xref: /openbmc/qemu/target/i386/tcg/translate.c (revision 988f2442)
1 /*
2  *  i386 translation
3  *
4  *  Copyright (c) 2003 Fabrice Bellard
5  *
6  * This library is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * This library is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
18  */
19 #include "qemu/osdep.h"
20 
21 #include "qemu/host-utils.h"
22 #include "cpu.h"
23 #include "disas/disas.h"
24 #include "exec/exec-all.h"
25 #include "tcg/tcg-op.h"
26 #include "tcg/tcg-op-gvec.h"
27 #include "exec/cpu_ldst.h"
28 #include "exec/translator.h"
29 #include "fpu/softfloat.h"
30 
31 #include "exec/helper-proto.h"
32 #include "exec/helper-gen.h"
33 #include "helper-tcg.h"
34 
35 #include "exec/log.h"
36 
37 #define HELPER_H "helper.h"
38 #include "exec/helper-info.c.inc"
39 #undef  HELPER_H
40 
41 
42 #define PREFIX_REPZ   0x01
43 #define PREFIX_REPNZ  0x02
44 #define PREFIX_LOCK   0x04
45 #define PREFIX_DATA   0x08
46 #define PREFIX_ADR    0x10
47 #define PREFIX_VEX    0x20
48 #define PREFIX_REX    0x40
49 
50 #ifdef TARGET_X86_64
51 # define ctztl  ctz64
52 # define clztl  clz64
53 #else
54 # define ctztl  ctz32
55 # define clztl  clz32
56 #endif
57 
58 /* For a switch indexed by MODRM, match all memory operands for a given OP.  */
59 #define CASE_MODRM_MEM_OP(OP) \
60     case (0 << 6) | (OP << 3) | 0 ... (0 << 6) | (OP << 3) | 7: \
61     case (1 << 6) | (OP << 3) | 0 ... (1 << 6) | (OP << 3) | 7: \
62     case (2 << 6) | (OP << 3) | 0 ... (2 << 6) | (OP << 3) | 7
63 
64 #define CASE_MODRM_OP(OP) \
65     case (0 << 6) | (OP << 3) | 0 ... (0 << 6) | (OP << 3) | 7: \
66     case (1 << 6) | (OP << 3) | 0 ... (1 << 6) | (OP << 3) | 7: \
67     case (2 << 6) | (OP << 3) | 0 ... (2 << 6) | (OP << 3) | 7: \
68     case (3 << 6) | (OP << 3) | 0 ... (3 << 6) | (OP << 3) | 7
69 
70 //#define MACRO_TEST   1
71 
72 /* global register indexes */
73 static TCGv cpu_cc_dst, cpu_cc_src, cpu_cc_src2;
74 static TCGv cpu_eip;
75 static TCGv_i32 cpu_cc_op;
76 static TCGv cpu_regs[CPU_NB_REGS];
77 static TCGv cpu_seg_base[6];
78 static TCGv_i64 cpu_bndl[4];
79 static TCGv_i64 cpu_bndu[4];
80 
81 typedef struct DisasContext {
82     DisasContextBase base;
83 
84     target_ulong pc;       /* pc = eip + cs_base */
85     target_ulong cs_base;  /* base of CS segment */
86     target_ulong pc_save;
87 
88     MemOp aflag;
89     MemOp dflag;
90 
91     int8_t override; /* -1 if no override, else R_CS, R_DS, etc */
92     uint8_t prefix;
93 
94     bool has_modrm;
95     uint8_t modrm;
96 
97 #ifndef CONFIG_USER_ONLY
98     uint8_t cpl;   /* code priv level */
99     uint8_t iopl;  /* i/o priv level */
100 #endif
101     uint8_t vex_l;  /* vex vector length */
102     uint8_t vex_v;  /* vex vvvv register, without 1's complement.  */
103     uint8_t popl_esp_hack; /* for correct popl with esp base handling */
104     uint8_t rip_offset; /* only used in x86_64, but left for simplicity */
105 
106 #ifdef TARGET_X86_64
107     uint8_t rex_r;
108     uint8_t rex_x;
109     uint8_t rex_b;
110 #endif
111     bool vex_w; /* used by AVX even on 32-bit processors */
112     bool jmp_opt; /* use direct block chaining for direct jumps */
113     bool repz_opt; /* optimize jumps within repz instructions */
114     bool cc_op_dirty;
115 
116     CCOp cc_op;  /* current CC operation */
117     int mem_index; /* select memory access functions */
118     uint32_t flags; /* all execution flags */
119     int cpuid_features;
120     int cpuid_ext_features;
121     int cpuid_ext2_features;
122     int cpuid_ext3_features;
123     int cpuid_7_0_ebx_features;
124     int cpuid_7_0_ecx_features;
125     int cpuid_7_1_eax_features;
126     int cpuid_xsave_features;
127 
128     /* TCG local temps */
129     TCGv cc_srcT;
130     TCGv A0;
131     TCGv T0;
132     TCGv T1;
133 
134     /* TCG local register indexes (only used inside old micro ops) */
135     TCGv tmp0;
136     TCGv tmp4;
137     TCGv_i32 tmp2_i32;
138     TCGv_i32 tmp3_i32;
139     TCGv_i64 tmp1_i64;
140 
141     sigjmp_buf jmpbuf;
142     TCGOp *prev_insn_end;
143 } DisasContext;
144 
145 #define DISAS_EOB_ONLY         DISAS_TARGET_0
146 #define DISAS_EOB_NEXT         DISAS_TARGET_1
147 #define DISAS_EOB_INHIBIT_IRQ  DISAS_TARGET_2
148 #define DISAS_JUMP             DISAS_TARGET_3
149 
150 /* The environment in which user-only runs is constrained. */
151 #ifdef CONFIG_USER_ONLY
152 #define PE(S)     true
153 #define CPL(S)    3
154 #define IOPL(S)   0
155 #define SVME(S)   false
156 #define GUEST(S)  false
157 #else
158 #define PE(S)     (((S)->flags & HF_PE_MASK) != 0)
159 #define CPL(S)    ((S)->cpl)
160 #define IOPL(S)   ((S)->iopl)
161 #define SVME(S)   (((S)->flags & HF_SVME_MASK) != 0)
162 #define GUEST(S)  (((S)->flags & HF_GUEST_MASK) != 0)
163 #endif
164 #if defined(CONFIG_USER_ONLY) && defined(TARGET_X86_64)
165 #define VM86(S)   false
166 #define CODE32(S) true
167 #define SS32(S)   true
168 #define ADDSEG(S) false
169 #else
170 #define VM86(S)   (((S)->flags & HF_VM_MASK) != 0)
171 #define CODE32(S) (((S)->flags & HF_CS32_MASK) != 0)
172 #define SS32(S)   (((S)->flags & HF_SS32_MASK) != 0)
173 #define ADDSEG(S) (((S)->flags & HF_ADDSEG_MASK) != 0)
174 #endif
175 #if !defined(TARGET_X86_64)
176 #define CODE64(S) false
177 #elif defined(CONFIG_USER_ONLY)
178 #define CODE64(S) true
179 #else
180 #define CODE64(S) (((S)->flags & HF_CS64_MASK) != 0)
181 #endif
182 #if defined(CONFIG_USER_ONLY) || defined(TARGET_X86_64)
183 #define LMA(S)    (((S)->flags & HF_LMA_MASK) != 0)
184 #else
185 #define LMA(S)    false
186 #endif
187 
188 #ifdef TARGET_X86_64
189 #define REX_PREFIX(S)  (((S)->prefix & PREFIX_REX) != 0)
190 #define REX_W(S)       ((S)->vex_w)
191 #define REX_R(S)       ((S)->rex_r + 0)
192 #define REX_X(S)       ((S)->rex_x + 0)
193 #define REX_B(S)       ((S)->rex_b + 0)
194 #else
195 #define REX_PREFIX(S)  false
196 #define REX_W(S)       false
197 #define REX_R(S)       0
198 #define REX_X(S)       0
199 #define REX_B(S)       0
200 #endif
201 
202 /*
203  * Many sysemu-only helpers are not reachable for user-only.
204  * Define stub generators here, so that we need not either sprinkle
205  * ifdefs through the translator, nor provide the helper function.
206  */
207 #define STUB_HELPER(NAME, ...) \
208     static inline void gen_helper_##NAME(__VA_ARGS__) \
209     { qemu_build_not_reached(); }
210 
211 #ifdef CONFIG_USER_ONLY
212 STUB_HELPER(clgi, TCGv_env env)
213 STUB_HELPER(flush_page, TCGv_env env, TCGv addr)
214 STUB_HELPER(hlt, TCGv_env env, TCGv_i32 pc_ofs)
215 STUB_HELPER(inb, TCGv ret, TCGv_env env, TCGv_i32 port)
216 STUB_HELPER(inw, TCGv ret, TCGv_env env, TCGv_i32 port)
217 STUB_HELPER(inl, TCGv ret, TCGv_env env, TCGv_i32 port)
218 STUB_HELPER(monitor, TCGv_env env, TCGv addr)
219 STUB_HELPER(mwait, TCGv_env env, TCGv_i32 pc_ofs)
220 STUB_HELPER(outb, TCGv_env env, TCGv_i32 port, TCGv_i32 val)
221 STUB_HELPER(outw, TCGv_env env, TCGv_i32 port, TCGv_i32 val)
222 STUB_HELPER(outl, TCGv_env env, TCGv_i32 port, TCGv_i32 val)
223 STUB_HELPER(rdmsr, TCGv_env env)
224 STUB_HELPER(read_crN, TCGv ret, TCGv_env env, TCGv_i32 reg)
225 STUB_HELPER(get_dr, TCGv ret, TCGv_env env, TCGv_i32 reg)
226 STUB_HELPER(set_dr, TCGv_env env, TCGv_i32 reg, TCGv val)
227 STUB_HELPER(stgi, TCGv_env env)
228 STUB_HELPER(svm_check_intercept, TCGv_env env, TCGv_i32 type)
229 STUB_HELPER(vmload, TCGv_env env, TCGv_i32 aflag)
230 STUB_HELPER(vmmcall, TCGv_env env)
231 STUB_HELPER(vmrun, TCGv_env env, TCGv_i32 aflag, TCGv_i32 pc_ofs)
232 STUB_HELPER(vmsave, TCGv_env env, TCGv_i32 aflag)
233 STUB_HELPER(write_crN, TCGv_env env, TCGv_i32 reg, TCGv val)
234 STUB_HELPER(wrmsr, TCGv_env env)
235 #endif
236 
237 static void gen_eob(DisasContext *s);
238 static void gen_jr(DisasContext *s);
239 static void gen_jmp_rel(DisasContext *s, MemOp ot, int diff, int tb_num);
240 static void gen_jmp_rel_csize(DisasContext *s, int diff, int tb_num);
241 static void gen_op(DisasContext *s1, int op, MemOp ot, int d);
242 static void gen_exception_gpf(DisasContext *s);
243 
244 /* i386 arith/logic operations */
245 enum {
246     OP_ADDL,
247     OP_ORL,
248     OP_ADCL,
249     OP_SBBL,
250     OP_ANDL,
251     OP_SUBL,
252     OP_XORL,
253     OP_CMPL,
254 };
255 
256 /* i386 shift ops */
257 enum {
258     OP_ROL,
259     OP_ROR,
260     OP_RCL,
261     OP_RCR,
262     OP_SHL,
263     OP_SHR,
264     OP_SHL1, /* undocumented */
265     OP_SAR = 7,
266 };
267 
268 enum {
269     JCC_O,
270     JCC_B,
271     JCC_Z,
272     JCC_BE,
273     JCC_S,
274     JCC_P,
275     JCC_L,
276     JCC_LE,
277 };
278 
279 enum {
280     /* I386 int registers */
281     OR_EAX,   /* MUST be even numbered */
282     OR_ECX,
283     OR_EDX,
284     OR_EBX,
285     OR_ESP,
286     OR_EBP,
287     OR_ESI,
288     OR_EDI,
289 
290     OR_TMP0 = 16,    /* temporary operand register */
291     OR_TMP1,
292     OR_A0, /* temporary register used when doing address evaluation */
293 };
294 
295 enum {
296     USES_CC_DST  = 1,
297     USES_CC_SRC  = 2,
298     USES_CC_SRC2 = 4,
299     USES_CC_SRCT = 8,
300 };
301 
302 /* Bit set if the global variable is live after setting CC_OP to X.  */
303 static const uint8_t cc_op_live[CC_OP_NB] = {
304     [CC_OP_DYNAMIC] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
305     [CC_OP_EFLAGS] = USES_CC_SRC,
306     [CC_OP_MULB ... CC_OP_MULQ] = USES_CC_DST | USES_CC_SRC,
307     [CC_OP_ADDB ... CC_OP_ADDQ] = USES_CC_DST | USES_CC_SRC,
308     [CC_OP_ADCB ... CC_OP_ADCQ] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
309     [CC_OP_SUBB ... CC_OP_SUBQ] = USES_CC_DST | USES_CC_SRC | USES_CC_SRCT,
310     [CC_OP_SBBB ... CC_OP_SBBQ] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
311     [CC_OP_LOGICB ... CC_OP_LOGICQ] = USES_CC_DST,
312     [CC_OP_INCB ... CC_OP_INCQ] = USES_CC_DST | USES_CC_SRC,
313     [CC_OP_DECB ... CC_OP_DECQ] = USES_CC_DST | USES_CC_SRC,
314     [CC_OP_SHLB ... CC_OP_SHLQ] = USES_CC_DST | USES_CC_SRC,
315     [CC_OP_SARB ... CC_OP_SARQ] = USES_CC_DST | USES_CC_SRC,
316     [CC_OP_BMILGB ... CC_OP_BMILGQ] = USES_CC_DST | USES_CC_SRC,
317     [CC_OP_ADCX] = USES_CC_DST | USES_CC_SRC,
318     [CC_OP_ADOX] = USES_CC_SRC | USES_CC_SRC2,
319     [CC_OP_ADCOX] = USES_CC_DST | USES_CC_SRC | USES_CC_SRC2,
320     [CC_OP_CLR] = 0,
321     [CC_OP_POPCNT] = USES_CC_SRC,
322 };
323 
324 static void set_cc_op(DisasContext *s, CCOp op)
325 {
326     int dead;
327 
328     if (s->cc_op == op) {
329         return;
330     }
331 
332     /* Discard CC computation that will no longer be used.  */
333     dead = cc_op_live[s->cc_op] & ~cc_op_live[op];
334     if (dead & USES_CC_DST) {
335         tcg_gen_discard_tl(cpu_cc_dst);
336     }
337     if (dead & USES_CC_SRC) {
338         tcg_gen_discard_tl(cpu_cc_src);
339     }
340     if (dead & USES_CC_SRC2) {
341         tcg_gen_discard_tl(cpu_cc_src2);
342     }
343     if (dead & USES_CC_SRCT) {
344         tcg_gen_discard_tl(s->cc_srcT);
345     }
346 
347     if (op == CC_OP_DYNAMIC) {
348         /* The DYNAMIC setting is translator only, and should never be
349            stored.  Thus we always consider it clean.  */
350         s->cc_op_dirty = false;
351     } else {
352         /* Discard any computed CC_OP value (see shifts).  */
353         if (s->cc_op == CC_OP_DYNAMIC) {
354             tcg_gen_discard_i32(cpu_cc_op);
355         }
356         s->cc_op_dirty = true;
357     }
358     s->cc_op = op;
359 }
360 
361 static void gen_update_cc_op(DisasContext *s)
362 {
363     if (s->cc_op_dirty) {
364         tcg_gen_movi_i32(cpu_cc_op, s->cc_op);
365         s->cc_op_dirty = false;
366     }
367 }
368 
369 #ifdef TARGET_X86_64
370 
371 #define NB_OP_SIZES 4
372 
373 #else /* !TARGET_X86_64 */
374 
375 #define NB_OP_SIZES 3
376 
377 #endif /* !TARGET_X86_64 */
378 
379 #if HOST_BIG_ENDIAN
380 #define REG_B_OFFSET (sizeof(target_ulong) - 1)
381 #define REG_H_OFFSET (sizeof(target_ulong) - 2)
382 #define REG_W_OFFSET (sizeof(target_ulong) - 2)
383 #define REG_L_OFFSET (sizeof(target_ulong) - 4)
384 #define REG_LH_OFFSET (sizeof(target_ulong) - 8)
385 #else
386 #define REG_B_OFFSET 0
387 #define REG_H_OFFSET 1
388 #define REG_W_OFFSET 0
389 #define REG_L_OFFSET 0
390 #define REG_LH_OFFSET 4
391 #endif
392 
393 /* In instruction encodings for byte register accesses the
394  * register number usually indicates "low 8 bits of register N";
395  * however there are some special cases where N 4..7 indicates
396  * [AH, CH, DH, BH], ie "bits 15..8 of register N-4". Return
397  * true for this special case, false otherwise.
398  */
399 static inline bool byte_reg_is_xH(DisasContext *s, int reg)
400 {
401     /* Any time the REX prefix is present, byte registers are uniform */
402     if (reg < 4 || REX_PREFIX(s)) {
403         return false;
404     }
405     return true;
406 }
407 
408 /* Select the size of a push/pop operation.  */
409 static inline MemOp mo_pushpop(DisasContext *s, MemOp ot)
410 {
411     if (CODE64(s)) {
412         return ot == MO_16 ? MO_16 : MO_64;
413     } else {
414         return ot;
415     }
416 }
417 
418 /* Select the size of the stack pointer.  */
419 static inline MemOp mo_stacksize(DisasContext *s)
420 {
421     return CODE64(s) ? MO_64 : SS32(s) ? MO_32 : MO_16;
422 }
423 
424 /* Select only size 64 else 32.  Used for SSE operand sizes.  */
425 static inline MemOp mo_64_32(MemOp ot)
426 {
427 #ifdef TARGET_X86_64
428     return ot == MO_64 ? MO_64 : MO_32;
429 #else
430     return MO_32;
431 #endif
432 }
433 
434 /* Select size 8 if lsb of B is clear, else OT.  Used for decoding
435    byte vs word opcodes.  */
436 static inline MemOp mo_b_d(int b, MemOp ot)
437 {
438     return b & 1 ? ot : MO_8;
439 }
440 
441 /* Select size 8 if lsb of B is clear, else OT capped at 32.
442    Used for decoding operand size of port opcodes.  */
443 static inline MemOp mo_b_d32(int b, MemOp ot)
444 {
445     return b & 1 ? (ot == MO_16 ? MO_16 : MO_32) : MO_8;
446 }
447 
448 /* Compute the result of writing t0 to the OT-sized register REG.
449  *
450  * If DEST is NULL, store the result into the register and return the
451  * register's TCGv.
452  *
453  * If DEST is not NULL, store the result into DEST and return the
454  * register's TCGv.
455  */
456 static TCGv gen_op_deposit_reg_v(DisasContext *s, MemOp ot, int reg, TCGv dest, TCGv t0)
457 {
458     switch(ot) {
459     case MO_8:
460         if (byte_reg_is_xH(s, reg)) {
461             dest = dest ? dest : cpu_regs[reg - 4];
462             tcg_gen_deposit_tl(dest, cpu_regs[reg - 4], t0, 8, 8);
463             return cpu_regs[reg - 4];
464         }
465         dest = dest ? dest : cpu_regs[reg];
466         tcg_gen_deposit_tl(dest, cpu_regs[reg], t0, 0, 8);
467         break;
468     case MO_16:
469         dest = dest ? dest : cpu_regs[reg];
470         tcg_gen_deposit_tl(dest, cpu_regs[reg], t0, 0, 16);
471         break;
472     case MO_32:
473         /* For x86_64, this sets the higher half of register to zero.
474            For i386, this is equivalent to a mov. */
475         dest = dest ? dest : cpu_regs[reg];
476         tcg_gen_ext32u_tl(dest, t0);
477         break;
478 #ifdef TARGET_X86_64
479     case MO_64:
480         dest = dest ? dest : cpu_regs[reg];
481         tcg_gen_mov_tl(dest, t0);
482         break;
483 #endif
484     default:
485         g_assert_not_reached();
486     }
487     return cpu_regs[reg];
488 }
489 
490 static void gen_op_mov_reg_v(DisasContext *s, MemOp ot, int reg, TCGv t0)
491 {
492     gen_op_deposit_reg_v(s, ot, reg, NULL, t0);
493 }
494 
495 static inline
496 void gen_op_mov_v_reg(DisasContext *s, MemOp ot, TCGv t0, int reg)
497 {
498     if (ot == MO_8 && byte_reg_is_xH(s, reg)) {
499         tcg_gen_extract_tl(t0, cpu_regs[reg - 4], 8, 8);
500     } else {
501         tcg_gen_mov_tl(t0, cpu_regs[reg]);
502     }
503 }
504 
505 static void gen_add_A0_im(DisasContext *s, int val)
506 {
507     tcg_gen_addi_tl(s->A0, s->A0, val);
508     if (!CODE64(s)) {
509         tcg_gen_ext32u_tl(s->A0, s->A0);
510     }
511 }
512 
513 static inline void gen_op_jmp_v(DisasContext *s, TCGv dest)
514 {
515     tcg_gen_mov_tl(cpu_eip, dest);
516     s->pc_save = -1;
517 }
518 
519 static inline
520 void gen_op_add_reg_im(DisasContext *s, MemOp size, int reg, int32_t val)
521 {
522     tcg_gen_addi_tl(s->tmp0, cpu_regs[reg], val);
523     gen_op_mov_reg_v(s, size, reg, s->tmp0);
524 }
525 
526 static inline void gen_op_add_reg(DisasContext *s, MemOp size, int reg, TCGv val)
527 {
528     tcg_gen_add_tl(s->tmp0, cpu_regs[reg], val);
529     gen_op_mov_reg_v(s, size, reg, s->tmp0);
530 }
531 
532 static inline void gen_op_ld_v(DisasContext *s, int idx, TCGv t0, TCGv a0)
533 {
534     tcg_gen_qemu_ld_tl(t0, a0, s->mem_index, idx | MO_LE);
535 }
536 
537 static inline void gen_op_st_v(DisasContext *s, int idx, TCGv t0, TCGv a0)
538 {
539     tcg_gen_qemu_st_tl(t0, a0, s->mem_index, idx | MO_LE);
540 }
541 
542 static inline void gen_op_st_rm_T0_A0(DisasContext *s, int idx, int d)
543 {
544     if (d == OR_TMP0) {
545         gen_op_st_v(s, idx, s->T0, s->A0);
546     } else {
547         gen_op_mov_reg_v(s, idx, d, s->T0);
548     }
549 }
550 
551 static void gen_update_eip_cur(DisasContext *s)
552 {
553     assert(s->pc_save != -1);
554     if (tb_cflags(s->base.tb) & CF_PCREL) {
555         tcg_gen_addi_tl(cpu_eip, cpu_eip, s->base.pc_next - s->pc_save);
556     } else if (CODE64(s)) {
557         tcg_gen_movi_tl(cpu_eip, s->base.pc_next);
558     } else {
559         tcg_gen_movi_tl(cpu_eip, (uint32_t)(s->base.pc_next - s->cs_base));
560     }
561     s->pc_save = s->base.pc_next;
562 }
563 
564 static void gen_update_eip_next(DisasContext *s)
565 {
566     assert(s->pc_save != -1);
567     if (tb_cflags(s->base.tb) & CF_PCREL) {
568         tcg_gen_addi_tl(cpu_eip, cpu_eip, s->pc - s->pc_save);
569     } else if (CODE64(s)) {
570         tcg_gen_movi_tl(cpu_eip, s->pc);
571     } else {
572         tcg_gen_movi_tl(cpu_eip, (uint32_t)(s->pc - s->cs_base));
573     }
574     s->pc_save = s->pc;
575 }
576 
577 static int cur_insn_len(DisasContext *s)
578 {
579     return s->pc - s->base.pc_next;
580 }
581 
582 static TCGv_i32 cur_insn_len_i32(DisasContext *s)
583 {
584     return tcg_constant_i32(cur_insn_len(s));
585 }
586 
587 static TCGv_i32 eip_next_i32(DisasContext *s)
588 {
589     assert(s->pc_save != -1);
590     /*
591      * This function has two users: lcall_real (always 16-bit mode), and
592      * iret_protected (16, 32, or 64-bit mode).  IRET only uses the value
593      * when EFLAGS.NT is set, which is illegal in 64-bit mode, which is
594      * why passing a 32-bit value isn't broken.  To avoid using this where
595      * we shouldn't, return -1 in 64-bit mode so that execution goes into
596      * the weeds quickly.
