 [4] = X86_OP_ENTRYw(XSAVE, M,y, cpuid(XSAVE) p_00), 290 [5] = X86_OP_ENTRYr(XRSTOR, M,y, cpuid(XSAVE) p_00), 291 [6] = X86_OP_ENTRYw(XSAVEOPT, M,b, cpuid(XSAVEOPT) p_00), 292 [7] = X86_OP_ENTRYw(NOP, M,b, cpuid(CLFLUSH) p_00), 293 }; 294 295 static const X86OpEntry group15_mem_66[8] = { 296 [6] = X86_OP_ENTRYw(NOP, M,b, cpuid(CLWB)), 297 [7] = X86_OP_ENTRYw(NOP, M,b, cpuid(CLFLUSHOPT)), 298 }; 299 300 uint8_t modrm = get_modrm(s, env); 301 int op = (modrm >> 3) & 7; 302 303 if ((modrm >> 6) == 3) { 304 *entry = group15_reg[op]; 305 } else if (s->prefix & PREFIX_DATA) { 306 *entry = group15_mem_66[op]; 307 } else { 308 *entry = group15_mem[op]; 309 } 310} 311 312static void decode_group17(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 313{ 314 static const X86GenFunc group17_gen[8] = { 315 NULL, gen_BLSR, gen_BLSMSK, gen_BLSI, 316 }; 317 int op = (get_modrm(s, env) >> 3) & 7; 318 entry->gen = group17_gen[op]; 319} 320 321static void decode_group12(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 322{ 323 static const X86OpEntry opcodes_group12[8] = { 324 {}, 325 {}, 326 X86_OP_ENTRY3(PSRLW_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 327 {}, 328 X86_OP_ENTRY3(PSRAW_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 329 {}, 330 X86_OP_ENTRY3(PSLLW_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 331 {}, 332 }; 333 334 int op = (get_modrm(s, env) >> 3) & 7; 335 *entry = opcodes_group12[op]; 336} 337 338static void decode_group13(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 339{ 340 static const X86OpEntry opcodes_group13[8] = { 341 {}, 342 {}, 343 X86_OP_ENTRY3(PSRLD_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 344 {}, 345 X86_OP_ENTRY3(PSRAD_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 346 {}, 347 X86_OP_ENTRY3(PSLLD_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 348 {}, 349 }; 350 351 int op = (get_modrm(s, env) >> 3) & 7; 352 *entry = opcodes_group13[op]; 353} 354 355static void decode_group14(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 356{ 357 static const X86OpEntry opcodes_group14[8] = { 358 /* grp14 */ 359 {}, 360 {}, 361 X86_OP_ENTRY3(PSRLQ_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 362 X86_OP_ENTRY3(PSRLDQ_i, H,x, U,x, I,b, vex7 avx2_256 p_66), 363 {}, 364 {}, 365 X86_OP_ENTRY3(PSLLQ_i, H,x, U,x, I,b, vex7 mmx avx2_256 p_00_66), 366 X86_OP_ENTRY3(PSLLDQ_i, H,x, U,x, I,b, vex7 avx2_256 p_66), 367 }; 368 369 int op = (get_modrm(s, env) >> 3) & 7; 370 *entry = opcodes_group14[op]; 371} 372 373static void decode_0F6F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 374{ 375 static const X86OpEntry opcodes_0F6F[4] = { 376 X86_OP_ENTRY3(MOVDQ, P,q, None,None, Q,q, vex5 mmx), /* movq */ 377 X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex1), /* movdqa */ 378 X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* movdqu */ 379 {}, 380 }; 381 *entry = *decode_by_prefix(s, opcodes_0F6F); 382} 383 384static void decode_0F70(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 385{ 386 static const X86OpEntry pshufw[4] = { 387 X86_OP_ENTRY3(PSHUFW, P,q, Q,q, I,b, vex4 mmx), 388 X86_OP_ENTRY3(PSHUFD, V,x, W,x, I,b, vex4 avx2_256), 389 X86_OP_ENTRY3(PSHUFHW, V,x, W,x, I,b, vex4 avx2_256), 390 X86_OP_ENTRY3(PSHUFLW, V,x, W,x, I,b, vex4 avx2_256), 391 }; 392 393 *entry = *decode_by_prefix(s, pshufw); 394} 395 396static void decode_0F77(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 397{ 398 if (!(s->prefix & PREFIX_VEX)) { 399 entry->gen = gen_EMMS; 400 } else if (!s->vex_l) { 401 entry->gen = gen_VZEROUPPER; 402 entry->vex_class = 8; 403 } else { 404 entry->gen = gen_VZEROALL; 405 entry->vex_class = 8; 406 } 407} 408 409static void decode_0F78(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 410{ 411 static const X86OpEntry opcodes_0F78[4] = { 412 {}, 413 X86_OP_ENTRY3(EXTRQ_i, V,x, None,None, I,w, cpuid(SSE4A)), /* AMD extension */ 414 {}, 415 X86_OP_ENTRY3(INSERTQ_i, V,x, U,x, I,w, cpuid(SSE4A)), /* AMD extension */ 416 }; 417 *entry = *decode_by_prefix(s, opcodes_0F78); 418} 419 420static void decode_0F79(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 421{ 422 if (s->prefix & PREFIX_REPNZ) { 423 entry->gen = gen_INSERTQ_r; /* AMD extension */ 424 } else if (s->prefix & PREFIX_DATA) { 425 entry->gen = gen_EXTRQ_r; /* AMD extension */ 426 } else { 427 entry->gen = NULL; 428 }; 429} 430 431static void decode_0F7E(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 432{ 433 static const X86OpEntry opcodes_0F7E[4] = { 434 X86_OP_ENTRY3(MOVD_from, E,y, None,None, P,y, vex5 mmx), 435 X86_OP_ENTRY3(MOVD_from, E,y, None,None, V,y, vex5), 436 X86_OP_ENTRY3(MOVQ, V,x, None,None, W,q, vex5), /* wrong dest Vy on SDM! */ 437 {}, 438 }; 439 *entry = *decode_by_prefix(s, opcodes_0F7E); 440} 441 442static void decode_0F7F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 443{ 444 static const X86OpEntry opcodes_0F7F[4] = { 445 X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex5 mmx), /* movq */ 446 X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex1), /* movdqa */ 447 X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4_unal), /* movdqu */ 448 {}, 449 }; 450 *entry = *decode_by_prefix(s, opcodes_0F7F); 451} 452 453static void decode_0FB8(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 454{ 455 static const X86OpEntry popcnt = 456 X86_OP_ENTRYwr(POPCNT, G,v, E,v, cpuid(POPCNT) zextT0); 457 458 if (s->prefix & PREFIX_REPZ) { 459 *entry = popcnt; 460 } else { 461 memset(entry, 0, sizeof(*entry)); 462 } 463} 464 465static void decode_0FBC(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 466{ 467 /* For BSF, pass 2op as the third operand so that we can use zextT0 */ 468 static const X86OpEntry opcodes_0FBC[4] = { 469 X86_OP_ENTRY3(BSF, G,v, E,v, 2op,v, zextT0), 470 X86_OP_ENTRY3(BSF, G,v, E,v, 2op,v, zextT0), /* 0x66 */ 471 X86_OP_ENTRYwr(TZCNT, G,v, E,v, zextT0), /* 0xf3 */ 472 X86_OP_ENTRY3(BSF, G,v, E,v, 2op,v, zextT0), /* 0xf2 */ 473 }; 474 if (!(s->cpuid_ext3_features & CPUID_EXT3_ABM)) { 475 *entry = opcodes_0FBC[0]; 476 } else { 477 *entry = *decode_by_prefix(s, opcodes_0FBC); 478 } 479} 480 481static void decode_0FBD(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 482{ 483 /* For BSR, pass 2op as the third operand so that we can use zextT0 */ 484 static const X86OpEntry opcodes_0FBD[4] = { 485 X86_OP_ENTRY3(BSR, G,v, E,v, 2op,v, zextT0), 486 X86_OP_ENTRY3(BSR, G,v, E,v, 2op,v, zextT0), /* 0x66 */ 487 X86_OP_ENTRYwr(LZCNT, G,v, E,v, zextT0), /* 0xf3 */ 488 X86_OP_ENTRY3(BSR, G,v, E,v, 2op,v, zextT0), /* 0xf2 */ 489 }; 490 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1)) { 491 *entry = opcodes_0FBD[0]; 492 } else { 493 *entry = *decode_by_prefix(s, opcodes_0FBD); 494 } 495} 496 497static void decode_0FD6(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 498{ 499 static const X86OpEntry movq[4] = { 500 {}, 501 X86_OP_ENTRY3(MOVQ, W,x, None, None, V,q, vex5), 502 X86_OP_ENTRY3(MOVq_dq, V,dq, None, None, N,q), 503 X86_OP_ENTRY3(MOVq_dq, P,q, None, None, U,q), 504 }; 505 506 *entry = *decode_by_prefix(s, movq); 507} 508 509static const X86OpEntry opcodes_0F38_00toEF[240] = { 510 [0x00] = X86_OP_ENTRY3(PSHUFB, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 511 [0x01] = X86_OP_ENTRY3(PHADDW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 512 [0x02] = X86_OP_ENTRY3(PHADDD, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 513 [0x03] = X86_OP_ENTRY3(PHADDSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 514 [0x04] = X86_OP_ENTRY3(PMADDUBSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 515 [0x05] = X86_OP_ENTRY3(PHSUBW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 516 [0x06] = X86_OP_ENTRY3(PHSUBD, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 517 [0x07] = X86_OP_ENTRY3(PHSUBSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 518 519 [0x10] = X86_OP_ENTRY2(PBLENDVB, V,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 520 [0x13] = X86_OP_ENTRY2(VCVTPH2PS, V,x, W,xh, vex11 chk(W0) cpuid(F16C) p_66), 521 [0x14] = X86_OP_ENTRY2(BLENDVPS, V,x, W,x, vex4 cpuid(SSE41) p_66), 522 [0x15] = X86_OP_ENTRY2(BLENDVPD, V,x, W,x, vex4 cpuid(SSE41) p_66), 523 /* Listed incorrectly as type 4 */ 524 [0x16] = X86_OP_ENTRY3(VPERMD, V,qq, H,qq, W,qq, vex6 chk(W0) cpuid(AVX2) p_66), /* vpermps */ 525 [0x17] = X86_OP_ENTRY3(VPTEST, None,None, V,x, W,x, vex4 cpuid(SSE41) p_66), 526 527 /* 528 * Source operand listed as Mq/Ux and similar in the manual; incorrectly listed 529 * as 128-bit only in 2-17. 530 */ 531 [0x20] = X86_OP_ENTRY3(VPMOVSXBW, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 532 [0x21] = X86_OP_ENTRY3(VPMOVSXBD, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 533 [0x22] = X86_OP_ENTRY3(VPMOVSXBQ, V,x, None,None, W,w, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 534 [0x23] = X86_OP_ENTRY3(VPMOVSXWD, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 535 [0x24] = X86_OP_ENTRY3(VPMOVSXWQ, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 536 [0x25] = X86_OP_ENTRY3(VPMOVSXDQ, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 537 538 /* Same as PMOVSX. */ 539 [0x30] = X86_OP_ENTRY3(VPMOVZXBW, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 540 [0x31] = X86_OP_ENTRY3(VPMOVZXBD, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 541 [0x32] = X86_OP_ENTRY3(VPMOVZXBQ, V,x, None,None, W,w, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 542 [0x33] = X86_OP_ENTRY3(VPMOVZXWD, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 543 [0x34] = X86_OP_ENTRY3(VPMOVZXWQ, V,x, None,None, W,d, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 544 [0x35] = X86_OP_ENTRY3(VPMOVZXDQ, V,x, None,None, W,q, vex5 cpuid(SSE41) avx_movx avx2_256 p_66), 545 [0x36] = X86_OP_ENTRY3(VPERMD, V,qq, H,qq, W,qq, vex6 chk(W0) cpuid(AVX2) p_66), 546 [0x37] = X86_OP_ENTRY3(PCMPGTQ, V,x, H,x, W,x, vex4 cpuid(SSE42) avx2_256 p_66), 547 548 [0x40] = X86_OP_ENTRY3(PMULLD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 549 [0x41] = X86_OP_ENTRY3(VPHMINPOSUW, V,dq, None,None, W,dq, vex4 cpuid(SSE41) p_66), 550 /* Listed incorrectly as type 4 */ 551 [0x45] = X86_OP_ENTRY3(VPSRLV, V,x, H,x, W,x, vex6 cpuid(AVX2) p_66), 552 [0x46] = X86_OP_ENTRY3(VPSRAV, V,x, H,x, W,x, vex6 chk(W0) cpuid(AVX2) p_66), 553 [0x47] = X86_OP_ENTRY3(VPSLLV, V,x, H,x, W,x, vex6 cpuid(AVX2) p_66), 554 555 [0x90] = X86_OP_ENTRY3(VPGATHERD, V,x, H,x, M,d, vex12 cpuid(AVX2) p_66), /* vpgatherdd/q */ 556 [0x91] = X86_OP_ENTRY3(VPGATHERQ, V,x, H,x, M,q, vex12 cpuid(AVX2) p_66), /* vpgatherqd/q */ 557 [0x92] = X86_OP_ENTRY3(VPGATHERD, V,x, H,x, M,d, vex12 cpuid(AVX2) p_66), /* vgatherdps/d */ 558 [0x93] = X86_OP_ENTRY3(VPGATHERQ, V,x, H,x, M,q, vex12 cpuid(AVX2) p_66), /* vgatherqps/d */ 559 560 /* Should be exception type 2 but they do not have legacy SSE equivalents? */ 561 [0x96] = X86_OP_ENTRY3(VFMADDSUB132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 562 [0x97] = X86_OP_ENTRY3(VFMSUBADD132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 563 564 [0xa6] = X86_OP_ENTRY3(VFMADDSUB213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 565 [0xa7] = X86_OP_ENTRY3(VFMSUBADD213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 566 567 [0xb6] = X86_OP_ENTRY3(VFMADDSUB231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 568 [0xb7] = X86_OP_ENTRY3(VFMSUBADD231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 569 570 [0x08] = X86_OP_ENTRY3(PSIGNB, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 571 [0x09] = X86_OP_ENTRY3(PSIGNW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 572 [0x0a] = X86_OP_ENTRY3(PSIGND, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 573 [0x0b] = X86_OP_ENTRY3(PMULHRSW, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 574 /* Listed incorrectly as type 4 */ 575 [0x0c] = X86_OP_ENTRY3(VPERMILPS, V,x, H,x, W,x, vex6 chk(W0) cpuid(AVX) p_00_66), 576 [0x0d] = X86_OP_ENTRY3(VPERMILPD, V,x, H,x, W,x, vex6 chk(W0) cpuid(AVX) p_66), 577 [0x0e] = X86_OP_ENTRY3(VTESTPS, None,None, V,x, W,x, vex6 chk(W0) cpuid(AVX) p_66), 578 [0x0f] = X86_OP_ENTRY3(VTESTPD, None,None, V,x, W,x, vex6 