597      */
598     if (CODE64(s)) {
599         return tcg_constant_i32(-1);
600     }
601     if (tb_cflags(s->base.tb) & CF_PCREL) {
602         TCGv_i32 ret = tcg_temp_new_i32();
603         tcg_gen_trunc_tl_i32(ret, cpu_eip);
604         tcg_gen_addi_i32(ret, ret, s->pc - s->pc_save);
605         return ret;
606     } else {
607         return tcg_constant_i32(s->pc - s->cs_base);
608     }
609 }
610 
611 static TCGv eip_next_tl(DisasContext *s)
612 {
613     assert(s->pc_save != -1);
614     if (tb_cflags(s->base.tb) & CF_PCREL) {
615         TCGv ret = tcg_temp_new();
616         tcg_gen_addi_tl(ret, cpu_eip, s->pc - s->pc_save);
617         return ret;
618     } else if (CODE64(s)) {
619         return tcg_constant_tl(s->pc);
620     } else {
621         return tcg_constant_tl((uint32_t)(s->pc - s->cs_base));
622     }
623 }
624 
625 static TCGv eip_cur_tl(DisasContext *s)
626 {
627     assert(s->pc_save != -1);
628     if (tb_cflags(s->base.tb) & CF_PCREL) {
629         TCGv ret = tcg_temp_new();
630         tcg_gen_addi_tl(ret, cpu_eip, s->base.pc_next - s->pc_save);
631         return ret;
632     } else if (CODE64(s)) {
633         return tcg_constant_tl(s->base.pc_next);
634     } else {
635         return tcg_constant_tl((uint32_t)(s->base.pc_next - s->cs_base));
636     }
637 }
638 
639 /* Compute SEG:REG into DEST.  SEG is selected from the override segment
640    (OVR_SEG) and the default segment (DEF_SEG).  OVR_SEG may be -1 to
641    indicate no override.  */
642 static void gen_lea_v_seg_dest(DisasContext *s, MemOp aflag, TCGv dest, TCGv a0,
643                                int def_seg, int ovr_seg)
644 {
645     switch (aflag) {
646 #ifdef TARGET_X86_64
647     case MO_64:
648         if (ovr_seg < 0) {
649             tcg_gen_mov_tl(dest, a0);
650             return;
651         }
652         break;
653 #endif
654     case MO_32:
655         /* 32 bit address */
656         if (ovr_seg < 0 && ADDSEG(s)) {
657             ovr_seg = def_seg;
658         }
659         if (ovr_seg < 0) {
660             tcg_gen_ext32u_tl(dest, a0);
661             return;
662         }
663         break;
664     case MO_16:
665         /* 16 bit address */
666         tcg_gen_ext16u_tl(dest, a0);
667         a0 = dest;
668         if (ovr_seg < 0) {
669             if (ADDSEG(s)) {
670                 ovr_seg = def_seg;
671             } else {
672                 return;
673             }
674         }
675         break;
676     default:
677         g_assert_not_reached();
678     }
679 
680     if (ovr_seg >= 0) {
681         TCGv seg = cpu_seg_base[ovr_seg];
682 
683         if (aflag == MO_64) {
684             tcg_gen_add_tl(dest, a0, seg);
685         } else if (CODE64(s)) {
686             tcg_gen_ext32u_tl(dest, a0);
687             tcg_gen_add_tl(dest, dest, seg);
688         } else {
689             tcg_gen_add_tl(dest, a0, seg);
690             tcg_gen_ext32u_tl(dest, dest);
691         }
692     }
693 }
694 
695 static void gen_lea_v_seg(DisasContext *s, MemOp aflag, TCGv a0,
696                           int def_seg, int ovr_seg)
697 {
698     gen_lea_v_seg_dest(s, aflag, s->A0, a0, def_seg, ovr_seg);
699 }
700 
701 static inline void gen_string_movl_A0_ESI(DisasContext *s)
702 {
703     gen_lea_v_seg(s, s->aflag, cpu_regs[R_ESI], R_DS, s->override);
704 }
705 
706 static inline void gen_string_movl_A0_EDI(DisasContext *s)
707 {
708     gen_lea_v_seg(s, s->aflag, cpu_regs[R_EDI], R_ES, -1);
709 }
710 
711 static inline TCGv gen_compute_Dshift(DisasContext *s, MemOp ot)
712 {
713     TCGv dshift = tcg_temp_new();
714     tcg_gen_ld32s_tl(dshift, tcg_env, offsetof(CPUX86State, df));
715     tcg_gen_shli_tl(dshift, dshift, ot);
716     return dshift;
717 };
718 
719 static TCGv gen_ext_tl(TCGv dst, TCGv src, MemOp size, bool sign)
720 {
721     if (size == MO_TL) {
722         return src;
723     }
724     if (!dst) {
725         dst = tcg_temp_new();
726     }
727     tcg_gen_ext_tl(dst, src, size | (sign ? MO_SIGN : 0));
728     return dst;
729 }
730 
731 static void gen_extu(MemOp ot, TCGv reg)
732 {
733     gen_ext_tl(reg, reg, ot, false);
734 }
735 
736 static void gen_exts(MemOp ot, TCGv reg)
737 {
738     gen_ext_tl(reg, reg, ot, true);
739 }
740 
741 static void gen_op_j_ecx(DisasContext *s, TCGCond cond, TCGLabel *label1)
742 {
743     TCGv tmp = gen_ext_tl(NULL, cpu_regs[R_ECX], s->aflag, false);
744 
745     tcg_gen_brcondi_tl(cond, tmp, 0, label1);
746 }
747 
748 static inline void gen_op_jz_ecx(DisasContext *s, TCGLabel *label1)
749 {
750     gen_op_j_ecx(s, TCG_COND_EQ, label1);
751 }
752 
753 static inline void gen_op_jnz_ecx(DisasContext *s, TCGLabel *label1)
754 {
755     gen_op_j_ecx(s, TCG_COND_NE, label1);
756 }
757 
758 static void gen_helper_in_func(MemOp ot, TCGv v, TCGv_i32 n)
759 {
760     switch (ot) {
761     case MO_8:
762         gen_helper_inb(v, tcg_env, n);
763         break;
764     case MO_16:
765         gen_helper_inw(v, tcg_env, n);
766         break;
767     case MO_32:
768         gen_helper_inl(v, tcg_env, n);
769         break;
770     default:
771         g_assert_not_reached();
772     }
773 }
774 
775 static void gen_helper_out_func(MemOp ot, TCGv_i32 v, TCGv_i32 n)
776 {
777     switch (ot) {
778     case MO_8:
779         gen_helper_outb(tcg_env, v, n);
780         break;
781     case MO_16:
782         gen_helper_outw(tcg_env, v, n);
783         break;
784     case MO_32:
785         gen_helper_outl(tcg_env, v, n);
786         break;
787     default:
788         g_assert_not_reached();
789     }
790 }
791 
792 /*
793  * Validate that access to [port, port + 1<<ot) is allowed.
794  * Raise #GP, or VMM exit if not.
795  */
796 static bool gen_check_io(DisasContext *s, MemOp ot, TCGv_i32 port,
797                          uint32_t svm_flags)
798 {
799 #ifdef CONFIG_USER_ONLY
800     /*
801      * We do not implement the ioperm(2) syscall, so the TSS check
802      * will always fail.
803      */
804     gen_exception_gpf(s);
805     return false;
806 #else
807     if (PE(s) && (CPL(s) > IOPL(s) || VM86(s))) {
808         gen_helper_check_io(tcg_env, port, tcg_constant_i32(1 << ot));
809     }
810     if (GUEST(s)) {
811         gen_update_cc_op(s);
812         gen_update_eip_cur(s);
813         if (s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) {
814             svm_flags |= SVM_IOIO_REP_MASK;
815         }
816         svm_flags |= 1 << (SVM_IOIO_SIZE_SHIFT + ot);
817         gen_helper_svm_check_io(tcg_env, port,
818                                 tcg_constant_i32(svm_flags),
819                                 cur_insn_len_i32(s));
820     }
821     return true;
822 #endif
823 }
824 
825 static void gen_movs(DisasContext *s, MemOp ot)
826 {
827     TCGv dshift;
828 
829     gen_string_movl_A0_ESI(s);
830     gen_op_ld_v(s, ot, s->T0, s->A0);
831     gen_string_movl_A0_EDI(s);
832     gen_op_st_v(s, ot, s->T0, s->A0);
833 
834     dshift = gen_compute_Dshift(s, ot);
835     gen_op_add_reg(s, s->aflag, R_ESI, dshift);
836     gen_op_add_reg(s, s->aflag, R_EDI, dshift);
837 }
838 
839 static void gen_op_update1_cc(DisasContext *s)
840 {
841     tcg_gen_mov_tl(cpu_cc_dst, s->T0);
842 }
843 
844 static void gen_op_update2_cc(DisasContext *s)
845 {
846     tcg_gen_mov_tl(cpu_cc_src, s->T1);
847     tcg_gen_mov_tl(cpu_cc_dst, s->T0);
848 }
849 
850 static void gen_op_update3_cc(DisasContext *s, TCGv reg)
851 {
852     tcg_gen_mov_tl(cpu_cc_src2, reg);
853     tcg_gen_mov_tl(cpu_cc_src, s->T1);
854     tcg_gen_mov_tl(cpu_cc_dst, s->T0);
855 }
856 
857 static inline void gen_op_testl_T0_T1_cc(DisasContext *s)
858 {
859     tcg_gen_and_tl(cpu_cc_dst, s->T0, s->T1);
860 }
861 
862 static void gen_op_update_neg_cc(DisasContext *s)
863 {
864     tcg_gen_mov_tl(cpu_cc_dst, s->T0);
865     tcg_gen_neg_tl(cpu_cc_src, s->T0);
866     tcg_gen_movi_tl(s->cc_srcT, 0);
867 }
868 
869 /* compute all eflags to reg */
870 static void gen_mov_eflags(DisasContext *s, TCGv reg)
871 {
872     TCGv dst, src1, src2;
873     TCGv_i32 cc_op;
874     int live, dead;
875 
876     if (s->cc_op == CC_OP_EFLAGS) {
877         tcg_gen_mov_tl(reg, cpu_cc_src);
878         return;
879     }
880     if (s->cc_op == CC_OP_CLR) {
881         tcg_gen_movi_tl(reg, CC_Z | CC_P);
882         return;
883     }
884 
885     dst = cpu_cc_dst;
886     src1 = cpu_cc_src;
887     src2 = cpu_cc_src2;
888 
889     /* Take care to not read values that are not live.  */
890     live = cc_op_live[s->cc_op] & ~USES_CC_SRCT;
891     dead = live ^ (USES_CC_DST | USES_CC_SRC | USES_CC_SRC2);
892     if (dead) {
893         TCGv zero = tcg_constant_tl(0);
894         if (dead & USES_CC_DST) {
895             dst = zero;
896         }
897         if (dead & USES_CC_SRC) {
898             src1 = zero;
899         }
900         if (dead & USES_CC_SRC2) {
901             src2 = zero;
902         }
903     }
904 
905     if (s->cc_op != CC_OP_DYNAMIC) {
906         cc_op = tcg_constant_i32(s->cc_op);
907     } else {
908         cc_op = cpu_cc_op;
909     }
910     gen_helper_cc_compute_all(reg, dst, src1, src2, cc_op);
911 }
912 
913 /* compute all eflags to cc_src */
914 static void gen_compute_eflags(DisasContext *s)
915 {
916     gen_mov_eflags(s, cpu_cc_src);
917     set_cc_op(s, CC_OP_EFLAGS);
918 }
919 
920 typedef struct CCPrepare {
921     TCGCond cond;
922     TCGv reg;
923     TCGv reg2;
924     target_ulong imm;
925     target_ulong mask;
926     bool use_reg2;
927     bool no_setcond;
928 } CCPrepare;
929 
930 /* compute eflags.C to reg */
931 static CCPrepare gen_prepare_eflags_c(DisasContext *s, TCGv reg)
932 {
933     TCGv t0, t1;
934     int size, shift;
935 
936     switch (s->cc_op) {
937     case CC_OP_SUBB ... CC_OP_SUBQ:
938         /* (DATA_TYPE)CC_SRCT < (DATA_TYPE)CC_SRC */
939         size = s->cc_op - CC_OP_SUBB;
940         t1 = gen_ext_tl(s->tmp0, cpu_cc_src, size, false);
941         /* If no temporary was used, be careful not to alias t1 and t0.  */
942         t0 = t1 == cpu_cc_src ? s->tmp0 : reg;
943         tcg_gen_mov_tl(t0, s->cc_srcT);
944         gen_extu(size, t0);
945         goto add_sub;
946 
947     case CC_OP_ADDB ... CC_OP_ADDQ:
948         /* (DATA_TYPE)CC_DST < (DATA_TYPE)CC_SRC */
949         size = s->cc_op - CC_OP_ADDB;
950         t1 = gen_ext_tl(s->tmp0, cpu_cc_src, size, false);
951         t0 = gen_ext_tl(reg, cpu_cc_dst, size, false);
952     add_sub:
953         return (CCPrepare) { .cond = TCG_COND_LTU, .reg = t0,
954                              .reg2 = t1, .mask = -1, .use_reg2 = true };
955 
956     case CC_OP_LOGICB ... CC_OP_LOGICQ:
957     case CC_OP_CLR:
958     case CC_OP_POPCNT:
959         return (CCPrepare) { .cond = TCG_COND_NEVER, .mask = -1 };
960 
961     case CC_OP_INCB ... CC_OP_INCQ:
962     case CC_OP_DECB ... CC_OP_DECQ:
963         return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
964                              .mask = -1, .no_setcond = true };
965 
966     case CC_OP_SHLB ... CC_OP_SHLQ:
967         /* (CC_SRC >> (DATA_BITS - 1)) & 1 */
968         size = s->cc_op - CC_OP_SHLB;
969         shift = (8 << size) - 1;
970         return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
971                              .mask = (target_ulong)1 << shift };
972 
973     case CC_OP_MULB ... CC_OP_MULQ:
974         return (CCPrepare) { .cond = TCG_COND_NE,
975                              .reg = cpu_cc_src, .mask = -1 };
976 
977     case CC_OP_BMILGB ... CC_OP_BMILGQ:
978         size = s->cc_op - CC_OP_BMILGB;
979         t0 = gen_ext_tl(reg, cpu_cc_src, size, false);
980         return (CCPrepare) { .cond = TCG_COND_EQ, .reg = t0, .mask = -1 };
981 
982     case CC_OP_ADCX:
983     case CC_OP_ADCOX:
984         return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_dst,
985                              .mask = -1, .no_setcond = true };
986 
987     case CC_OP_EFLAGS:
988     case CC_OP_SARB ... CC_OP_SARQ:
989         /* CC_SRC & 1 */
990         return (CCPrepare) { .cond = TCG_COND_NE,
991                              .reg = cpu_cc_src, .mask = CC_C };
992 
993     default:
994        /* The need to compute only C from CC_OP_DYNAMIC is important
995           in efficiently implementing e.g. INC at the start of a TB.  */
996        gen_update_cc_op(s);
997        gen_helper_cc_compute_c(reg, cpu_cc_dst, cpu_cc_src,
998                                cpu_cc_src2, cpu_cc_op);
999        return (CCPrepare) { .cond = TCG_COND_NE, .reg = reg,
1000                             .mask = -1, .no_setcond = true };
1001     }
1002 }
1003 
1004 /* compute eflags.P to reg */
1005 static CCPrepare gen_prepare_eflags_p(DisasContext *s, TCGv reg)
1006 {
1007     gen_compute_eflags(s);
1008     return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
1009                          .mask = CC_P };
1010 }
1011 
1012 /* compute eflags.S to reg */
1013 static CCPrepare gen_prepare_eflags_s(DisasContext *s, TCGv reg)
1014 {
1015     switch (s->cc_op) {
1016     case CC_OP_DYNAMIC:
1017         gen_compute_eflags(s);
1018         /* FALLTHRU */
1019     case CC_OP_EFLAGS:
1020     case CC_OP_ADCX:
1021     case CC_OP_ADOX:
1022     case CC_OP_ADCOX:
1023         return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
1024                              .mask = CC_S };
1025     case CC_OP_CLR:
1026     case CC_OP_POPCNT:
1027         return (CCPrepare) { .cond = TCG_COND_NEVER, .mask = -1 };
1028     default:
1029         {
1030             MemOp size = (s->cc_op - CC_OP_ADDB) & 3;
1031             TCGv t0 = gen_ext_tl(reg, cpu_cc_dst, size, true);
1032             return (CCPrepare) { .cond = TCG_COND_LT, .reg = t0, .mask = -1 };
1033         }
1034     }
1035 }
1036 
1037 /* compute eflags.O to reg */
1038 static CCPrepare gen_prepare_eflags_o(DisasContext *s, TCGv reg)
1039 {
1040     switch (s->cc_op) {
1041     case CC_OP_ADOX:
1042     case CC_OP_ADCOX:
1043         return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src2,
1044                              .mask = -1, .no_setcond = true };
1045     case CC_OP_CLR:
1046     case CC_OP_POPCNT:
1047         return (CCPrepare) { .cond = TCG_COND_NEVER, .mask = -1 };
1048     case CC_OP_MULB ... CC_OP_MULQ:
1049         return (CCPrepare) { .cond = TCG_COND_NE,
1050                              .reg = cpu_cc_src, .mask = -1 };
1051     default:
1052         gen_compute_eflags(s);
1053         return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
1054                              .mask = CC_O };
1055     }
1056 }
1057 
1058 /* compute eflags.Z to reg */
1059 static CCPrepare gen_prepare_eflags_z(DisasContext *s, TCGv reg)
1060 {
1061     switch (s->cc_op) {
1062     case CC_OP_DYNAMIC:
1063         gen_compute_eflags(s);
1064         /* FALLTHRU */
1065     case CC_OP_EFLAGS:
1066     case CC_OP_ADCX:
1067     case CC_OP_ADOX:
1068     case CC_OP_ADCOX:
1069         return (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
1070                              .mask = CC_Z };
1071     case CC_OP_CLR:
1072         return (CCPrepare) { .cond = TCG_COND_ALWAYS, .mask = -1 };
1073     case CC_OP_POPCNT:
1074         return (CCPrepare) { .cond = TCG_COND_EQ, .reg = cpu_cc_src,
1075                              .mask = -1 };
1076     default:
1077         {
1078             MemOp size = (s->cc_op - CC_OP_ADDB) & 3;
1079             TCGv t0 = gen_ext_tl(reg, cpu_cc_dst, size, false);
1080             return (CCPrepare) { .cond = TCG_COND_EQ, .reg = t0, .mask = -1 };
1081         }
1082     }
1083 }
1084 
1085 /* perform a conditional store into register 'reg' according to jump opcode
1086    value 'b'. In the fast case, T0 is guaranteed not to be used. */
1087 static CCPrepare gen_prepare_cc(DisasContext *s, int b, TCGv reg)
1088 {
1089     int inv, jcc_op, cond;
1090     MemOp size;
1091     CCPrepare cc;
1092     TCGv t0;
1093 
1094     inv = b & 1;
1095     jcc_op = (b >> 1) & 7;
1096 
1097     switch (s->cc_op) {
1098     case CC_OP_SUBB ... CC_OP_SUBQ:
1099         /* We optimize relational operators for the cmp/jcc case.  */
1100         size = s->cc_op - CC_OP_SUBB;
1101         switch (jcc_op) {
1102         case JCC_BE:
1103             tcg_gen_mov_tl(s->tmp4, s->cc_srcT);
1104             gen_extu(size, s->tmp4);
1105             t0 = gen_ext_tl(s->tmp0, cpu_cc_src, size, false);
1106             cc = (CCPrepare) { .cond = TCG_COND_LEU, .reg = s->tmp4,
1107                                .reg2 = t0, .mask = -1, .use_reg2 = true };
1108             break;
1109 
1110         case JCC_L:
1111             cond = TCG_COND_LT;
1112             goto fast_jcc_l;
1113         case JCC_LE:
1114             cond = TCG_COND_LE;
1115         fast_jcc_l:
1116             tcg_gen_mov_tl(s->tmp4, s->cc_srcT);
1117             gen_exts(size, s->tmp4);
1118             t0 = gen_ext_tl(s->tmp0, cpu_cc_src, size, true);
1119             cc = (CCPrepare) { .cond = cond, .reg = s->tmp4,
1120                                .reg2 = t0, .mask = -1, .use_reg2 = true };
1121             break;
1122 
1123         default:
1124             goto slow_jcc;
1125         }
1126         break;
1127 
1128     default:
1129     slow_jcc:
1130         /* This actually generates good code for JC, JZ and JS.  */
1131         switch (jcc_op) {
1132         case JCC_O:
1133             cc = gen_prepare_eflags_o(s, reg);
1134             break;
1135         case JCC_B:
1136             cc = gen_prepare_eflags_c(s, reg);
1137             break;
1138         case JCC_Z:
1139             cc = gen_prepare_eflags_z(s, reg);
1140             break;
1141         case JCC_BE:
1142             gen_compute_eflags(s);
1143             cc = (CCPrepare) { .cond = TCG_COND_NE, .reg = cpu_cc_src,
1144                                .mask = CC_Z | CC_C };
1145             break;
1146         case JCC_S:
1147             cc = gen_prepare_eflags_s(s, reg);
1148             break;
1149         case JCC_P:
1150             cc = gen_prepare_eflags_p(s, reg);
1151             break;
1152         case JCC_L:
1153             gen_compute_eflags(s);
1154             if (reg == cpu_cc_src) {
1155                 reg = s->tmp0;
1156             }
1157             tcg_gen_addi_tl(reg, cpu_cc_src, CC_O - CC_S);
1158             cc = (CCPrepare) { .cond = TCG_COND_NE, .reg = reg,
1159                                .mask = CC_O };
1160             break;
1161         default:
1162         case JCC_LE:
1163             gen_compute_eflags(s);
1164             if (reg == cpu_cc_src) {
1165                 reg = s->tmp0;
1166             }
1167             tcg_gen_addi_tl(reg, cpu_cc_src, CC_O - CC_S);
1168             cc = (CCPrepare) { .cond = TCG_COND_NE, .reg = reg,
1169                                .mask = CC_O | CC_Z };
1170             break;
1171         }
1172         break;
1173     }
1174 
1175     if (inv) {
1176         cc.cond = tcg_invert_cond(cc.cond);
1177     }
1178     return cc;
1179 }
1180 
1181 static void gen_setcc1(DisasContext *s, int b, TCGv reg)
1182 {
1183     CCPrepare cc = gen_prepare_cc(s, b, reg);
1184 
1185     if (cc.no_setcond) {
1186         if (cc.cond == TCG_COND_EQ) {
1187             tcg_gen_xori_tl(reg, cc.reg, 1);
1188         } else {
1189             tcg_gen_mov_tl(reg, cc.reg);
1190         }
1191         return;
1192     }
1193 
1194     if (cc.cond == TCG_COND_NE && !cc.use_reg2 && cc.imm == 0 &&
1195         cc.mask != 0 && (cc.mask & (cc.mask - 1)) == 0) {
1196         tcg_gen_shri_tl(reg, cc.reg, ctztl(cc.mask));
1197         tcg_gen_andi_tl(reg, reg, 1);
1198         return;
1199     }
1200     if (cc.mask != -1) {
1201         tcg_gen_andi_tl(reg, cc.reg, cc.mask);
1202         cc.reg = reg;
1203     }
1204     if (cc.use_reg2) {
1205         tcg_gen_setcond_tl(cc.cond, reg, cc.reg, cc.reg2);
1206     } else {
1207         tcg_gen_setcondi_tl(cc.cond, reg, cc.reg, cc.imm);
1208     }
1209 }
1210 
1211 static inline void gen_compute_eflags_c(DisasContext *s, TCGv reg)
1212 {
1213     gen_setcc1(s, JCC_B << 1, reg);
1214 }
1215 
1216 /* generate a conditional jump to label 'l1' according to jump opcode
1217    value 'b'. In the fast case, T0 is guaranteed not to be used. */
1218 static inline void gen_jcc1_noeob(DisasContext *s, int b, TCGLabel *l1)
1219 {
1220     CCPrepare cc = gen_prepare_cc(s, b, s->T0);
1221 
1222     if (cc.mask != -1) {
1223         tcg_gen_andi_tl(s->T0, cc.reg, cc.mask);
1224         cc.reg = s->T0;
1225     }
1226     if (cc.use_reg2) {
1227         tcg_gen_brcond_tl(cc.cond, cc.reg, cc.reg2, l1);
1228     } else {
1229         tcg_gen_brcondi_tl(cc.cond, cc.reg, cc.imm, l1);
1230     }
1231 }
1232 
1233 /* Generate a conditional jump to label 'l1' according to jump opcode
1234    value 'b'. In the fast case, T0 is guaranteed not to be used.