chk(W0) cpuid(AVX) p_66), 579 580 [0x18] = X86_OP_ENTRY3(VPBROADCASTD, V,x, None,None, W,d, vex6 chk(W0) cpuid(AVX) p_66), /* vbroadcastss */ 581 [0x19] = X86_OP_ENTRY3(VPBROADCASTQ, V,qq, None,None, W,q, vex6 chk(W0) cpuid(AVX) p_66), /* vbroadcastsd */ 582 [0x1a] = X86_OP_ENTRY3(VBROADCASTx128, V,qq, None,None, WM,dq,vex6 chk(W0) cpuid(AVX) p_66), 583 [0x1c] = X86_OP_ENTRY3(PABSB, V,x, None,None, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 584 [0x1d] = X86_OP_ENTRY3(PABSW, V,x, None,None, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 585 [0x1e] = X86_OP_ENTRY3(PABSD, V,x, None,None, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 586 587 [0x28] = X86_OP_ENTRY3(PMULDQ, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 588 [0x29] = X86_OP_ENTRY3(PCMPEQQ, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 589 [0x2a] = X86_OP_ENTRY3(MOVDQ, V,x, None,None, WM,x, vex1 cpuid(SSE41) avx2_256 p_66), /* movntdqa */ 590 [0x2b] = X86_OP_ENTRY3(VPACKUSDW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 591 [0x2c] = X86_OP_ENTRY3(VMASKMOVPS, V,x, H,x, WM,x, vex6 chk(W0) cpuid(AVX) p_66), 592 [0x2d] = X86_OP_ENTRY3(VMASKMOVPD, V,x, H,x, WM,x, vex6 chk(W0) cpuid(AVX) p_66), 593 /* Incorrectly listed as Mx,Hx,Vx in the manual */ 594 [0x2e] = X86_OP_ENTRY3(VMASKMOVPS_st, M,x, V,x, H,x, vex6 chk(W0) cpuid(AVX) p_66), 595 [0x2f] = X86_OP_ENTRY3(VMASKMOVPD_st, M,x, V,x, H,x, vex6 chk(W0) cpuid(AVX) p_66), 596 597 [0x38] = X86_OP_ENTRY3(PMINSB, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 598 [0x39] = X86_OP_ENTRY3(PMINSD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 599 [0x3a] = X86_OP_ENTRY3(PMINUW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 600 [0x3b] = X86_OP_ENTRY3(PMINUD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 601 [0x3c] = X86_OP_ENTRY3(PMAXSB, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 602 [0x3d] = X86_OP_ENTRY3(PMAXSD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 603 [0x3e] = X86_OP_ENTRY3(PMAXUW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 604 [0x3f] = X86_OP_ENTRY3(PMAXUD, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 605 606 /* VPBROADCASTQ not listed as W0 in table 2-16 */ 607 [0x58] = X86_OP_ENTRY3(VPBROADCASTD, V,x, None,None, W,d, vex6 chk(W0) cpuid(AVX2) p_66), 608 [0x59] = X86_OP_ENTRY3(VPBROADCASTQ, V,x, None,None, W,q, vex6 chk(W0) cpuid(AVX2) p_66), 609 [0x5a] = X86_OP_ENTRY3(VBROADCASTx128, V,qq, None,None, WM,dq,vex6 chk(W0) cpuid(AVX2) p_66), 610 611 [0x78] = X86_OP_ENTRY3(VPBROADCASTB, V,x, None,None, W,b, vex6 chk(W0) cpuid(AVX2) p_66), 612 [0x79] = X86_OP_ENTRY3(VPBROADCASTW, V,x, None,None, W,w, vex6 chk(W0) cpuid(AVX2) p_66), 613 614 [0x8c] = X86_OP_ENTRY3(VPMASKMOV, V,x, H,x, WM,x, vex6 cpuid(AVX2) p_66), 615 [0x8e] = X86_OP_ENTRY3(VPMASKMOV_st, M,x, V,x, H,x, vex6 cpuid(AVX2) p_66), 616 617 /* Should be exception type 2 or 3 but they do not have legacy SSE equivalents? */ 618 [0x98] = X86_OP_ENTRY3(VFMADD132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 619 [0x99] = X86_OP_ENTRY3(VFMADD132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 620 [0x9a] = X86_OP_ENTRY3(VFMSUB132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 621 [0x9b] = X86_OP_ENTRY3(VFMSUB132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 622 [0x9c] = X86_OP_ENTRY3(VFNMADD132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 623 [0x9d] = X86_OP_ENTRY3(VFNMADD132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 624 [0x9e] = X86_OP_ENTRY3(VFNMSUB132Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 625 [0x9f] = X86_OP_ENTRY3(VFNMSUB132Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 626 627 [0xa8] = X86_OP_ENTRY3(VFMADD213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 628 [0xa9] = X86_OP_ENTRY3(VFMADD213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 629 [0xaa] = X86_OP_ENTRY3(VFMSUB213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 630 [0xab] = X86_OP_ENTRY3(VFMSUB213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 631 [0xac] = X86_OP_ENTRY3(VFNMADD213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 632 [0xad] = X86_OP_ENTRY3(VFNMADD213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 633 [0xae] = X86_OP_ENTRY3(VFNMSUB213Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 634 [0xaf] = X86_OP_ENTRY3(VFNMSUB213Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 635 636 [0xb8] = X86_OP_ENTRY3(VFMADD231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 637 [0xb9] = X86_OP_ENTRY3(VFMADD231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 638 [0xba] = X86_OP_ENTRY3(VFMSUB231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 639 [0xbb] = X86_OP_ENTRY3(VFMSUB231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 640 [0xbc] = X86_OP_ENTRY3(VFNMADD231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 641 [0xbd] = X86_OP_ENTRY3(VFNMADD231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 642 [0xbe] = X86_OP_ENTRY3(VFNMSUB231Px, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 643 [0xbf] = X86_OP_ENTRY3(VFNMSUB231Sx, V,x, H,x, W,x, vex6 cpuid(FMA) p_66), 644 645 [0xc8] = X86_OP_ENTRY2(SHA1NEXTE, V,dq, W,dq, cpuid(SHA_NI)), 646 [0xc9] = X86_OP_ENTRY2(SHA1MSG1, V,dq, W,dq, cpuid(SHA_NI)), 647 [0xca] = X86_OP_ENTRY2(SHA1MSG2, V,dq, W,dq, cpuid(SHA_NI)), 648 [0xcb] = X86_OP_ENTRY2(SHA256RNDS2, V,dq, W,dq, cpuid(SHA_NI)), 649 [0xcc] = X86_OP_ENTRY2(SHA256MSG1, V,dq, W,dq, cpuid(SHA_NI)), 650 [0xcd] = X86_OP_ENTRY2(SHA256MSG2, V,dq, W,dq, cpuid(SHA_NI)), 651 652 [0xdb] = X86_OP_ENTRY3(VAESIMC, V,dq, None,None, W,dq, vex4 cpuid(AES) p_66), 653 [0xdc] = X86_OP_ENTRY3(VAESENC, V,x, H,x, W,x, vex4 cpuid(AES) p_66), 654 [0xdd] = X86_OP_ENTRY3(VAESENCLAST, V,x, H,x, W,x, vex4 cpuid(AES) p_66), 655 [0xde] = X86_OP_ENTRY3(VAESDEC, V,x, H,x, W,x, vex4 cpuid(AES) p_66), 656 [0xdf] = X86_OP_ENTRY3(VAESDECLAST, V,x, H,x, W,x, vex4 cpuid(AES) p_66), 657 658 /* 659 * REG selects srcdest2 operand, VEX.vvvv selects src3. VEX class not found 660 * in manual, assumed to be 13 from the VEX.L0 constraint. 661 */ 662 [0xe0] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 663 [0xe1] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 664 [0xe2] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 665 [0xe3] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 666 [0xe4] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 667 [0xe5] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 668 [0xe6] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 669 [0xe7] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 670 671 [0xe8] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 672 [0xe9] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 673 [0xea] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 674 [0xeb] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 675 [0xec] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 676 [0xed] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 677 [0xee] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 678 [0xef] = X86_OP_ENTRY3(CMPccXADD, M,y, G,y, B,y, vex13 xchg chk(o64) cpuid(CMPCCXADD) p_66), 679}; 680 681/* five rows for no prefix, 66, F3, F2, 66+F2 */ 682static const X86OpEntry opcodes_0F38_F0toFF[16][5] = { 683 [0] = { 684 X86_OP_ENTRYwr(MOVBE, G,y, M,y, cpuid(MOVBE)), 685 X86_OP_ENTRYwr(MOVBE, G,w, M,w, cpuid(MOVBE)), 686 {}, 687 X86_OP_ENTRY2(CRC32, G,d, E,b, cpuid(SSE42)), 688 X86_OP_ENTRY2(CRC32, G,d, E,b, cpuid(SSE42)), 689 }, 690 [1] = { 691 X86_OP_ENTRYwr(MOVBE, M,y, G,y, cpuid(MOVBE)), 692 X86_OP_ENTRYwr(MOVBE, M,w, G,w, cpuid(MOVBE)), 693 {}, 694 X86_OP_ENTRY2(CRC32, G,d, E,y, cpuid(SSE42)), 695 X86_OP_ENTRY2(CRC32, G,d, E,w, cpuid(SSE42)), 696 }, 697 [2] = { 698 X86_OP_ENTRY3(ANDN, G,y, B,y, E,y, vex13 cpuid(BMI1)), 699 {}, 700 {}, 701 {}, 702 {}, 703 }, 704 [3] = { 705 X86_OP_GROUP3(group17, B,y, None,None, E,y, vex13 cpuid(BMI1)), 706 {}, 707 {}, 708 {}, 709 {}, 710 }, 711 [5] = { 712 X86_OP_ENTRY3(BZHI, G,y, E,y, B,y, vex13 cpuid(BMI1)), 713 {}, 714 X86_OP_ENTRY3(PEXT, G,y, B,y, E,y, vex13 zextT0 cpuid(BMI2)), 715 X86_OP_ENTRY3(PDEP, G,y, B,y, E,y, vex13 zextT0 cpuid(BMI2)), 716 {}, 717 }, 718 [6] = { 719 {}, 720 X86_OP_ENTRY2(ADCX, G,y, E,y, cpuid(ADX)), 721 X86_OP_ENTRY2(ADOX, G,y, E,y, cpuid(ADX)), 722 X86_OP_ENTRY3(MULX, /* B,y, */ G,y, E,y, 2,y, vex13 cpuid(BMI2)), 723 {}, 724 }, 725 [7] = { 726 X86_OP_ENTRY3(BEXTR, G,y, E,y, B,y, vex13 zextT0 cpuid(BMI1)), 727 X86_OP_ENTRY3(SHLX, G,y, E,y, B,y, vex13 cpuid(BMI1)), 728 X86_OP_ENTRY3(SARX, G,y, E,y, B,y, vex13 sextT0 cpuid(BMI1)), 729 X86_OP_ENTRY3(SHRX, G,y, E,y, B,y, vex13 zextT0 cpuid(BMI1)), 730 {}, 731 }, 732}; 733 734static void decode_0F38(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 735{ 736 *b = x86_ldub_code(env, s); 737 if (*b < 0xf0) { 738 *entry = opcodes_0F38_00toEF[*b]; 739 } else { 740 int row = 0; 741 if (s->prefix & PREFIX_REPZ) { 742 /* The REPZ (F3) prefix has priority over 66 */ 743 row = 2; 744 } else { 745 row += s->prefix & PREFIX_REPNZ ? 3 : 0; 746 row += s->prefix & PREFIX_DATA ? 1 : 0; 747 } 748 *entry = opcodes_0F38_F0toFF[*b & 15][row]; 749 } 750} 751 752static void decode_VINSERTPS(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 753{ 754 static const X86OpEntry 755 vinsertps_reg = X86_OP_ENTRY4(VINSERTPS_r, V,dq, H,dq, U,dq, vex5 cpuid(SSE41) p_66), 756 vinsertps_mem = X86_OP_ENTRY4(VINSERTPS_m, V,dq, H,dq, M,d, vex5 cpuid(SSE41) p_66); 757 758 int modrm = get_modrm(s, env); 759 *entry = (modrm >> 6) == 3 ? vinsertps_reg : vinsertps_mem; 760} 761 762static const X86OpEntry opcodes_0F3A[256] = { 763 /* 764 * These are VEX-only, but incorrectly listed in the manual as exception type 4. 765 * Also the "qq" instructions are sometimes omitted by Table 2-17, but are VEX256 766 * only. 767 */ 768 [0x00] = X86_OP_ENTRY3(VPERMQ, V,qq, W,qq, I,b, vex6 chk(W1) cpuid(AVX2) p_66), 769 [0x01] = X86_OP_ENTRY3(VPERMQ, V,qq, W,qq, I,b, vex6 chk(W1) cpuid(AVX2) p_66), /* VPERMPD */ 770 [0x02] = X86_OP_ENTRY4(VBLENDPS, V,x, H,x, W,x, vex6 chk(W0) cpuid(AVX2) p_66), /* VPBLENDD */ 771 [0x04] = X86_OP_ENTRY3(VPERMILPS_i, V,x, W,x, I,b, vex6 chk(W0) cpuid(AVX) p_66), 772 [0x05] = X86_OP_ENTRY3(VPERMILPD_i, V,x, W,x, I,b, vex6 chk(W0) cpuid(AVX) p_66), 773 [0x06] = X86_OP_ENTRY4(VPERM2x128, V,qq, H,qq, W,qq, vex6 chk(W0) cpuid(AVX) p_66), 774 775 [0x14] = X86_OP_ENTRY3(PEXTRB, E,b, V,dq, I,b, vex5 cpuid(SSE41) op0_Rd p_66), 776 [0x15] = X86_OP_ENTRY3(PEXTRW, E,w, V,dq, I,b, vex5 cpuid(SSE41) op0_Rd p_66), 777 [0x16] = X86_OP_ENTRY3(PEXTR, E,y, V,dq, I,b, vex5 cpuid(SSE41) p_66), 778 [0x17] = X86_OP_ENTRY3(VEXTRACTPS, E,d, V,dq, I,b, vex5 cpuid(SSE41) p_66), 779 [0x1d] = X86_OP_ENTRY3(VCVTPS2PH, W,xh, V,x, I,b, vex11 chk(W0) cpuid(F16C) p_66), 780 781 [0x20] = X86_OP_ENTRY4(PINSRB, V,dq, H,dq, E,b, vex5 cpuid(SSE41) op2_Ry p_66), 782 [0x21] = X86_OP_GROUP0(VINSERTPS), 783 [0x22] = X86_OP_ENTRY4(PINSR, V,dq, H,dq, E,y, vex5 cpuid(SSE41) p_66), 784 785 [0x40] = X86_OP_ENTRY4(VDDPS, V,x, H,x, W,x, vex2 cpuid(SSE41) p_66), 786 [0x41] = X86_OP_ENTRY4(VDDPD, V,dq, H,dq, W,dq, vex2 cpuid(SSE41) p_66), 787 [0x42] = X86_OP_ENTRY4(VMPSADBW, V,x, H,x, W,x, vex2 cpuid(SSE41) avx2_256 p_66), 788 [0x44] = X86_OP_ENTRY4(PCLMULQDQ, V,dq, H,dq, W,dq, vex4 cpuid(PCLMULQDQ) p_66), 789 [0x46] = X86_OP_ENTRY4(VPERM2x128, V,qq, H,qq, W,qq, vex6 chk(W0) cpuid(AVX2) p_66), 790 791 [0x60] = X86_OP_ENTRY4(PCMPESTRM, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), 792 [0x61] = X86_OP_ENTRY4(PCMPESTRI, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), 793 [0x62] = X86_OP_ENTRY4(PCMPISTRM, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), 794 [0x63] = X86_OP_ENTRY4(PCMPISTRI, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66), 795 796 [0x08] = X86_OP_ENTRY3(VROUNDPS, V,x, W,x, I,b, vex2 cpuid(SSE41) p_66), 797 [0x09] = X86_OP_ENTRY3(VROUNDPD, V,x, W,x, I,b, vex2 cpuid(SSE41) p_66), 798 /* 799 * Not listed as four operand in the manual. Also writes and reads 128-bits 800 * from the first two operands due to the V operand picking higher entries of 801 * the H operand; the "Vss,Hss,Wss" description from the manual is incorrect. 802 * For other unary operations such as VSQRTSx this is hidden by the "REPScalar" 803 * value of vex_special, because the table lists the operand types of VSQRTPx. 804 */ 805 [0x0a] = X86_OP_ENTRY4(VROUNDSS, V,x, H,x, W,ss, vex3 cpuid(SSE41) p_66), 806 [0x0b] = X86_OP_ENTRY4(VROUNDSD, V,x, H,x, W,sd, vex3 cpuid(SSE41) p_66), 807 [0x0c] = X86_OP_ENTRY4(VBLENDPS, V,x, H,x, W,x, vex4 cpuid(SSE41) p_66), 808 [0x0d] = X86_OP_ENTRY4(VBLENDPD, V,x, H,x, W,x, vex4 cpuid(SSE41) p_66), 809 [0x0e] = X86_OP_ENTRY4(VPBLENDW, V,x, H,x, W,x, vex4 cpuid(SSE41) avx2_256 p_66), 810 [0x0f] = X86_OP_ENTRY4(PALIGNR, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx avx2_256 p_00_66), 811 812 [0x18] = X86_OP_ENTRY4(VINSERTx128, V,qq, H,qq, W,qq, vex6 chk(W0) cpuid(AVX) p_66), 813 [0x19] = X86_OP_ENTRY3(VEXTRACTx128, W,dq, V,qq, I,b, vex6 chk(W0) cpuid(AVX) p_66), 814 815 [0x38] = X86_OP_ENTRY4(VINSERTx128, V,qq, H,qq, W,qq, vex6 chk(W0) cpuid(AVX2) p_66), 816 [0x39] = X86_OP_ENTRY3(VEXTRACTx128, W,dq, V,qq, I,b, vex6 chk(W0) cpuid(AVX2) p_66), 817 818 /* Listed incorrectly as type 4 */ 819 [0x4a] = X86_OP_ENTRY4(VBLENDVPS, V,x, H,x, W,x, vex6 chk(W0) cpuid(AVX) p_66), 820 [0x4b] = X86_OP_ENTRY4(VBLENDVPD, V,x, H,x, W,x, vex6 chk(W0) cpuid(AVX) p_66), 821 [0x4c] = X86_OP_ENTRY4(VPBLENDVB, V,x, H,x, W,x, vex6 chk(W0) cpuid(AVX) p_66 avx2_256), 822 823 [0xcc] = X86_OP_ENTRY3(SHA1RNDS4, V,dq, W,dq, I,b, cpuid(SHA_NI)), 824 825 [0xdf] = X86_OP_ENTRY3(VAESKEYGEN, V,dq, W,dq, I,b, vex4 cpuid(AES) p_66), 826 827 [0xF0] = X86_OP_ENTRY3(RORX, G,y, E,y, I,b, vex13 cpuid(BMI2) p_f2), 828}; 829 830static void decode_0F3A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 831{ 832 *b = x86_ldub_code(env, s); 833 *entry = opcodes_0F3A[*b]; 834} 835 836/* 837 * There are some mistakes in the operands in the manual, and the load/store/register 838 * cases are easiest to keep separate, so the entries for 10-17 follow simplicity and 839 * efficiency of implementation rather than copying what the manual says. 840 * 841 * In particular: 842 * 843 * 1) "VMOVSS m32, xmm1" and "VMOVSD m64, xmm1" do not support VEX.vvvv != 1111b, 844 * but this is not mentioned in the tables. 845 * 846 * 2) MOVHLPS, MOVHPS, MOVHPD, MOVLPD, MOVLPS read the high quadword of one of their 847 * operands, which must therefore be dq; MOVLPD and MOVLPS also write the high 848 * quadword of the V operand. 849 */ 850static void decode_0F10(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 851{ 852 static const X86OpEntry opcodes_0F10_reg[4] = { 853 X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPS */ 854 X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPD */ 855 X86_OP_ENTRY3(VMOVSS, V,x, H,x, W,x, vex5), 856 X86_OP_ENTRY3(VMOVLPx, V,x, H,x, W,x, vex5), /* MOVSD */ 857 }; 858 859 static const X86OpEntry opcodes_0F10_mem[4] = { 860 X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPS */ 861 X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex4_unal), /* MOVUPD */ 862 X86_OP_ENTRY3(VMOVSS_ld, V,x, H,x, M,ss, vex5), 863 X86_OP_ENTRY3(VMOVSD_ld, V,x, H,x, M,sd, vex5), 864 }; 865 866 if ((get_modrm(s, env) >> 6) == 3) { 867 *entry = *decode_by_prefix(s, opcodes_0F10_reg); 868 } else { 869 *entry = *decode_by_prefix(s, opcodes_0F10_mem); 870 } 871} 872 873static void decode_0F11(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 874{ 875 static const X86OpEntry opcodes_0F11_reg[4] = { 876 X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPS */ 877 X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPD */ 878 X86_OP_ENTRY3(VMOVSS, W,x, H,x, V,x, vex5), 879 X86_OP_ENTRY3(VMOVLPx, W,x, H,x, V,q, vex5), /* MOVSD */ 880 }; 881 882 static const X86OpEntry opcodes_0F11_mem[4] = { 883 X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPS */ 884 X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex4), /* MOVUPD */ 885 X86_OP_ENTRY3(VMOVSS_st, M,ss, None,None, V,x, vex5), 886 X86_OP_ENTRY3(VMOVLPx_st, M,sd, None,None, V,x, vex5), /* MOVSD */ 887 }; 888 889 if ((get_modrm(s, env) >> 6) == 3) { 890 *entry = *decode_by_prefix(s, opcodes_0F11_reg); 891 } else { 892 *entry = *decode_by_prefix(s, opcodes_0F11_mem); 893 } 894} 895 896static void decode_0F12(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 897{ 898 static const X86OpEntry opcodes_0F12_mem[4] = { 899 /* 900 * Use dq for operand for compatibility with gen_MOVSD and 901 * to allow VEX128 only. 902 */ 903 X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq, M,q, vex5), /* MOVLPS */ 904 X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq, M,q, vex5), /* MOVLPD */ 905 X86_OP_ENTRY3(VMOVSLDUP, V,x, None,None, W,x, vex4 cpuid(SSE3)), 906 X86_OP_ENTRY3(VMOVDDUP, V,x, None,None, WM,q, vex5 cpuid(SSE3)), /* qq if VEX.256 */ 907 }; 908 static const X86OpEntry opcodes_0F12_reg[4] = { 909 X86_OP_ENTRY3(VMOVHLPS, V,dq, H,dq, U,dq, vex7), 910 X86_OP_ENTRY3(VMOVLPx, W,x, H,x, U,q, vex5), /* MOVLPD */ 911 X86_OP_ENTRY3(VMOVSLDUP, V,x, None,None, U,x, vex4 cpuid(SSE3)), 912 X86_OP_ENTRY3(VMOVDDUP, V,x, None,None, U,x, vex5 cpuid(SSE3)), 913 }; 914 915 if ((get_modrm(s, env) >> 6) == 3) { 916 *entry = *decode_by_prefix(s, opcodes_0F12_reg); 917 } else { 918 *entry = *decode_by_prefix(s, opcodes_0F12_mem); 919 if ((s->prefix & PREFIX_REPNZ) && s->vex_l) { 920 entry->s2 = X86_SIZE_qq; 921 } 922 } 923} 924 925static void decode_0F16(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 926{ 927 static const X86OpEntry opcodes_0F16_mem[4] = { 928 /* 929 * Operand 1 technically only reads the low 64 bits, but uses dq so that 930 * it is easier to check for op0 == op1 in an endianness-neutral manner. 931 */ 932 X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq, M,q, vex5), /* MOVHPS */ 933 X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq, M,q, vex5), /* MOVHPD */ 934 X86_OP_ENTRY3(VMOVSHDUP, V,x, None,None, W,x, vex4 cpuid(SSE3)), 935 {}, 936 }; 937 static const X86OpEntry opcodes_0F16_reg[4] = { 938 /* Same as above, operand 1 could be Hq if it wasn't for big-endian. */ 939 X86_OP_ENTRY3(VMOVLHPS, V,dq, H,dq, U,q, vex7), 940 X86_OP_ENTRY3(VMOVHPx, V,x, H,x, U,x, vex5), /* MOVHPD */ 941 X86_OP_ENTRY3(VMOVSHDUP, V,x, None,None, U,x, vex4 cpuid(SSE3)), 942 {}, 943 }; 944 945 if ((get_modrm(s, env) >> 6) == 3) { 946 *entry = *decode_by_prefix(s, opcodes_0F16_reg); 947 } else { 948 *entry = *decode_by_prefix(s, opcodes_0F16_mem); 949 } 950} 951 952static void decode_0F2A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 953{ 954 static const X86OpEntry opcodes_0F2A[4] = { 955 X86_OP_ENTRY3(CVTPI2Px, V,x, None,None, Q,q), 956 X86_OP_ENTRY3(CVTPI2Px, V,x, None,None, Q,q), 957 X86_OP_ENTRY3(VCVTSI2Sx, V,x, H,x, E,y, vex3), 958 X86_OP_ENTRY3(VCVTSI2Sx, V,x, H,x, E,y, vex3), 959 }; 960 *entry = *decode_by_prefix(s, opcodes_0F2A); 961} 962 963static void decode_0F2B(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 964{ 965 static const X86OpEntry opcodes_0F2B[4] = { 966 X86_OP_ENTRY3(MOVDQ, M,x, None,None, V,x, vex1), /* MOVNTPS */ 967 X86_OP_ENTRY3(MOVDQ, M,x, None,None, V,x, vex1), /* MOVNTPD */ 968 /* AMD extensions */ 969 X86_OP_ENTRY3(VMOVSS_st, M,ss, None,None, V,x, vex4 cpuid(SSE4A)), /* MOVNTSS */ 970 X86_OP_ENTRY3(VMOVLPx_st, M,sd, None,None, V,x, vex4 cpuid(SSE4A)), /* MOVNTSD */ 971 }; 972 973 *entry = *decode_by_prefix(s, opcodes_0F2B); 974} 975 976static void decode_0F2C(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 977{ 978 static const X86OpEntry opcodes_0F2C[4] = { 979 /* Listed as ps/pd in the manual, but CVTTPS2PI only reads 64-bit. */ 980 X86_OP_ENTRY3(CVTTPx2PI, P,q, None,None, W,q), 981 X86_OP_ENTRY3(CVTTPx2PI, P,q, None,None, W,dq), 982 X86_OP_ENTRY3(VCVTTSx2SI, G,y, None,None, W,ss, vex3), 983 X86_OP_ENTRY3(VCVTTSx2SI, G,y, None,None, W,sd, vex3), 984 }; 985 *entry = *decode_by_prefix(s, opcodes_0F2C); 986} 987 988static void decode_0F2D(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 989{ 990 static const X86OpEntry opcodes_0F2D[4] = { 991 /* Listed as ps/pd in the manual, but CVTPS2PI only reads 64-bit. */ 992 X86_OP_ENTRY3(CVTPx2PI, P,q, None,None, W,q), 993 X86_OP_ENTRY3(CVTPx2PI, P,q, None,None, W,dq), 994 X86_OP_ENTRY3(VCVTSx2SI, G,y, None,None, W,ss, vex3), 995 X86_OP_ENTRY3(VCVTSx2SI, G,y, None,None, W,sd, vex3), 996 }; 997 *entry = *decode_by_prefix(s, opcodes_0F2D); 998} 999 1000static void decode_VxCOMISx(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1001{ 1002 /* 1003 * VUCOMISx and VCOMISx are different and use no-prefix and 0x66 for SS and SD 1004 * respectively. Scalar values usually are associated with 0xF2 and 0xF3, for 1005 * which X86_VEX_REPScalar exists, but here it has to be decoded by hand. 1006 */ 1007 entry->s1 = entry->s2 = (s->prefix & PREFIX_DATA ? X86_SIZE_sd : X86_SIZE_ss); 1008 entry->gen = (*b == 0x2E ? gen_VUCOMI : gen_VCOMI); 1009} 1010 1011static void decode_sse_unary(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1012{ 1013 if (!(s->prefix & (PREFIX_REPZ | PREFIX_REPNZ))) { 1014 entry->op1 = X86_TYPE_None; 1015 entry->s1 = X86_SIZE_None; 1016 } 1017 switch (*b) { 1018 case 0x51: entry->gen = gen_VSQRT; break; 1019 case 0x52: entry->gen = gen_VRSQRT; break; 1020 case 0x53: entry->gen = gen_VRCP; break; 1021 } 1022} 1023 1024static void decode_0F5A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1025{ 1026 static const X86OpEntry opcodes_0F5A[4] = { 1027 X86_OP_ENTRY2(VCVTPS2PD, V,x, W,xh, vex2), /* VCVTPS2PD */ 1028 X86_OP_ENTRY2(VCVTPD2PS, V,x, W,x, vex2), /* VCVTPD2PS */ 1029 X86_OP_ENTRY3(VCVTSS2SD, V,x, H,x, W,x, vex2_rep3), /* VCVTSS2SD */ 1030 X86_OP_ENTRY3(VCVTSD2SS, V,x, H,x, W,x, vex2_rep3), /* VCVTSD2SS */ 1031 }; 1032 *entry = *decode_by_prefix(s, opcodes_0F5A); 1033} 1034 1035static void decode_0F5B(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1036{ 1037 static const X86OpEntry opcodes_0F5B[4] = { 1038 X86_OP_ENTRY2(VCVTDQ2PS, V,x, W,x, vex2), 1039 X86_OP_ENTRY2(VCVTPS2DQ, V,x, W,x, vex2), 1040 X86_OP_ENTRY2(VCVTTPS2DQ, V,x, W,x, vex2), 1041 {}, 1042 }; 1043 *entry = *decode_by_prefix(s, opcodes_0F5B); 1044} 1045 1046static void decode_0FE6(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1047{ 1048 static const X86OpEntry opcodes_0FE6[4] = { 1049 {}, 1050 X86_OP_ENTRY2(VCVTTPD2DQ, V,x, W,x, vex2), 1051 X86_OP_ENTRY2(VCVTDQ2PD, V,x, W,x, vex5), 1052 X86_OP_ENTRY2(VCVTPD2DQ, V,x, W,x, vex2), 1053 }; 1054 *entry = *decode_by_prefix(s, opcodes_0FE6); 1055} 1056 1057/* 1058 * These ignore the mod bits (assume (modrm&0xc0)==0xc0), so group the 1059 * pre-decode tweak here for all MOVs from/to CR and DR. 1060 * 1061 * AMD documentation (24594.pdf) and testing of Intel 386 and 486 1062 * processors all show that the mod bits are assumed to be 1's, 1063 * regardless of actual values. 1064 */ 1065static void decode_MOV_CR_DR(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1066{ 1067 /* 1068 */ 1069 get_modrm(s, env); 1070 s->modrm |= 0xC0; 1071 1072 entry->gen = gen_MOV; 1073} 1074 1075static const X86OpEntry opcodes_0F[256] = { 1076 [0x02] = X86_OP_ENTRYwr(LAR, G,v, E,w, chk(prot)), 1077 [0x03] = X86_OP_ENTRYwr(LSL, G,v, E,w, chk(prot)), 1078 [0x05] = X86_OP_ENTRY0(SYSCALL, chk(o64_intel)), 1079 [0x06] = X86_OP_ENTRY0(CLTS, chk(cpl0) svm(WRITE_CR0)), 1080 [0x07] = X86_OP_ENTRY0(SYSRET, chk3(o64_intel, prot, cpl0)), 1081 1082 [0x10] = X86_OP_GROUP0(0F10), 1083 [0x11] = X86_OP_GROUP0(0F11), 1084 [0x12] = X86_OP_GROUP0(0F12), 1085 [0x13] = X86_OP_ENTRY3(VMOVLPx_st, M,q, None,None, V,q, vex5 p_00_66), 1086 [0x14] = X86_OP_ENTRY3(VUNPCKLPx, V,x, H,x, W,x, vex4 p_00_66), 1087 [0x15] = X86_OP_ENTRY3(VUNPCKHPx, V,x, H,x, W,x, vex4 p_00_66), 1088 [0x16] = X86_OP_GROUP0(0F16), 1089 /* Incorrectly listed as Mq,Vq in the manual */ 1090 [0x17] = X86_OP_ENTRY3(VMOVHPx_st, M,q, None,None, V,dq, vex5 p_00_66), 1091 1092 /* 1093 * Incorrectly listed as using "d" operand type in the manual. In reality 1094 * there's no 16-bit version (like y) and it does not use REX.W (like d64). 