1235    A translation block must end soon.  */
1236 static inline void gen_jcc1(DisasContext *s, int b, TCGLabel *l1)
1237 {
1238     CCPrepare cc = gen_prepare_cc(s, b, s->T0);
1239 
1240     gen_update_cc_op(s);
1241     if (cc.mask != -1) {
1242         tcg_gen_andi_tl(s->T0, cc.reg, cc.mask);
1243         cc.reg = s->T0;
1244     }
1245     set_cc_op(s, CC_OP_DYNAMIC);
1246     if (cc.use_reg2) {
1247         tcg_gen_brcond_tl(cc.cond, cc.reg, cc.reg2, l1);
1248     } else {
1249         tcg_gen_brcondi_tl(cc.cond, cc.reg, cc.imm, l1);
1250     }
1251 }
1252 
1253 /* XXX: does not work with gdbstub "ice" single step - not a
1254    serious problem */
1255 static TCGLabel *gen_jz_ecx_string(DisasContext *s)
1256 {
1257     TCGLabel *l1 = gen_new_label();
1258     TCGLabel *l2 = gen_new_label();
1259     gen_op_jnz_ecx(s, l1);
1260     gen_set_label(l2);
1261     gen_jmp_rel_csize(s, 0, 1);
1262     gen_set_label(l1);
1263     return l2;
1264 }
1265 
1266 static void gen_stos(DisasContext *s, MemOp ot)
1267 {
1268     gen_string_movl_A0_EDI(s);
1269     gen_op_st_v(s, ot, s->T0, s->A0);
1270     gen_op_add_reg(s, s->aflag, R_EDI, gen_compute_Dshift(s, ot));
1271 }
1272 
1273 static void gen_lods(DisasContext *s, MemOp ot)
1274 {
1275     gen_string_movl_A0_ESI(s);
1276     gen_op_ld_v(s, ot, s->T0, s->A0);
1277     gen_op_mov_reg_v(s, ot, R_EAX, s->T0);
1278     gen_op_add_reg(s, s->aflag, R_ESI, gen_compute_Dshift(s, ot));
1279 }
1280 
1281 static void gen_scas(DisasContext *s, MemOp ot)
1282 {
1283     gen_string_movl_A0_EDI(s);
1284     gen_op_ld_v(s, ot, s->T1, s->A0);
1285     tcg_gen_mov_tl(cpu_cc_src, s->T1);
1286     tcg_gen_mov_tl(s->cc_srcT, s->T0);
1287     tcg_gen_sub_tl(cpu_cc_dst, s->T0, s->T1);
1288     set_cc_op(s, CC_OP_SUBB + ot);
1289 
1290     gen_op_add_reg(s, s->aflag, R_EDI, gen_compute_Dshift(s, ot));
1291 }
1292 
1293 static void gen_cmps(DisasContext *s, MemOp ot)
1294 {
1295     TCGv dshift;
1296 
1297     gen_string_movl_A0_EDI(s);
1298     gen_op_ld_v(s, ot, s->T1, s->A0);
1299     gen_string_movl_A0_ESI(s);
1300     gen_op(s, OP_CMPL, ot, OR_TMP0);
1301 
1302     dshift = gen_compute_Dshift(s, ot);
1303     gen_op_add_reg(s, s->aflag, R_ESI, dshift);
1304     gen_op_add_reg(s, s->aflag, R_EDI, dshift);
1305 }
1306 
1307 static void gen_bpt_io(DisasContext *s, TCGv_i32 t_port, int ot)
1308 {
1309     if (s->flags & HF_IOBPT_MASK) {
1310 #ifdef CONFIG_USER_ONLY
1311         /* user-mode cpu should not be in IOBPT mode */
1312         g_assert_not_reached();
1313 #else
1314         TCGv_i32 t_size = tcg_constant_i32(1 << ot);
1315         TCGv t_next = eip_next_tl(s);
1316         gen_helper_bpt_io(tcg_env, t_port, t_size, t_next);
1317 #endif /* CONFIG_USER_ONLY */
1318     }
1319 }
1320 
1321 static void gen_ins(DisasContext *s, MemOp ot)
1322 {
1323     gen_string_movl_A0_EDI(s);
1324     /* Note: we must do this dummy write first to be restartable in
1325        case of page fault. */
1326     tcg_gen_movi_tl(s->T0, 0);
1327     gen_op_st_v(s, ot, s->T0, s->A0);
1328     tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_EDX]);
1329     tcg_gen_andi_i32(s->tmp2_i32, s->tmp2_i32, 0xffff);
1330     gen_helper_in_func(ot, s->T0, s->tmp2_i32);
1331     gen_op_st_v(s, ot, s->T0, s->A0);
1332     gen_op_add_reg(s, s->aflag, R_EDI, gen_compute_Dshift(s, ot));
1333     gen_bpt_io(s, s->tmp2_i32, ot);
1334 }
1335 
1336 static void gen_outs(DisasContext *s, MemOp ot)
1337 {
1338     gen_string_movl_A0_ESI(s);
1339     gen_op_ld_v(s, ot, s->T0, s->A0);
1340 
1341     tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_EDX]);
1342     tcg_gen_andi_i32(s->tmp2_i32, s->tmp2_i32, 0xffff);
1343     tcg_gen_trunc_tl_i32(s->tmp3_i32, s->T0);
1344     gen_helper_out_func(ot, s->tmp2_i32, s->tmp3_i32);
1345     gen_op_add_reg(s, s->aflag, R_ESI, gen_compute_Dshift(s, ot));
1346     gen_bpt_io(s, s->tmp2_i32, ot);
1347 }
1348 
1349 /* Generate jumps to current or next instruction */
1350 static void gen_repz(DisasContext *s, MemOp ot,
1351                      void (*fn)(DisasContext *s, MemOp ot))
1352 {
1353     TCGLabel *l2;
1354     gen_update_cc_op(s);
1355     l2 = gen_jz_ecx_string(s);
1356     fn(s, ot);
1357     gen_op_add_reg_im(s, s->aflag, R_ECX, -1);
1358     /*
1359      * A loop would cause two single step exceptions if ECX = 1
1360      * before rep string_insn
1361      */
1362     if (s->repz_opt) {
1363         gen_op_jz_ecx(s, l2);
1364     }
1365     gen_jmp_rel_csize(s, -cur_insn_len(s), 0);
1366 }
1367 
1368 #define GEN_REPZ(op) \
1369     static inline void gen_repz_ ## op(DisasContext *s, MemOp ot) \
1370     { gen_repz(s, ot, gen_##op); }
1371 
1372 static void gen_repz2(DisasContext *s, MemOp ot, int nz,
1373                       void (*fn)(DisasContext *s, MemOp ot))
1374 {
1375     TCGLabel *l2;
1376     gen_update_cc_op(s);
1377     l2 = gen_jz_ecx_string(s);
1378     fn(s, ot);
1379     gen_op_add_reg_im(s, s->aflag, R_ECX, -1);
1380     gen_update_cc_op(s);
1381     gen_jcc1(s, (JCC_Z << 1) | (nz ^ 1), l2);
1382     if (s->repz_opt) {
1383         gen_op_jz_ecx(s, l2);
1384     }
1385     gen_jmp_rel_csize(s, -cur_insn_len(s), 0);
1386 }
1387 
1388 #define GEN_REPZ2(op) \
1389     static inline void gen_repz_ ## op(DisasContext *s, MemOp ot, int nz) \
1390     { gen_repz2(s, ot, nz, gen_##op); }
1391 
1392 GEN_REPZ(movs)
1393 GEN_REPZ(stos)
1394 GEN_REPZ(lods)
1395 GEN_REPZ(ins)
1396 GEN_REPZ(outs)
1397 GEN_REPZ2(scas)
1398 GEN_REPZ2(cmps)
1399 
1400 static void gen_helper_fp_arith_ST0_FT0(int op)
1401 {
1402     switch (op) {
1403     case 0:
1404         gen_helper_fadd_ST0_FT0(tcg_env);
1405         break;
1406     case 1:
1407         gen_helper_fmul_ST0_FT0(tcg_env);
1408         break;
1409     case 2:
1410         gen_helper_fcom_ST0_FT0(tcg_env);
1411         break;
1412     case 3:
1413         gen_helper_fcom_ST0_FT0(tcg_env);
1414         break;
1415     case 4:
1416         gen_helper_fsub_ST0_FT0(tcg_env);
1417         break;
1418     case 5:
1419         gen_helper_fsubr_ST0_FT0(tcg_env);
1420         break;
1421     case 6:
1422         gen_helper_fdiv_ST0_FT0(tcg_env);
1423         break;
1424     case 7:
1425         gen_helper_fdivr_ST0_FT0(tcg_env);
1426         break;
1427     }
1428 }
1429 
1430 /* NOTE the exception in "r" op ordering */
1431 static void gen_helper_fp_arith_STN_ST0(int op, int opreg)
1432 {
1433     TCGv_i32 tmp = tcg_constant_i32(opreg);
1434     switch (op) {
1435     case 0:
1436         gen_helper_fadd_STN_ST0(tcg_env, tmp);
1437         break;
1438     case 1:
1439         gen_helper_fmul_STN_ST0(tcg_env, tmp);
1440         break;
1441     case 4:
1442         gen_helper_fsubr_STN_ST0(tcg_env, tmp);
1443         break;
1444     case 5:
1445         gen_helper_fsub_STN_ST0(tcg_env, tmp);
1446         break;
1447     case 6:
1448         gen_helper_fdivr_STN_ST0(tcg_env, tmp);
1449         break;
1450     case 7:
1451         gen_helper_fdiv_STN_ST0(tcg_env, tmp);
1452         break;
1453     }
1454 }
1455 
1456 static void gen_exception(DisasContext *s, int trapno)
1457 {
1458     gen_update_cc_op(s);
1459     gen_update_eip_cur(s);
1460     gen_helper_raise_exception(tcg_env, tcg_constant_i32(trapno));
1461     s->base.is_jmp = DISAS_NORETURN;
1462 }
1463 
1464 /* Generate #UD for the current instruction.  The assumption here is that
1465    the instruction is known, but it isn't allowed in the current cpu mode.  */
1466 static void gen_illegal_opcode(DisasContext *s)
1467 {
1468     gen_exception(s, EXCP06_ILLOP);
1469 }
1470 
1471 /* Generate #GP for the current instruction. */
1472 static void gen_exception_gpf(DisasContext *s)
1473 {
1474     gen_exception(s, EXCP0D_GPF);
1475 }
1476 
1477 /* Check for cpl == 0; if not, raise #GP and return false. */
1478 static bool check_cpl0(DisasContext *s)
1479 {
1480     if (CPL(s) == 0) {
1481         return true;
1482     }
1483     gen_exception_gpf(s);
1484     return false;
1485 }
1486 
1487 /* If vm86, check for iopl == 3; if not, raise #GP and return false. */
1488 static bool check_vm86_iopl(DisasContext *s)
1489 {
1490     if (!VM86(s) || IOPL(s) == 3) {
1491         return true;
1492     }
1493     gen_exception_gpf(s);
1494     return false;
1495 }
1496 
1497 /* Check for iopl allowing access; if not, raise #GP and return false. */
1498 static bool check_iopl(DisasContext *s)
1499 {
1500     if (VM86(s) ? IOPL(s) == 3 : CPL(s) <= IOPL(s)) {
1501         return true;
1502     }
1503     gen_exception_gpf(s);
1504     return false;
1505 }
1506 
1507 /* if d == OR_TMP0, it means memory operand (address in A0) */
1508 static void gen_op(DisasContext *s1, int op, MemOp ot, int d)
1509 {
1510     if (d != OR_TMP0) {
1511         if (s1->prefix & PREFIX_LOCK) {
1512             /* Lock prefix when destination is not memory.  */
1513             gen_illegal_opcode(s1);
1514             return;
1515         }
1516         gen_op_mov_v_reg(s1, ot, s1->T0, d);
1517     } else if (!(s1->prefix & PREFIX_LOCK)) {
1518         gen_op_ld_v(s1, ot, s1->T0, s1->A0);
1519     }
1520     switch(op) {
1521     case OP_ADCL:
1522         gen_compute_eflags_c(s1, s1->tmp4);
1523         if (s1->prefix & PREFIX_LOCK) {
1524             tcg_gen_add_tl(s1->T0, s1->tmp4, s1->T1);
1525             tcg_gen_atomic_add_fetch_tl(s1->T0, s1->A0, s1->T0,
1526                                         s1->mem_index, ot | MO_LE);
1527         } else {
1528             tcg_gen_add_tl(s1->T0, s1->T0, s1->T1);
1529             tcg_gen_add_tl(s1->T0, s1->T0, s1->tmp4);
1530             gen_op_st_rm_T0_A0(s1, ot, d);
1531         }
1532         gen_op_update3_cc(s1, s1->tmp4);
1533         set_cc_op(s1, CC_OP_ADCB + ot);
1534         break;
1535     case OP_SBBL:
1536         gen_compute_eflags_c(s1, s1->tmp4);
1537         if (s1->prefix & PREFIX_LOCK) {
1538             tcg_gen_add_tl(s1->T0, s1->T1, s1->tmp4);
1539             tcg_gen_neg_tl(s1->T0, s1->T0);
1540             tcg_gen_atomic_add_fetch_tl(s1->T0, s1->A0, s1->T0,
1541                                         s1->mem_index, ot | MO_LE);
1542         } else {
1543             tcg_gen_sub_tl(s1->T0, s1->T0, s1->T1);
1544             tcg_gen_sub_tl(s1->T0, s1->T0, s1->tmp4);
1545             gen_op_st_rm_T0_A0(s1, ot, d);
1546         }
1547         gen_op_update3_cc(s1, s1->tmp4);
1548         set_cc_op(s1, CC_OP_SBBB + ot);
1549         break;
1550     case OP_ADDL:
1551         if (s1->prefix & PREFIX_LOCK) {
1552             tcg_gen_atomic_add_fetch_tl(s1->T0, s1->A0, s1->T1,
1553                                         s1->mem_index, ot | MO_LE);
1554         } else {
1555             tcg_gen_add_tl(s1->T0, s1->T0, s1->T1);
1556             gen_op_st_rm_T0_A0(s1, ot, d);
1557         }
1558         gen_op_update2_cc(s1);
1559         set_cc_op(s1, CC_OP_ADDB + ot);
1560         break;
1561     case OP_SUBL:
1562         if (s1->prefix & PREFIX_LOCK) {
1563             tcg_gen_neg_tl(s1->T0, s1->T1);
1564             tcg_gen_atomic_fetch_add_tl(s1->cc_srcT, s1->A0, s1->T0,
1565                                         s1->mem_index, ot | MO_LE);
1566             tcg_gen_sub_tl(s1->T0, s1->cc_srcT, s1->T1);
1567         } else {
1568             tcg_gen_mov_tl(s1->cc_srcT, s1->T0);
1569             tcg_gen_sub_tl(s1->T0, s1->T0, s1->T1);
1570             gen_op_st_rm_T0_A0(s1, ot, d);
1571         }
1572         gen_op_update2_cc(s1);
1573         set_cc_op(s1, CC_OP_SUBB + ot);
1574         break;
1575     default:
1576     case OP_ANDL:
1577         if (s1->prefix & PREFIX_LOCK) {
1578             tcg_gen_atomic_and_fetch_tl(s1->T0, s1->A0, s1->T1,
1579                                         s1->mem_index, ot | MO_LE);
1580         } else {
1581             tcg_gen_and_tl(s1->T0, s1->T0, s1->T1);
1582             gen_op_st_rm_T0_A0(s1, ot, d);
1583         }
1584         gen_op_update1_cc(s1);
1585         set_cc_op(s1, CC_OP_LOGICB + ot);
1586         break;
1587     case OP_ORL:
1588         if (s1->prefix & PREFIX_LOCK) {
1589             tcg_gen_atomic_or_fetch_tl(s1->T0, s1->A0, s1->T1,
1590                                        s1->mem_index, ot | MO_LE);
1591         } else {
1592             tcg_gen_or_tl(s1->T0, s1->T0, s1->T1);
1593             gen_op_st_rm_T0_A0(s1, ot, d);
1594         }
1595         gen_op_update1_cc(s1);
1596         set_cc_op(s1, CC_OP_LOGICB + ot);
1597         break;
1598     case OP_XORL:
1599         if (s1->prefix & PREFIX_LOCK) {
1600             tcg_gen_atomic_xor_fetch_tl(s1->T0, s1->A0, s1->T1,
1601                                         s1->mem_index, ot | MO_LE);
1602         } else {
1603             tcg_gen_xor_tl(s1->T0, s1->T0, s1->T1);
1604             gen_op_st_rm_T0_A0(s1, ot, d);
1605         }
1606         gen_op_update1_cc(s1);
1607         set_cc_op(s1, CC_OP_LOGICB + ot);
1608         break;
1609     case OP_CMPL:
1610         tcg_gen_mov_tl(cpu_cc_src, s1->T1);
1611         tcg_gen_mov_tl(s1->cc_srcT, s1->T0);
1612         tcg_gen_sub_tl(cpu_cc_dst, s1->T0, s1->T1);
1613         set_cc_op(s1, CC_OP_SUBB + ot);
1614         break;
1615     }
1616 }
1617 
1618 /* if d == OR_TMP0, it means memory operand (address in A0) */
1619 static void gen_inc(DisasContext *s1, MemOp ot, int d, int c)
1620 {
1621     if (s1->prefix & PREFIX_LOCK) {
1622         if (d != OR_TMP0) {
1623             /* Lock prefix when destination is not memory */
1624             gen_illegal_opcode(s1);
1625             return;
1626         }
1627         tcg_gen_movi_tl(s1->T0, c > 0 ? 1 : -1);
1628         tcg_gen_atomic_add_fetch_tl(s1->T0, s1->A0, s1->T0,
1629                                     s1->mem_index, ot | MO_LE);
1630     } else {
1631         if (d != OR_TMP0) {
1632             gen_op_mov_v_reg(s1, ot, s1->T0, d);
1633         } else {
1634             gen_op_ld_v(s1, ot, s1->T0, s1->A0);
1635         }
1636         tcg_gen_addi_tl(s1->T0, s1->T0, (c > 0 ? 1 : -1));
1637         gen_op_st_rm_T0_A0(s1, ot, d);
1638     }
1639 
1640     gen_compute_eflags_c(s1, cpu_cc_src);
1641     tcg_gen_mov_tl(cpu_cc_dst, s1->T0);
1642     set_cc_op(s1, (c > 0 ? CC_OP_INCB : CC_OP_DECB) + ot);
1643 }
1644 
1645 static void gen_shift_flags(DisasContext *s, MemOp ot, TCGv result,
1646                             TCGv shm1, TCGv count, bool is_right)
1647 {
1648     TCGv_i32 z32, s32, oldop;
1649     TCGv z_tl;
1650 
1651     /* Store the results into the CC variables.  If we know that the
1652        variable must be dead, store unconditionally.  Otherwise we'll
1653        need to not disrupt the current contents.  */
1654     z_tl = tcg_constant_tl(0);
1655     if (cc_op_live[s->cc_op] & USES_CC_DST) {
1656         tcg_gen_movcond_tl(TCG_COND_NE, cpu_cc_dst, count, z_tl,
1657                            result, cpu_cc_dst);
1658     } else {
1659         tcg_gen_mov_tl(cpu_cc_dst, result);
1660     }
1661     if (cc_op_live[s->cc_op] & USES_CC_SRC) {
1662         tcg_gen_movcond_tl(TCG_COND_NE, cpu_cc_src, count, z_tl,
1663                            shm1, cpu_cc_src);
1664     } else {
1665         tcg_gen_mov_tl(cpu_cc_src, shm1);
1666     }
1667 
1668     /* Get the two potential CC_OP values into temporaries.  */
1669     tcg_gen_movi_i32(s->tmp2_i32, (is_right ? CC_OP_SARB : CC_OP_SHLB) + ot);
1670     if (s->cc_op == CC_OP_DYNAMIC) {
1671         oldop = cpu_cc_op;
1672     } else {
1673         tcg_gen_movi_i32(s->tmp3_i32, s->cc_op);
1674         oldop = s->tmp3_i32;
1675     }
1676 
1677     /* Conditionally store the CC_OP value.  */
1678     z32 = tcg_constant_i32(0);
1679     s32 = tcg_temp_new_i32();
1680     tcg_gen_trunc_tl_i32(s32, count);
1681     tcg_gen_movcond_i32(TCG_COND_NE, cpu_cc_op, s32, z32, s->tmp2_i32, oldop);
1682 
1683     /* The CC_OP value is no longer predictable.  */
1684     set_cc_op(s, CC_OP_DYNAMIC);
1685 }
1686 
1687 static void gen_shift_rm_T1(DisasContext *s, MemOp ot, int op1,
1688                             int is_right, int is_arith)
1689 {
1690     target_ulong mask = (ot == MO_64 ? 0x3f : 0x1f);
1691 
1692     /* load */
1693     if (op1 == OR_TMP0) {
1694         gen_op_ld_v(s, ot, s->T0, s->A0);
1695     } else {
1696         gen_op_mov_v_reg(s, ot, s->T0, op1);
1697     }
1698 
1699     tcg_gen_andi_tl(s->T1, s->T1, mask);
1700     tcg_gen_subi_tl(s->tmp0, s->T1, 1);
1701 
1702     if (is_right) {
1703         if (is_arith) {
1704             gen_exts(ot, s->T0);
1705             tcg_gen_sar_tl(s->tmp0, s->T0, s->tmp0);
1706             tcg_gen_sar_tl(s->T0, s->T0, s->T1);
1707         } else {
1708             gen_extu(ot, s->T0);
1709             tcg_gen_shr_tl(s->tmp0, s->T0, s->tmp0);
1710             tcg_gen_shr_tl(s->T0, s->T0, s->T1);
1711         }
1712     } else {
1713         tcg_gen_shl_tl(s->tmp0, s->T0, s->tmp0);
1714         tcg_gen_shl_tl(s->T0, s->T0, s->T1);
1715     }
1716 
1717     /* store */
1718     gen_op_st_rm_T0_A0(s, ot, op1);
1719 
1720     gen_shift_flags(s, ot, s->T0, s->tmp0, s->T1, is_right);
1721 }
1722 
1723 static void gen_shift_rm_im(DisasContext *s, MemOp ot, int op1, int op2,
1724                             int is_right, int is_arith)
1725 {
1726     int mask = (ot == MO_64 ? 0x3f : 0x1f);
1727 
1728     /* load */
1729     if (op1 == OR_TMP0)
1730         gen_op_ld_v(s, ot, s->T0, s->A0);
1731     else
1732         gen_op_mov_v_reg(s, ot, s->T0, op1);
1733 
1734     op2 &= mask;
1735     if (op2 != 0) {
1736         if (is_right) {
1737             if (is_arith) {
1738                 gen_exts(ot, s->T0);
1739                 tcg_gen_sari_tl(s->tmp4, s->T0, op2 - 1);
1740                 tcg_gen_sari_tl(s->T0, s->T0, op2);
1741             } else {
1742                 gen_extu(ot, s->T0);
1743                 tcg_gen_shri_tl(s->tmp4, s->T0, op2 - 1);
1744                 tcg_gen_shri_tl(s->T0, s->T0, op2);
1745             }
1746         } else {
1747             tcg_gen_shli_tl(s->tmp4, s->T0, op2 - 1);
1748             tcg_gen_shli_tl(s->T0, s->T0, op2);
1749         }
1750     }
1751 
1752     /* store */
1753     gen_op_st_rm_T0_A0(s, ot, op1);
1754 
1755     /* update eflags if non zero shift */
1756     if (op2 != 0) {
1757         tcg_gen_mov_tl(cpu_cc_src, s->tmp4);
1758         tcg_gen_mov_tl(cpu_cc_dst, s->T0);
1759         set_cc_op(s, (is_right ? CC_OP_SARB : CC_OP_SHLB) + ot);
1760     }
1761 }
1762 
1763 static void gen_rot_rm_T1(DisasContext *s, MemOp ot, int op1, int is_right)
1764 {
1765     target_ulong mask = (ot == MO_64 ? 0x3f : 0x1f);
1766     TCGv_i32 t0, t1;
1767 
1768     /* load */
1769     if (op1 == OR_TMP0) {
1770         gen_op_ld_v(s, ot, s->T0, s->A0);
1771     } else {
1772         gen_op_mov_v_reg(s, ot, s->T0, op1);
1773     }
1774 
1775     tcg_gen_andi_tl(s->T1, s->T1, mask);
1776 
1777     switch (ot) {
1778     case MO_8:
1779         /* Replicate the 8-bit input so that a 32-bit rotate works.  */
1780         tcg_gen_ext8u_tl(s->T0, s->T0);
1781         tcg_gen_muli_tl(s->T0, s->T0, 0x01010101);
1782         goto do_long;
1783     case MO_16:
1784         /* Replicate the 16-bit input so that a 32-bit rotate works.  */
1785         tcg_gen_deposit_tl(s->T0, s->T0, s->T0, 16, 16);
1786         goto do_long;
1787     do_long:
1788 #ifdef TARGET_X86_64
1789     case MO_32:
1790         tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
1791         tcg_gen_trunc_tl_i32(s->tmp3_i32, s->T1);
1792         if (is_right) {
1793             tcg_gen_rotr_i32(s->tmp2_i32, s->tmp2_i32, s->tmp3_i32);
1794         } else {
1795             tcg_gen_rotl_i32(s->tmp2_i32, s->tmp2_i32, s->tmp3_i32);
1796         }
1797         tcg_gen_extu_i32_tl(s->T0, s->tmp2_i32);
1798         break;
1799 #endif
1800     default:
1801         if (is_right) {
1802             tcg_gen_rotr_tl(s->T0, s->T0, s->T1);
1803         } else {
1804             tcg_gen_rotl_tl(s->T0, s->T0, s->T1);
1805         }
1806         break;
1807     }
1808 
1809     /* store */
1810     gen_op_st_rm_T0_A0(s, ot, op1);
1811 
1812     /* We'll need the flags computed into CC_SRC.  */
1813     gen_compute_eflags(s);
1814 
1815     /* The value that was "rotated out" is now present at the other end
1816        of the word.  Compute C into CC_DST and O into CC_SRC2.  Note that
1817        since we've computed the flags into CC_SRC, these variables are
1818        currently dead.  */
1819     if (is_right) {
1820         tcg_gen_shri_tl(cpu_cc_src2, s->T0, mask - 1);
1821         tcg_gen_shri_tl(cpu_cc_dst, s->T0, mask);
1822         tcg_gen_andi_tl(cpu_cc_dst, cpu_cc_dst, 1);
1823     } else {
1824         tcg_gen_shri_tl(cpu_cc_src2, s->T0, mask);
1825         tcg_gen_andi_tl(cpu_cc_dst, s->T0, 1);
1826     }
1827     tcg_gen_andi_tl(cpu_cc_src2, cpu_cc_src2, 1);
1828     tcg_gen_xor_tl(cpu_cc_src2, cpu_cc_src2, cpu_cc_dst);
1829 
1830     /* Now conditionally store the new CC_OP value.  If the shift count
1831        is 0 we keep the CC_OP_EFLAGS setting so that only CC_SRC is live.
1832        Otherwise reuse CC_OP_ADCOX which have the C and O flags split out
1833        exactly as we computed above.  */
1834     t0 = tcg_constant_i32(0);
1835     t1 = tcg_temp_new_i32();
1836     tcg_gen_trunc_tl_i32(t1, s->T1);
1837     tcg_gen_movi_i32(s->tmp2_i32, CC_OP_ADCOX);
1838     tcg_gen_movi_i32(s->tmp3_i32, CC_OP_EFLAGS);
1839     tcg_gen_movcond_i32(TCG_COND_NE, cpu_cc_op, t1, t0,
1840                         s->tmp2_i32, s->tmp3_i32);
1841 
1842     /* The CC_OP value is no longer predictable.  */
1843     set_cc_op(s, CC_OP_DYNAMIC);
1844 }
1845 
1846 static void gen_rot_rm_im(DisasContext *s, MemOp ot, int op1, int op2,
1847                           int is_right)
1848 {
1849     int mask = (ot == MO_64 ? 0x3f : 0x1f);
1850     int shift;
1851 
1852     /* load */
1853     if (op1 == OR_TMP0) {
1854         gen_op_ld_v(s, ot, s->T0, s->A0);
1855     } else {
1856         gen_op_mov_v_reg(s, ot, s->T0, op1);
1857     }
1858 
1859     op2 &= mask;
1860     if (op2 != 0) {
1861         switch (ot) {
1862 #ifdef TARGET_X86_64
1863         case MO_32:
1864             tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
1865             if (is_right) {
1866                 tcg_gen_rotri_i32(s->tmp2_i32, s->tmp2_i32, op2);
1867             } else {
1868                 tcg_gen_rotli_i32(s->tmp2_i32, s->tmp2_i32, op2);
1869             }
1870             tcg_gen_extu_i32_tl(s->T0, s->tmp2_i32);
1871             break;
1872 #endif
1873         default:
1874             if (is_right) {
1875                 tcg_gen_rotri_tl(s->T0, s->T0, op2);
1876             } else {
1877                 tcg_gen_rotli_tl(s->T0, s->T0, op2);
1878             }
1879             break;
1880         case MO_8:
1881             mask = 7;
1882             goto do_shifts;
1883         case MO_16:
1884             mask = 15;
1885         do_shifts:
1886             shift = op2 & mask;
1887             if (is_right) {
1888                 shift = mask + 1 - shift;
1889             }
1890             gen_extu(ot, s->T0);
1891             tcg_gen_shli_tl(s->tmp0, s->T0, shift);
1892             tcg_gen_shri_tl(s->T0, s->T0, mask + 1 - shift);
1893             tcg_gen_or_tl(s->T0, s->T0, s->tmp0);
1894             break;
1895         }
1896     }
1897 
1898     /* store */
1899     gen_op_st_rm_T0_A0(s, ot, op1);
1900 
1901     if (op2 != 0) {
1902         /* Compute the flags into CC_SRC.  */
1903         gen_compute_eflags(s);
1904 
1905         /* The value that was "rotated out" is now present at the other end
1906            of the word.  Compute C into CC_DST and O into CC_SRC2.  Note that
1907            since we've computed the flags into CC_SRC, these variables are
1908            currently dead.  */
1909         if (is_right) {
1910             tcg_gen_shri_tl(cpu_cc_src2, s->T0, mask - 1);
1911             tcg_gen_shri_tl(cpu_cc_dst, s->T0, mask);
1912             tcg_gen_andi_tl(cpu_cc_dst, cpu_cc_dst, 1);
1913         } else {
1914             tcg_gen_shri_tl(cpu_cc_src2, s->T0, mask);
1915             tcg_gen_andi_tl(cpu_cc_dst, s->T0, 1);
1916         }
1917         tcg_gen_andi_tl(cpu_cc_src2, cpu_cc_src2, 1);
1918         tcg_gen_xor_tl(cpu_cc_src2, cpu_cc_src2, cpu_cc_dst);
1919         set_cc_op(s, CC_OP_ADCOX);
1920     }
1921 }
1922 
1923 /* XXX: add faster immediate = 1 case */
1924 static void gen_rotc_rm_T1(DisasContext *s, MemOp ot, int op1,
1925                            int is_right)
1926 {
1927     gen_compute_eflags(s);
1928     assert(s->cc_op == CC_OP_EFLAGS);
1929 
1930     /* load */
1931     if (op1 == OR_TMP0)
1932         gen_op_ld_v(s, ot, s->T0, s->A0);
1933     else
1934         gen_op_mov_v_reg(s, ot, s->T0, op1);
1935 
1936     if (is_right) {
1937         switch (ot) {
1938         case MO_8:
1939             gen_helper_rcrb(s->T0, tcg_env, s->T0, s->T1);
1940             break;
1941         case MO_16:
1942             gen_helper_rcrw(s->T0, tcg_env, s->T0, s->T1);
1943             break;
1944         case MO_32:
1945             gen_helper_rcrl(s->T0, tcg_env, s->T0, s->T1);
1946             break;
1947 #ifdef TARGET_X86_64
1948         case MO_64:
1949             gen_helper_rcrq(s->T0, tcg_env, s->T0, s->T1);
1950             break;
1951 #endif
1952         default:
1953             g_assert_not_reached();
1954         }
1955     } else {
1956         switch (ot) {
1957         case MO_8:
1958             gen_helper_rclb(s->T0, tcg_env, s->T0, s->T1);
1959             break;
1960         case MO_16:
1961             gen_helper_rclw(s->T0, tcg_env, s->T0, s->T1);
1962             break;
1963         case MO_32:
1964             gen_helper_rcll(s->T0, tcg_env, s->T0, s->T1);
1965             break;
1966 #ifdef TARGET_X86_64
1967         case MO_64:
1968             gen_helper_rclq(s->T0, tcg_env, s->T0, s->T1);
1969             break;
1970 #endif
1971         default:
1972             g_assert_not_reached();
1973         }
1974     }
1975     /* store */
1976     gen_op_st_rm_T0_A0(s, ot, op1);
1977 }
1978 
1979 /* XXX: add faster immediate case */
1980 static void gen_shiftd_rm_T1(DisasContext *s, MemOp ot, int op1,
1981                              bool is_right, TCGv count_in)
1982 {
1983     target_ulong mask = (ot == MO_64 ? 63 : 31);
1984     TCGv count;
1985 
1986     /* load */
1987     if (op1 == OR_TMP0) {
1988         gen_op_ld_v(s, ot, s->T0, s->A0);
1989     } else {
1990         gen_op_mov_v_reg(s, ot, s->T0, op1);
1991     }
1992 
1993     count = tcg_temp_new();
1994     tcg_gen_andi_tl(count, count_in, mask);
1995 
1996     switch (ot) {
1997     case MO_16:
1998         /* Note: we implement the Intel behaviour for shift count > 16.
1999            This means "shrdw C, B, A" shifts A:B:A >> C.  Build the B:A
2000            portion by constructing it as a 32-bit value.  */
2001         if (is_right) {
2002             tcg_gen_deposit_tl(s->tmp0, s->T0, s->T1, 16, 16);
2003             tcg_gen_mov_tl(s->T1, s->T0);
2004             tcg_gen_mov_tl(s->T0, s->tmp0);
2005         } else {
2006             tcg_gen_deposit_tl(s->T1, s->T0, s->T1, 16, 16);
2007         }
2008         /*
2009          * If TARGET_X86_64 defined then fall through into MO_32 case,
2010          * otherwise fall through default case.
2011          */
2012     case MO_32:
2013 #ifdef TARGET_X86_64
2014         /* Concatenate the two 32-bit values and use a 64-bit shift.  */
2015         tcg_gen_subi_tl(s->tmp0, count, 1);
2016         if (is_right) {
2017             tcg_gen_concat_tl_i64(s->T0, s->T0, s->T1);
2018             tcg_gen_shr_i64(s->tmp0, s->T0, s->tmp0);
2019             tcg_gen_shr_i64(s->T0, s->T0, count);
2020         } else {
2021             tcg_gen_concat_tl_i64(s->T0, s->T1, s->T0);
2022             tcg_gen_shl_i64(s->tmp0, s->T0, s->tmp0);
2023             tcg_gen_shl_i64(s->T0, s->T0, count);
2024             tcg_gen_shri_i64(s->tmp0, s->tmp0, 32);
2025             tcg_gen_shri_i64(s->T0, s->T0, 32);
2026         }
2027         break;
2028 #endif
2029     default:
2030         tcg_gen_subi_tl(s->tmp0, count, 1);
2031         if (is_right) {
2032             tcg_gen_shr_tl(s->tmp0, s->T0, s->tmp0);
2033 
2034             tcg_gen_subfi_tl(s->tmp4, mask + 1, count);
2035             tcg_gen_shr_tl(s->T0, s->T0, count);
2036             tcg_gen_shl_tl(s->T1, s->T1, s->tmp4);
2037         } else {
2038             tcg_gen_shl_tl(s->tmp0, s->T0, s->tmp0);
2039             if (ot == MO_16) {
2040                 /* Only needed if count > 16, for Intel behaviour.  */
2041                 tcg_gen_subfi_tl(s->tmp4, 33, count);
2042                 tcg_gen_shr_tl(s->tmp4, s->T1, s->tmp4);
2043                 tcg_gen_or_tl(s->tmp0, s->tmp0, s->tmp4);
2044             }
2045 
2046             tcg_gen_subfi_tl(s->tmp4, mask + 1, count);
2047             tcg_gen_shl_tl(s->T0, s->T0, count);
2048             tcg_gen_shr_tl(s->T1, s->T1, s->tmp4);
2049         }
2050         tcg_gen_movi_tl(s->tmp4, 0);
2051         tcg_gen_movcond_tl(TCG_COND_EQ, s->T1, count, s->tmp4,
2052                            s->tmp4, s->T1);
2053         tcg_gen_or_tl(s->T0, s->T0, s->T1);
2054         break;
2055     }
2056 
2057     /* store */
2058     gen_op_st_rm_T0_A0(s, ot, op1);
2059 
2060     gen_shift_flags(s, ot, s->T0, s->tmp0, count, is_right);
2061 }
2062 
2063 static void gen_shift(DisasContext *s1, int op, MemOp ot, int d, int s)
2064 {
2065     if (s != OR_TMP1)
2066         gen_op_mov_v_reg(s1, ot, s1->T1, s);
2067     switch(op) {
2068     case OP_ROL:
2069         gen_rot_rm_T1(s1, ot, d, 0);
2070         break;
2071     case OP_ROR:
2072         gen_rot_rm_T1(s1, ot, d, 1);
2073         break;
2074     case OP_SHL:
2075     case OP_SHL1:
2076         gen_shift_rm_T1(s1, ot, d, 0, 0);
2077         break;
2078     case OP_SHR:
2079         gen_shift_rm_T1(s1, ot, d, 1, 0);
2080         break;
2081     case OP_SAR:
2082         gen_shift_rm_T1(s1, ot, d, 1, 1);
2083         break;
2084     case OP_RCL:
2085         gen_rotc_rm_T1(s1, ot, d, 0);
2086         break;
2087     case OP_RCR:
2088         gen_rotc_rm_T1(s1, ot, d, 1);
2089         break;
2090     }
2091 }
2092 
2093 static void gen_shifti(DisasContext *s1, int op, MemOp ot, int d, int c)
2094 {
2095     switch(op) {
2096     case OP_ROL:
2097         gen_rot_rm_im(s1, ot, d, c, 0);
2098         break;
2099     case OP_ROR:
2100         gen_rot_rm_im(s1, ot, d, c, 1);
2101         break;
2102     case OP_SHL:
2103     case OP_SHL1:
2104         gen_shift_rm_im(s1, ot, d, c, 0, 0);
2105         break;
2106     case OP_SHR:
2107         gen_shift_rm_im(s1, ot, d, c, 1, 0);
2108         break;
2109     case OP_SAR:
2110         gen_shift_rm_im(s1, ot, d, c, 1, 1);
2111         break;
2112     default:
2113         /* currently not optimized */
2114         tcg_gen_movi_tl(s1->T1, c);
2115         gen_shift(s1, op, ot, d, OR_TMP1);
2116         break;
2117     }
2118 }
2119 
2120 #define X86_MAX_INSN_LENGTH 15
2121 
2122 static uint64_t advance_pc(CPUX86State *env, DisasContext *s, int num_bytes)
2123 {
2124     uint64_t pc = s->pc;
2125 
2126     /* This is a subsequent insn that crosses a page boundary.  */
2127     if (s->base.num_insns > 1 &&
2128         !is_same_page(&s->base, s->pc + num_bytes - 1)) {
2129         siglongjmp(s->jmpbuf, 2);
2130     }
2131 
2132     s->pc += num_bytes;
2133     if (unlikely(cur_insn_len(s) > X86_MAX_INSN_LENGTH)) {
2134         /* If the instruction's 16th byte is on a different page than the 1st, a
2135          * page fault on the second page wins over the general protection fault
2136          * caused by the instruction being too long.
2137          * This can happen even if the operand is only one byte long!
2138          */
2139         if (((s->pc - 1) ^ (pc - 1)) & TARGET_PAGE_MASK) {
2140             volatile uint8_t unused =
2141                 cpu_ldub_code(env, (s->pc - 1) & TARGET_PAGE_MASK);
2142             (void) unused;
2143         }
2144         siglongjmp(s->jmpbuf, 1);
2145     }
2146 
2147     return pc;
2148 }
2149 
2150 static inline uint8_t x86_ldub_code(CPUX86State *env, DisasContext *s)
2151 {
2152     return translator_ldub(env, &s->base, advance_pc(env, s, 1));
2153 }
2154 
2155 static inline int16_t x86_ldsw_code(CPUX86State *env, DisasContext *s)
2156 {
2157     return translator_lduw(env, &s->base, advance_pc(env, s, 2));
2158 }
2159 
2160 static inline uint16_t x86_lduw_code(CPUX86State *env, DisasContext *s)
2161 {
2162     return translator_lduw(env, &s->base, advance_pc(env, s, 2));
2163 }
2164 
2165 static inline uint32_t x86_ldl_code(CPUX86State *env, DisasContext *s)
2166 {
2167     return translator_ldl(env, &s->base, advance_pc(env, s, 4));
2168 }
2169 
2170 #ifdef TARGET_X86_64
2171 static inline uint64_t x86_ldq_code(CPUX86State *env, DisasContext *s)
2172 {
2173     return translator_ldq(env, &s->base, advance_pc(env, s, 8));
2174 }
2175 #endif
2176 
2177 /* Decompose an address.  */
2178 
2179 typedef struct AddressParts {
2180     int def_seg;
2181     int base;
2182     int index;
2183     int scale;
2184     target_long disp;
2185 } AddressParts;
2186 
2187 static AddressParts gen_lea_modrm_0(CPUX86State *env, DisasContext *s,
2188                                     int modrm)
2189 {
2190     int def_seg, base, index, scale, mod, rm;
2191     target_long disp;
2192     bool havesib;
2193 
2194     def_seg = R_DS;
2195     index = -1;
2196     scale = 0;
2197     disp = 0;
2198 
2199     mod = (modrm >> 6) & 3;
2200     rm = modrm & 7;
2201     base = rm | REX_B(s);
2202 
2203     if (mod == 3) {
2204         /* Normally filtered out earlier, but including this path
2205            simplifies multi-byte nop, as well as bndcl, bndcu, bndcn.  */
2206         goto done;
2207     }
2208 
2209     switch (s->aflag) {
2210     case MO_64:
2211     case MO_32:
2212         havesib = 0;
2213         if (rm == 4) {
2214             int code = x86_ldub_code(env, s);
2215             scale = (code >> 6) & 3;
2216             index = ((code >> 3) & 7) | REX_X(s);
2217             if (index == 4) {
2218                 index = -1;  /* no index */
2219             }
2220             base = (code & 7) | REX_B(s);
2221             havesib = 1;
2222         }
2223 
2224         switch (mod) {
2225         case 0:
2226             if ((base & 7) == 5) {
2227                 base = -1;
2228                 disp = (int32_t)x86_ldl_code(env, s);
2229                 if (CODE64(s) && !havesib) {
2230                     base = -2;
2231                     disp += s->pc + s->rip_offset;
2232                 }
2233             }
2234             break;
2235         case 1:
2236             disp = (int8_t)x86_ldub_code(env, s);
2237             break;
2238         default:
2239         case 2:
2240             disp = (int32_t)x86_ldl_code(env, s);
2241             break;
2242         }
2243 
2244         /* For correct popl handling with esp.  */
2245         if (base == R_ESP && s->popl_esp_hack) {
2246             disp += s->popl_esp_hack;
2247         }
2248         if (base == R_EBP || base == R_ESP) {
2249             def_seg = R_SS;
2250         }
2251         break;
2252 
2253     case MO_16:
2254         if (mod == 0) {
2255             if (rm == 6) {
2256                 base = -1;
2257                 disp = x86_lduw_code(env, s);
2258                 break;
2259             }
2260         } else if (mod == 1) {
2261             disp = (int8_t)x86_ldub_code(env, s);
2262         } else {
2263             disp = (int16_t)x86_lduw_code(env, s);
2264         }
2265 
2266         switch (rm) {
2267         case 0:
2268             base = R_EBX;
2269             index = R_ESI;
2270             break;
2271         case 1:
2272             base = R_EBX;
2273             index = R_EDI;
2274             break;
2275         case 2:
2276             base = R_EBP;
2277             index = R_ESI;
2278             def_seg = R_SS;
2279             break;
2280         case 3:
2281             base = R_EBP;
2282             index = R_EDI;
2283             def_seg = R_SS;
2284             break;
2285         case 4:
2286             base = R_ESI;
2287             break;
2288         case 5:
2289             base = R_EDI;
2290             break;
2291         case 6:
2292             base = R_EBP;
2293             def_seg = R_SS;
2294             break;
2295         default:
2296         case 7:
2297             base = R_EBX;
2298             break;
2299         }
2300         break;
2301 
2302     default:
2303         g_assert_not_reached();
2304     }
2305 
2306  done:
2307     return (AddressParts){ def_seg, base, index, scale, disp };
2308 }
2309 
2310 /* Compute the address, with a minimum number of TCG ops.  */
2311 static TCGv gen_lea_modrm_1(DisasContext *s, AddressParts a, bool is_vsib)
2312 {
2313     TCGv ea = NULL;
2314 
2315     if (a.index >= 0 && !is_vsib) {
2316         if (a.scale == 0) {
2317             ea = cpu_regs[a.index];
2318         } else {
2319             tcg_gen_shli_tl(s->A0, cpu_regs[a.index], a.scale);
2320             ea = s->A0;
2321         }
2322         if (a.base >= 0) {
2323             tcg_gen_add_tl(s->A0, ea, cpu_regs[a.base]);
2324             ea = s->A0;
2325         }
2326     } else if (a.base >= 0) {
2327         ea = cpu_regs[a.base];
2328     }
2329     if (!ea) {
2330         if (tb_cflags(s->base.tb) & CF_PCREL && a.base == -2) {
2331             /* With cpu_eip ~= pc_save, the expression is pc-relative. */
2332             tcg_gen_addi_tl(s->A0, cpu_eip, a.disp - s->pc_save);
2333         } else {
2334             tcg_gen_movi_tl(s->A0, a.disp);
2335         }
2336         ea = s->A0;
2337     } else if (a.disp != 0) {
2338         tcg_gen_addi_tl(s->A0, ea, a.disp);
2339         ea = s->A0;
2340     }
2341 
2342     return ea;
2343 }
2344 
2345 static void gen_lea_modrm(CPUX86State *env, DisasContext *s, int modrm)
2346 {
2347     AddressParts a = gen_lea_modrm_0(env, s, modrm);
2348     TCGv ea = gen_lea_modrm_1(s, a, false);
2349     gen_lea_v_seg(s, s->aflag, ea, a.def_seg, s->override);
2350 }
2351 
2352 static void gen_nop_modrm(CPUX86State *env, DisasContext *s, int modrm)
2353 {
2354     (void)gen_lea_modrm_0(env, s, modrm);
2355 }
2356 
2357 /* Used for BNDCL, BNDCU, BNDCN.  */
2358 static void gen_bndck(CPUX86State *env, DisasContext *s, int modrm,
2359                       TCGCond cond, TCGv_i64 bndv)
2360 {
2361     AddressParts a = gen_lea_modrm_0(env, s, modrm);
2362     TCGv ea = gen_lea_modrm_1(s, a, false);
2363 
2364     tcg_gen_extu_tl_i64(s->tmp1_i64, ea);
2365     if (!CODE64(s)) {
2366         tcg_gen_ext32u_i64(s->tmp1_i64, s->tmp1_i64);
2367     }
2368     tcg_gen_setcond_i64(cond, s->tmp1_i64, s->tmp1_i64, bndv);
2369     tcg_gen_extrl_i64_i32(s->tmp2_i32, s->tmp1_i64);
2370     gen_helper_bndck(tcg_env, s->tmp2_i32);
2371 }
2372 
2373 /* used for LEA and MOV AX, mem */
2374 static void gen_add_A0_ds_seg(DisasContext *s)
2375 {
2376     gen_lea_v_seg(s, s->aflag, s->A0, R_DS, s->override);
2377 }
2378 
2379 /* generate modrm memory load or store of 'reg'. TMP0 is used if reg ==
2380    OR_TMP0 */
2381 static void gen_ldst_modrm(CPUX86State *env, DisasContext *s, int modrm,
2382                            MemOp ot, int reg, int is_store)
2383 {
2384     int mod, rm;
2385 
2386     mod = (modrm >> 6) & 3;
2387     rm = (modrm & 7) | REX_B(s);
2388     if (mod == 3) {
2389         if (is_store) {
2390             if (reg != OR_TMP0)
2391                 gen_op_mov_v_reg(s, ot, s->T0, reg);
2392             gen_op_mov_reg_v(s, ot, rm, s->T0);
2393         } else {
2394             gen_op_mov_v_reg(s, ot, s->T0, rm);
2395             if (reg != OR_TMP0)
2396                 gen_op_mov_reg_v(s, ot, reg, s->T0);
2397         }
2398     } else {
2399         gen_lea_modrm(env, s, modrm);
2400         if (is_store) {
2401             if (reg != OR_TMP0)
2402                 gen_op_mov_v_reg(s, ot, s->T0, reg);
2403             gen_op_st_v(s, ot, s->T0, s->A0);
2404         } else {
2405             gen_op_ld_v(s, ot, s->T0, s->A0);
2406             if (reg != OR_TMP0)
2407                 gen_op_mov_reg_v(s, ot, reg, s->T0);
2408         }
2409     }
2410 }
2411 
2412 static target_ulong insn_get_addr(CPUX86State *env, DisasContext *s, MemOp ot)
2413 {
2414     target_ulong ret;
2415 
2416     switch (ot) {
2417     case MO_8:
2418         ret = x86_ldub_code(env, s);
2419         break;
2420     case MO_16:
2421         ret = x86_lduw_code(env, s);
2422         break;
2423     case MO_32:
2424         ret = x86_ldl_code(env, s);
2425         break;
2426 #ifdef TARGET_X86_64
2427     case MO_64:
2428         ret = x86_ldq_code(env, s);
2429         break;
2430 #endif
2431     default:
2432         g_assert_not_reached();
2433     }
2434     return ret;
2435 }
2436 
2437 static inline uint32_t insn_get(CPUX86State *env, DisasContext *s, MemOp ot)
2438 {
2439     uint32_t ret;
2440 
2441     switch (ot) {
2442     case MO_8:
2443         ret = x86_ldub_code(env, s);
2444         break;
2445     case MO_16:
2446         ret = x86_lduw_code(env, s);
2447         break;
2448     case MO_32:
2449 #ifdef TARGET_X86_64
2450     case MO_64:
2451 #endif
2452         ret = x86_ldl_code(env, s);
2453         break;
2454     default:
2455         g_assert_not_reached();
2456     }
2457     return ret;
2458 }
2459 
2460 static target_long insn_get_signed(CPUX86State *env, DisasContext *s, MemOp ot)
2461 {
2462     target_long ret;
2463 
2464     switch (ot) {
2465     case MO_8:
2466         ret = (int8_t) x86_ldub_code(env, s);
2467         break;
2468     case MO_16:
2469         ret = (int16_t) x86_lduw_code(env, s);
2470         break;
2471     case MO_32:
2472         ret = (int32_t) x86_ldl_code(env, s);
2473         break;
2474 #ifdef TARGET_X86_64
2475     case MO_64:
2476         ret = x86_ldq_code(env, s);
2477         break;
2478 #endif
2479     default:
2480         g_assert_not_reached();
2481     }
2482     return ret;
2483 }
2484 
2485 static inline int insn_const_size(MemOp ot)
2486 {
2487     if (ot <= MO_32) {
2488         return 1 << ot;
2489     } else {
2490         return 4;
2491     }
2492 }
2493 
2494 static void gen_jcc(DisasContext *s, int b, int diff)
2495 {
2496     TCGLabel *l1 = gen_new_label();
2497 
2498     gen_jcc1(s, b, l1);
2499     gen_jmp_rel_csize(s, 0, 1);
2500     gen_set_label(l1);
2501     gen_jmp_rel(s, s->dflag, diff, 0);
2502 }
2503 
2504 static void gen_cmovcc1(DisasContext *s, int b, TCGv dest, TCGv src)
2505 {
2506     CCPrepare cc = gen_prepare_cc(s, b, s->T1);
2507 
2508     if (cc.mask != -1) {
2509         TCGv t0 = tcg_temp_new();
2510         tcg_gen_andi_tl(t0, cc.reg, cc.mask);
2511         cc.reg = t0;
2512     }
2513     if (!cc.use_reg2) {
2514         cc.reg2 = tcg_constant_tl(cc.imm);
2515     }
2516 
2517     tcg_gen_movcond_tl(cc.cond, dest, cc.reg, cc.reg2, src, dest);
2518 }