1095 */ 1096 [0x20] = X86_OP_GROUPwr(MOV_CR_DR, R,y_d64, C,y_d64, chk(cpl0) svm(READ_CR0)), 1097 [0x21] = X86_OP_GROUPwr(MOV_CR_DR, R,y_d64, D,y_d64, chk(cpl0) svm(READ_DR0)), 1098 [0x22] = X86_OP_GROUPwr(MOV_CR_DR, C,y_d64, R,y_d64, zextT0 chk(cpl0) svm(WRITE_CR0)), 1099 [0x23] = X86_OP_GROUPwr(MOV_CR_DR, D,y_d64, R,y_d64, zextT0 chk(cpl0) svm(WRITE_DR0)), 1100 1101 [0x30] = X86_OP_ENTRY0(WRMSR, chk(cpl0)), 1102 [0x31] = X86_OP_ENTRY0(RDTSC), 1103 [0x32] = X86_OP_ENTRY0(RDMSR, chk(cpl0)), 1104 [0x33] = X86_OP_ENTRY0(RDPMC), 1105 [0x34] = X86_OP_ENTRY0(SYSENTER, chk2(i64_amd, prot_or_vm86)), 1106 [0x35] = X86_OP_ENTRY0(SYSEXIT, chk3(i64_amd, prot, cpl0)), 1107 1108 [0x40] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1109 [0x41] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1110 [0x42] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1111 [0x43] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1112 [0x44] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1113 [0x45] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1114 [0x46] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1115 [0x47] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1116 1117 [0x50] = X86_OP_ENTRY3(MOVMSK, G,y, None,None, U,x, vex7 p_00_66), 1118 [0x51] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), /* sqrtps */ 1119 [0x52] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex4_rep5 p_00_f3), /* rsqrtps */ 1120 [0x53] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex4_rep5 p_00_f3), /* rcpps */ 1121 [0x54] = X86_OP_ENTRY3(PAND, V,x, H,x, W,x, vex4 p_00_66), /* vand */ 1122 [0x55] = X86_OP_ENTRY3(PANDN, V,x, H,x, W,x, vex4 p_00_66), /* vandn */ 1123 [0x56] = X86_OP_ENTRY3(POR, V,x, H,x, W,x, vex4 p_00_66), /* vor */ 1124 [0x57] = X86_OP_ENTRY3(PXOR, V,x, H,x, W,x, vex4 p_00_66), /* vxor */ 1125 1126 [0x60] = X86_OP_ENTRY3(PUNPCKLBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1127 [0x61] = X86_OP_ENTRY3(PUNPCKLWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1128 [0x62] = X86_OP_ENTRY3(PUNPCKLDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1129 [0x63] = X86_OP_ENTRY3(PACKSSWB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1130 [0x64] = X86_OP_ENTRY3(PCMPGTB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1131 [0x65] = X86_OP_ENTRY3(PCMPGTW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1132 [0x66] = X86_OP_ENTRY3(PCMPGTD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1133 [0x67] = X86_OP_ENTRY3(PACKUSWB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1134 1135 [0x70] = X86_OP_GROUP0(0F70), 1136 [0x71] = X86_OP_GROUP0(group12), 1137 [0x72] = X86_OP_GROUP0(group13), 1138 [0x73] = X86_OP_GROUP0(group14), 1139 [0x74] = X86_OP_ENTRY3(PCMPEQB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1140 [0x75] = X86_OP_ENTRY3(PCMPEQW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1141 [0x76] = X86_OP_ENTRY3(PCMPEQD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1142 [0x77] = X86_OP_GROUP0(0F77), 1143 1144 [0x80] = X86_OP_ENTRYr(Jcc, J,z_f64), 1145 [0x81] = X86_OP_ENTRYr(Jcc, J,z_f64), 1146 [0x82] = X86_OP_ENTRYr(Jcc, J,z_f64), 1147 [0x83] = X86_OP_ENTRYr(Jcc, J,z_f64), 1148 [0x84] = X86_OP_ENTRYr(Jcc, J,z_f64), 1149 [0x85] = X86_OP_ENTRYr(Jcc, J,z_f64), 1150 [0x86] = X86_OP_ENTRYr(Jcc, J,z_f64), 1151 [0x87] = X86_OP_ENTRYr(Jcc, J,z_f64), 1152 1153 [0x90] = X86_OP_ENTRYw(SETcc, E,b), 1154 [0x91] = X86_OP_ENTRYw(SETcc, E,b), 1155 [0x92] = X86_OP_ENTRYw(SETcc, E,b), 1156 [0x93] = X86_OP_ENTRYw(SETcc, E,b), 1157 [0x94] = X86_OP_ENTRYw(SETcc, E,b), 1158 [0x95] = X86_OP_ENTRYw(SETcc, E,b), 1159 [0x96] = X86_OP_ENTRYw(SETcc, E,b), 1160 [0x97] = X86_OP_ENTRYw(SETcc, E,b), 1161 1162 [0xa0] = X86_OP_ENTRYr(PUSH, FS, w), 1163 [0xa1] = X86_OP_ENTRYw(POP, FS, w), 1164 [0xa2] = X86_OP_ENTRY0(CPUID), 1165 [0xa4] = X86_OP_ENTRY4(SHLD, E,v, 2op,v, G,v), 1166 [0xa5] = X86_OP_ENTRY3(SHLD, E,v, 2op,v, G,v), 1167 1168 [0xb0] = X86_OP_ENTRY2(CMPXCHG,E,b, G,b, lock), 1169 [0xb1] = X86_OP_ENTRY2(CMPXCHG,E,v, G,v, lock), 1170 [0xb2] = X86_OP_ENTRY3(LSS, G,v, EM,p, None, None), 1171 [0xb4] = X86_OP_ENTRY3(LFS, G,v, EM,p, None, None), 1172 [0xb5] = X86_OP_ENTRY3(LGS, G,v, EM,p, None, None), 1173 [0xb6] = X86_OP_ENTRY3(MOV, G,v, E,b, None, None, zextT0), /* MOVZX */ 1174 [0xb7] = X86_OP_ENTRY3(MOV, G,v, E,w, None, None, zextT0), /* MOVZX */ 1175 1176 [0xc0] = X86_OP_ENTRY2(XADD, E,b, G,b, lock), 1177 [0xc1] = X86_OP_ENTRY2(XADD, E,v, G,v, lock), 1178 [0xc2] = X86_OP_ENTRY4(VCMP, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), 1179 [0xc3] = X86_OP_ENTRY3(MOV, EM,y,G,y, None,None, cpuid(SSE2)), /* MOVNTI */ 1180 [0xc4] = X86_OP_ENTRY4(PINSRW, V,dq,H,dq,E,w, vex5 mmx p_00_66), 1181 [0xc5] = X86_OP_ENTRY3(PEXTRW, G,d, U,dq,I,b, vex5 mmx p_00_66), 1182 [0xc6] = X86_OP_ENTRY4(VSHUF, V,x, H,x, W,x, vex4 p_00_66), 1183 1184 [0xd0] = X86_OP_ENTRY3(VADDSUB, V,x, H,x, W,x, vex2 cpuid(SSE3) p_66_f2), 1185 [0xd1] = X86_OP_ENTRY3(PSRLW_r, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1186 [0xd2] = X86_OP_ENTRY3(PSRLD_r, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1187 [0xd3] = X86_OP_ENTRY3(PSRLQ_r, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1188 [0xd4] = X86_OP_ENTRY3(PADDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1189 [0xd5] = X86_OP_ENTRY3(PMULLW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1190 [0xd6] = X86_OP_GROUP0(0FD6), 1191 [0xd7] = X86_OP_ENTRY3(PMOVMSKB, G,d, None,None, U,x, vex7 mmx avx2_256 p_00_66), 1192 1193 [0xe0] = X86_OP_ENTRY3(PAVGB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1194 [0xe1] = X86_OP_ENTRY3(PSRAW_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), 1195 [0xe2] = X86_OP_ENTRY3(PSRAD_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), 1196 [0xe3] = X86_OP_ENTRY3(PAVGW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1197 [0xe4] = X86_OP_ENTRY3(PMULHUW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1198 [0xe5] = X86_OP_ENTRY3(PMULHW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1199 [0xe6] = X86_OP_GROUP0(0FE6), 1200 [0xe7] = X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex1 mmx p_00_66), /* MOVNTQ/MOVNTDQ */ 1201 1202 [0xf0] = X86_OP_ENTRY3(MOVDQ, V,x, None,None, WM,x, vex4_unal cpuid(SSE3) p_f2), /* LDDQU */ 1203 [0xf1] = X86_OP_ENTRY3(PSLLW_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), 1204 [0xf2] = X86_OP_ENTRY3(PSLLD_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), 1205 [0xf3] = X86_OP_ENTRY3(PSLLQ_r, V,x, H,x, W,x, vex7 mmx avx2_256 p_00_66), 1206 [0xf4] = X86_OP_ENTRY3(PMULUDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1207 [0xf5] = X86_OP_ENTRY3(PMADDWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1208 [0xf6] = X86_OP_ENTRY3(PSADBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1209 [0xf7] = X86_OP_ENTRY3(MASKMOV, None,None, V,dq, U,dq, vex4_unal avx2_256 mmx p_00_66), 1210 1211 [0x08] = X86_OP_ENTRY0(NOP, svm(INVD)), 1212 [0x09] = X86_OP_ENTRY0(NOP, svm(WBINVD)), 1213 [0x0b] = X86_OP_ENTRY0(UD), /* UD2 */ 1214 [0x0d] = X86_OP_ENTRY1(NOP, M,v), /* 3DNow! prefetch */ 1215 [0x0e] = X86_OP_ENTRY0(EMMS, cpuid(3DNOW)), /* femms */ 1216 /* 1217 * 3DNow!'s opcode byte comes *after* modrm and displacements, making it 1218 * more like an Ib operand. Dispatch to the right helper in a single gen_* 1219 * function. 1220 */ 1221 [0x0f] = X86_OP_ENTRY3(3dnow, P,q, Q,q, I,b, cpuid(3DNOW)), 1222 1223 [0x18] = X86_OP_ENTRY1(NOP, nop,v), /* prefetch/reserved NOP */ 1224 [0x19] = X86_OP_ENTRY1(NOP, nop,v), /* reserved NOP */ 1225 [0x1c] = X86_OP_ENTRY1(NOP, nop,v), /* reserved NOP */ 1226 [0x1d] = X86_OP_ENTRY1(NOP, nop,v), /* reserved NOP */ 1227 [0x1e] = X86_OP_ENTRY1(NOP, nop,v), /* reserved NOP */ 1228 [0x1f] = X86_OP_ENTRY1(NOP, nop,v), /* NOP/reserved NOP */ 1229 1230 [0x28] = X86_OP_ENTRY3(MOVDQ, V,x, None,None, W,x, vex1 p_00_66), /* MOVAPS */ 1231 [0x29] = X86_OP_ENTRY3(MOVDQ, W,x, None,None, V,x, vex1 p_00_66), /* MOVAPS */ 1232 [0x2A] = X86_OP_GROUP0(0F2A), 1233 [0x2B] = X86_OP_GROUP0(0F2B), 1234 [0x2C] = X86_OP_GROUP0(0F2C), 1235 [0x2D] = X86_OP_GROUP0(0F2D), 1236 [0x2E] = X86_OP_GROUP3(VxCOMISx, None,None, V,x, W,x, vex3 p_00_66), /* VUCOMISS/SD */ 1237 [0x2F] = X86_OP_GROUP3(VxCOMISx, None,None, V,x, W,x, vex3 p_00_66), /* VCOMISS/SD */ 1238 1239 [0x38] = X86_OP_GROUP0(0F38), 1240 [0x3a] = X86_OP_GROUP0(0F3A), 1241 1242 [0x48] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1243 [0x49] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1244 [0x4a] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1245 [0x4b] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1246 [0x4c] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1247 [0x4d] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1248 [0x4e] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1249 [0x4f] = X86_OP_ENTRY2(CMOVcc, G,v, E,v, cpuid(CMOV)), 1250 1251 [0x58] = X86_OP_ENTRY3(VADD, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), 1252 [0x59] = X86_OP_ENTRY3(VMUL, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), 1253 [0x5a] = X86_OP_GROUP0(0F5A), 1254 [0x5b] = X86_OP_GROUP0(0F5B), 1255 [0x5c] = X86_OP_ENTRY3(VSUB, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), 1256 [0x5d] = X86_OP_ENTRY3(VMIN, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), 1257 [0x5e] = X86_OP_ENTRY3(VDIV, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), 1258 [0x5f] = X86_OP_ENTRY3(VMAX, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2), 1259 1260 [0x68] = X86_OP_ENTRY3(PUNPCKHBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1261 [0x69] = X86_OP_ENTRY3(PUNPCKHWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1262 [0x6a] = X86_OP_ENTRY3(PUNPCKHDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1263 [0x6b] = X86_OP_ENTRY3(PACKSSDW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1264 [0x6c] = X86_OP_ENTRY3(PUNPCKLQDQ, V,x, H,x, W,x, vex4 p_66 avx2_256), 1265 [0x6d] = X86_OP_ENTRY3(PUNPCKHQDQ, V,x, H,x, W,x, vex4 p_66 avx2_256), 1266 [0x6e] = X86_OP_ENTRY3(MOVD_to, V,x, None,None, E,y, vex5 mmx p_00_66), /* wrong dest Vy on SDM! */ 1267 [0x6f] = X86_OP_GROUP0(0F6F), 1268 1269 [0x78] = X86_OP_GROUP0(0F78), 1270 [0x79] = X86_OP_GROUP2(0F79, V,x, U,x, cpuid(SSE4A)), 1271 [0x7c] = X86_OP_ENTRY3(VHADD, V,x, H,x, W,x, vex2 cpuid(SSE3) p_66_f2), 1272 [0x7d] = X86_OP_ENTRY3(VHSUB, V,x, H,x, W,x, vex2 cpuid(SSE3) p_66_f2), 1273 [0x7e] = X86_OP_GROUP0(0F7E), 1274 [0x7f] = X86_OP_GROUP0(0F7F), 1275 1276 [0x88] = X86_OP_ENTRYr(Jcc, J,z_f64), 1277 [0x89] = X86_OP_ENTRYr(Jcc, J,z_f64), 1278 [0x8a] = X86_OP_ENTRYr(Jcc, J,z_f64), 1279 [0x8b] = X86_OP_ENTRYr(Jcc, J,z_f64), 1280 [0x8c] = X86_OP_ENTRYr(Jcc, J,z_f64), 1281 [0x8d] = X86_OP_ENTRYr(Jcc, J,z_f64), 1282 [0x8e] = X86_OP_ENTRYr(Jcc, J,z_f64), 1283 [0x8f] = X86_OP_ENTRYr(Jcc, J,z_f64), 1284 1285 [0x98] = X86_OP_ENTRYw(SETcc, E,b), 1286 [0x99] = X86_OP_ENTRYw(SETcc, E,b), 1287 [0x9a] = X86_OP_ENTRYw(SETcc, E,b), 1288 [0x9b] = X86_OP_ENTRYw(SETcc, E,b), 1289 [0x9c] = X86_OP_ENTRYw(SETcc, E,b), 1290 [0x9d] = X86_OP_ENTRYw(SETcc, E,b), 1291 [0x9e] = X86_OP_ENTRYw(SETcc, E,b), 1292 [0x9f] = X86_OP_ENTRYw(SETcc, E,b), 1293 1294 [0xa8] = X86_OP_ENTRYr(PUSH, GS, w), 1295 [0xa9] = X86_OP_ENTRYw(POP, GS, w), 1296 [0xaa] = X86_OP_ENTRY0(RSM, chk(smm) svm(RSM)), 1297 [0xac] = X86_OP_ENTRY4(SHRD, E,v, 2op,v, G,v), 1298 [0xad] = X86_OP_ENTRY3(SHRD, E,v, 2op,v, G,v), 1299 [0xae] = X86_OP_GROUP0(group15), 1300 /* 1301 * It's slightly more efficient to put Ev operand in T0 and allow gen_IMUL3 1302 * to assume sextT0. Multiplication is commutative anyway. 