2519 
2520 static inline void gen_op_movl_T0_seg(DisasContext *s, X86Seg seg_reg)
2521 {
2522     tcg_gen_ld32u_tl(s->T0, tcg_env,
2523                      offsetof(CPUX86State,segs[seg_reg].selector));
2524 }
2525 
2526 static inline void gen_op_movl_seg_T0_vm(DisasContext *s, X86Seg seg_reg)
2527 {
2528     tcg_gen_ext16u_tl(s->T0, s->T0);
2529     tcg_gen_st32_tl(s->T0, tcg_env,
2530                     offsetof(CPUX86State,segs[seg_reg].selector));
2531     tcg_gen_shli_tl(cpu_seg_base[seg_reg], s->T0, 4);
2532 }
2533 
2534 /* move T0 to seg_reg and compute if the CPU state may change. Never
2535    call this function with seg_reg == R_CS */
2536 static void gen_movl_seg_T0(DisasContext *s, X86Seg seg_reg)
2537 {
2538     if (PE(s) && !VM86(s)) {
2539         tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
2540         gen_helper_load_seg(tcg_env, tcg_constant_i32(seg_reg), s->tmp2_i32);
2541         /* abort translation because the addseg value may change or
2542            because ss32 may change. For R_SS, translation must always
2543            stop as a special handling must be done to disable hardware
2544            interrupts for the next instruction */
2545         if (seg_reg == R_SS) {
2546             s->base.is_jmp = DISAS_EOB_INHIBIT_IRQ;
2547         } else if (CODE32(s) && seg_reg < R_FS) {
2548             s->base.is_jmp = DISAS_EOB_NEXT;
2549         }
2550     } else {
2551         gen_op_movl_seg_T0_vm(s, seg_reg);
2552         if (seg_reg == R_SS) {
2553             s->base.is_jmp = DISAS_EOB_INHIBIT_IRQ;
2554         }
2555     }
2556 }
2557 
2558 static void gen_svm_check_intercept(DisasContext *s, uint32_t type)
2559 {
2560     /* no SVM activated; fast case */
2561     if (likely(!GUEST(s))) {
2562         return;
2563     }
2564     gen_helper_svm_check_intercept(tcg_env, tcg_constant_i32(type));
2565 }
2566 
2567 static inline void gen_stack_update(DisasContext *s, int addend)
2568 {
2569     gen_op_add_reg_im(s, mo_stacksize(s), R_ESP, addend);
2570 }
2571 
2572 /* Generate a push. It depends on ss32, addseg and dflag.  */
2573 static void gen_push_v(DisasContext *s, TCGv val)
2574 {
2575     MemOp d_ot = mo_pushpop(s, s->dflag);
2576     MemOp a_ot = mo_stacksize(s);
2577     int size = 1 << d_ot;
2578     TCGv new_esp = s->A0;
2579 
2580     tcg_gen_subi_tl(s->A0, cpu_regs[R_ESP], size);
2581 
2582     if (!CODE64(s)) {
2583         if (ADDSEG(s)) {
2584             new_esp = tcg_temp_new();
2585             tcg_gen_mov_tl(new_esp, s->A0);
2586         }
2587         gen_lea_v_seg(s, a_ot, s->A0, R_SS, -1);
2588     }
2589 
2590     gen_op_st_v(s, d_ot, val, s->A0);
2591     gen_op_mov_reg_v(s, a_ot, R_ESP, new_esp);
2592 }
2593 
2594 /* two step pop is necessary for precise exceptions */
2595 static MemOp gen_pop_T0(DisasContext *s)
2596 {
2597     MemOp d_ot = mo_pushpop(s, s->dflag);
2598 
2599     gen_lea_v_seg_dest(s, mo_stacksize(s), s->T0, cpu_regs[R_ESP], R_SS, -1);
2600     gen_op_ld_v(s, d_ot, s->T0, s->T0);
2601 
2602     return d_ot;
2603 }
2604 
2605 static inline void gen_pop_update(DisasContext *s, MemOp ot)
2606 {
2607     gen_stack_update(s, 1 << ot);
2608 }
2609 
2610 static inline void gen_stack_A0(DisasContext *s)
2611 {
2612     gen_lea_v_seg(s, SS32(s) ? MO_32 : MO_16, cpu_regs[R_ESP], R_SS, -1);
2613 }
2614 
2615 static void gen_pusha(DisasContext *s)
2616 {
2617     MemOp s_ot = SS32(s) ? MO_32 : MO_16;
2618     MemOp d_ot = s->dflag;
2619     int size = 1 << d_ot;
2620     int i;
2621 
2622     for (i = 0; i < 8; i++) {
2623         tcg_gen_addi_tl(s->A0, cpu_regs[R_ESP], (i - 8) * size);
2624         gen_lea_v_seg(s, s_ot, s->A0, R_SS, -1);
2625         gen_op_st_v(s, d_ot, cpu_regs[7 - i], s->A0);
2626     }
2627 
2628     gen_stack_update(s, -8 * size);
2629 }
2630 
2631 static void gen_popa(DisasContext *s)
2632 {
2633     MemOp s_ot = SS32(s) ? MO_32 : MO_16;
2634     MemOp d_ot = s->dflag;
2635     int size = 1 << d_ot;
2636     int i;
2637 
2638     for (i = 0; i < 8; i++) {
2639         /* ESP is not reloaded */
2640         if (7 - i == R_ESP) {
2641             continue;
2642         }
2643         tcg_gen_addi_tl(s->A0, cpu_regs[R_ESP], i * size);
2644         gen_lea_v_seg(s, s_ot, s->A0, R_SS, -1);
2645         gen_op_ld_v(s, d_ot, s->T0, s->A0);
2646         gen_op_mov_reg_v(s, d_ot, 7 - i, s->T0);
2647     }
2648 
2649     gen_stack_update(s, 8 * size);
2650 }
2651 
2652 static void gen_enter(DisasContext *s, int esp_addend, int level)
2653 {
2654     MemOp d_ot = mo_pushpop(s, s->dflag);
2655     MemOp a_ot = CODE64(s) ? MO_64 : SS32(s) ? MO_32 : MO_16;
2656     int size = 1 << d_ot;
2657 
2658     /* Push BP; compute FrameTemp into T1.  */
2659     tcg_gen_subi_tl(s->T1, cpu_regs[R_ESP], size);
2660     gen_lea_v_seg(s, a_ot, s->T1, R_SS, -1);
2661     gen_op_st_v(s, d_ot, cpu_regs[R_EBP], s->A0);
2662 
2663     level &= 31;
2664     if (level != 0) {
2665         int i;
2666 
2667         /* Copy level-1 pointers from the previous frame.  */
2668         for (i = 1; i < level; ++i) {
2669             tcg_gen_subi_tl(s->A0, cpu_regs[R_EBP], size * i);
2670             gen_lea_v_seg(s, a_ot, s->A0, R_SS, -1);
2671             gen_op_ld_v(s, d_ot, s->tmp0, s->A0);
2672 
2673             tcg_gen_subi_tl(s->A0, s->T1, size * i);
2674             gen_lea_v_seg(s, a_ot, s->A0, R_SS, -1);
2675             gen_op_st_v(s, d_ot, s->tmp0, s->A0);
2676         }
2677 
2678         /* Push the current FrameTemp as the last level.  */
2679         tcg_gen_subi_tl(s->A0, s->T1, size * level);
2680         gen_lea_v_seg(s, a_ot, s->A0, R_SS, -1);
2681         gen_op_st_v(s, d_ot, s->T1, s->A0);
2682     }
2683 
2684     /* Copy the FrameTemp value to EBP.  */
2685     gen_op_mov_reg_v(s, a_ot, R_EBP, s->T1);
2686 
2687     /* Compute the final value of ESP.  */
2688     tcg_gen_subi_tl(s->T1, s->T1, esp_addend + size * level);
2689     gen_op_mov_reg_v(s, a_ot, R_ESP, s->T1);
2690 }
2691 
2692 static void gen_leave(DisasContext *s)
2693 {
2694     MemOp d_ot = mo_pushpop(s, s->dflag);
2695     MemOp a_ot = mo_stacksize(s);
2696 
2697     gen_lea_v_seg(s, a_ot, cpu_regs[R_EBP], R_SS, -1);
2698     gen_op_ld_v(s, d_ot, s->T0, s->A0);
2699 
2700     tcg_gen_addi_tl(s->T1, cpu_regs[R_EBP], 1 << d_ot);
2701 
2702     gen_op_mov_reg_v(s, d_ot, R_EBP, s->T0);
2703     gen_op_mov_reg_v(s, a_ot, R_ESP, s->T1);
2704 }
2705 
2706 /* Similarly, except that the assumption here is that we don't decode
2707    the instruction at all -- either a missing opcode, an unimplemented
2708    feature, or just a bogus instruction stream.  */
2709 static void gen_unknown_opcode(CPUX86State *env, DisasContext *s)
2710 {
2711     gen_illegal_opcode(s);
2712 
2713     if (qemu_loglevel_mask(LOG_UNIMP)) {
2714         FILE *logfile = qemu_log_trylock();
2715         if (logfile) {
2716             target_ulong pc = s->base.pc_next, end = s->pc;
2717 
2718             fprintf(logfile, "ILLOPC: " TARGET_FMT_lx ":", pc);
2719             for (; pc < end; ++pc) {
2720                 fprintf(logfile, " %02x", cpu_ldub_code(env, pc));
2721             }
2722             fprintf(logfile, "\n");
2723             qemu_log_unlock(logfile);
2724         }
2725     }
2726 }
2727 
2728 /* an interrupt is different from an exception because of the
2729    privilege checks */
2730 static void gen_interrupt(DisasContext *s, int intno)
2731 {
2732     gen_update_cc_op(s);
2733     gen_update_eip_cur(s);
2734     gen_helper_raise_interrupt(tcg_env, tcg_constant_i32(intno),
2735                                cur_insn_len_i32(s));
2736     s->base.is_jmp = DISAS_NORETURN;
2737 }
2738 
2739 static void gen_set_hflag(DisasContext *s, uint32_t mask)
2740 {
2741     if ((s->flags & mask) == 0) {
2742         TCGv_i32 t = tcg_temp_new_i32();
2743         tcg_gen_ld_i32(t, tcg_env, offsetof(CPUX86State, hflags));
2744         tcg_gen_ori_i32(t, t, mask);
2745         tcg_gen_st_i32(t, tcg_env, offsetof(CPUX86State, hflags));
2746         s->flags |= mask;
2747     }
2748 }
2749 
2750 static void gen_reset_hflag(DisasContext *s, uint32_t mask)
2751 {
2752     if (s->flags & mask) {
2753         TCGv_i32 t = tcg_temp_new_i32();
2754         tcg_gen_ld_i32(t, tcg_env, offsetof(CPUX86State, hflags));
2755         tcg_gen_andi_i32(t, t, ~mask);
2756         tcg_gen_st_i32(t, tcg_env, offsetof(CPUX86State, hflags));
2757         s->flags &= ~mask;
2758     }
2759 }
2760 
2761 static void gen_set_eflags(DisasContext *s, target_ulong mask)
2762 {
2763     TCGv t = tcg_temp_new();
2764 
2765     tcg_gen_ld_tl(t, tcg_env, offsetof(CPUX86State, eflags));
2766     tcg_gen_ori_tl(t, t, mask);
2767     tcg_gen_st_tl(t, tcg_env, offsetof(CPUX86State, eflags));
2768 }
2769 
2770 static void gen_reset_eflags(DisasContext *s, target_ulong mask)
2771 {
2772     TCGv t = tcg_temp_new();
2773 
2774     tcg_gen_ld_tl(t, tcg_env, offsetof(CPUX86State, eflags));
2775     tcg_gen_andi_tl(t, t, ~mask);
2776     tcg_gen_st_tl(t, tcg_env, offsetof(CPUX86State, eflags));
2777 }
2778 
2779 /* Clear BND registers during legacy branches.  */
2780 static void gen_bnd_jmp(DisasContext *s)
2781 {
2782     /* Clear the registers only if BND prefix is missing, MPX is enabled,
2783        and if the BNDREGs are known to be in use (non-zero) already.
2784        The helper itself will check BNDPRESERVE at runtime.  */
2785     if ((s->prefix & PREFIX_REPNZ) == 0
2786         && (s->flags & HF_MPX_EN_MASK) != 0
2787         && (s->flags & HF_MPX_IU_MASK) != 0) {
2788         gen_helper_bnd_jmp(tcg_env);
2789     }
2790 }
2791 
2792 /* Generate an end of block. Trace exception is also generated if needed.
2793    If INHIBIT, set HF_INHIBIT_IRQ_MASK if it isn't already set.
2794    If RECHECK_TF, emit a rechecking helper for #DB, ignoring the state of
2795    S->TF.  This is used by the syscall/sysret insns.  */
2796 static void
2797 do_gen_eob_worker(DisasContext *s, bool inhibit, bool recheck_tf, bool jr)
2798 {
2799     gen_update_cc_op(s);
2800 
2801     /* If several instructions disable interrupts, only the first does it.  */
2802     if (inhibit && !(s->flags & HF_INHIBIT_IRQ_MASK)) {
2803         gen_set_hflag(s, HF_INHIBIT_IRQ_MASK);
2804     } else {
2805         gen_reset_hflag(s, HF_INHIBIT_IRQ_MASK);
2806     }
2807 
2808     if (s->base.tb->flags & HF_RF_MASK) {
2809         gen_reset_eflags(s, RF_MASK);
2810     }
2811     if (recheck_tf) {
2812         gen_helper_rechecking_single_step(tcg_env);
2813         tcg_gen_exit_tb(NULL, 0);
2814     } else if (s->flags & HF_TF_MASK) {
2815         gen_helper_single_step(tcg_env);
2816     } else if (jr) {
2817         tcg_gen_lookup_and_goto_ptr();
2818     } else {
2819         tcg_gen_exit_tb(NULL, 0);
2820     }
2821     s->base.is_jmp = DISAS_NORETURN;
2822 }
2823 
2824 static inline void
2825 gen_eob_worker(DisasContext *s, bool inhibit, bool recheck_tf)
2826 {
2827     do_gen_eob_worker(s, inhibit, recheck_tf, false);
2828 }
2829 
2830 /* End of block.
2831    If INHIBIT, set HF_INHIBIT_IRQ_MASK if it isn't already set.  */
2832 static void gen_eob_inhibit_irq(DisasContext *s, bool inhibit)
2833 {
2834     gen_eob_worker(s, inhibit, false);
2835 }
2836 
2837 /* End of block, resetting the inhibit irq flag.  */
2838 static void gen_eob(DisasContext *s)
2839 {
2840     gen_eob_worker(s, false, false);
2841 }
2842 
2843 /* Jump to register */
2844 static void gen_jr(DisasContext *s)
2845 {
2846     do_gen_eob_worker(s, false, false, true);
2847 }
2848 
2849 /* Jump to eip+diff, truncating the result to OT. */
2850 static void gen_jmp_rel(DisasContext *s, MemOp ot, int diff, int tb_num)
2851 {
2852     bool use_goto_tb = s->jmp_opt;
2853     target_ulong mask = -1;
2854     target_ulong new_pc = s->pc + diff;
2855     target_ulong new_eip = new_pc - s->cs_base;
2856 
2857     /* In 64-bit mode, operand size is fixed at 64 bits. */
2858     if (!CODE64(s)) {
2859         if (ot == MO_16) {
2860             mask = 0xffff;
2861             if (tb_cflags(s->base.tb) & CF_PCREL && CODE32(s)) {
2862                 use_goto_tb = false;
2863             }
2864         } else {
2865             mask = 0xffffffff;
2866         }
2867     }
2868     new_eip &= mask;
2869 
2870     gen_update_cc_op(s);
2871     set_cc_op(s, CC_OP_DYNAMIC);
2872 
2873     if (tb_cflags(s->base.tb) & CF_PCREL) {
2874         tcg_gen_addi_tl(cpu_eip, cpu_eip, new_pc - s->pc_save);
2875         /*
2876          * If we can prove the branch does not leave the page and we have
2877          * no extra masking to apply (data16 branch in code32, see above),
2878          * then we have also proven that the addition does not wrap.
2879          */
2880         if (!use_goto_tb || !is_same_page(&s->base, new_pc)) {
2881             tcg_gen_andi_tl(cpu_eip, cpu_eip, mask);
2882             use_goto_tb = false;
2883         }
2884     } else if (!CODE64(s)) {
2885         new_pc = (uint32_t)(new_eip + s->cs_base);
2886     }
2887 
2888     if (use_goto_tb && translator_use_goto_tb(&s->base, new_pc)) {
2889         /* jump to same page: we can use a direct jump */
2890         tcg_gen_goto_tb(tb_num);
2891         if (!(tb_cflags(s->base.tb) & CF_PCREL)) {
2892             tcg_gen_movi_tl(cpu_eip, new_eip);
2893         }
2894         tcg_gen_exit_tb(s->base.tb, tb_num);
2895         s->base.is_jmp = DISAS_NORETURN;
2896     } else {
2897         if (!(tb_cflags(s->base.tb) & CF_PCREL)) {
2898             tcg_gen_movi_tl(cpu_eip, new_eip);
2899         }
2900         if (s->jmp_opt) {
2901             gen_jr(s);   /* jump to another page */
2902         } else {
2903             gen_eob(s);  /* exit to main loop */
2904         }
2905     }
2906 }
2907 
2908 /* Jump to eip+diff, truncating to the current code size. */
2909 static void gen_jmp_rel_csize(DisasContext *s, int diff, int tb_num)
2910 {
2911     /* CODE64 ignores the OT argument, so we need not consider it. */
2912     gen_jmp_rel(s, CODE32(s) ? MO_32 : MO_16, diff, tb_num);
2913 }
2914 
2915 static inline void gen_ldq_env_A0(DisasContext *s, int offset)
2916 {
2917     tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0, s->mem_index, MO_LEUQ);
2918     tcg_gen_st_i64(s->tmp1_i64, tcg_env, offset);
2919 }
2920 
2921 static inline void gen_stq_env_A0(DisasContext *s, int offset)
2922 {
2923     tcg_gen_ld_i64(s->tmp1_i64, tcg_env, offset);
2924     tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0, s->mem_index, MO_LEUQ);
2925 }
2926 
2927 static inline void gen_ldo_env_A0(DisasContext *s, int offset, bool align)
2928 {
2929     MemOp atom = (s->cpuid_ext_features & CPUID_EXT_AVX
2930                   ? MO_ATOM_IFALIGN : MO_ATOM_IFALIGN_PAIR);
2931     MemOp mop = MO_128 | MO_LE | atom | (align ? MO_ALIGN_16 : 0);
2932     int mem_index = s->mem_index;
2933     TCGv_i128 t = tcg_temp_new_i128();
2934 
2935     tcg_gen_qemu_ld_i128(t, s->A0, mem_index, mop);
2936     tcg_gen_st_i128(t, tcg_env, offset);
2937 }
2938 
2939 static inline void gen_sto_env_A0(DisasContext *s, int offset, bool align)
2940 {
2941     MemOp atom = (s->cpuid_ext_features & CPUID_EXT_AVX
2942                   ? MO_ATOM_IFALIGN : MO_ATOM_IFALIGN_PAIR);
2943     MemOp mop = MO_128 | MO_LE | atom | (align ? MO_ALIGN_16 : 0);
2944     int mem_index = s->mem_index;
2945     TCGv_i128 t = tcg_temp_new_i128();
2946 
2947     tcg_gen_ld_i128(t, tcg_env, offset);
2948     tcg_gen_qemu_st_i128(t, s->A0, mem_index, mop);
2949 }
2950 
2951 static void gen_ldy_env_A0(DisasContext *s, int offset, bool align)
2952 {
2953     MemOp mop = MO_128 | MO_LE | MO_ATOM_IFALIGN_PAIR;
2954     int mem_index = s->mem_index;
2955     TCGv_i128 t0 = tcg_temp_new_i128();
2956     TCGv_i128 t1 = tcg_temp_new_i128();
2957 
2958     tcg_gen_qemu_ld_i128(t0, s->A0, mem_index, mop | (align ? MO_ALIGN_32 : 0));
2959     tcg_gen_addi_tl(s->tmp0, s->A0, 16);
2960     tcg_gen_qemu_ld_i128(t1, s->tmp0, mem_index, mop);
2961 
2962     tcg_gen_st_i128(t0, tcg_env, offset + offsetof(YMMReg, YMM_X(0)));
2963     tcg_gen_st_i128(t1, tcg_env, offset + offsetof(YMMReg, YMM_X(1)));
2964 }
2965 
2966 static void gen_sty_env_A0(DisasContext *s, int offset, bool align)
2967 {
2968     MemOp mop = MO_128 | MO_LE | MO_ATOM_IFALIGN_PAIR;
2969     int mem_index = s->mem_index;
2970     TCGv_i128 t = tcg_temp_new_i128();
2971 
2972     tcg_gen_ld_i128(t, tcg_env, offset + offsetof(YMMReg, YMM_X(0)));
2973     tcg_gen_qemu_st_i128(t, s->A0, mem_index, mop | (align ? MO_ALIGN_32 : 0));
2974     tcg_gen_addi_tl(s->tmp0, s->A0, 16);
2975     tcg_gen_ld_i128(t, tcg_env, offset + offsetof(YMMReg, YMM_X(1)));
2976     tcg_gen_qemu_st_i128(t, s->tmp0, mem_index, mop);
2977 }
2978 
2979 #include "decode-new.h"
2980 #include "emit.c.inc"
2981 #include "decode-new.c.inc"
2982 
2983 static void gen_cmpxchg8b(DisasContext *s, CPUX86State *env, int modrm)
2984 {
2985     TCGv_i64 cmp, val, old;
2986     TCGv Z;
2987 
2988     gen_lea_modrm(env, s, modrm);
2989 
2990     cmp = tcg_temp_new_i64();
2991     val = tcg_temp_new_i64();
2992     old = tcg_temp_new_i64();
2993 
2994     /* Construct the comparison values from the register pair. */
2995     tcg_gen_concat_tl_i64(cmp, cpu_regs[R_EAX], cpu_regs[R_EDX]);
2996     tcg_gen_concat_tl_i64(val, cpu_regs[R_EBX], cpu_regs[R_ECX]);
2997 
2998     /* Only require atomic with LOCK; non-parallel handled in generator. */
2999     if (s->prefix & PREFIX_LOCK) {
3000         tcg_gen_atomic_cmpxchg_i64(old, s->A0, cmp, val, s->mem_index, MO_TEUQ);
3001     } else {
3002         tcg_gen_nonatomic_cmpxchg_i64(old, s->A0, cmp, val,
3003                                       s->mem_index, MO_TEUQ);
3004     }
3005 
3006     /* Set tmp0 to match the required value of Z. */
3007     tcg_gen_setcond_i64(TCG_COND_EQ, cmp, old, cmp);
3008     Z = tcg_temp_new();
3009     tcg_gen_trunc_i64_tl(Z, cmp);
3010 
3011     /*
3012      * Extract the result values for the register pair.
3013      * For 32-bit, we may do this unconditionally, because on success (Z=1),
3014      * the old value matches the previous value in EDX:EAX.  For x86_64,
3015      * the store must be conditional, because we must leave the source
3016      * registers unchanged on success, and zero-extend the writeback
3017      * on failure (Z=0).
3018      */
3019     if (TARGET_LONG_BITS == 32) {
3020         tcg_gen_extr_i64_tl(cpu_regs[R_EAX], cpu_regs[R_EDX], old);
3021     } else {
3022         TCGv zero = tcg_constant_tl(0);
3023 
3024         tcg_gen_extr_i64_tl(s->T0, s->T1, old);
3025         tcg_gen_movcond_tl(TCG_COND_EQ, cpu_regs[R_EAX], Z, zero,
3026                            s->T0, cpu_regs[R_EAX]);
3027         tcg_gen_movcond_tl(TCG_COND_EQ, cpu_regs[R_EDX], Z, zero,
3028                            s->T1, cpu_regs[R_EDX]);
3029     }
3030 
3031     /* Update Z. */
3032     gen_compute_eflags(s);
3033     tcg_gen_deposit_tl(cpu_cc_src, cpu_cc_src, Z, ctz32(CC_Z), 1);
3034 }
3035 
3036 #ifdef TARGET_X86_64
3037 static void gen_cmpxchg16b(DisasContext *s, CPUX86State *env, int modrm)
3038 {
3039     MemOp mop = MO_TE | MO_128 | MO_ALIGN;
3040     TCGv_i64 t0, t1;
3041     TCGv_i128 cmp, val;
3042 
3043     gen_lea_modrm(env, s, modrm);
3044 
3045     cmp = tcg_temp_new_i128();
3046     val = tcg_temp_new_i128();
3047     tcg_gen_concat_i64_i128(cmp, cpu_regs[R_EAX], cpu_regs[R_EDX]);
3048     tcg_gen_concat_i64_i128(val, cpu_regs[R_EBX], cpu_regs[R_ECX]);
3049 
3050     /* Only require atomic with LOCK; non-parallel handled in generator. */
3051     if (s->prefix & PREFIX_LOCK) {
3052         tcg_gen_atomic_cmpxchg_i128(val, s->A0, cmp, val, s->mem_index, mop);
3053     } else {
3054         tcg_gen_nonatomic_cmpxchg_i128(val, s->A0, cmp, val, s->mem_index, mop);
3055     }
3056 
3057     tcg_gen_extr_i128_i64(s->T0, s->T1, val);
3058 
3059     /* Determine success after the fact. */
3060     t0 = tcg_temp_new_i64();
3061     t1 = tcg_temp_new_i64();
3062     tcg_gen_xor_i64(t0, s->T0, cpu_regs[R_EAX]);
3063     tcg_gen_xor_i64(t1, s->T1, cpu_regs[R_EDX]);
3064     tcg_gen_or_i64(t0, t0, t1);
3065 
3066     /* Update Z. */
3067     gen_compute_eflags(s);
3068     tcg_gen_setcondi_i64(TCG_COND_EQ, t0, t0, 0);
3069     tcg_gen_deposit_tl(cpu_cc_src, cpu_cc_src, t0, ctz32(CC_Z), 1);
3070 
3071     /*
3072      * Extract the result values for the register pair.  We may do this
3073      * unconditionally, because on success (Z=1), the old value matches
3074      * the previous value in RDX:RAX.
3075      */
3076     tcg_gen_mov_i64(cpu_regs[R_EAX], s->T0);
3077     tcg_gen_mov_i64(cpu_regs[R_EDX], s->T1);
3078 }
3079 #endif
3080 
3081 /* convert one instruction. s->base.is_jmp is set if the translation must
3082    be stopped. Return the next pc value */
3083 static bool disas_insn(DisasContext *s, CPUState *cpu)
3084 {
3085     CPUX86State *env = cpu_env(cpu);
3086     int b, prefixes;
3087     int shift;
3088     MemOp ot, aflag, dflag;
3089     int modrm, reg, rm, mod, op, opreg, val;
3090     bool orig_cc_op_dirty = s->cc_op_dirty;
3091     CCOp orig_cc_op = s->cc_op;
3092     target_ulong orig_pc_save = s->pc_save;
3093 
3094     s->pc = s->base.pc_next;
3095     s->override = -1;
3096 #ifdef TARGET_X86_64
3097     s->rex_r = 0;
3098     s->rex_x = 0;
3099     s->rex_b = 0;
3100 #endif
3101     s->rip_offset = 0; /* for relative ip address */
3102     s->vex_l = 0;
3103     s->vex_v = 0;
3104     s->vex_w = false;
3105     switch (sigsetjmp(s->jmpbuf, 0)) {
3106     case 0:
3107         break;
3108     case 1:
3109         gen_exception_gpf(s);
3110         return true;
3111     case 2:
3112         /* Restore state that may affect the next instruction. */
3113         s->pc = s->base.pc_next;
3114         /*
3115          * TODO: These save/restore can be removed after the table-based
3116          * decoder is complete; we will be decoding the insn completely
3117          * before any code generation that might affect these variables.
3118          */
3119         s->cc_op_dirty = orig_cc_op_dirty;
3120         s->cc_op = orig_cc_op;
3121         s->pc_save = orig_pc_save;
3122         /* END TODO */
3123         s->base.num_insns--;
3124         tcg_remove_ops_after(s->prev_insn_end);
3125         s->base.is_jmp = DISAS_TOO_MANY;
3126         return false;
3127     default:
3128         g_assert_not_reached();
3129     }
3130 
3131     prefixes = 0;
3132 
3133  next_byte:
3134     s->prefix = prefixes;
3135     b = x86_ldub_code(env, s);
3136     /* Collect prefixes.  */
3137     switch (b) {
3138     default:
3139         break;
3140     case 0x0f:
3141         b = x86_ldub_code(env, s) + 0x100;
3142         break;
3143     case 0xf3:
3144         prefixes |= PREFIX_REPZ;
3145         prefixes &= ~PREFIX_REPNZ;
3146         goto next_byte;
3147     case 0xf2:
3148         prefixes |= PREFIX_REPNZ;
3149         prefixes &= ~PREFIX_REPZ;
3150         goto next_byte;
3151     case 0xf0:
3152         prefixes |= PREFIX_LOCK;
3153         goto next_byte;
3154     case 0x2e:
3155         s->override = R_CS;
3156         goto next_byte;
3157     case 0x36:
3158         s->override = R_SS;
3159         goto next_byte;
3160     case 0x3e:
3161         s->override = R_DS;
3162         goto next_byte;
3163     case 0x26:
3164         s->override = R_ES;
3165         goto next_byte;
3166     case 0x64:
3167         s->override = R_FS;
3168         goto next_byte;
3169     case 0x65:
3170         s->override = R_GS;
3171         goto next_byte;
3172     case 0x66:
3173         prefixes |= PREFIX_DATA;
3174         goto next_byte;
3175     case 0x67:
3176         prefixes |= PREFIX_ADR;
3177         goto next_byte;
3178 #ifdef TARGET_X86_64
3179     case 0x40 ... 0x4f:
3180         if (CODE64(s)) {
3181             /* REX prefix */
3182             prefixes |= PREFIX_REX;
3183             s->vex_w = (b >> 3) & 1;
3184             s->rex_r = (b & 0x4) << 1;
3185             s->rex_x = (b & 0x2) << 2;
3186             s->rex_b = (b & 0x1) << 3;
3187             goto next_byte;
3188         }
3189         break;
3190 #endif
3191     case 0xc5: /* 2-byte VEX */
3192     case 0xc4: /* 3-byte VEX */
3193         if (CODE32(s) && !VM86(s)) {
3194             int vex2 = x86_ldub_code(env, s);
3195             s->pc--; /* rewind the advance_pc() x86_ldub_code() did */
3196 
3197             if (!CODE64(s) && (vex2 & 0xc0) != 0xc0) {
3198                 /* 4.1.4.6: In 32-bit mode, bits [7:6] must be 11b,
3199                    otherwise the instruction is LES or LDS.  */
3200                 break;
3201             }
3202             disas_insn_new(s, cpu, b);
3203             return s->pc;
3204         }
3205         break;
3206     }
3207 
3208     /* Post-process prefixes.  */
3209     if (CODE64(s)) {
3210         /* In 64-bit mode, the default data size is 32-bit.  Select 64-bit
3211            data with rex_w, and 16-bit data with 0x66; rex_w takes precedence
3212            over 0x66 if both are present.  */
3213         dflag = (REX_W(s) ? MO_64 : prefixes & PREFIX_DATA ? MO_16 : MO_32);
3214         /* In 64-bit mode, 0x67 selects 32-bit addressing.  */
3215         aflag = (prefixes & PREFIX_ADR ? MO_32 : MO_64);
3216     } else {
3217         /* In 16/32-bit mode, 0x66 selects the opposite data size.  */
3218         if (CODE32(s) ^ ((prefixes & PREFIX_DATA) != 0)) {
3219             dflag = MO_32;
3220         } else {
3221             dflag = MO_16;
3222         }
3223         /* In 16/32-bit mode, 0x67 selects the opposite addressing.  */
3224         if (CODE32(s) ^ ((prefixes & PREFIX_ADR) != 0)) {
3225             aflag = MO_32;
3226         }  else {
3227             aflag = MO_16;
3228         }
3229     }
3230 
3231     s->prefix = prefixes;
3232     s->aflag = aflag;
3233     s->dflag = dflag;
3234 
3235     /* now check op code */
3236     switch (b) {
3237         /**************************/
3238         /* arith & logic */
3239     case 0x00 ... 0x05:
3240     case 0x08 ... 0x0d:
3241     case 0x10 ... 0x15:
3242     case 0x18 ... 0x1d:
3243     case 0x20 ... 0x25:
3244     case 0x28 ... 0x2d:
3245     case 0x30 ... 0x35:
3246     case 0x38 ... 0x3d:
3247         {
3248             int f;
3249             op = (b >> 3) & 7;
3250             f = (b >> 1) & 3;
3251 
3252             ot = mo_b_d(b, dflag);
3253 
3254             switch(f) {
3255             case 0: /* OP Ev, Gv */
3256                 modrm = x86_ldub_code(env, s);
3257                 reg = ((modrm >> 3) & 7) | REX_R(s);
3258                 mod = (modrm >> 6) & 3;
3259                 rm = (modrm & 7) | REX_B(s);
3260                 if (mod != 3) {
3261                     gen_lea_modrm(env, s, modrm);
3262                     opreg = OR_TMP0;
3263                 } else if (op == OP_XORL && rm == reg) {
3264                 xor_zero:
3265                     /* xor reg, reg optimisation */
3266                     set_cc_op(s, CC_OP_CLR);
3267                     tcg_gen_movi_tl(s->T0, 0);
3268                     gen_op_mov_reg_v(s, ot, reg, s->T0);
3269                     break;
3270                 } else {
3271                     opreg = rm;
3272                 }
3273                 gen_op_mov_v_reg(s, ot, s->T1, reg);
3274                 gen_op(s, op, ot, opreg);
3275                 break;
3276             case 1: /* OP Gv, Ev */
3277                 modrm = x86_ldub_code(env, s);
3278                 mod = (modrm >> 6) & 3;
3279                 reg = ((modrm >> 3) & 7) | REX_R(s);
3280                 rm = (modrm & 7) | REX_B(s);
3281                 if (mod != 3) {
3282                     gen_lea_modrm(env, s, modrm);
3283                     gen_op_ld_v(s, ot, s->T1, s->A0);
3284                 } else if (op == OP_XORL && rm == reg) {
3285                     goto xor_zero;
3286                 } else {
3287                     gen_op_mov_v_reg(s, ot, s->T1, rm);
3288                 }
3289                 gen_op(s, op, ot, reg);
3290                 break;
3291             case 2: /* OP A, Iv */
3292                 val = insn_get(env, s, ot);
3293                 tcg_gen_movi_tl(s->T1, val);
3294                 gen_op(s, op, ot, OR_EAX);
3295                 break;
3296             }
3297         }
3298         break;
3299 
3300     case 0x82:
3301         if (CODE64(s))
3302             goto illegal_op;
3303         /* fall through */
3304     case 0x80: /* GRP1 */
3305     case 0x81:
3306     case 0x83:
3307         {
3308             ot = mo_b_d(b, dflag);
3309 
3310             modrm = x86_ldub_code(env, s);
3311             mod = (modrm >> 6) & 3;
3312             rm = (modrm & 7) | REX_B(s);
3313             op = (modrm >> 3) & 7;
3314 
3315             if (mod != 3) {
3316                 if (b == 0x83)
3317                     s->rip_offset = 1;
3318                 else
3319                     s->rip_offset = insn_const_size(ot);
3320                 gen_lea_modrm(env, s, modrm);
3321                 opreg = OR_TMP0;
3322             } else {
3323                 opreg = rm;
3324             }
3325 
3326             switch(b) {
3327             default:
3328             case 0x80:
3329             case 0x81:
3330             case 0x82:
3331                 val = insn_get(env, s, ot);
3332                 break;
3333             case 0x83:
3334                 val = (int8_t)insn_get(env, s, MO_8);
3335                 break;
3336             }
3337             tcg_gen_movi_tl(s->T1, val);
3338             gen_op(s, op, ot, opreg);
3339         }
3340         break;
3341 
3342         /**************************/
3343         /* inc, dec, and other misc arith */
3344     case 0x40 ... 0x47: /* inc Gv */
3345         ot = dflag;
3346         gen_inc(s, ot, OR_EAX + (b & 7), 1);
3347         break;
3348     case 0x48 ... 0x4f: /* dec Gv */
3349         ot = dflag;
3350         gen_inc(s, ot, OR_EAX + (b & 7), -1);
3351         break;
3352     case 0xf6: /* GRP3 */
3353     case 0xf7:
3354         ot = mo_b_d(b, dflag);
3355 
3356         modrm = x86_ldub_code(env, s);
3357         mod = (modrm >> 6) & 3;
3358         rm = (modrm & 7) | REX_B(s);
3359         op = (modrm >> 3) & 7;
3360         if (mod != 3) {
3361             if (op == 0) {
3362                 s->rip_offset = insn_const_size(ot);
3363             }
3364             gen_lea_modrm(env, s, modrm);
3365             /* For those below that handle locked memory, don't load here.  */
3366             if (!(s->prefix & PREFIX_LOCK)
3367                 || op != 2) {
3368                 gen_op_ld_v(s, ot, s->T0, s->A0);
3369             }
3370         } else {
3371             gen_op_mov_v_reg(s, ot, s->T0, rm);
3372         }
3373 
3374         switch(op) {
3375         case 0: /* test */
3376             val = insn_get(env, s, ot);
3377             tcg_gen_movi_tl(s->T1, val);
3378             gen_op_testl_T0_T1_cc(s);
3379             set_cc_op(s, CC_OP_LOGICB + ot);
3380             break;
3381         case 2: /* not */
3382             if (s->prefix & PREFIX_LOCK) {
3383                 if (mod == 3) {
3384                     goto illegal_op;
3385                 }
3386                 tcg_gen_movi_tl(s->T0, ~0);
3387                 tcg_gen_atomic_xor_fetch_tl(s->T0, s->A0, s->T0,
3388                                             s->mem_index, ot | MO_LE);
3389             } else {
3390                 tcg_gen_not_tl(s->T0, s->T0);
3391                 if (mod != 3) {
3392                     gen_op_st_v(s, ot, s->T0, s->A0);
3393                 } else {
3394                     gen_op_mov_reg_v(s, ot, rm, s->T0);
3395                 }
3396             }
3397             break;
3398         case 3: /* neg */
3399             if (s->prefix & PREFIX_LOCK) {
3400                 TCGLabel *label1;
3401                 TCGv a0, t0, t1, t2;
3402 
3403                 if (mod == 3) {
3404                     goto illegal_op;
3405                 }
3406                 a0 = s->A0;
3407                 t0 = s->T0;
3408                 label1 = gen_new_label();
3409 
3410                 gen_set_label(label1);
3411                 t1 = tcg_temp_new();
3412                 t2 = tcg_temp_new();
3413                 tcg_gen_mov_tl(t2, t0);
3414                 tcg_gen_neg_tl(t1, t0);
3415                 tcg_gen_atomic_cmpxchg_tl(t0, a0, t0, t1,
3416                                           s->mem_index, ot | MO_LE);
3417                 tcg_gen_brcond_tl(TCG_COND_NE, t0, t2, label1);
3418 
3419                 tcg_gen_neg_tl(s->T0, t0);
3420             } else {
3421                 tcg_gen_neg_tl(s->T0, s->T0);
3422                 if (mod != 3) {
3423                     gen_op_st_v(s, ot, s->T0, s->A0);
3424                 } else {
3425                     gen_op_mov_reg_v(s, ot, rm, s->T0);
3426                 }
3427             }
3428             gen_op_update_neg_cc(s);
3429             set_cc_op(s, CC_OP_SUBB + ot);
3430             break;
3431         case 4: /* mul */
3432             switch(ot) {
3433             case MO_8:
3434                 gen_op_mov_v_reg(s, MO_8, s->T1, R_EAX);
3435                 tcg_gen_ext8u_tl(s->T0, s->T0);
3436                 tcg_gen_ext8u_tl(s->T1, s->T1);
3437                 /* XXX: use 32 bit mul which could be faster */
3438                 tcg_gen_mul_tl(s->T0, s->T0, s->T1);
3439                 gen_op_mov_reg_v(s, MO_16, R_EAX, s->T0);
3440                 tcg_gen_mov_tl(cpu_cc_dst, s->T0);
3441                 tcg_gen_andi_tl(cpu_cc_src, s->T0, 0xff00);
3442                 set_cc_op(s, CC_OP_MULB);
3443                 break;
3444             case MO_16:
3445                 gen_op_mov_v_reg(s, MO_16, s->T1, R_EAX);
3446                 tcg_gen_ext16u_tl(s->T0, s->T0);
3447                 tcg_gen_ext16u_tl(s->T1, s->T1);
3448                 /* XXX: use 32 bit mul which could be faster */
3449                 tcg_gen_mul_tl(s->T0, s->T0, s->T1);
3450                 gen_op_mov_reg_v(s, MO_16, R_EAX, s->T0);
3451                 tcg_gen_mov_tl(cpu_cc_dst, s->T0);
3452                 tcg_gen_shri_tl(s->T0, s->T0, 16);
3453                 gen_op_mov_reg_v(s, MO_16, R_EDX, s->T0);
3454                 tcg_gen_mov_tl(cpu_cc_src, s->T0);
3455                 set_cc_op(s, CC_OP_MULW);
3456                 break;
3457             default:
3458             case MO_32:
3459                 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
3460                 tcg_gen_trunc_tl_i32(s->tmp3_i32, cpu_regs[R_EAX]);
3461                 tcg_gen_mulu2_i32(s->tmp2_i32, s->tmp3_i32,
3462                                   s->tmp2_i32, s->tmp3_i32);
3463                 tcg_gen_extu_i32_tl(cpu_regs[R_EAX], s->tmp2_i32);
3464                 tcg_gen_extu_i32_tl(cpu_regs[R_EDX], s->tmp3_i32);
3465                 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
3466                 tcg_gen_mov_tl(cpu_cc_src, cpu_regs[R_EDX]);
3467                 set_cc_op(s, CC_OP_MULL);
3468                 break;
3469 #ifdef TARGET_X86_64
3470             case MO_64:
3471                 tcg_gen_mulu2_i64(cpu_regs[R_EAX], cpu_regs[R_EDX],
3472                                   s->T0, cpu_regs[R_EAX]);
3473                 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
3474                 tcg_gen_mov_tl(cpu_cc_src, cpu_regs[R_EDX]);
3475                 set_cc_op(s, CC_OP_MULQ);
3476                 break;
3477 #endif
3478             }
3479             break;
3480         case 5: /* imul */
3481             switch(ot) {
3482             case MO_8:
3483                 gen_op_mov_v_reg(s, MO_8, s->T1, R_EAX);
3484                 tcg_gen_ext8s_tl(s->T0, s->T0);
3485                 tcg_gen_ext8s_tl(s->T1, s->T1);
3486                 /* XXX: use 32 bit mul which could be faster */
3487                 tcg_gen_mul_tl(s->T0, s->T0, s->T1);
3488                 gen_op_mov_reg_v(s, MO_16, R_EAX, s->T0);
3489                 tcg_gen_mov_tl(cpu_cc_dst, s->T0);
3490                 tcg_gen_ext8s_tl(s->tmp0, s->T0);
3491                 tcg_gen_sub_tl(cpu_cc_src, s->T0, s->tmp0);
3492                 set_cc_op(s, CC_OP_MULB);
3493                 break;
3494             case MO_16:
3495                 gen_op_mov_v_reg(s, MO_16, s->T1, R_EAX);
3496                 tcg_gen_ext16s_tl(s->T0, s->T0);
3497                 tcg_gen_ext16s_tl(s->T1, s->T1);
3498                 /* XXX: use 32 bit mul which could be faster */
3499                 tcg_gen_mul_tl(s->T0, s->T0, s->T1);
3500                 gen_op_mov_reg_v(s, MO_16, R_EAX, s->T0);
3501                 tcg_gen_mov_tl(cpu_cc_dst, s->T0);
3502                 tcg_gen_ext16s_tl(s->tmp0, s->T0);
3503                 tcg_gen_sub_tl(cpu_cc_src, s->T0, s->tmp0);
3504                 tcg_gen_shri_tl(s->T0, s->T0, 16);
3505                 gen_op_mov_reg_v(s, MO_16, R_EDX, s->T0);
3506                 set_cc_op(s, CC_OP_MULW);
3507                 break;
3508             default:
3509             case MO_32:
3510                 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
3511                 tcg_gen_trunc_tl_i32(s->tmp3_i32, cpu_regs[R_EAX]);
3512                 tcg_gen_muls2_i32(s->tmp2_i32, s->tmp3_i32,
3513                                   s->tmp2_i32, s->tmp3_i32);
3514                 tcg_gen_extu_i32_tl(cpu_regs[R_EAX], s->tmp2_i32);
3515                 tcg_gen_extu_i32_tl(cpu_regs[R_EDX], s->tmp3_i32);
3516                 tcg_gen_sari_i32(s->tmp2_i32, s->tmp2_i32, 31);
3517                 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
3518                 tcg_gen_sub_i32(s->tmp2_i32, s->tmp2_i32, s->tmp3_i32);
3519                 tcg_gen_extu_i32_tl(cpu_cc_src, s->tmp2_i32);
3520                 set_cc_op(s, CC_OP_MULL);
3521                 break;
3522 #ifdef TARGET_X86_64
3523             case MO_64:
3524                 tcg_gen_muls2_i64(cpu_regs[R_EAX], cpu_regs[R_EDX],
3525                                   s->T0, cpu_regs[R_EAX]);
3526                 tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[R_EAX]);
3527                 tcg_gen_sari_tl(cpu_cc_src, cpu_regs[R_EAX], 63);
3528                 tcg_gen_sub_tl(cpu_cc_src, cpu_cc_src, cpu_regs[R_EDX]);
3529                 set_cc_op(s, CC_OP_MULQ);
3530                 break;
3531 #endif
3532             }
3533             break;
3534         case 6: /* div */
3535             switch(ot) {
3536             case MO_8:
3537                 gen_helper_divb_AL(tcg_env, s->T0);
3538                 break;
3539             case MO_16:
3540                 gen_helper_divw_AX(tcg_env, s->T0);
3541                 break;
3542             default:
3543             case MO_32:
3544                 gen_helper_divl_EAX(tcg_env, s->T0);
3545                 break;
3546 #ifdef TARGET_X86_64
3547             case MO_64:
3548                 gen_helper_divq_EAX(tcg_env, s->T0);
3549                 break;
3550 #endif
3551             }
3552             break;
3553         case 7: /* idiv */
3554             switch(ot) {
3555             case MO_8:
3556                 gen_helper_idivb_AL(tcg_env, s->T0);
3557                 break;
3558             case MO_16:
3559                 gen_helper_idivw_AX(tcg_env, s->T0);
3560                 break;
3561             default:
3562             case MO_32:
3563                 gen_helper_idivl_EAX(tcg_env, s->T0);
3564                 break;
3565 #ifdef TARGET_X86_64
3566             case MO_64:
3567                 gen_helper_idivq_EAX(tcg_env, s->T0);
3568                 break;
3569 #endif
3570             }
3571             break;
3572         default:
3573             goto unknown_op;
3574         }
3575         break;
3576 
3577     case 0xfe: /* GRP4 */
3578     case 0xff: /* GRP5 */
3579         ot = mo_b_d(b, dflag);
3580 
3581         modrm = x86_ldub_code(env, s);
3582         mod = (modrm >> 6) & 3;
3583         rm = (modrm & 7) | REX_B(s);
3584         op = (modrm >> 3) & 7;
3585         if (op >= 2 && b == 0xfe) {
3586             goto unknown_op;
3587         }
3588         if (CODE64(s)) {
3589             if (op == 2 || op == 4) {
3590                 /* operand size for jumps is 64 bit */
3591                 ot = MO_64;
3592             } else if (op == 3 || op == 5) {
3593                 ot = dflag != MO_16 ? MO_32 + REX_W(s) : MO_16;
3594             } else if (op == 6) {
3595                 /* default push size is 64 bit */
3596                 ot = mo_pushpop(s, dflag);
3597             }
3598         }
3599         if (mod != 3) {
3600             gen_lea_modrm(env, s, modrm);
3601             if (op >= 2 && op != 3 && op != 5)
3602                 gen_op_ld_v(s, ot, s->T0, s->A0);
3603         } else {
3604             gen_op_mov_v_reg(s, ot, s->T0, rm);
3605         }
3606 
3607         switch(op) {
3608         case 0: /* inc Ev */
3609             if (mod != 3)
3610                 opreg = OR_TMP0;
3611             else
3612                 opreg = rm;
3613             gen_inc(s, ot, opreg, 1);
3614             break;
3615         case 1: /* dec Ev */
3616             if (mod != 3)
3617                 opreg = OR_TMP0;
3618             else
3619                 opreg = rm;
3620             gen_inc(s, ot, opreg, -1);
3621             break;
3622         case 2: /* call Ev */
3623             /* XXX: optimize if memory (no 'and' is necessary) */
3624             if (dflag == MO_16) {
3625                 tcg_gen_ext16u_tl(s->T0, s->T0);
3626             }
3627             gen_push_v(s, eip_next_tl(s));
3628             gen_op_jmp_v(s, s->T0);
3629             gen_bnd_jmp(s);
3630             s->base.is_jmp = DISAS_JUMP;
3631             break;
3632         case 3: /* lcall Ev */
3633             if (mod == 3) {
3634                 goto illegal_op;
3635             }
3636             gen_op_ld_v(s, ot, s->T1, s->A0);
3637             gen_add_A0_im(s, 1 << ot);
3638             gen_op_ld_v(s, MO_16, s->T0, s->A0);
3639         do_lcall:
3640             if (PE(s) && !VM86(s)) {
3641                 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
3642                 gen_helper_lcall_protected(tcg_env, s->tmp2_i32, s->T1,
3643                                            tcg_constant_i32(dflag - 1),
3644                                            eip_next_tl(s));
3645             } else {
3646                 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
3647                 tcg_gen_trunc_tl_i32(s->tmp3_i32, s->T1);
3648                 gen_helper_lcall_real(tcg_env, s->tmp2_i32, s->tmp3_i32,
3649                                       tcg_constant_i32(dflag - 1),
3650                                       eip_next_i32(s));
3651             }
3652             s->base.is_jmp = DISAS_JUMP;
3653             break;
3654         case 4: /* jmp Ev */
3655             if (dflag == MO_16) {
3656                 tcg_gen_ext16u_tl(s->T0, s->T0);
3657             }
3658             gen_op_jmp_v(s, s->T0);
3659             gen_bnd_jmp(s);
3660             s->base.is_jmp = DISAS_JUMP;
3661             break;
3662         case 5: /* ljmp Ev */
3663             if (mod == 3) {
3664                 goto illegal_op;
3665             }
3666             gen_op_ld_v(s, ot, s->T1, s->A0);
3667             gen_add_A0_im(s, 1 << ot);
3668             gen_op_ld_v(s, MO_16, s->T0, s->A0);
3669         do_ljmp:
3670             if (PE(s) && !VM86(s)) {
3671                 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
3672                 gen_helper_ljmp_protected(tcg_env, s->tmp2_i32, s->T1,
3673                                           eip_next_tl(s));
3674             } else {
3675                 gen_op_movl_seg_T0_vm(s, R_CS);
3676                 gen_op_jmp_v(s, s->T1);
3677             }
3678             s->base.is_jmp = DISAS_JUMP;
3679             break;
3680         case 6: /* push Ev */
3681             gen_push_v(s, s->T0);
3682             break;
3683         default:
3684             goto unknown_op;
3685         }
3686         break;
3687 
3688     case 0x84: /* test Ev, Gv */
3689     case 0x85:
3690         ot = mo_b_d(b, dflag);
3691 
3692         modrm = x86_ldub_code(env, s);
3693         reg = ((modrm >> 3) & 7) | REX_R(s);
3694 
3695         gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3696         gen_op_mov_v_reg(s, ot, s->T1, reg);
3697         gen_op_testl_T0_T1_cc(s);
3698         set_cc_op(s, CC_OP_LOGICB + ot);
3699         break;
3700 
3701     case 0xa8: /* test eAX, Iv */
3702     case 0xa9:
3703         ot = mo_b_d(b, dflag);
3704         val = insn_get(env, s, ot);
3705 
3706         gen_op_mov_v_reg(s, ot, s->T0, OR_EAX);
3707         tcg_gen_movi_tl(s->T1, val);
3708         gen_op_testl_T0_T1_cc(s);
3709         set_cc_op(s, CC_OP_LOGICB + ot);
3710         break;
3711 
3712     case 0x98: /* CWDE/CBW */
3713         switch (dflag) {
3714 #ifdef TARGET_X86_64
3715         case MO_64:
3716             gen_op_mov_v_reg(s, MO_32, s->T0, R_EAX);
3717             tcg_gen_ext32s_tl(s->T0, s->T0);
3718             gen_op_mov_reg_v(s, MO_64, R_EAX, s->T0);
3719             break;
3720 #endif
3721         case MO_32:
3722             gen_op_mov_v_reg(s, MO_16, s->T0, R_EAX);
3723             tcg_gen_ext16s_tl(s->T0, s->T0);
3724             gen_op_mov_reg_v(s, MO_32, R_EAX, s->T0);
3725             break;
3726         case MO_16:
3727             gen_op_mov_v_reg(s, MO_8, s->T0, R_EAX);
3728             tcg_gen_ext8s_tl(s->T0, s->T0);
3729             gen_op_mov_reg_v(s, MO_16, R_EAX, s->T0);
3730             break;
3731         default:
3732             g_assert_not_reached();
3733         }
3734         break;
3735     case 0x99: /* CDQ/CWD */
3736         switch (dflag) {
3737 #ifdef TARGET_X86_64
3738         case MO_64:
3739             gen_op_mov_v_reg(s, MO_64, s->T0, R_EAX);
3740             tcg_gen_sari_tl(s->T0, s->T0, 63);
3741             gen_op_mov_reg_v(s, MO_64, R_EDX, s->T0);
3742             break;
3743 #endif
3744         case MO_32:
3745             gen_op_mov_v_reg(s, MO_32, s->T0, R_EAX);
3746             tcg_gen_ext32s_tl(s->T0, s->T0);
3747             tcg_gen_sari_tl(s->T0, s->T0, 31);
3748             gen_op_mov_reg_v(s, MO_32, R_EDX, s->T0);
3749             break;
3750         case MO_16:
3751             gen_op_mov_v_reg(s, MO_16, s->T0, R_EAX);
3752             tcg_gen_ext16s_tl(s->T0, s->T0);
3753             tcg_gen_sari_tl(s->T0, s->T0, 15);
3754             gen_op_mov_reg_v(s, MO_16, R_EDX, s->T0);
3755             break;
3756         default:
3757             g_assert_not_reached();
3758         }
3759         break;
3760     case 0x1af: /* imul Gv, Ev */
3761     case 0x69: /* imul Gv, Ev, I */
3762     case 0x6b:
3763         ot = dflag;
3764         modrm = x86_ldub_code(env, s);
3765         reg = ((modrm >> 3) & 7) | REX_R(s);
3766         if (b == 0x69)
3767             s->rip_offset = insn_const_size(ot);
3768         else if (b == 0x6b)
3769             s->rip_offset = 1;
3770         gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
3771         if (b == 0x69) {
3772             val = insn_get(env, s, ot);
3773             tcg_gen_movi_tl(s->T1, val);
3774         } else if (b == 0x6b) {
3775             val = (int8_t)insn_get(env, s, MO_8);
3776             tcg_gen_movi_tl(s->T1, val);
3777         } else {
3778             gen_op_mov_v_reg(s, ot, s->T1, reg);
3779         }
3780         switch (ot) {
3781 #ifdef TARGET_X86_64
3782         case MO_64:
3783             tcg_gen_muls2_i64(cpu_regs[reg], s->T1, s->T0, s->T1);
3784             tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[reg]);
3785             tcg_gen_sari_tl(cpu_cc_src, cpu_cc_dst, 63);
3786             tcg_gen_sub_tl(cpu_cc_src, cpu_cc_src, s->T1);
3787             break;
3788 #endif
3789         case MO_32:
3790             tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
3791             tcg_gen_trunc_tl_i32(s->tmp3_i32, s->T1);
3792             tcg_gen_muls2_i32(s->tmp2_i32, s->tmp3_i32,
3793                               s->tmp2_i32, s->tmp3_i32);
3794             tcg_gen_extu_i32_tl(cpu_regs[reg], s->tmp2_i32);
3795             tcg_gen_sari_i32(s->tmp2_i32, s->tmp2_i32, 31);
3796             tcg_gen_mov_tl(cpu_cc_dst, cpu_regs[reg]);
3797             tcg_gen_sub_i32(s->tmp2_i32, s->tmp2_i32, s->tmp3_i32);
3798             tcg_gen_extu_i32_tl(cpu_cc_src, s->tmp2_i32);
3799             break;
3800         default:
3801             tcg_gen_ext16s_tl(s->T0, s->T0);
3802             tcg_gen_ext16s_tl(s->T1, s->T1);
3803             /* XXX: use 32 bit mul which could be faster */
3804             tcg_gen_mul_tl(s->T0, s->T0, s->T1);
3805             tcg_gen_mov_tl(cpu_cc_dst, s->T0);
3806             tcg_gen_ext16s_tl(s->tmp0, s->T0);
3807             tcg_gen_sub_tl(cpu_cc_src, s->T0, s->tmp0);
3808             gen_op_mov_reg_v(s, ot, reg, s->T0);
3809             break;
3810         }
3811         set_cc_op(s, CC_OP_MULB + ot);
3812         break;
3813     case 0x1c0:
3814     case 0x1c1: /* xadd Ev, Gv */
3815         ot = mo_b_d(b, dflag);
3816         modrm = x86_ldub_code(env, s);
3817         reg = ((modrm >> 3) & 7) | REX_R(s);
3818         mod = (modrm >> 6) & 3;
3819         gen_op_mov_v_reg(s, ot, s->T0, reg);
3820         if (mod == 3) {
3821             rm = (modrm & 7) | REX_B(s);
3822             gen_op_mov_v_reg(s, ot, s->T1, rm);
3823             tcg_gen_add_tl(s->T0, s->T0, s->T1);
3824             gen_op_mov_reg_v(s, ot, reg, s->T1);
3825             gen_op_mov_reg_v(s, ot, rm, s->T0);
3826         } else {
3827             gen_lea_modrm(env, s, modrm);
3828             if (s->prefix & PREFIX_LOCK) {
3829                 tcg_gen_atomic_fetch_add_tl(s->T1, s->A0, s->T0,
3830                                             s->mem_index, ot | MO_LE);
3831                 tcg_gen_add_tl(s->T0, s->T0, s->T1);
3832             } else {
3833                 gen_op_ld_v(s, ot, s->T1, s->A0);
3834                 tcg_gen_add_tl(s->T0, s->T0, s->T1);
3835                 gen_op_st_v(s, ot, s->T0, s->A0);
3836             }
3837             gen_op_mov_reg_v(s, ot, reg, s->T1);
3838         }
3839         gen_op_update2_cc(s);
3840         set_cc_op(s, CC_OP_ADDB + ot);
3841         break;
3842     case 0x1b0:
3843     case 0x1b1: /* cmpxchg Ev, Gv */
3844         {
3845             TCGv oldv, newv, cmpv, dest;
3846 
3847             ot = mo_b_d(b, dflag);
3848             modrm = x86_ldub_code(env, s);
3849             reg = ((modrm >> 3) & 7) | REX_R(s);
3850             mod = (modrm >> 6) & 3;
3851             oldv = tcg_temp_new();
3852             newv = tcg_temp_new();
3853             cmpv = tcg_temp_new();
3854             gen_op_mov_v_reg(s, ot, newv, reg);
3855             tcg_gen_mov_tl(cmpv, cpu_regs[R_EAX]);
3856             gen_extu(ot, cmpv);
3857             if (s->prefix & PREFIX_LOCK) {
3858                 if (mod == 3) {
3859                     goto illegal_op;
3860                 }
3861                 gen_lea_modrm(env, s, modrm);
3862                 tcg_gen_atomic_cmpxchg_tl(oldv, s->A0, cmpv, newv,
3863                                           s->mem_index, ot | MO_LE);
3864             } else {
3865                 if (mod == 3) {
3866                     rm = (modrm & 7) | REX_B(s);
3867                     gen_op_mov_v_reg(s, ot, oldv, rm);
3868                     gen_extu(ot, oldv);
3869 
3870                     /*
3871                      * Unlike the memory case, where "the destination operand receives
3872                      * a write cycle without regard to the result of the comparison",
3873                      * rm must not be touched altogether if the write fails, including
3874                      * not zero-extending it on 64-bit processors.  So, precompute
3875                      * the result of a successful writeback and perform the movcond
3876                      * directly on cpu_regs.  Also need to write accumulator first, in
3877                      * case rm is part of RAX too.