1303 */ 1304 [0xaf] = X86_OP_ENTRY3(IMUL3, G,v, E,v, 2op,v, sextT0), 1305 1306 [0xb8] = X86_OP_GROUP0(0FB8), 1307 /* decoded as modrm, which is visible as a difference between page fault and #UD */ 1308 [0xb9] = X86_OP_ENTRYr(UD, nop,v), /* UD1 */ 1309 [0xbc] = X86_OP_GROUP0(0FBC), 1310 [0xbd] = X86_OP_GROUP0(0FBD), 1311 [0xbe] = X86_OP_ENTRY3(MOV, G,v, E,b, None, None, sextT0), /* MOVSX */ 1312 [0xbf] = X86_OP_ENTRY3(MOV, G,v, E,w, None, None, sextT0), /* MOVSX */ 1313 1314 [0xc8] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1315 [0xc9] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1316 [0xca] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1317 [0xcb] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1318 [0xcc] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1319 [0xcd] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1320 [0xce] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1321 [0xcf] = X86_OP_ENTRY1(BSWAP, LoBits,y), 1322 1323 /* Incorrectly missing from 2-17 */ 1324 [0xd8] = X86_OP_ENTRY3(PSUBUSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1325 [0xd9] = X86_OP_ENTRY3(PSUBUSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1326 [0xda] = X86_OP_ENTRY3(PMINUB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1327 [0xdb] = X86_OP_ENTRY3(PAND, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1328 [0xdc] = X86_OP_ENTRY3(PADDUSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1329 [0xdd] = X86_OP_ENTRY3(PADDUSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1330 [0xde] = X86_OP_ENTRY3(PMAXUB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1331 [0xdf] = X86_OP_ENTRY3(PANDN, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1332 1333 [0xe8] = X86_OP_ENTRY3(PSUBSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1334 [0xe9] = X86_OP_ENTRY3(PSUBSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1335 [0xea] = X86_OP_ENTRY3(PMINSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1336 [0xeb] = X86_OP_ENTRY3(POR, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1337 [0xec] = X86_OP_ENTRY3(PADDSB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1338 [0xed] = X86_OP_ENTRY3(PADDSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1339 [0xee] = X86_OP_ENTRY3(PMAXSW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1340 [0xef] = X86_OP_ENTRY3(PXOR, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1341 1342 [0xf8] = X86_OP_ENTRY3(PSUBB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1343 [0xf9] = X86_OP_ENTRY3(PSUBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1344 [0xfa] = X86_OP_ENTRY3(PSUBD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1345 [0xfb] = X86_OP_ENTRY3(PSUBQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1346 [0xfc] = X86_OP_ENTRY3(PADDB, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1347 [0xfd] = X86_OP_ENTRY3(PADDW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1348 [0xfe] = X86_OP_ENTRY3(PADDD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66), 1349 [0xff] = X86_OP_ENTRYr(UD, nop,v), /* UD0 */ 1350}; 1351 1352static void do_decode_0F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1353{ 1354 *entry = opcodes_0F[*b]; 1355} 1356 1357static void decode_0F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1358{ 1359 *b = x86_ldub_code(env, s); 1360 do_decode_0F(s, env, entry, b); 1361} 1362 1363static void decode_63(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1364{ 1365 static const X86OpEntry arpl = X86_OP_ENTRY2(ARPL, E,w, G,w, chk(prot)); 1366 static const X86OpEntry mov = X86_OP_ENTRY3(MOV, G,v, E,v, None, None); 1367 static const X86OpEntry movsxd = X86_OP_ENTRY3(MOV, G,v, E,d, None, None, sextT0); 1368 if (!CODE64(s)) { 1369 *entry = arpl; 1370 } else if (REX_W(s)) { 1371 *entry = movsxd; 1372 } else { 1373 *entry = mov; 1374 } 1375} 1376 1377static void decode_group1(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1378{ 1379 static const X86GenFunc group1_gen[8] = { 1380 gen_ADD, gen_OR, gen_ADC, gen_SBB, gen_AND, gen_SUB, gen_XOR, gen_SUB, 1381 }; 1382 int op = (get_modrm(s, env) >> 3) & 7; 1383 entry->gen = group1_gen[op]; 1384 1385 if (op == 7) { 1386 /* prevent writeback for CMP */ 1387 entry->op1 = entry->op0; 1388 entry->op0 = X86_TYPE_None; 1389 entry->s0 = X86_SIZE_None; 1390 } else { 1391 entry->special = X86_SPECIAL_HasLock; 1392 } 1393} 1394 1395static void decode_group1A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1396{ 1397 int op = (get_modrm(s, env) >> 3) & 7; 1398 if (op != 0) { 1399 /* could be XOP prefix too */ 1400 *entry = UNKNOWN_OPCODE; 1401 } else { 1402 entry->gen = gen_POP; 1403 /* The address must use the value of ESP after the pop. */ 1404 s->popl_esp_hack = 1 << mo_pushpop(s, s->dflag); 1405 } 1406} 1407 1408static void decode_group2(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1409{ 1410 static const X86GenFunc group2_gen[8] = { 1411 gen_ROL, gen_ROR, gen_RCL, gen_RCR, 1412 gen_SHL, gen_SHR, gen_SHL /* SAL, undocumented */, gen_SAR, 1413 }; 1414 int op = (get_modrm(s, env) >> 3) & 7; 1415 entry->gen = group2_gen[op]; 1416 if (op == 7) { 1417 entry->special = X86_SPECIAL_SExtT0; 1418 } else { 1419 entry->special = X86_SPECIAL_ZExtT0; 1420 } 1421} 1422 1423static void decode_group3(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1424{ 1425 static const X86OpEntry opcodes_grp3[16] = { 1426 /* 0xf6 */ 1427 [0x00] = X86_OP_ENTRYrr(AND, E,b, I,b), 1428 [0x02] = X86_OP_ENTRY1(NOT, E,b, lock), 1429 [0x03] = X86_OP_ENTRY1(NEG, E,b, lock), 1430 [0x04] = X86_OP_ENTRYrr(MUL, E,b, 0,b, zextT0), 1431 [0x05] = X86_OP_ENTRYrr(IMUL,E,b, 0,b, sextT0), 1432 [0x06] = X86_OP_ENTRYr(DIV, E,b), 1433 [0x07] = X86_OP_ENTRYr(IDIV, E,b), 1434 1435 /* 0xf7 */ 1436 [0x08] = X86_OP_ENTRYrr(AND, E,v, I,z), 1437 [0x0a] = X86_OP_ENTRY1(NOT, E,v, lock), 1438 [0x0b] = X86_OP_ENTRY1(NEG, E,v, lock), 1439 [0x0c] = X86_OP_ENTRYrr(MUL, E,v, 0,v, zextT0), 1440 [0x0d] = X86_OP_ENTRYrr(IMUL,E,v, 0,v, sextT0), 1441 [0x0e] = X86_OP_ENTRYr(DIV, E,v), 1442 [0x0f] = X86_OP_ENTRYr(IDIV, E,v), 1443 }; 1444 1445 int w = (*b & 1); 1446 int reg = (get_modrm(s, env) >> 3) & 7; 1447 1448 *entry = opcodes_grp3[(w << 3) | reg]; 1449} 1450 1451static void decode_group4_5(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1452{ 1453 static const X86OpEntry opcodes_grp4_5[16] = { 1454 /* 0xfe */ 1455 [0x00] = X86_OP_ENTRY1(INC, E,b, lock), 1456 [0x01] = X86_OP_ENTRY1(DEC, E,b, lock), 1457 1458 /* 0xff */ 1459 [0x08] = X86_OP_ENTRY1(INC, E,v, lock), 1460 [0x09] = X86_OP_ENTRY1(DEC, E,v, lock), 1461 [0x0a] = X86_OP_ENTRYr(CALL_m, E,f64, zextT0), 1462 [0x0b] = X86_OP_ENTRYr(CALLF_m, M,p), 1463 [0x0c] = X86_OP_ENTRYr(JMP_m, E,f64, zextT0), 1464 [0x0d] = X86_OP_ENTRYr(JMPF_m, M,p), 1465 [0x0e] = X86_OP_ENTRYr(PUSH, E,f64), 1466 }; 1467 1468 int w = (*b & 1); 1469 int reg = (get_modrm(s, env) >> 3) & 7; 1470 1471 *entry = opcodes_grp4_5[(w << 3) | reg]; 1472} 1473 1474 1475static void decode_group11(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1476{ 1477 int op = (get_modrm(s, env) >> 3) & 7; 1478 if (op != 0) { 1479 *entry = UNKNOWN_OPCODE; 1480 } else { 1481 entry->gen = gen_MOV; 1482 } 1483} 1484 1485static void decode_90(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1486{ 1487 static X86OpEntry pause = X86_OP_ENTRY0(PAUSE, svm(PAUSE)); 1488 static X86OpEntry nop = X86_OP_ENTRY0(NOP); 1489 static X86OpEntry xchg_ax = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v); 1490 1491 if (REX_B(s)) { 1492 *entry = xchg_ax; 1493 } else { 1494 *entry = (s->prefix & PREFIX_REPZ) ? pause : nop; 1495 } 1496} 1497 1498static const X86OpEntry opcodes_root[256] = { 1499 [0x00] = X86_OP_ENTRY2(ADD, E,b, G,b, lock), 1500 [0x01] = X86_OP_ENTRY2(ADD, E,v, G,v, lock), 1501 [0x02] = X86_OP_ENTRY2(ADD, G,b, E,b, lock), 1502 [0x03] = X86_OP_ENTRY2(ADD, G,v, E,v, lock), 1503 [0x04] = X86_OP_ENTRY2(ADD, 0,b, I,b, lock), /* AL, Ib */ 1504 [0x05] = X86_OP_ENTRY2(ADD, 0,v, I,z, lock), /* rAX, Iz */ 1505 [0x06] = X86_OP_ENTRYr(PUSH, ES, w, chk(i64)), 1506 [0x07] = X86_OP_ENTRYw(POP, ES, w, chk(i64)), 1507 1508 [0x10] = X86_OP_ENTRY2(ADC, E,b, G,b, lock), 1509 [0x11] = X86_OP_ENTRY2(ADC, E,v, G,v, lock), 1510 [0x12] = X86_OP_ENTRY2(ADC, G,b, E,b, lock), 1511 [0x13] = X86_OP_ENTRY2(ADC, G,v, E,v, lock), 1512 [0x14] = X86_OP_ENTRY2(ADC, 0,b, I,b, lock), /* AL, Ib */ 1513 [0x15] = X86_OP_ENTRY2(ADC, 0,v, I,z, lock), /* rAX, Iz */ 1514 [0x16] = X86_OP_ENTRYr(PUSH, SS, w, chk(i64)), 1515 [0x17] = X86_OP_ENTRYw(POP, SS, w, chk(i64)), 1516 1517 [0x20] = X86_OP_ENTRY2(AND, E,b, G,b, lock), 1518 [0x21] = X86_OP_ENTRY2(AND, E,v, G,v, lock), 1519 [0x22] = X86_OP_ENTRY2(AND, G,b, E,b, lock), 1520 [0x23] = X86_OP_ENTRY2(AND, G,v, E,v, lock), 1521 [0x24] = X86_OP_ENTRY2(AND, 0,b, I,b, lock), /* AL, Ib */ 1522 [0x25] = X86_OP_ENTRY2(AND, 0,v, I,z, lock), /* rAX, Iz */ 1523 [0x26] = {}, 1524 [0x27] = X86_OP_ENTRY0(DAA, chk(i64)), 1525 1526 [0x30] = X86_OP_ENTRY2(XOR, E,b, G,b, lock), 1527 [0x31] = X86_OP_ENTRY2(XOR, E,v, G,v, lock), 1528 [0x32] = X86_OP_ENTRY2(XOR, G,b, E,b, lock), 1529 [0x33] = X86_OP_ENTRY2(XOR, G,v, E,v, lock), 1530 [0x34] = X86_OP_ENTRY2(XOR, 0,b, I,b, lock), /* AL, Ib */ 1531 [0x35] = X86_OP_ENTRY2(XOR, 0,v, I,z, lock), /* rAX, Iz */ 1532 [0x36] = {}, 1533 [0x37] = X86_OP_ENTRY0(AAA, chk(i64)), 1534 1535 [0x40] = X86_OP_ENTRY1(INC, 0,v, chk(i64)), 1536 [0x41] = X86_OP_ENTRY1(INC, 1,v, chk(i64)), 1537 [0x42] = X86_OP_ENTRY1(INC, 2,v, chk(i64)), 1538 [0x43] = X86_OP_ENTRY1(INC, 3,v, chk(i64)), 1539 [0x44] = X86_OP_ENTRY1(INC, 4,v, chk(i64)), 1540 [0x45] = X86_OP_ENTRY1(INC, 5,v, chk(i64)), 1541 [0x46] = X86_OP_ENTRY1(INC, 6,v, chk(i64)), 1542 [0x47] = X86_OP_ENTRY1(INC, 7,v, chk(i64)), 1543 1544 [0x50] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1545 [0x51] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1546 [0x52] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1547 [0x53] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1548 [0x54] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1549 [0x55] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1550 [0x56] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1551 [0x57] = X86_OP_ENTRYr(PUSH, LoBits,d64), 1552 1553 [0x60] = X86_OP_ENTRY0(PUSHA, chk(i64)), 1554 [0x61] = X86_OP_ENTRY0(POPA, chk(i64)), 1555 [0x62] = X86_OP_ENTRYrr(BOUND, G,v, M,a, chk(i64)), 1556 [0x63] = X86_OP_GROUP0(63), 1557 [0x64] = {}, 1558 [0x65] = {}, 1559 [0x66] = {}, 1560 [0x67] = {}, 1561 1562 [0x70] = X86_OP_ENTRYr(Jcc, J,b), 1563 [0x71] = X86_OP_ENTRYr(Jcc, J,b), 1564 [0x72] = X86_OP_ENTRYr(Jcc, J,b), 1565 [0x73] = X86_OP_ENTRYr(Jcc, J,b), 1566 [0x74] = X86_OP_ENTRYr(Jcc, J,b), 1567 [0x75] = X86_OP_ENTRYr(Jcc, J,b), 1568 [0x76] = X86_OP_ENTRYr(Jcc, J,b), 1569 [0x77] = X86_OP_ENTRYr(Jcc, J,b), 1570 1571 [0x80] = X86_OP_GROUP2(group1, E,b, I,b), 1572 [0x81] = X86_OP_GROUP2(group1, E,v, I,z), 1573 [0x82] = X86_OP_GROUP2(group1, E,b, I,b, chk(i64)), 1574 [0x83] = X86_OP_GROUP2(group1, E,v, I,b), 1575 [0x84] = X86_OP_ENTRYrr(AND, E,b, G,b), 1576 [0x85] = X86_OP_ENTRYrr(AND, E,v, G,v), 1577 [0x86] = X86_OP_ENTRY2(XCHG, E,b, G,b, xchg), 1578 [0x87] = X86_OP_ENTRY2(XCHG, E,v, G,v, xchg), 1579 1580 [0x90] = X86_OP_GROUP0(90), 1581 [0x91] = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v), 1582 [0x92] = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v), 1583 [0x93] = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v), 1584 [0x94] = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v), 1585 [0x95] = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v), 1586 [0x96] = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v), 1587 [0x97] = X86_OP_ENTRY2(XCHG, 0,v, LoBits,v), 1588 1589 [0xA0] = X86_OP_ENTRY3(MOV, 0,b, O,b, None, None), /* AL, Ob */ 1590 [0xA1] = X86_OP_ENTRY3(MOV, 0,v, O,v, None, None), /* rAX, Ov */ 1591 [0xA2] = X86_OP_ENTRY3(MOV, O,b, 0,b, None, None), /* Ob, AL */ 1592 [0xA3] = X86_OP_ENTRY3(MOV, O,v, 0,v, None, None), /* Ov, rAX */ 1593 [0xA4] = X86_OP_ENTRYrr(MOVS, Y,b, X,b), 1594 [0xA5] = X86_OP_ENTRYrr(MOVS, Y,v, X,v), 1595 [0xA6] = X86_OP_ENTRYrr(CMPS, Y,b, X,b), 1596 [0xA7] = X86_OP_ENTRYrr(CMPS, Y,v, X,v), 1597 1598 [0xB0] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1599 [0xB1] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1600 [0xB2] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1601 [0xB3] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1602 [0xB4] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1603 [0xB5] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1604 [0xB6] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1605 [0xB7] = X86_OP_ENTRY3(MOV, LoBits,b, I,b, None, None), 1606 1607 [0xC0] = X86_OP_GROUP2(group2, E,b, I,b), 1608 [0xC1] = X86_OP_GROUP2(group2, E,v, I,b), 1609 [0xC2] = X86_OP_ENTRYr(RET, I,w), 1610 [0xC3] = X86_OP_ENTRY0(RET), 1611 [0xC4] = X86_OP_ENTRY3(LES, G,z, EM,p, None, None, chk(i64)), 1612 [0xC5] = X86_OP_ENTRY3(LDS, G,z, EM,p, None, None, chk(i64)), 1613 [0xC6] = X86_OP_GROUP3(group11, E,b, I,b, None, None), /* reg=000b */ 1614 [0xC7] = X86_OP_GROUP3(group11, E,v, I,z, None, None), /* reg=000b */ 1615 1616 [0xD0] = X86_OP_GROUP1(group2, E,b), 1617 [0xD1] = X86_OP_GROUP1(group2, E,v), 1618 [0xD2] = X86_OP_GROUP2(group2, E,b, 1,b), /* CL */ 1619 [0xD3] = X86_OP_GROUP2(group2, E,v, 1,b), /* CL */ 1620 [0xD4] = X86_OP_ENTRY2(AAM, 0,w, I,b), 1621 [0xD5] = X86_OP_ENTRY2(AAD, 0,w, I,b), 1622 [0xD6] = X86_OP_ENTRYw(SALC, 0,b), 1623 [0xD7] = X86_OP_ENTRY1(XLAT, 0,b, zextT0), /* AL read/written */ 1624 1625 [0xE0] = X86_OP_ENTRYr(LOOPNE, J,b), /* implicit: CX with aflag size */ 1626 [0xE1] = X86_OP_ENTRYr(LOOPE, J,b), /* implicit: CX with aflag size */ 1627 [0xE2] = X86_OP_ENTRYr(LOOP, J,b), /* implicit: CX with aflag size */ 1628 [0xE3] = X86_OP_ENTRYr(JCXZ, J,b), /* implicit: CX with aflag size */ 1629 [0xE4] = X86_OP_ENTRYwr(IN, 0,b, I_unsigned,b), /* AL */ 1630 [0xE5] = X86_OP_ENTRYwr(IN, 0,z, I_unsigned,b), /* AX/EAX */ 1631 [0xE6] = X86_OP_ENTRYrr(OUT, 0,b, I_unsigned,b), /* AL */ 1632 [0xE7] = X86_OP_ENTRYrr(OUT, 0,z, I_unsigned,b), /* AX/EAX */ 1633 1634 [0xF1] = X86_OP_ENTRY0(INT1, svm(ICEBP)), 1635 [0xF4] = X86_OP_ENTRY0(HLT, chk(cpl0) svm(HLT)), 1636 [0xF5] = X86_OP_ENTRY0(CMC), 1637 [0xF6] = X86_OP_GROUP1(group3, E,b), 1638 [0xF7] = X86_OP_GROUP1(group3, E,v), 1639 1640 [0x08] = X86_OP_ENTRY2(OR, E,b, G,b, lock), 1641 [0x09] = X86_OP_ENTRY2(OR, E,v, G,v, lock), 1642 [0x0A] = X86_OP_ENTRY2(OR, G,b, E,b, lock), 1643 [0x0B] = X86_OP_ENTRY2(OR, G,v, E,v, lock), 1644 [0x0C] = X86_OP_ENTRY2(OR, 0,b, I,b, lock), /* AL, Ib */ 1645 [0x0D] = X86_OP_ENTRY2(OR, 0,v, I,z, lock), /* rAX, Iz */ 1646 [0x0E] = X86_OP_ENTRYr(PUSH, CS, w, chk(i64)), 1647 [0x0F] = X86_OP_GROUP0(0F), 1648 1649 [0x18] = X86_OP_ENTRY2(SBB, E,b, G,b, lock), 1650 [0x19] = X86_OP_ENTRY2(SBB, E,v, G,v, lock), 1651 [0x1A] = X86_OP_ENTRY2(SBB, G,b, E,b, lock), 1652 [0x1B] = X86_OP_ENTRY2(SBB, G,v, E,v, lock), 1653 [0x1C] = X86_OP_ENTRY2(SBB, 0,b, I,b, lock), /* AL, Ib */ 1654 [0x1D] = X86_OP_ENTRY2(SBB, 0,v, I,z, lock), /* rAX, Iz */ 1655 [0x1E] = X86_OP_ENTRYr(PUSH, DS, w, chk(i64)), 1656 [0x1F] = X86_OP_ENTRYw(POP, DS, w, chk(i64)), 1657 1658 [0x28] = X86_OP_ENTRY2(SUB, E,b, G,b, lock), 1659 [0x29] = X86_OP_ENTRY2(SUB, E,v, G,v, lock), 1660 [0x2A] = X86_OP_ENTRY2(SUB, G,b, E,b, lock), 1661 [0x2B] = X86_OP_ENTRY2(SUB, G,v, E,v, lock), 1662 [0x2C] = X86_OP_ENTRY2(SUB, 0,b, I,b, lock), /* AL, Ib */ 1663 [0x2D] = X86_OP_ENTRY2(SUB, 0,v, I,z, lock), /* rAX, Iz */ 1664 [0x2E] = {}, 1665 [0x2F] = X86_OP_ENTRY0(DAS, chk(i64)), 1666 1667 [0x38] = X86_OP_ENTRYrr(SUB, E,b, G,b), 1668 [0x39] = X86_OP_ENTRYrr(SUB, E,v, G,v), 1669 [0x3A] = X86_OP_ENTRYrr(SUB, G,b, E,b), 1670 [0x3B] = X86_OP_ENTRYrr(SUB, G,v, E,v), 1671 [0x3C] = X86_OP_ENTRYrr(SUB, 0,b, I,b), /* AL, Ib */ 1672 [0x3D] = X86_OP_ENTRYrr(SUB, 0,v, I,z), /* rAX, Iz */ 1673 [0x3E] = {}, 1674 [0x3F] = X86_OP_ENTRY0(AAS, chk(i64)), 1675 1676 [0x48] = X86_OP_ENTRY1(DEC, 0,v, chk(i64)), 1677 [0x49] = X86_OP_ENTRY1(DEC, 1,v, chk(i64)), 1678 [0x4A] = X86_OP_ENTRY1(DEC, 2,v, chk(i64)), 1679 [0x4B] = X86_OP_ENTRY1(DEC, 3,v, chk(i64)), 1680 [0x4C] = X86_OP_ENTRY1(DEC, 4,v, chk(i64)), 1681 [0x4D] = X86_OP_ENTRY1(DEC, 5,v, chk(i64)), 1682 [0x4E] = X86_OP_ENTRY1(DEC, 6,v, chk(i64)), 1683 [0x4F] = X86_OP_ENTRY1(DEC, 7,v, chk(i64)), 1684 1685 [0x58] = X86_OP_ENTRYw(POP, LoBits,d64), 1686 [0x59] = X86_OP_ENTRYw(POP, LoBits,d64), 1687 [0x5A] = X86_OP_ENTRYw(POP, LoBits,d64), 1688 [0x5B] = X86_OP_ENTRYw(POP, LoBits,d64), 1689 [0x5C] = X86_OP_ENTRYw(POP, LoBits,d64), 1690 [0x5D] = X86_OP_ENTRYw(POP, LoBits,d64), 1691 [0x5E] = X86_OP_ENTRYw(POP, LoBits,d64), 1692 [0x5F] = X86_OP_ENTRYw(POP, LoBits,d64), 1693 1694 [0x68] = X86_OP_ENTRYr(PUSH, I,z), 1695 [0x69] = X86_OP_ENTRY3(IMUL3, G,v, E,v, I,z, sextT0), 1696 [0x6A] = X86_OP_ENTRYr(PUSH, I,b), 1697 [0x6B] = X86_OP_ENTRY3(IMUL3, G,v, E,v, I,b, sextT0), 1698 [0x6C] = X86_OP_ENTRYrr(INS, Y,b, 2,w), /* DX */ 1699 [0x6D] = X86_OP_ENTRYrr(INS, Y,z, 2,w), /* DX */ 1700 [0x6E] = X86_OP_ENTRYrr(OUTS, X,b, 2,w), /* DX */ 1701 [0x6F] = X86_OP_ENTRYrr(OUTS, X,z, 2,w), /* DX */ 1702 1703 [0x78] = X86_OP_ENTRYr(Jcc, J,b), 1704 [0x79] = X86_OP_ENTRYr(Jcc, J,b), 1705 [0x7A] = X86_OP_ENTRYr(Jcc, J,b), 1706 [0x7B] = X86_OP_ENTRYr(Jcc, J,b), 1707 [0x7C] = X86_OP_ENTRYr(Jcc, J,b), 1708 [0x7D] = X86_OP_ENTRYr(Jcc, J,b), 1709 [0x7E] = X86_OP_ENTRYr(Jcc, J,b), 1710 [0x7F] = X86_OP_ENTRYr(Jcc, J,b), 1711 1712 [0x88] = X86_OP_ENTRYwr(MOV, E,b, G,b), 1713 [0x89] = X86_OP_ENTRYwr(MOV, E,v, G,v), 1714 [0x8A] = X86_OP_ENTRYwr(MOV, G,b, E,b), 1715 [0x8B] = X86_OP_ENTRYwr(MOV, G,v, E,v), 1716 /* Missing in Table A-2: memory destination is always 16-bit. */ 1717 [0x8C] = X86_OP_ENTRYwr(MOV, E,v, S,w, op0_Mw), 1718 [0x8D] = X86_OP_ENTRYwr(LEA, G,v, M,v, nolea), 1719 [0x8E] = X86_OP_ENTRYwr(MOV, S,w, E,w), 1720 [0x8F] = X86_OP_GROUPw(group1A, E,d64), 1721 1722 [0x98] = X86_OP_ENTRY1(CBW, 0,v), /* rAX */ 1723 [0x99] = X86_OP_ENTRYwr(CWD, 2,v, 0,v), /* rDX, rAX */ 1724 [0x9A] = X86_OP_ENTRYrr(CALLF, I_unsigned,p, I_unsigned,w, chk(i64)), 1725 [0x9B] = X86_OP_ENTRY0(WAIT), 1726 [0x9C] = X86_OP_ENTRY0(PUSHF, chk(vm86_iopl) svm(PUSHF)), 1727 [0x9D] = X86_OP_ENTRY0(POPF, chk(vm86_iopl) svm(POPF)), 1728 [0x9E] = X86_OP_ENTRY0(SAHF), 1729 [0x9F] = X86_OP_ENTRY0(LAHF), 1730 1731 [0xA8] = X86_OP_ENTRYrr(AND, 0,b, I,b), /* AL, Ib */ 1732 [0xA9] = X86_OP_ENTRYrr(AND, 0,v, I,z), /* rAX, Iz */ 1733 [0xAA] = X86_OP_ENTRYwr(STOS, Y,b, 0,b), 1734 [0xAB] = X86_OP_ENTRYwr(STOS, Y,v, 0,v), 1735 /* Manual writeback because REP LODS (!) has to write EAX/RAX after every LODS. */ 1736 [0xAC] = X86_OP_ENTRYr(LODS, X,b), 1737 [0xAD] = X86_OP_ENTRYr(LODS, X,v), 1738 [0xAE] = X86_OP_ENTRYrr(SCAS, 0,b, Y,b), 1739 [0xAF] = X86_OP_ENTRYrr(SCAS, 0,v, Y,v), 1740 1741 [0xB8] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1742 [0xB9] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1743 [0xBA] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1744 [0xBB] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1745 [0xBC] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1746 [0xBD] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1747 [0xBE] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1748 [0xBF] = X86_OP_ENTRYwr(MOV, LoBits,v, I,v), 1749 1750 [0xC8] = X86_OP_ENTRYrr(ENTER, I,w, I,b), 1751 [0xC9] = X86_OP_ENTRY1(LEAVE, A,d64), 1752 [0xCA] = X86_OP_ENTRYr(RETF, I,w), 1753 [0xCB] = X86_OP_ENTRY0(RETF), 1754 [0xCC] = X86_OP_ENTRY0(INT3), 1755 [0xCD] = X86_OP_ENTRYr(INT, I,b, chk(vm86_iopl)), 1756 [0xCE] = X86_OP_ENTRY0(INTO), 1757 [0xCF] = X86_OP_ENTRY0(IRET, chk(vm86_iopl) svm(IRET)), 1758 1759 [0xE8] = X86_OP_ENTRYr(CALL, J,z_f64), 1760 [0xE9] = X86_OP_ENTRYr(JMP, J,z_f64), 1761 [0xEA] = X86_OP_ENTRYrr(JMPF, I_unsigned,p, I_unsigned,w, chk(i64)), 1762 [0xEB] = X86_OP_ENTRYr(JMP, J,b), 1763 [0xEC] = X86_OP_ENTRYwr(IN, 0,b, 2,w), /* AL, DX */ 1764 [0xED] = X86_OP_ENTRYwr(IN, 0,z, 2,w), /* AX/EAX, DX */ 1765 [0xEE] = X86_OP_ENTRYrr(OUT, 0,b, 2,w), /* DX, AL */ 1766 [0xEF] = X86_OP_ENTRYrr(OUT, 0,z, 2,w), /* DX, AX/EAX */ 1767 1768 [0xF8] = X86_OP_ENTRY0(CLC), 1769 [0xF9] = X86_OP_ENTRY0(STC), 1770 [0xFA] = X86_OP_ENTRY0(CLI, chk(iopl)), 1771 [0xFB] = X86_OP_ENTRY0(STI, chk(iopl)), 1772 [0xFC] = X86_OP_ENTRY0(CLD), 1773 [0xFD] = X86_OP_ENTRY0(STD), 1774 [0xFE] = X86_OP_GROUP1(group4_5, E,b), 1775 [0xFF] = X86_OP_GROUP1(group4_5, E,v), 1776}; 1777 1778#undef mmx 1779#undef vex1 1780#undef vex2 1781#undef vex3 1782#undef vex4 1783#undef vex4_unal 1784#undef vex5 1785#undef vex6 1786#undef vex7 1787#undef vex8 1788#undef vex11 1789#undef vex12 1790#undef vex13 1791 1792/* 1793 * Decode the fixed part of the opcode and place the last 1794 * in b. 1795 */ 1796static void decode_root(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b) 1797{ 1798 *entry = opcodes_root[*b]; 1799} 1800 1801 1802static int decode_modrm(DisasContext *s, CPUX86State *env, 1803 X86DecodedInsn *decode, X86DecodedOp *op) 1804{ 1805 int modrm = get_modrm(s, env); 1806 if ((modrm >> 6) == 3) { 1807 op->n = (modrm & 7); 1808 if (op->unit != X86_OP_MMX) { 1809 op->n |= REX_B(s); 1810 } 1811 } else { 1812 op->has_ea = true; 1813 op->n = -1; 1814 decode->mem = gen_lea_modrm_0(env, s, modrm, 1815 decode->e.vex_class == 12); 1816 } 1817 return modrm; 1818} 1819 1820static bool decode_op_size(DisasContext *s, X86OpEntry *e, X86OpSize size, MemOp *ot) 1821{ 1822 switch (size) { 1823 case X86_SIZE_b: /* byte */ 1824 *ot = MO_8; 1825 return true; 1826 1827 case X86_SIZE_d: /* 32-bit */ 1828 case X86_SIZE_ss: /* SSE/AVX scalar single precision */ 1829 *ot = MO_32; 1830 return true; 1831 1832 case X86_SIZE_p: /* Far pointer, return offset size */ 1833 case X86_SIZE_s: /* Descriptor, return offset size */ 1834 case X86_SIZE_v: /* 16/32/64-bit, based on operand size */ 1835 *ot = s->dflag; 1836 return true; 1837 1838 case X86_SIZE_pi: /* MMX */ 1839 case X86_SIZE_q: /* 64-bit */ 1840 case X86_SIZE_sd: /* SSE/AVX scalar double precision */ 1841 *ot = MO_64; 1842 return true; 1843 1844 case X86_SIZE_w: /* 16-bit */ 1845 *ot = MO_16; 1846 return true; 1847 1848 case X86_SIZE_y: /* 32/64-bit, based on operand size */ 1849 *ot = s->dflag == MO_16 ? MO_32 : s->dflag; 1850 return true; 1851 1852 case X86_SIZE_y_d64: /* Full (not 16-bit) register access */ 1853 *ot = CODE64(s) ? MO_64 : MO_32; 1854 return true; 1855 1856 case X86_SIZE_z: /* 16-bit for 16-bit operand size, else 32-bit */ 1857 *ot = s->dflag == MO_16 ? MO_16 : MO_32; 1858 return true; 1859 1860 case X86_SIZE_z_f64: /* 32-bit for 32-bit operand size or 64-bit mode, else 16-bit */ 1861 *ot = !CODE64(s) && s->dflag == MO_16 ? MO_16 : MO_32; 1862 return true; 1863 1864 case X86_SIZE_dq: /* SSE/AVX 128-bit */ 1865 if (e->special == X86_SPECIAL_MMX && 1866 !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { 1867 *ot = MO_64; 1868 return true; 1869 } 1870 if (s->vex_l && e->s0 != X86_SIZE_qq && e->s1 != X86_SIZE_qq) { 1871 return false; 1872 } 1873 *ot = MO_128; 1874 return true; 1875 1876 case X86_SIZE_qq: /* AVX 256-bit */ 1877 if (!s->vex_l) { 1878 return false; 1879 } 1880 *ot = MO_256; 1881 return true; 1882 1883 case X86_SIZE_x: /* 128/256-bit, based on operand size */ 1884 if (e->special == X86_SPECIAL_MMX && 1885 !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { 1886 *ot = MO_64; 1887 return true; 1888 } 1889 /* fall through */ 1890 case X86_SIZE_ps: /* SSE/AVX packed single precision */ 1891 case X86_SIZE_pd: /* SSE/AVX packed double precision */ 1892 *ot = s->vex_l ? MO_256 : MO_128; 1893 return true; 1894 1895 case X86_SIZE_xh: /* SSE/AVX packed half register */ 1896 *ot = s->vex_l ? MO_128 : MO_64; 1897 return true; 1898 1899 case X86_SIZE_d64: /* Default to 64-bit in 64-bit mode */ 1900 *ot = CODE64(s) && s->dflag == MO_32 ? MO_64 : s->dflag; 1901 return true; 1902 1903 case X86_SIZE_f64: /* Ignore size override prefix in 64-bit mode */ 1904 *ot = CODE64(s) ? MO_64 : s->dflag; 1905 return true; 1906 1907 default: 1908 *ot = -1; 1909 return true; 1910 } 1911} 1912 1913static bool decode_op(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode, 1914 X86DecodedOp *op, X86OpType type, int b) 1915{ 1916 int modrm; 1917 1918 switch (type) { 1919 case X86_TYPE_None: /* Implicit or absent */ 1920 case X86_TYPE_A: /* Implicit */ 1921 case X86_TYPE_F: /* EFLAGS/RFLAGS */ 1922 case X86_TYPE_X: /* string source */ 1923 case X86_TYPE_Y: /* string destination */ 1924 break; 1925 1926 case X86_TYPE_B: /* VEX.vvvv selects a GPR */ 1927 op->unit = X86_OP_INT; 1928 op->n = s->vex_v; 1929 break; 1930 1931 case X86_TYPE_C: /* REG in the modrm byte selects a control register */ 1932 op->unit = X86_OP_CR; 1933 op->n = ((get_modrm(s, env) >> 3) & 7) | REX_R(s); 1934 if (op->n == 0 && (s->prefix & PREFIX_LOCK) && 1935 (s->cpuid_ext3_features & CPUID_EXT3_CR8LEG)) { 1936 op->n = 8; 1937 s->prefix &= ~PREFIX_LOCK; 1938 } 1939 if (op->n != 0 && op->n != 2 && op->n != 3 && op->n != 4 && op->n != 8) { 1940 return false; 1941 } 1942 if (decode->e.intercept) { 1943 decode->e.intercept += op->n; 1944 } 1945 break; 1946 1947 case X86_TYPE_D: /* REG in the modrm byte selects a debug register */ 1948 op->unit = X86_OP_DR; 1949 op->n = ((get_modrm(s, env) >> 3) & 7) | REX_R(s); 1950 if (op->n >= 8) { 1951 /* 1952 * illegal opcode. The DR4 and DR5 case is checked in the generated 1953 * code instead, to save on hflags bits. 1954 */ 1955 return false; 1956 } 1957 if (decode->e.intercept) { 1958 decode->e.intercept += op->n; 1959 } 1960 break; 1961 1962 case X86_TYPE_G: /* REG in the modrm byte selects a GPR */ 1963 op->unit = X86_OP_INT; 1964 goto get_reg; 1965 1966 case X86_TYPE_S: /* reg selects a segment register */ 1967 op->unit = X86_OP_SEG; 1968 goto get_reg; 1969 1970 case X86_TYPE_P: 1971 op->unit = X86_OP_MMX; 1972 goto get_reg; 1973 1974 case X86_TYPE_V: /* reg in the modrm byte selects an XMM/YMM register */ 1975 if (decode->e.special == X86_SPECIAL_MMX && 1976 !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { 1977 op->unit = X86_OP_MMX; 1978 } else { 1979 op->unit = X86_OP_SSE; 1980 } 1981 get_reg: 1982 op->n = ((get_modrm(s, env) >> 3) & 7); 1983 if (op->unit != X86_OP_MMX) { 1984 op->n |= REX_R(s); 1985 } 1986 break; 1987 1988 case X86_TYPE_E: /* ALU modrm operand */ 1989 op->unit = X86_OP_INT; 1990 goto get_modrm; 1991 1992 case X86_TYPE_Q: /* MMX modrm operand */ 1993 op->unit = X86_OP_MMX; 1994 goto get_modrm; 1995 1996 case X86_TYPE_W: /* XMM/YMM modrm operand */ 1997 if (decode->e.special == X86_SPECIAL_MMX && 1998 !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { 1999 op->unit = X86_OP_MMX; 2000 } else { 2001 op->unit = X86_OP_SSE; 2002 } 2003 goto get_modrm; 2004 2005 case X86_TYPE_N: /* R/M in the modrm byte selects an MMX register */ 2006 op->unit = X86_OP_MMX; 2007 goto get_modrm_reg; 2008 2009 case X86_TYPE_U: /* R/M in the modrm byte selects an XMM/YMM register */ 2010 if (decode->e.special == X86_SPECIAL_MMX && 2011 !(s->prefix & (PREFIX_DATA | PREFIX_REPZ | PREFIX_REPNZ))) { 2012 op->unit = X86_OP_MMX; 2013 } else { 2014 op->unit = X86_OP_SSE; 2015 } 2016 goto get_modrm_reg; 2017 2018 case X86_TYPE_R: /* R/M in the modrm byte selects a register */ 2019 op->unit = X86_OP_INT; 2020 get_modrm_reg: 2021 modrm = get_modrm(s, env); 2022 if ((modrm >> 6) != 3) { 2023 return false; 2024 } 2025 goto get_modrm; 2026 2027 case X86_TYPE_WM: /* modrm byte selects an XMM/YMM memory operand */ 2028 op->unit = X86_OP_SSE; 2029 goto get_modrm_mem; 2030 2031 case X86_TYPE_EM: /* modrm byte selects an ALU memory operand */ 2032 op->unit = X86_OP_INT; 2033 /* fall through */ 2034 case X86_TYPE_M: /* modrm byte selects a memory operand */ 2035 get_modrm_mem: 2036 modrm = get_modrm(s, env); 2037 if ((modrm >> 6) == 3) { 2038 return false; 2039 } 2040 /* fall through */ 2041 case X86_TYPE_nop: /* modrm operand decoded but not fetched */ 2042 get_modrm: 2043 decode_modrm(s, env, decode, op); 2044 break; 2045 2046 case X86_TYPE_O: /* Absolute address encoded in the instruction */ 2047 op->unit = X86_OP_INT; 2048 op->has_ea = true; 2049 op->n = -1; 2050 decode->mem = (AddressParts) { 2051 .def_seg = R_DS, 2052 .base = -1, 2053 .index = -1, 2054 .disp = insn_get_addr(env, s, s->aflag) 2055 }; 2056 break; 2057 2058 case X86_TYPE_H: /* For AVX, VEX.vvvv selects an XMM/YMM register */ 2059 if ((s->prefix & PREFIX_VEX)) { 2060 op->unit = X86_OP_SSE; 2061 op->n = s->vex_v; 2062 break; 2063 } 2064 if (op == &decode->op[0]) { 2065 /* shifts place the destination in VEX.vvvv, use modrm */ 2066 return decode_op(s, env, decode, op, decode->e.op1, b); 2067 } else { 2068 return decode_op(s, env, decode, op, decode->e.op0, b); 2069 } 2070 2071 case X86_TYPE_I: /* Immediate */ 2072 case X86_TYPE_J: /* Relative offset for a jump */ 2073 op->unit = X86_OP_IMM; 2074 decode->immediate = op->imm = insn_get_signed(env, s, op->ot); 2075 break; 2076 2077 case X86_TYPE_I_unsigned: /* Immediate */ 2078 op->unit = X86_OP_IMM; 2079 decode->immediate = op->imm = insn_get(env, s, op->ot); 2080 break; 2081 2082 case X86_TYPE_L: /* The upper 4 bits of the immediate select a 128-bit register */ 2083 op->n = insn_get(env, s, op->ot) >> 4; 2084 break; 2085 2086 case X86_TYPE_2op: 2087 *op = decode->op[0]; 2088 break; 2089 2090 case X86_TYPE_LoBits: 2091 op->n = (b & 7) | REX_B(s); 2092 op->unit = X86_OP_INT; 2093 break; 2094 2095 case X86_TYPE_0 ... X86_TYPE_7: 2096 op->n = type - X86_TYPE_0; 2097 op->unit = X86_OP_INT; 2098 break; 2099 2100 case X86_TYPE_ES ... X86_TYPE_GS: 2101 op->n = type - X86_TYPE_ES; 2102 op->unit = X86_OP_SEG; 2103 break; 2104 } 2105 2106 return true; 2107} 2108 2109static bool validate_sse_prefix(DisasContext *s, X86OpEntry *e) 2110{ 2111 uint16_t sse_prefixes; 2112 2113 if (!e->valid_prefix) { 2114 return true; 2115 } 2116 if (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) { 2117 /* In SSE instructions, 0xF3 and 0xF2 cancel 0x66. */ 2118 s->prefix &= ~PREFIX_DATA; 2119 } 2120 2121 /* Now, either zero or one bit is set in sse_prefixes. */ 2122 sse_prefixes = s->prefix & (PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA); 2123 return e->valid_prefix & (1 << sse_prefixes); 2124} 2125 2126static bool decode_insn(DisasContext *s, CPUX86State *env, X86DecodeFunc decode_func, 2127 X86DecodedInsn *decode) 2128{ 2129 X86OpEntry *e = &decode->e; 2130 2131 decode_func(s, env, e, &decode->b); 2132 while (e->is_decode) { 2133 e->is_decode = false; 2134 e->decode(s, env, e, &decode->b); 2135 } 2136 2137 if (!validate_sse_prefix(s, e)) { 2138 return false; 2139 } 2140 2141 /* First compute size of operands in order to initialize s->rip_offset. */ 2142 if (e->op0 != X86_TYPE_None) { 2143 if (!decode_op_size(s, e, e->s0, &decode->op[0].ot)) { 2144 return false; 2145 } 2146 if (e->op0 == X86_TYPE_I) { 2147 s->rip_offset += 1 << decode->op[0].ot; 2148 } 2149 } 2150 if (e->op1 != X86_TYPE_None) { 2151 if (!decode_op_size(s, e, e->s1, &decode->op[1].ot)) { 2152 return false; 2153 } 2154 if (e->op1 == X86_TYPE_I) { 2155 s->rip_offset += 1 << decode->op[1].ot; 2156 } 2157 } 2158 if (e->op2 != X86_TYPE_None) { 2159 if (!decode_op_size(s, e, e->s2, &decode->op[2].ot)) { 2160 return false; 2161 } 2162 if (e->op2 == X86_TYPE_I) { 2163 s->rip_offset += 1 << decode->op[2].ot; 2164 } 2165 } 2166 if (e->op3 != X86_TYPE_None) { 2167 /* 2168 * A couple instructions actually use the extra immediate byte for an Lx 2169 * register operand; those are handled in the gen_* functions as one off. 2170 */ 2171 assert(e->op3 == X86_TYPE_I && e->s3 == X86_SIZE_b); 2172 s->rip_offset += 1; 2173 } 2174 2175 if (e->op0 != X86_TYPE_None && 2176 !decode_op(s, env, decode, &decode->op[0], e->op0, decode->b)) { 2177 return false; 2178 } 2179 2180 if (e->op1 != X86_TYPE_None && 2181 !decode_op(s, env, decode, &decode->op[1], e->op1, decode->b)) { 2182 return false; 2183 } 2184 2185 if (e->op2 != X86_TYPE_None && 2186 !decode_op(s, env, decode, &decode->op[2], e->op2, decode->b)) { 2187 return false; 2188 } 2189 2190 if (e->op3 != X86_TYPE_None) { 2191 decode->immediate = insn_get_signed(env, s, MO_8); 2192 } 2193 2194 return true; 2195} 2196 2197static bool has_cpuid_feature(DisasContext *s, X86CPUIDFeature cpuid) 2198{ 2199 switch (cpuid) { 2200 case X86_FEAT_None: 2201 return true; 2202 case X86_FEAT_CMOV: 2203 return (s->cpuid_features & CPUID_CMOV); 2204 case X86_FEAT_CLFLUSH: 2205 return (s->cpuid_features & CPUID_CLFLUSH); 2206 case X86_FEAT_FXSR: 2207 return (s->cpuid_features & CPUID_FXSR); 2208 case X86_FEAT_F16C: 2209 return (s->cpuid_ext_features & CPUID_EXT_F16C); 2210 case X86_FEAT_FMA: 2211 return (s->cpuid_ext_features & CPUID_EXT_FMA); 2212 case X86_FEAT_MOVBE: 2213 return (s->cpuid_ext_features & CPUID_EXT_MOVBE); 2214 case X86_FEAT_PCLMULQDQ: 2215 return (s->cpuid_ext_features & CPUID_EXT_PCLMULQDQ); 2216 case X86_FEAT_POPCNT: 2217 return (s->cpuid_ext_features & CPUID_EXT_POPCNT); 2218 case X86_FEAT_SSE: 2219 return (s->cpuid_features & CPUID_SSE); 2220 case X86_FEAT_SSE2: 2221 return (s->cpuid_features & CPUID_SSE2); 2222 case X86_FEAT_SSE3: 2223 return (s->cpuid_ext_features & CPUID_EXT_SSE3); 2224 case X86_FEAT_SSSE3: 2225 return (s->cpuid_ext_features & CPUID_EXT_SSSE3); 2226 case X86_FEAT_SSE41: 2227 return (s->cpuid_ext_features & CPUID_EXT_SSE41); 2228 case X86_FEAT_SSE42: 2229 return (s->cpuid_ext_features & CPUID_EXT_SSE42); 2230 case X86_FEAT_AES: 2231 if (!(s->cpuid_ext_features & CPUID_EXT_AES)) { 2232 return false; 2233 } else if (!(s->prefix & PREFIX_VEX)) { 2234 return true; 2235 } else if (!(s->cpuid_ext_features & CPUID_EXT_AVX)) { 2236 return false; 2237 } else { 2238 return !s->vex_l || (s->cpuid_7_0_ecx_features & CPUID_7_0_ECX_VAES); 2239 } 2240 2241 case X86_FEAT_AVX: 2242 return (s->cpuid_ext_features & CPUID_EXT_AVX); 2243 case X86_FEAT_XSAVE: 2244 return (s->cpuid_ext_features & CPUID_EXT_XSAVE); 2245 2246 case X86_FEAT_3DNOW: 2247 return (s->cpuid_ext2_features & CPUID_EXT2_3DNOW); 2248 case X86_FEAT_SSE4A: 2249 return (s->cpuid_ext3_features & CPUID_EXT3_SSE4A); 2250 2251 case X86_FEAT_ADX: 2252 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_ADX); 2253 case X86_FEAT_BMI1: 2254 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1); 2255 case X86_FEAT_BMI2: 2256 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI2); 2257 case X86_FEAT_AVX2: 2258 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_AVX2); 2259 case X86_FEAT_CLFLUSHOPT: 2260 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_CLFLUSHOPT); 2261 case X86_FEAT_CLWB: 2262 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_CLWB); 2263 case X86_FEAT_FSGSBASE: 2264 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_FSGSBASE); 2265 case X86_FEAT_SHA_NI: 2266 return (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_SHA_NI); 2267 2268 case X86_FEAT_CMPCCXADD: 2269 return (s->cpuid_7_1_eax_features & CPUID_7_1_EAX_CMPCCXADD); 2270 2271 case X86_FEAT_XSAVEOPT: 2272 return (s->cpuid_xsave_features & CPUID_XSAVE_XSAVEOPT); 2273 } 2274 g_assert_not_reached(); 2275} 2276 2277static bool validate_vex(DisasContext *s, X86DecodedInsn *decode) 2278{ 2279 X86OpEntry *e = &decode->e; 2280 2281 switch (e->vex_special) { 2282 case X86_VEX_REPScalar: 2283 /* 2284 * Instructions which differ between 00/66 and F2/F3 in the 2285 * exception classification and the size of the memory operand. 2286 */ 2287 assert(e->vex_class == 1 || e->vex_class == 2 || e->vex_class == 4); 2288 if (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) { 2289 e->vex_class = e->vex_class < 4 ? 3 : 5; 2290 if (s->vex_l) { 2291 goto illegal; 2292 } 2293 assert(decode->e.s2 == X86_SIZE_x); 2294 if (decode->op[2].has_ea) { 2295 decode->op[2].ot = s->prefix & PREFIX_REPZ ? MO_32 : MO_64; 2296 } 2297 } 2298 break; 2299 2300 case X86_VEX_SSEUnaligned: 2301 /* handled in sse_needs_alignment. */ 2302 break; 2303 2304 case X86_VEX_AVX2_256: 2305 if ((s->prefix & PREFIX_VEX) && s->vex_l && !has_cpuid_feature(s, X86_FEAT_AVX2)) { 2306 goto illegal; 2307 } 2308 } 2309 2310 switch (e->vex_class) { 2311 case 0: 2312 if (s->prefix & PREFIX_VEX) { 2313 goto illegal; 2314 } 2315 return true; 2316 case 1: 2317 case 2: 2318 case 3: 2319 case 4: 2320 case 5: 2321 case 7: 2322 if (s->prefix & PREFIX_VEX) { 2323 if (!(s->flags & HF_AVX_EN_MASK)) { 2324 goto illegal; 2325 } 2326 } else if (e->special != X86_SPECIAL_MMX || 2327 (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA))) { 2328 if (!(s->flags & HF_OSFXSR_MASK)) { 2329 goto illegal; 2330 } 2331 } 2332 break; 2333 case 12: 2334 /* Must have a VSIB byte and no address prefix. */ 2335 assert(s->has_modrm); 2336 if ((s->modrm & 7) != 4 || s->aflag == MO_16) { 2337 goto illegal; 2338 } 2339 2340 /* Check no overlap between registers. */ 2341 if (!decode->op[0].has_ea && 2342 (decode->op[0].n == decode->mem.index || decode->op[0].n == decode->op[1].n)) { 2343 goto illegal; 2344 } 2345 assert(!decode->op[1].has_ea); 2346 if (decode->op[1].n == decode->mem.index) { 2347 goto illegal; 2348 } 2349 if (!decode->op[2].has_ea && 2350 (decode->op[2].n == decode->mem.index || decode->op[2].n == decode->op[1].n)) { 2351 goto illegal; 2352 } 2353 /* fall through */ 2354 case 6: 2355 case 11: 2356 if (!(s->prefix & PREFIX_VEX)) { 2357 goto illegal; 2358 } 2359 if (!(s->flags & HF_AVX_EN_MASK)) { 2360 goto illegal; 2361 } 2362 break; 2363 case 8: 2364 /* Non-VEX case handled in decode_0F77. */ 2365 assert(s->prefix & PREFIX_VEX); 2366 if (!(s->flags & HF_AVX_EN_MASK)) { 2367 goto illegal; 2368 } 2369 break; 2370 case 13: 2371 if (!(s->prefix & PREFIX_VEX)) { 2372 goto illegal; 2373 } 2374 if (s->vex_l) { 2375 goto illegal; 2376 } 2377 /* All integer instructions use VEX.vvvv, so exit. */ 2378 return true; 2379 } 2380 2381 if (s->vex_v != 0 && 2382 e->op0 != X86_TYPE_H && e->op0 != X86_TYPE_B && 2383 e->op1 != X86_TYPE_H && e->op1 != X86_TYPE_B && 2384 e->op2 != X86_TYPE_H && e->op2 != X86_TYPE_B) { 2385 goto illegal; 2386 } 2387 2388 if (s->flags & HF_TS_MASK) { 2389 goto nm_exception; 2390 } 2391 if (s->flags & HF_EM_MASK) { 2392 goto illegal; 2393 } 2394 2395 if (e->check) { 2396 if (e->check & X86_CHECK_VEX128) { 2397 if (s->vex_l) { 2398 goto illegal; 2399 } 2400 } 2401 if (e->check & X86_CHECK_W0) { 2402 if (s->vex_w) { 2403 goto illegal; 2404 } 2405 } 2406 if (e->check & X86_CHECK_W1) { 2407 if (!s->vex_w) { 2408 goto illegal; 2409 } 2410 } 2411 } 2412 return true; 2413 2414nm_exception: 2415 gen_NM_exception(s); 2416 return false; 2417illegal: 2418 gen_illegal_opcode(s); 2419 return false; 2420} 2421 2422/* 2423 * Convert one instruction. s->base.is_jmp is set if the translation must 2424 * be stopped. 2425 */ 2426static void disas_insn(DisasContext *s, CPUState *cpu) 2427{ 2428 CPUX86State *env = cpu_env(cpu); 2429 X86DecodedInsn decode; 2430 X86DecodeFunc decode_func = decode_root; 2431 uint8_t cc_live, b; 2432 2433 s->pc = s->base.pc_next; 2434 s->override = -1; 2435 s->popl_esp_hack = 0; 2436#ifdef TARGET_X86_64 2437 s->rex_r = 0; 2438 s->rex_x = 0; 2439 s->rex_b = 0; 2440#endif 2441 s->rip_offset = 0; /* for relative ip address */ 2442 s->vex_l = 0; 2443 s->vex_v = 0; 2444 s->vex_w = false; 2445 s->has_modrm = false; 2446 s->prefix = 0; 2447 2448 next_byte: 2449 b = x86_ldub_code(env, s); 2450 2451 /* Collect prefixes. */ 2452 switch (b) { 2453 case 0xf3: 2454 s->prefix |= PREFIX_REPZ; 2455 s->prefix &= ~PREFIX_REPNZ; 2456 goto next_byte; 2457 case 0xf2: 2458 s->prefix |= PREFIX_REPNZ; 2459 s->prefix &= ~PREFIX_REPZ; 2460 goto next_byte; 2461 case 0xf0: 2462 s->prefix |= PREFIX_LOCK; 2463 goto next_byte; 2464 case 0x2e: 2465 s->override = R_CS; 2466 goto next_byte; 2467 case 0x36: 2468 s->override = R_SS; 2469 goto next_byte; 2470 case 0x3e: 2471 s->override = R_DS; 2472 goto next_byte; 2473 case 0x26: 2474 s->override = R_ES; 2475 goto next_byte; 2476 case 0x64: 2477 s->override = R_FS; 2478 goto next_byte; 2479 case 0x65: 2480 s->override = R_GS; 2481 goto next_byte; 2482 case 0x66: 2483 s->prefix |= PREFIX_DATA; 2484 goto next_byte; 2485 case 0x67: 2486 s->prefix |= PREFIX_ADR; 2487 goto next_byte; 2488#ifdef TARGET_X86_64 2489 case 0x40 ... 0x4f: 2490 if (CODE64(s)) { 2491 /* REX prefix */ 2492 s->prefix |= PREFIX_REX; 2493 s->vex_w = (b >> 3) & 1; 2494 s->rex_r = (b & 0x4) << 1; 2495 s->rex_x = (b & 0x2) << 2; 2496 s->rex_b = (b & 0x1) << 3; 2497 goto next_byte; 2498 } 2499 break; 2500#endif 2501 case 0xc5: /* 2-byte VEX */ 2502 case 0xc4: /* 3-byte VEX */ 2503 /* 2504 * VEX prefixes cannot be used except in 32-bit mode. 2505 * Otherwise the instruction is LES or LDS. 2506 */ 2507 if (CODE32(s) && !VM86(s)) { 2508 static const int pp_prefix[4] = { 2509 0, PREFIX_DATA, PREFIX_REPZ, PREFIX_REPNZ 2510 }; 2511 int vex3, vex2 = x86_ldub_code(env, s); 2512 2513 if (!CODE64(s) && (vex2 & 0xc0) != 0xc0) { 2514 /* 2515 * 4.1.4.6: In 32-bit mode, bits [7:6] must be 11b, 2516 * otherwise the instruction is LES or LDS. 2517 */ 2518 s->pc--; /* rewind the advance_pc() x86_ldub_code() did */ 2519 break; 2520 } 2521 2522 /* 4.1.1-4.1.3: No preceding lock, 66, f2, f3, or rex prefixes. */ 2523 if (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ 2524 | PREFIX_LOCK | PREFIX_DATA | PREFIX_REX)) { 2525 goto illegal_op; 2526 } 2527#ifdef TARGET_X86_64 2528 s->rex_r = (~vex2 >> 4) & 8; 2529#endif 2530 if (b == 0xc5) { 2531 /* 2-byte VEX prefix: RVVVVlpp, implied 0f leading opcode byte */ 2532 vex3 = vex2; 2533 decode_func = decode_0F; 2534 } else { 2535 /* 3-byte VEX prefix: RXBmmmmm wVVVVlpp */ 2536 vex3 = x86_ldub_code(env, s); 2537#ifdef TARGET_X86_64 2538 s->rex_x = (~vex2 >> 3) & 8; 2539 s->rex_b = (~vex2 >> 2) & 8; 2540#endif 2541 s->vex_w = (vex3 >> 7) & 1; 2542 switch (vex2 & 0x1f) { 2543 case 0x01: /* Implied 0f leading opcode bytes. */ 2544 decode_func = decode_0F; 2545 break; 2546 case 0x02: /* Implied 0f 38 leading opcode bytes. */ 2547 decode_func = decode_0F38; 2548 break; 2549 case 0x03: /* Implied 0f 3a leading opcode bytes. */ 2550 decode_func = decode_0F3A; 2551 break; 2552 default: /* Reserved for future use. */ 2553 goto unknown_op; 2554 } 2555 } 2556 s->vex_v = (~vex3 >> 3) & 0xf; 2557 s->vex_l = (vex3 >> 2) & 1; 2558 s->prefix |= pp_prefix[vex3 & 3] | PREFIX_VEX; 2559 } 2560 break; 2561 default: 2562 break; 2563 } 2564 2565 /* Post-process prefixes. */ 2566 if (CODE64(s)) { 2567 /* 2568 * In 64-bit mode, the default data size is 32-bit. Select 64-bit 2569 * data with rex_w, and 16-bit data with 0x66; rex_w takes precedence 2570 * over 0x66 if both are present. 2571 */ 2572 s->dflag = (REX_W(s) ? MO_64 : s->prefix & PREFIX_DATA ? MO_16 : MO_32); 2573 /* In 64-bit mode, 0x67 selects 32-bit addressing. */ 2574 s->aflag = (s->prefix & PREFIX_ADR ? MO_32 : MO_64); 2575 } else { 2576 /* In 16/32-bit mode, 0x66 selects the opposite data size. */ 2577 if (CODE32(s) ^ ((s->prefix & PREFIX_DATA) != 0)) { 2578 s->dflag = MO_32; 2579 } else { 2580 s->dflag = MO_16; 2581 } 2582 /* In 16/32-bit mode, 0x67 selects the opposite addressing. */ 2583 if (CODE32(s) ^ ((s->prefix & PREFIX_ADR) != 0)) { 2584 s->aflag = MO_32; 2585 } else { 2586 s->aflag = MO_16; 2587 } 2588 } 2589 2590 /* Go back to old decoder for unconverted opcodes. */ 2591 if (!(s->prefix & PREFIX_VEX)) { 2592 if ((b & ~7) == 0xd8) { 2593 if (!disas_insn_x87(s, cpu, b)) { 2594 goto unknown_op; 2595 } 2596 return; 2597 } 2598 2599 if (b == 0x0f) { 2600 b = x86_ldub_code(env, s); 2601 switch (b) { 2602 case 0x00 ... 0x01: /* mostly privileged instructions */ 2603 case 0x1a ... 0x1b: /* MPX */ 2604 case 0xa3: /* bt */ 2605 case 0xab: /* bts */ 2606 case 0xb3: /* btr */ 2607 case 0xba ... 0xbb: /* grp8, btc */ 2608 case 0xc7: /* grp9 */ 2609 disas_insn_old(s, cpu, b + 0x100); 2610 return; 2611 default: 2612 decode_func = do_decode_0F; 2613 break; 2614 } 2615 } 2616 } 2617 2618 memset(&decode, 0, sizeof(decode)); 2619 decode.cc_op = -1; 2620 decode.b = b; 2621 if (!decode_insn(s, env, decode_func, &decode)) { 2622 goto illegal_op; 2623 } 2624 if (!decode.e.gen) { 2625 goto unknown_op; 2626 } 2627 2628 if (!has_cpuid_feature(s, decode.e.cpuid)) { 2629 goto illegal_op; 2630 } 2631 2632 /* Checks that result in #UD come first. */ 2633 if (decode.e.check) { 2634 if (CODE64(s)) { 2635 if (decode.e.check & X86_CHECK_i64) { 2636 goto illegal_op; 2637 } 2638 if ((decode.e.check & X86_CHECK_i64_amd) && env->cpuid_vendor1 != CPUID_VENDOR_INTEL_1) { 2639 goto illegal_op; 2640 } 2641 } else { 2642 if (decode.e.check & X86_CHECK_o64) { 2643 goto illegal_op; 2644 } 2645 if ((decode.e.check & X86_CHECK_o64_intel) && env->cpuid_vendor1 == CPUID_VENDOR_INTEL_1) { 2646 goto illegal_op; 2647 } 2648 } 2649 if (decode.e.check & X86_CHECK_prot_or_vm86) { 2650 if (!PE(s)) { 2651 goto illegal_op; 2652 } 2653 } 2654 if (decode.e.check & X86_CHECK_no_vm86) { 2655 if (VM86(s)) { 2656 goto illegal_op; 2657 } 2658 } 2659 } 2660 2661 switch (decode.e.special) { 2662 case X86_SPECIAL_None: 2663 break; 2664 2665 case X86_SPECIAL_Locked: 2666 if (decode.op[0].has_ea) { 2667 s->prefix |= PREFIX_LOCK; 2668 } 2669 decode.e.special = X86_SPECIAL_HasLock; 2670 /* fallthrough */ 2671 case X86_SPECIAL_HasLock: 2672 break; 2673 2674 case X86_SPECIAL_Op0_Rd: 2675 assert(decode.op[0].unit == X86_OP_INT); 2676 if (!decode.op[0].has_ea) { 2677 decode.op[0].ot = MO_32; 2678 } 2679 break; 2680 2681 case X86_SPECIAL_Op2_Ry: 2682 assert(decode.op[2].unit == X86_OP_INT); 2683 if (!decode.op[2].has_ea) { 2684 decode.op[2].ot = s->dflag == MO_16 ? MO_32 : s->dflag; 2685 } 2686 break; 2687 2688 case X86_SPECIAL_AVXExtMov: 2689 if (!decode.op[2].has_ea) { 2690 decode.op[2].ot = s->vex_l ? MO_256 : MO_128; 2691 } else if (s->vex_l) { 2692 decode.op[2].ot++; 2693 } 2694 break; 2695 2696 case X86_SPECIAL_SExtT0: 2697 case X86_SPECIAL_ZExtT0: 2698 /* Handled in gen_load. */ 2699 assert(decode.op[1].unit == X86_OP_INT); 2700 break; 2701 2702 case X86_SPECIAL_Op0_Mw: 2703 assert(decode.op[0].unit == X86_OP_INT); 2704 if (decode.op[0].has_ea) { 2705 decode.op[0].ot = MO_16; 2706 } 2707 break; 2708 2709 default: 2710 break; 2711 } 2712 2713 if (s->prefix & PREFIX_LOCK) { 2714 if (decode.e.special != X86_SPECIAL_HasLock || !decode.op[0].has_ea) { 2715 goto illegal_op; 2716 } 2717 } 2718 2719 if (!validate_vex(s, &decode)) { 2720 return; 2721 } 2722 2723 /* 2724 * Checks that result in #GP or VMEXIT come second. Intercepts are 2725 * generally checked after non-memory exceptions (i.e. after all 2726 * exceptions if there is no memory operand). Exceptions are 2727 * vm86 checks (INTn, IRET, PUSHF/POPF), RSM and XSETBV (!). 2728 * 2729 * XSETBV will check for CPL0 in the gen_* function instead of using chk(). 2730 */ 2731 if (decode.e.check & X86_CHECK_cpl0) { 2732 if (CPL(s) != 0) { 2733 goto gp_fault; 2734 } 2735 } 2736 if (decode.e.has_intercept && unlikely(GUEST(s))) { 2737 gen_helper_svm_check_intercept(tcg_env, 2738 tcg_constant_i32(decode.e.intercept)); 2739 } 2740 if (decode.e.check) { 2741 if ((decode.e.check & X86_CHECK_smm) && !(s->flags & HF_SMM_MASK)) { 2742 goto illegal_op; 2743 } 2744 if ((decode.e.check & X86_CHECK_vm86_iopl) && VM86(s)) { 2745 if (IOPL(s) < 3) { 2746 goto gp_fault; 2747 } 2748 } else if (decode.e.check & X86_CHECK_cpl_iopl) { 2749 if (IOPL(s) < CPL(s)) { 2750 goto gp_fault; 2751 } 2752 } 2753 } 2754 2755 if (decode.e.special == X86_SPECIAL_MMX && 2756 !(s->prefix & (PREFIX_REPZ | PREFIX_REPNZ | PREFIX_DATA))) { 2757 gen_helper_enter_mmx(tcg_env); 2758 } 2759 2760 if (decode.e.special != X86_SPECIAL_NoLoadEA && 2761 (decode.op[0].has_ea || decode.op[1].has_ea || decode.op[2].has_ea)) { 2762 gen_load_ea(s, &decode.mem, decode.e.vex_class == 12); 2763 } 2764 if (s->prefix & PREFIX_LOCK) { 2765 gen_load(s, &decode, 2, s->T1); 2766 decode.e.gen(s, &decode); 2767 } else { 2768 if (decode.op[0].unit == X86_OP_MMX) { 2769 compute_mmx_offset(&decode.op[0]); 2770 } else if (decode.op[0].unit == X86_OP_SSE) { 2771 compute_xmm_offset(&decode.op[0]); 2772 } 2773 gen_load(s, &decode, 1, s->T0); 2774 gen_load(s, &decode, 2, s->T1); 2775 decode.e.gen(s, &decode); 2776 gen_writeback(s, &decode, 0, s->T0); 2777 } 2778 2779 /* 2780 * Write back flags after last memory access. Some older ALU instructions, as 2781 * well as SSE instructions, write flags in the gen_* function, but that can 2782 * cause incorrect tracking of CC_OP for instructions that write to both memory 2783 * and flags. 2784 */ 2785 if (decode.cc_op != -1) { 2786 if (decode.cc_dst) { 2787 tcg_gen_mov_tl(cpu_cc_dst, decode.cc_dst); 2788 } 2789 if (decode.cc_src) { 2790 tcg_gen_mov_tl(cpu_cc_src, decode.cc_src); 2791 } 2792 if (decode.cc_src2) { 2793 tcg_gen_mov_tl(cpu_cc_src2, decode.cc_src2); 2794 } 2795 if (decode.cc_op == CC_OP_DYNAMIC) { 2796 tcg_gen_mov_i32(cpu_cc_op, decode.cc_op_dynamic); 2797 } 2798 set_cc_op(s, decode.cc_op); 2799 cc_live = cc_op_live[decode.cc_op]; 2800 } else { 2801 cc_live = 0; 2802 } 2803 if (decode.cc_op != CC_OP_DYNAMIC) { 2804 assert(!decode.cc_op_dynamic); 2805 assert(!!decode.cc_dst == !!(cc_live & USES_CC_DST)); 2806 assert(!!decode.cc_src == !!(cc_live & USES_CC_SRC)); 2807 assert(!!decode.cc_src2 == !!(cc_live & USES_CC_SRC2)); 2808 } 2809 2810 return; 2811 gp_fault: 2812 gen_exception_gpf(s); 2813 return; 2814 illegal_op: 2815 gen_illegal_opcode(s); 2816 return; 2817 unknown_op: 2818 gen_unknown_opcode(env, s); 2819} 2820