3878                      */
3879                     dest = gen_op_deposit_reg_v(s, ot, rm, newv, newv);
3880                     tcg_gen_movcond_tl(TCG_COND_EQ, dest, oldv, cmpv, newv, dest);
3881                 } else {
3882                     gen_lea_modrm(env, s, modrm);
3883                     gen_op_ld_v(s, ot, oldv, s->A0);
3884 
3885                     /*
3886                      * Perform an unconditional store cycle like physical cpu;
3887                      * must be before changing accumulator to ensure
3888                      * idempotency if the store faults and the instruction
3889                      * is restarted
3890                      */
3891                     tcg_gen_movcond_tl(TCG_COND_EQ, newv, oldv, cmpv, newv, oldv);
3892                     gen_op_st_v(s, ot, newv, s->A0);
3893                 }
3894             }
3895 	    /*
3896 	     * Write EAX only if the cmpxchg fails; reuse newv as the destination,
3897 	     * since it's dead here.
3898 	     */
3899             dest = gen_op_deposit_reg_v(s, ot, R_EAX, newv, oldv);
3900             tcg_gen_movcond_tl(TCG_COND_EQ, dest, oldv, cmpv, dest, newv);
3901             tcg_gen_mov_tl(cpu_cc_src, oldv);
3902             tcg_gen_mov_tl(s->cc_srcT, cmpv);
3903             tcg_gen_sub_tl(cpu_cc_dst, cmpv, oldv);
3904             set_cc_op(s, CC_OP_SUBB + ot);
3905         }
3906         break;
3907     case 0x1c7: /* cmpxchg8b */
3908         modrm = x86_ldub_code(env, s);
3909         mod = (modrm >> 6) & 3;
3910         switch ((modrm >> 3) & 7) {
3911         case 1: /* CMPXCHG8, CMPXCHG16 */
3912             if (mod == 3) {
3913                 goto illegal_op;
3914             }
3915 #ifdef TARGET_X86_64
3916             if (dflag == MO_64) {
3917                 if (!(s->cpuid_ext_features & CPUID_EXT_CX16)) {
3918                     goto illegal_op;
3919                 }
3920                 gen_cmpxchg16b(s, env, modrm);
3921                 break;
3922             }
3923 #endif
3924             if (!(s->cpuid_features & CPUID_CX8)) {
3925                 goto illegal_op;
3926             }
3927             gen_cmpxchg8b(s, env, modrm);
3928             break;
3929 
3930         case 7: /* RDSEED, RDPID with f3 prefix */
3931             if (mod != 3 ||
3932                 (s->prefix & (PREFIX_LOCK | PREFIX_REPNZ))) {
3933                 goto illegal_op;
3934             }
3935             if (s->prefix & PREFIX_REPZ) {
3936                 if (!(s->cpuid_ext_features & CPUID_7_0_ECX_RDPID)) {
3937                     goto illegal_op;
3938                 }
3939                 gen_helper_rdpid(s->T0, tcg_env);
3940                 rm = (modrm & 7) | REX_B(s);
3941                 gen_op_mov_reg_v(s, dflag, rm, s->T0);
3942                 break;
3943             } else {
3944                 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_RDSEED)) {
3945                     goto illegal_op;
3946                 }
3947                 goto do_rdrand;
3948             }
3949 
3950         case 6: /* RDRAND */
3951             if (mod != 3 ||
3952                 (s->prefix & (PREFIX_LOCK | PREFIX_REPZ | PREFIX_REPNZ)) ||
3953                 !(s->cpuid_ext_features & CPUID_EXT_RDRAND)) {
3954                 goto illegal_op;
3955             }
3956         do_rdrand:
3957             translator_io_start(&s->base);
3958             gen_helper_rdrand(s->T0, tcg_env);
3959             rm = (modrm & 7) | REX_B(s);
3960             gen_op_mov_reg_v(s, dflag, rm, s->T0);
3961             set_cc_op(s, CC_OP_EFLAGS);
3962             break;
3963 
3964         default:
3965             goto illegal_op;
3966         }
3967         break;
3968 
3969         /**************************/
3970         /* push/pop */
3971     case 0x50 ... 0x57: /* push */
3972         gen_op_mov_v_reg(s, MO_32, s->T0, (b & 7) | REX_B(s));
3973         gen_push_v(s, s->T0);
3974         break;
3975     case 0x58 ... 0x5f: /* pop */
3976         ot = gen_pop_T0(s);
3977         /* NOTE: order is important for pop %sp */
3978         gen_pop_update(s, ot);
3979         gen_op_mov_reg_v(s, ot, (b & 7) | REX_B(s), s->T0);
3980         break;
3981     case 0x60: /* pusha */
3982         if (CODE64(s))
3983             goto illegal_op;
3984         gen_pusha(s);
3985         break;
3986     case 0x61: /* popa */
3987         if (CODE64(s))
3988             goto illegal_op;
3989         gen_popa(s);
3990         break;
3991     case 0x68: /* push Iv */
3992     case 0x6a:
3993         ot = mo_pushpop(s, dflag);
3994         if (b == 0x68)
3995             val = insn_get(env, s, ot);
3996         else
3997             val = (int8_t)insn_get(env, s, MO_8);
3998         tcg_gen_movi_tl(s->T0, val);
3999         gen_push_v(s, s->T0);
4000         break;
4001     case 0x8f: /* pop Ev */
4002         modrm = x86_ldub_code(env, s);
4003         mod = (modrm >> 6) & 3;
4004         ot = gen_pop_T0(s);
4005         if (mod == 3) {
4006             /* NOTE: order is important for pop %sp */
4007             gen_pop_update(s, ot);
4008             rm = (modrm & 7) | REX_B(s);
4009             gen_op_mov_reg_v(s, ot, rm, s->T0);
4010         } else {
4011             /* NOTE: order is important too for MMU exceptions */
4012             s->popl_esp_hack = 1 << ot;
4013             gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
4014             s->popl_esp_hack = 0;
4015             gen_pop_update(s, ot);
4016         }
4017         break;
4018     case 0xc8: /* enter */
4019         {
4020             int level;
4021             val = x86_lduw_code(env, s);
4022             level = x86_ldub_code(env, s);
4023             gen_enter(s, val, level);
4024         }
4025         break;
4026     case 0xc9: /* leave */
4027         gen_leave(s);
4028         break;
4029     case 0x06: /* push es */
4030     case 0x0e: /* push cs */
4031     case 0x16: /* push ss */
4032     case 0x1e: /* push ds */
4033         if (CODE64(s))
4034             goto illegal_op;
4035         gen_op_movl_T0_seg(s, b >> 3);
4036         gen_push_v(s, s->T0);
4037         break;
4038     case 0x1a0: /* push fs */
4039     case 0x1a8: /* push gs */
4040         gen_op_movl_T0_seg(s, (b >> 3) & 7);
4041         gen_push_v(s, s->T0);
4042         break;
4043     case 0x07: /* pop es */
4044     case 0x17: /* pop ss */
4045     case 0x1f: /* pop ds */
4046         if (CODE64(s))
4047             goto illegal_op;
4048         reg = b >> 3;
4049         ot = gen_pop_T0(s);
4050         gen_movl_seg_T0(s, reg);
4051         gen_pop_update(s, ot);
4052         break;
4053     case 0x1a1: /* pop fs */
4054     case 0x1a9: /* pop gs */
4055         ot = gen_pop_T0(s);
4056         gen_movl_seg_T0(s, (b >> 3) & 7);
4057         gen_pop_update(s, ot);
4058         break;
4059 
4060         /**************************/
4061         /* mov */
4062     case 0x88:
4063     case 0x89: /* mov Gv, Ev */
4064         ot = mo_b_d(b, dflag);
4065         modrm = x86_ldub_code(env, s);
4066         reg = ((modrm >> 3) & 7) | REX_R(s);
4067 
4068         /* generate a generic store */
4069         gen_ldst_modrm(env, s, modrm, ot, reg, 1);
4070         break;
4071     case 0xc6:
4072     case 0xc7: /* mov Ev, Iv */
4073         ot = mo_b_d(b, dflag);
4074         modrm = x86_ldub_code(env, s);
4075         mod = (modrm >> 6) & 3;
4076         if (mod != 3) {
4077             s->rip_offset = insn_const_size(ot);
4078             gen_lea_modrm(env, s, modrm);
4079         }
4080         val = insn_get(env, s, ot);
4081         tcg_gen_movi_tl(s->T0, val);
4082         if (mod != 3) {
4083             gen_op_st_v(s, ot, s->T0, s->A0);
4084         } else {
4085             gen_op_mov_reg_v(s, ot, (modrm & 7) | REX_B(s), s->T0);
4086         }
4087         break;
4088     case 0x8a:
4089     case 0x8b: /* mov Ev, Gv */
4090         ot = mo_b_d(b, dflag);
4091         modrm = x86_ldub_code(env, s);
4092         reg = ((modrm >> 3) & 7) | REX_R(s);
4093 
4094         gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
4095         gen_op_mov_reg_v(s, ot, reg, s->T0);
4096         break;
4097     case 0x8e: /* mov seg, Gv */
4098         modrm = x86_ldub_code(env, s);
4099         reg = (modrm >> 3) & 7;
4100         if (reg >= 6 || reg == R_CS)
4101             goto illegal_op;
4102         gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
4103         gen_movl_seg_T0(s, reg);
4104         break;
4105     case 0x8c: /* mov Gv, seg */
4106         modrm = x86_ldub_code(env, s);
4107         reg = (modrm >> 3) & 7;
4108         mod = (modrm >> 6) & 3;
4109         if (reg >= 6)
4110             goto illegal_op;
4111         gen_op_movl_T0_seg(s, reg);
4112         ot = mod == 3 ? dflag : MO_16;
4113         gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
4114         break;
4115 
4116     case 0x1b6: /* movzbS Gv, Eb */
4117     case 0x1b7: /* movzwS Gv, Eb */
4118     case 0x1be: /* movsbS Gv, Eb */
4119     case 0x1bf: /* movswS Gv, Eb */
4120         {
4121             MemOp d_ot;
4122             MemOp s_ot;
4123 
4124             /* d_ot is the size of destination */
4125             d_ot = dflag;
4126             /* ot is the size of source */
4127             ot = (b & 1) + MO_8;
4128             /* s_ot is the sign+size of source */
4129             s_ot = b & 8 ? MO_SIGN | ot : ot;
4130 
4131             modrm = x86_ldub_code(env, s);
4132             reg = ((modrm >> 3) & 7) | REX_R(s);
4133             mod = (modrm >> 6) & 3;
4134             rm = (modrm & 7) | REX_B(s);
4135 
4136             if (mod == 3) {
4137                 if (s_ot == MO_SB && byte_reg_is_xH(s, rm)) {
4138                     tcg_gen_sextract_tl(s->T0, cpu_regs[rm - 4], 8, 8);
4139                 } else {
4140                     gen_op_mov_v_reg(s, ot, s->T0, rm);
4141                     switch (s_ot) {
4142                     case MO_UB:
4143                         tcg_gen_ext8u_tl(s->T0, s->T0);
4144                         break;
4145                     case MO_SB:
4146                         tcg_gen_ext8s_tl(s->T0, s->T0);
4147                         break;
4148                     case MO_UW:
4149                         tcg_gen_ext16u_tl(s->T0, s->T0);
4150                         break;
4151                     default:
4152                     case MO_SW:
4153                         tcg_gen_ext16s_tl(s->T0, s->T0);
4154                         break;
4155                     }
4156                 }
4157                 gen_op_mov_reg_v(s, d_ot, reg, s->T0);
4158             } else {
4159                 gen_lea_modrm(env, s, modrm);
4160                 gen_op_ld_v(s, s_ot, s->T0, s->A0);
4161                 gen_op_mov_reg_v(s, d_ot, reg, s->T0);
4162             }
4163         }
4164         break;
4165 
4166     case 0x8d: /* lea */
4167         modrm = x86_ldub_code(env, s);
4168         mod = (modrm >> 6) & 3;
4169         if (mod == 3)
4170             goto illegal_op;
4171         reg = ((modrm >> 3) & 7) | REX_R(s);
4172         {
4173             AddressParts a = gen_lea_modrm_0(env, s, modrm);
4174             TCGv ea = gen_lea_modrm_1(s, a, false);
4175             gen_lea_v_seg(s, s->aflag, ea, -1, -1);
4176             gen_op_mov_reg_v(s, dflag, reg, s->A0);
4177         }
4178         break;
4179 
4180     case 0xa0: /* mov EAX, Ov */
4181     case 0xa1:
4182     case 0xa2: /* mov Ov, EAX */
4183     case 0xa3:
4184         {
4185             target_ulong offset_addr;
4186 
4187             ot = mo_b_d(b, dflag);
4188             offset_addr = insn_get_addr(env, s, s->aflag);
4189             tcg_gen_movi_tl(s->A0, offset_addr);
4190             gen_add_A0_ds_seg(s);
4191             if ((b & 2) == 0) {
4192                 gen_op_ld_v(s, ot, s->T0, s->A0);
4193                 gen_op_mov_reg_v(s, ot, R_EAX, s->T0);
4194             } else {
4195                 gen_op_mov_v_reg(s, ot, s->T0, R_EAX);
4196                 gen_op_st_v(s, ot, s->T0, s->A0);
4197             }
4198         }
4199         break;
4200     case 0xd7: /* xlat */
4201         tcg_gen_mov_tl(s->A0, cpu_regs[R_EBX]);
4202         tcg_gen_ext8u_tl(s->T0, cpu_regs[R_EAX]);
4203         tcg_gen_add_tl(s->A0, s->A0, s->T0);
4204         gen_add_A0_ds_seg(s);
4205         gen_op_ld_v(s, MO_8, s->T0, s->A0);
4206         gen_op_mov_reg_v(s, MO_8, R_EAX, s->T0);
4207         break;
4208     case 0xb0 ... 0xb7: /* mov R, Ib */
4209         val = insn_get(env, s, MO_8);
4210         tcg_gen_movi_tl(s->T0, val);
4211         gen_op_mov_reg_v(s, MO_8, (b & 7) | REX_B(s), s->T0);
4212         break;
4213     case 0xb8 ... 0xbf: /* mov R, Iv */
4214 #ifdef TARGET_X86_64
4215         if (dflag == MO_64) {
4216             uint64_t tmp;
4217             /* 64 bit case */
4218             tmp = x86_ldq_code(env, s);
4219             reg = (b & 7) | REX_B(s);
4220             tcg_gen_movi_tl(s->T0, tmp);
4221             gen_op_mov_reg_v(s, MO_64, reg, s->T0);
4222         } else
4223 #endif
4224         {
4225             ot = dflag;
4226             val = insn_get(env, s, ot);
4227             reg = (b & 7) | REX_B(s);
4228             tcg_gen_movi_tl(s->T0, val);
4229             gen_op_mov_reg_v(s, ot, reg, s->T0);
4230         }
4231         break;
4232 
4233     case 0x91 ... 0x97: /* xchg R, EAX */
4234     do_xchg_reg_eax:
4235         ot = dflag;
4236         reg = (b & 7) | REX_B(s);
4237         rm = R_EAX;
4238         goto do_xchg_reg;
4239     case 0x86:
4240     case 0x87: /* xchg Ev, Gv */
4241         ot = mo_b_d(b, dflag);
4242         modrm = x86_ldub_code(env, s);
4243         reg = ((modrm >> 3) & 7) | REX_R(s);
4244         mod = (modrm >> 6) & 3;
4245         if (mod == 3) {
4246             rm = (modrm & 7) | REX_B(s);
4247         do_xchg_reg:
4248             gen_op_mov_v_reg(s, ot, s->T0, reg);
4249             gen_op_mov_v_reg(s, ot, s->T1, rm);
4250             gen_op_mov_reg_v(s, ot, rm, s->T0);
4251             gen_op_mov_reg_v(s, ot, reg, s->T1);
4252         } else {
4253             gen_lea_modrm(env, s, modrm);
4254             gen_op_mov_v_reg(s, ot, s->T0, reg);
4255             /* for xchg, lock is implicit */
4256             tcg_gen_atomic_xchg_tl(s->T1, s->A0, s->T0,
4257                                    s->mem_index, ot | MO_LE);
4258             gen_op_mov_reg_v(s, ot, reg, s->T1);
4259         }
4260         break;
4261     case 0xc4: /* les Gv */
4262         /* In CODE64 this is VEX3; see above.  */
4263         op = R_ES;
4264         goto do_lxx;
4265     case 0xc5: /* lds Gv */
4266         /* In CODE64 this is VEX2; see above.  */
4267         op = R_DS;
4268         goto do_lxx;
4269     case 0x1b2: /* lss Gv */
4270         op = R_SS;
4271         goto do_lxx;
4272     case 0x1b4: /* lfs Gv */
4273         op = R_FS;
4274         goto do_lxx;
4275     case 0x1b5: /* lgs Gv */
4276         op = R_GS;
4277     do_lxx:
4278         ot = dflag != MO_16 ? MO_32 : MO_16;
4279         modrm = x86_ldub_code(env, s);
4280         reg = ((modrm >> 3) & 7) | REX_R(s);
4281         mod = (modrm >> 6) & 3;
4282         if (mod == 3)
4283             goto illegal_op;
4284         gen_lea_modrm(env, s, modrm);
4285         gen_op_ld_v(s, ot, s->T1, s->A0);
4286         gen_add_A0_im(s, 1 << ot);
4287         /* load the segment first to handle exceptions properly */
4288         gen_op_ld_v(s, MO_16, s->T0, s->A0);
4289         gen_movl_seg_T0(s, op);
4290         /* then put the data */
4291         gen_op_mov_reg_v(s, ot, reg, s->T1);
4292         break;
4293 
4294         /************************/
4295         /* shifts */
4296     case 0xc0:
4297     case 0xc1:
4298         /* shift Ev,Ib */
4299         shift = 2;
4300     grp2:
4301         {
4302             ot = mo_b_d(b, dflag);
4303             modrm = x86_ldub_code(env, s);
4304             mod = (modrm >> 6) & 3;
4305             op = (modrm >> 3) & 7;
4306 
4307             if (mod != 3) {
4308                 if (shift == 2) {
4309                     s->rip_offset = 1;
4310                 }
4311                 gen_lea_modrm(env, s, modrm);
4312                 opreg = OR_TMP0;
4313             } else {
4314                 opreg = (modrm & 7) | REX_B(s);
4315             }
4316 
4317             /* simpler op */
4318             if (shift == 0) {
4319                 gen_shift(s, op, ot, opreg, OR_ECX);
4320             } else {
4321                 if (shift == 2) {
4322                     shift = x86_ldub_code(env, s);
4323                 }
4324                 gen_shifti(s, op, ot, opreg, shift);
4325             }
4326         }
4327         break;
4328     case 0xd0:
4329     case 0xd1:
4330         /* shift Ev,1 */
4331         shift = 1;
4332         goto grp2;
4333     case 0xd2:
4334     case 0xd3:
4335         /* shift Ev,cl */
4336         shift = 0;
4337         goto grp2;
4338 
4339     case 0x1a4: /* shld imm */
4340         op = 0;
4341         shift = 1;
4342         goto do_shiftd;
4343     case 0x1a5: /* shld cl */
4344         op = 0;
4345         shift = 0;
4346         goto do_shiftd;
4347     case 0x1ac: /* shrd imm */
4348         op = 1;
4349         shift = 1;
4350         goto do_shiftd;
4351     case 0x1ad: /* shrd cl */
4352         op = 1;
4353         shift = 0;
4354     do_shiftd:
4355         ot = dflag;
4356         modrm = x86_ldub_code(env, s);
4357         mod = (modrm >> 6) & 3;
4358         rm = (modrm & 7) | REX_B(s);
4359         reg = ((modrm >> 3) & 7) | REX_R(s);
4360         if (mod != 3) {
4361             gen_lea_modrm(env, s, modrm);
4362             opreg = OR_TMP0;
4363         } else {
4364             opreg = rm;
4365         }
4366         gen_op_mov_v_reg(s, ot, s->T1, reg);
4367 
4368         if (shift) {
4369             TCGv imm = tcg_constant_tl(x86_ldub_code(env, s));
4370             gen_shiftd_rm_T1(s, ot, opreg, op, imm);
4371         } else {
4372             gen_shiftd_rm_T1(s, ot, opreg, op, cpu_regs[R_ECX]);
4373         }
4374         break;
4375 
4376         /************************/
4377         /* floats */
4378     case 0xd8 ... 0xdf:
4379         {
4380             bool update_fip = true;
4381 
4382             if (s->flags & (HF_EM_MASK | HF_TS_MASK)) {
4383                 /* if CR0.EM or CR0.TS are set, generate an FPU exception */
4384                 /* XXX: what to do if illegal op ? */
4385                 gen_exception(s, EXCP07_PREX);
4386                 break;
4387             }
4388             modrm = x86_ldub_code(env, s);
4389             mod = (modrm >> 6) & 3;
4390             rm = modrm & 7;
4391             op = ((b & 7) << 3) | ((modrm >> 3) & 7);
4392             if (mod != 3) {
4393                 /* memory op */
4394                 AddressParts a = gen_lea_modrm_0(env, s, modrm);
4395                 TCGv ea = gen_lea_modrm_1(s, a, false);
4396                 TCGv last_addr = tcg_temp_new();
4397                 bool update_fdp = true;
4398 
4399                 tcg_gen_mov_tl(last_addr, ea);
4400                 gen_lea_v_seg(s, s->aflag, ea, a.def_seg, s->override);
4401 
4402                 switch (op) {
4403                 case 0x00 ... 0x07: /* fxxxs */
4404                 case 0x10 ... 0x17: /* fixxxl */
4405                 case 0x20 ... 0x27: /* fxxxl */
4406                 case 0x30 ... 0x37: /* fixxx */
4407                     {
4408                         int op1;
4409                         op1 = op & 7;
4410 
4411                         switch (op >> 4) {
4412                         case 0:
4413                             tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0,
4414                                                 s->mem_index, MO_LEUL);
4415                             gen_helper_flds_FT0(tcg_env, s->tmp2_i32);
4416                             break;
4417                         case 1:
4418                             tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0,
4419                                                 s->mem_index, MO_LEUL);
4420                             gen_helper_fildl_FT0(tcg_env, s->tmp2_i32);
4421                             break;
4422                         case 2:
4423                             tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0,
4424                                                 s->mem_index, MO_LEUQ);
4425                             gen_helper_fldl_FT0(tcg_env, s->tmp1_i64);
4426                             break;
4427                         case 3:
4428                         default:
4429                             tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0,
4430                                                 s->mem_index, MO_LESW);
4431                             gen_helper_fildl_FT0(tcg_env, s->tmp2_i32);
4432                             break;
4433                         }
4434 
4435                         gen_helper_fp_arith_ST0_FT0(op1);
4436                         if (op1 == 3) {
4437                             /* fcomp needs pop */
4438                             gen_helper_fpop(tcg_env);
4439                         }
4440                     }
4441                     break;
4442                 case 0x08: /* flds */
4443                 case 0x0a: /* fsts */
4444                 case 0x0b: /* fstps */
4445                 case 0x18 ... 0x1b: /* fildl, fisttpl, fistl, fistpl */
4446                 case 0x28 ... 0x2b: /* fldl, fisttpll, fstl, fstpl */
4447                 case 0x38 ... 0x3b: /* filds, fisttps, fists, fistps */
4448                     switch (op & 7) {
4449                     case 0:
4450                         switch (op >> 4) {
4451                         case 0:
4452                             tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0,
4453                                                 s->mem_index, MO_LEUL);
4454                             gen_helper_flds_ST0(tcg_env, s->tmp2_i32);
4455                             break;
4456                         case 1:
4457                             tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0,
4458                                                 s->mem_index, MO_LEUL);
4459                             gen_helper_fildl_ST0(tcg_env, s->tmp2_i32);
4460                             break;
4461                         case 2:
4462                             tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0,
4463                                                 s->mem_index, MO_LEUQ);
4464                             gen_helper_fldl_ST0(tcg_env, s->tmp1_i64);
4465                             break;
4466                         case 3:
4467                         default:
4468                             tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0,
4469                                                 s->mem_index, MO_LESW);
4470                             gen_helper_fildl_ST0(tcg_env, s->tmp2_i32);
4471                             break;
4472                         }
4473                         break;
4474                     case 1:
4475                         /* XXX: the corresponding CPUID bit must be tested ! */
4476                         switch (op >> 4) {
4477                         case 1:
4478                             gen_helper_fisttl_ST0(s->tmp2_i32, tcg_env);
4479                             tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0,
4480                                                 s->mem_index, MO_LEUL);
4481                             break;
4482                         case 2:
4483                             gen_helper_fisttll_ST0(s->tmp1_i64, tcg_env);
4484                             tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0,
4485                                                 s->mem_index, MO_LEUQ);
4486                             break;
4487                         case 3:
4488                         default:
4489                             gen_helper_fistt_ST0(s->tmp2_i32, tcg_env);
4490                             tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0,
4491                                                 s->mem_index, MO_LEUW);
4492                             break;
4493                         }
4494                         gen_helper_fpop(tcg_env);
4495                         break;
4496                     default:
4497                         switch (op >> 4) {
4498                         case 0:
4499                             gen_helper_fsts_ST0(s->tmp2_i32, tcg_env);
4500                             tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0,
4501                                                 s->mem_index, MO_LEUL);
4502                             break;
4503                         case 1:
4504                             gen_helper_fistl_ST0(s->tmp2_i32, tcg_env);
4505                             tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0,
4506                                                 s->mem_index, MO_LEUL);
4507                             break;
4508                         case 2:
4509                             gen_helper_fstl_ST0(s->tmp1_i64, tcg_env);
4510                             tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0,
4511                                                 s->mem_index, MO_LEUQ);
4512                             break;
4513                         case 3:
4514                         default:
4515                             gen_helper_fist_ST0(s->tmp2_i32, tcg_env);
4516                             tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0,
4517                                                 s->mem_index, MO_LEUW);
4518                             break;
4519                         }
4520                         if ((op & 7) == 3) {
4521                             gen_helper_fpop(tcg_env);
4522                         }
4523                         break;
4524                     }
4525                     break;
4526                 case 0x0c: /* fldenv mem */
4527                     gen_helper_fldenv(tcg_env, s->A0,
4528                                       tcg_constant_i32(dflag - 1));
4529                     update_fip = update_fdp = false;
4530                     break;
4531                 case 0x0d: /* fldcw mem */
4532                     tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0,
4533                                         s->mem_index, MO_LEUW);
4534                     gen_helper_fldcw(tcg_env, s->tmp2_i32);
4535                     update_fip = update_fdp = false;
4536                     break;
4537                 case 0x0e: /* fnstenv mem */
4538                     gen_helper_fstenv(tcg_env, s->A0,
4539                                       tcg_constant_i32(dflag - 1));
4540                     update_fip = update_fdp = false;
4541                     break;
4542                 case 0x0f: /* fnstcw mem */
4543                     gen_helper_fnstcw(s->tmp2_i32, tcg_env);
4544                     tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0,
4545                                         s->mem_index, MO_LEUW);
4546                     update_fip = update_fdp = false;
4547                     break;
4548                 case 0x1d: /* fldt mem */
4549                     gen_helper_fldt_ST0(tcg_env, s->A0);
4550                     break;
4551                 case 0x1f: /* fstpt mem */
4552                     gen_helper_fstt_ST0(tcg_env, s->A0);
4553                     gen_helper_fpop(tcg_env);
4554                     break;
4555                 case 0x2c: /* frstor mem */
4556                     gen_helper_frstor(tcg_env, s->A0,
4557                                       tcg_constant_i32(dflag - 1));
4558                     update_fip = update_fdp = false;
4559                     break;
4560                 case 0x2e: /* fnsave mem */
4561                     gen_helper_fsave(tcg_env, s->A0,
4562                                      tcg_constant_i32(dflag - 1));
4563                     update_fip = update_fdp = false;
4564                     break;
4565                 case 0x2f: /* fnstsw mem */
4566                     gen_helper_fnstsw(s->tmp2_i32, tcg_env);
4567                     tcg_gen_qemu_st_i32(s->tmp2_i32, s->A0,
4568                                         s->mem_index, MO_LEUW);
4569                     update_fip = update_fdp = false;
4570                     break;
4571                 case 0x3c: /* fbld */
4572                     gen_helper_fbld_ST0(tcg_env, s->A0);
4573                     break;
4574                 case 0x3e: /* fbstp */
4575                     gen_helper_fbst_ST0(tcg_env, s->A0);
4576                     gen_helper_fpop(tcg_env);
4577                     break;
4578                 case 0x3d: /* fildll */
4579                     tcg_gen_qemu_ld_i64(s->tmp1_i64, s->A0,
4580                                         s->mem_index, MO_LEUQ);
4581                     gen_helper_fildll_ST0(tcg_env, s->tmp1_i64);
4582                     break;
4583                 case 0x3f: /* fistpll */
4584                     gen_helper_fistll_ST0(s->tmp1_i64, tcg_env);
4585                     tcg_gen_qemu_st_i64(s->tmp1_i64, s->A0,
4586                                         s->mem_index, MO_LEUQ);
4587                     gen_helper_fpop(tcg_env);
4588                     break;
4589                 default:
4590                     goto unknown_op;
4591                 }
4592 
4593                 if (update_fdp) {
4594                     int last_seg = s->override >= 0 ? s->override : a.def_seg;
4595 
4596                     tcg_gen_ld_i32(s->tmp2_i32, tcg_env,
4597                                    offsetof(CPUX86State,
4598                                             segs[last_seg].selector));
4599                     tcg_gen_st16_i32(s->tmp2_i32, tcg_env,
4600                                      offsetof(CPUX86State, fpds));
4601                     tcg_gen_st_tl(last_addr, tcg_env,
4602                                   offsetof(CPUX86State, fpdp));
4603                 }
4604             } else {
4605                 /* register float ops */
4606                 opreg = rm;
4607 
4608                 switch (op) {
4609                 case 0x08: /* fld sti */
4610                     gen_helper_fpush(tcg_env);
4611                     gen_helper_fmov_ST0_STN(tcg_env,
4612                                             tcg_constant_i32((opreg + 1) & 7));
4613                     break;
4614                 case 0x09: /* fxchg sti */
4615                 case 0x29: /* fxchg4 sti, undocumented op */
4616                 case 0x39: /* fxchg7 sti, undocumented op */
4617                     gen_helper_fxchg_ST0_STN(tcg_env, tcg_constant_i32(opreg));
4618                     break;
4619                 case 0x0a: /* grp d9/2 */
4620                     switch (rm) {
4621                     case 0: /* fnop */
4622                         /*
4623                          * check exceptions (FreeBSD FPU probe)
4624                          * needs to be treated as I/O because of ferr_irq
4625                          */
4626                         translator_io_start(&s->base);
4627                         gen_helper_fwait(tcg_env);
4628                         update_fip = false;
4629                         break;
4630                     default:
4631                         goto unknown_op;
4632                     }
4633                     break;
4634                 case 0x0c: /* grp d9/4 */
4635                     switch (rm) {
4636                     case 0: /* fchs */
4637                         gen_helper_fchs_ST0(tcg_env);
4638                         break;
4639                     case 1: /* fabs */
4640                         gen_helper_fabs_ST0(tcg_env);
4641                         break;
4642                     case 4: /* ftst */
4643                         gen_helper_fldz_FT0(tcg_env);
4644                         gen_helper_fcom_ST0_FT0(tcg_env);
4645                         break;
4646                     case 5: /* fxam */
4647                         gen_helper_fxam_ST0(tcg_env);
4648                         break;
4649                     default:
4650                         goto unknown_op;
4651                     }
4652                     break;
4653                 case 0x0d: /* grp d9/5 */
4654                     {
4655                         switch (rm) {
4656                         case 0:
4657                             gen_helper_fpush(tcg_env);
4658                             gen_helper_fld1_ST0(tcg_env);
4659                             break;
4660                         case 1:
4661                             gen_helper_fpush(tcg_env);
4662                             gen_helper_fldl2t_ST0(tcg_env);
4663                             break;
4664                         case 2:
4665                             gen_helper_fpush(tcg_env);
4666                             gen_helper_fldl2e_ST0(tcg_env);
4667                             break;
4668                         case 3:
4669                             gen_helper_fpush(tcg_env);
4670                             gen_helper_fldpi_ST0(tcg_env);
4671                             break;
4672                         case 4:
4673                             gen_helper_fpush(tcg_env);
4674                             gen_helper_fldlg2_ST0(tcg_env);
4675                             break;
4676                         case 5:
4677                             gen_helper_fpush(tcg_env);
4678                             gen_helper_fldln2_ST0(tcg_env);
4679                             break;
4680                         case 6:
4681                             gen_helper_fpush(tcg_env);
4682                             gen_helper_fldz_ST0(tcg_env);
4683                             break;
4684                         default:
4685                             goto unknown_op;
4686                         }
4687                     }
4688                     break;
4689                 case 0x0e: /* grp d9/6 */
4690                     switch (rm) {
4691                     case 0: /* f2xm1 */
4692                         gen_helper_f2xm1(tcg_env);
4693                         break;
4694                     case 1: /* fyl2x */
4695                         gen_helper_fyl2x(tcg_env);
4696                         break;
4697                     case 2: /* fptan */
4698                         gen_helper_fptan(tcg_env);
4699                         break;
4700                     case 3: /* fpatan */
4701                         gen_helper_fpatan(tcg_env);
4702                         break;
4703                     case 4: /* fxtract */
4704                         gen_helper_fxtract(tcg_env);
4705                         break;
4706                     case 5: /* fprem1 */
4707                         gen_helper_fprem1(tcg_env);
4708                         break;
4709                     case 6: /* fdecstp */
4710                         gen_helper_fdecstp(tcg_env);
4711                         break;
4712                     default:
4713                     case 7: /* fincstp */
4714                         gen_helper_fincstp(tcg_env);
4715                         break;
4716                     }
4717                     break;
4718                 case 0x0f: /* grp d9/7 */
4719                     switch (rm) {
4720                     case 0: /* fprem */
4721                         gen_helper_fprem(tcg_env);
4722                         break;
4723                     case 1: /* fyl2xp1 */
4724                         gen_helper_fyl2xp1(tcg_env);
4725                         break;
4726                     case 2: /* fsqrt */
4727                         gen_helper_fsqrt(tcg_env);
4728                         break;
4729                     case 3: /* fsincos */
4730                         gen_helper_fsincos(tcg_env);
4731                         break;
4732                     case 5: /* fscale */
4733                         gen_helper_fscale(tcg_env);
4734                         break;
4735                     case 4: /* frndint */
4736                         gen_helper_frndint(tcg_env);
4737                         break;
4738                     case 6: /* fsin */
4739                         gen_helper_fsin(tcg_env);
4740                         break;
4741                     default:
4742                     case 7: /* fcos */
4743                         gen_helper_fcos(tcg_env);
4744                         break;
4745                     }
4746                     break;
4747                 case 0x00: case 0x01: case 0x04 ... 0x07: /* fxxx st, sti */
4748                 case 0x20: case 0x21: case 0x24 ... 0x27: /* fxxx sti, st */
4749                 case 0x30: case 0x31: case 0x34 ... 0x37: /* fxxxp sti, st */
4750                     {
4751                         int op1;
4752 
4753                         op1 = op & 7;
4754                         if (op >= 0x20) {
4755                             gen_helper_fp_arith_STN_ST0(op1, opreg);
4756                             if (op >= 0x30) {
4757                                 gen_helper_fpop(tcg_env);
4758                             }
4759                         } else {
4760                             gen_helper_fmov_FT0_STN(tcg_env,
4761                                                     tcg_constant_i32(opreg));
4762                             gen_helper_fp_arith_ST0_FT0(op1);
4763                         }
4764                     }
4765                     break;
4766                 case 0x02: /* fcom */
4767                 case 0x22: /* fcom2, undocumented op */
4768                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4769                     gen_helper_fcom_ST0_FT0(tcg_env);
4770                     break;
4771                 case 0x03: /* fcomp */
4772                 case 0x23: /* fcomp3, undocumented op */
4773                 case 0x32: /* fcomp5, undocumented op */
4774                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4775                     gen_helper_fcom_ST0_FT0(tcg_env);
4776                     gen_helper_fpop(tcg_env);
4777                     break;
4778                 case 0x15: /* da/5 */
4779                     switch (rm) {
4780                     case 1: /* fucompp */
4781                         gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(1));
4782                         gen_helper_fucom_ST0_FT0(tcg_env);
4783                         gen_helper_fpop(tcg_env);
4784                         gen_helper_fpop(tcg_env);
4785                         break;
4786                     default:
4787                         goto unknown_op;
4788                     }
4789                     break;
4790                 case 0x1c:
4791                     switch (rm) {
4792                     case 0: /* feni (287 only, just do nop here) */
4793                         break;
4794                     case 1: /* fdisi (287 only, just do nop here) */
4795                         break;
4796                     case 2: /* fclex */
4797                         gen_helper_fclex(tcg_env);
4798                         update_fip = false;
4799                         break;
4800                     case 3: /* fninit */
4801                         gen_helper_fninit(tcg_env);
4802                         update_fip = false;
4803                         break;
4804                     case 4: /* fsetpm (287 only, just do nop here) */
4805                         break;
4806                     default:
4807                         goto unknown_op;
4808                     }
4809                     break;
4810                 case 0x1d: /* fucomi */
4811                     if (!(s->cpuid_features & CPUID_CMOV)) {
4812                         goto illegal_op;
4813                     }
4814                     gen_update_cc_op(s);
4815                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4816                     gen_helper_fucomi_ST0_FT0(tcg_env);
4817                     set_cc_op(s, CC_OP_EFLAGS);
4818                     break;
4819                 case 0x1e: /* fcomi */
4820                     if (!(s->cpuid_features & CPUID_CMOV)) {
4821                         goto illegal_op;
4822                     }
4823                     gen_update_cc_op(s);
4824                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4825                     gen_helper_fcomi_ST0_FT0(tcg_env);
4826                     set_cc_op(s, CC_OP_EFLAGS);
4827                     break;
4828                 case 0x28: /* ffree sti */
4829                     gen_helper_ffree_STN(tcg_env, tcg_constant_i32(opreg));
4830                     break;
4831                 case 0x2a: /* fst sti */
4832                     gen_helper_fmov_STN_ST0(tcg_env, tcg_constant_i32(opreg));
4833                     break;
4834                 case 0x2b: /* fstp sti */
4835                 case 0x0b: /* fstp1 sti, undocumented op */
4836                 case 0x3a: /* fstp8 sti, undocumented op */
4837                 case 0x3b: /* fstp9 sti, undocumented op */
4838                     gen_helper_fmov_STN_ST0(tcg_env, tcg_constant_i32(opreg));
4839                     gen_helper_fpop(tcg_env);
4840                     break;
4841                 case 0x2c: /* fucom st(i) */
4842                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4843                     gen_helper_fucom_ST0_FT0(tcg_env);
4844                     break;
4845                 case 0x2d: /* fucomp st(i) */
4846                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4847                     gen_helper_fucom_ST0_FT0(tcg_env);
4848                     gen_helper_fpop(tcg_env);
4849                     break;
4850                 case 0x33: /* de/3 */
4851                     switch (rm) {
4852                     case 1: /* fcompp */
4853                         gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(1));
4854                         gen_helper_fcom_ST0_FT0(tcg_env);
4855                         gen_helper_fpop(tcg_env);
4856                         gen_helper_fpop(tcg_env);
4857                         break;
4858                     default:
4859                         goto unknown_op;
4860                     }
4861                     break;
4862                 case 0x38: /* ffreep sti, undocumented op */
4863                     gen_helper_ffree_STN(tcg_env, tcg_constant_i32(opreg));
4864                     gen_helper_fpop(tcg_env);
4865                     break;
4866                 case 0x3c: /* df/4 */
4867                     switch (rm) {
4868                     case 0:
4869                         gen_helper_fnstsw(s->tmp2_i32, tcg_env);
4870                         tcg_gen_extu_i32_tl(s->T0, s->tmp2_i32);
4871                         gen_op_mov_reg_v(s, MO_16, R_EAX, s->T0);
4872                         break;
4873                     default:
4874                         goto unknown_op;
4875                     }
4876                     break;
4877                 case 0x3d: /* fucomip */
4878                     if (!(s->cpuid_features & CPUID_CMOV)) {
4879                         goto illegal_op;
4880                     }
4881                     gen_update_cc_op(s);
4882                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4883                     gen_helper_fucomi_ST0_FT0(tcg_env);
4884                     gen_helper_fpop(tcg_env);
4885                     set_cc_op(s, CC_OP_EFLAGS);
4886                     break;
4887                 case 0x3e: /* fcomip */
4888                     if (!(s->cpuid_features & CPUID_CMOV)) {
4889                         goto illegal_op;
4890                     }
4891                     gen_update_cc_op(s);
4892                     gen_helper_fmov_FT0_STN(tcg_env, tcg_constant_i32(opreg));
4893                     gen_helper_fcomi_ST0_FT0(tcg_env);
4894                     gen_helper_fpop(tcg_env);
4895                     set_cc_op(s, CC_OP_EFLAGS);
4896                     break;
4897                 case 0x10 ... 0x13: /* fcmovxx */
4898                 case 0x18 ... 0x1b:
4899                     {
4900                         int op1;
4901                         TCGLabel *l1;
4902                         static const uint8_t fcmov_cc[8] = {
4903                             (JCC_B << 1),
4904                             (JCC_Z << 1),
4905                             (JCC_BE << 1),
4906                             (JCC_P << 1),
4907                         };
4908 
4909                         if (!(s->cpuid_features & CPUID_CMOV)) {
4910                             goto illegal_op;
4911                         }
4912                         op1 = fcmov_cc[op & 3] | (((op >> 3) & 1) ^ 1);
4913                         l1 = gen_new_label();
4914                         gen_jcc1_noeob(s, op1, l1);
4915                         gen_helper_fmov_ST0_STN(tcg_env,
4916                                                 tcg_constant_i32(opreg));
4917                         gen_set_label(l1);
4918                     }
4919                     break;
4920                 default:
4921                     goto unknown_op;
4922                 }
4923             }
4924 
4925             if (update_fip) {
4926                 tcg_gen_ld_i32(s->tmp2_i32, tcg_env,
4927                                offsetof(CPUX86State, segs[R_CS].selector));
4928                 tcg_gen_st16_i32(s->tmp2_i32, tcg_env,
4929                                  offsetof(CPUX86State, fpcs));
4930                 tcg_gen_st_tl(eip_cur_tl(s),
4931                               tcg_env, offsetof(CPUX86State, fpip));
4932             }
4933         }
4934         break;
4935         /************************/
4936         /* string ops */
4937 
4938     case 0xa4: /* movsS */
4939     case 0xa5:
4940         ot = mo_b_d(b, dflag);
4941         if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
4942             gen_repz_movs(s, ot);
4943         } else {
4944             gen_movs(s, ot);
4945         }
4946         break;
4947 
4948     case 0xaa: /* stosS */
4949     case 0xab:
4950         ot = mo_b_d(b, dflag);
4951         gen_op_mov_v_reg(s, MO_32, s->T0, R_EAX);
4952         if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
4953             gen_repz_stos(s, ot);
4954         } else {
4955             gen_stos(s, ot);
4956         }
4957         break;
4958     case 0xac: /* lodsS */
4959     case 0xad:
4960         ot = mo_b_d(b, dflag);
4961         if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
4962             gen_repz_lods(s, ot);
4963         } else {
4964             gen_lods(s, ot);
4965         }
4966         break;
4967     case 0xae: /* scasS */
4968     case 0xaf:
4969         ot = mo_b_d(b, dflag);
4970         gen_op_mov_v_reg(s, MO_32, s->T0, R_EAX);
4971         if (prefixes & PREFIX_REPNZ) {
4972             gen_repz_scas(s, ot, 1);
4973         } else if (prefixes & PREFIX_REPZ) {
4974             gen_repz_scas(s, ot, 0);
4975         } else {
4976             gen_scas(s, ot);
4977         }
4978         break;
4979 
4980     case 0xa6: /* cmpsS */
4981     case 0xa7:
4982         ot = mo_b_d(b, dflag);
4983         if (prefixes & PREFIX_REPNZ) {
4984             gen_repz_cmps(s, ot, 1);
4985         } else if (prefixes & PREFIX_REPZ) {
4986             gen_repz_cmps(s, ot, 0);
4987         } else {
4988             gen_cmps(s, ot);
4989         }
4990         break;
4991     case 0x6c: /* insS */
4992     case 0x6d:
4993         ot = mo_b_d32(b, dflag);
4994         tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_EDX]);
4995         tcg_gen_ext16u_i32(s->tmp2_i32, s->tmp2_i32);
4996         if (!gen_check_io(s, ot, s->tmp2_i32,
4997                           SVM_IOIO_TYPE_MASK | SVM_IOIO_STR_MASK)) {
4998             break;
4999         }
5000         translator_io_start(&s->base);
5001         if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
5002             gen_repz_ins(s, ot);
5003         } else {
5004             gen_ins(s, ot);
5005         }
5006         break;
5007     case 0x6e: /* outsS */
5008     case 0x6f:
5009         ot = mo_b_d32(b, dflag);
5010         tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_EDX]);
5011         tcg_gen_ext16u_i32(s->tmp2_i32, s->tmp2_i32);
5012         if (!gen_check_io(s, ot, s->tmp2_i32, SVM_IOIO_STR_MASK)) {
5013             break;
5014         }
5015         translator_io_start(&s->base);
5016         if (prefixes & (PREFIX_REPZ | PREFIX_REPNZ)) {
5017             gen_repz_outs(s, ot);
5018         } else {
5019             gen_outs(s, ot);
5020         }
5021         break;
5022 
5023         /************************/
5024         /* port I/O */
5025 
5026     case 0xe4:
5027     case 0xe5:
5028         ot = mo_b_d32(b, dflag);
5029         val = x86_ldub_code(env, s);
5030         tcg_gen_movi_i32(s->tmp2_i32, val);
5031         if (!gen_check_io(s, ot, s->tmp2_i32, SVM_IOIO_TYPE_MASK)) {
5032             break;
5033         }
5034         translator_io_start(&s->base);
5035         gen_helper_in_func(ot, s->T1, s->tmp2_i32);
5036         gen_op_mov_reg_v(s, ot, R_EAX, s->T1);
5037         gen_bpt_io(s, s->tmp2_i32, ot);
5038         break;
5039     case 0xe6:
5040     case 0xe7:
5041         ot = mo_b_d32(b, dflag);
5042         val = x86_ldub_code(env, s);
5043         tcg_gen_movi_i32(s->tmp2_i32, val);
5044         if (!gen_check_io(s, ot, s->tmp2_i32, 0)) {
5045             break;
5046         }
5047         translator_io_start(&s->base);
5048         gen_op_mov_v_reg(s, ot, s->T1, R_EAX);
5049         tcg_gen_trunc_tl_i32(s->tmp3_i32, s->T1);
5050         gen_helper_out_func(ot, s->tmp2_i32, s->tmp3_i32);
5051         gen_bpt_io(s, s->tmp2_i32, ot);
5052         break;
5053     case 0xec:
5054     case 0xed:
5055         ot = mo_b_d32(b, dflag);
5056         tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_EDX]);
5057         tcg_gen_ext16u_i32(s->tmp2_i32, s->tmp2_i32);
5058         if (!gen_check_io(s, ot, s->tmp2_i32, SVM_IOIO_TYPE_MASK)) {
5059             break;
5060         }
5061         translator_io_start(&s->base);
5062         gen_helper_in_func(ot, s->T1, s->tmp2_i32);
5063         gen_op_mov_reg_v(s, ot, R_EAX, s->T1);
5064         gen_bpt_io(s, s->tmp2_i32, ot);
5065         break;
5066     case 0xee:
5067     case 0xef:
5068         ot = mo_b_d32(b, dflag);
5069         tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_EDX]);
5070         tcg_gen_ext16u_i32(s->tmp2_i32, s->tmp2_i32);
5071         if (!gen_check_io(s, ot, s->tmp2_i32, 0)) {
5072             break;
5073         }
5074         translator_io_start(&s->base);
5075         gen_op_mov_v_reg(s, ot, s->T1, R_EAX);
5076         tcg_gen_trunc_tl_i32(s->tmp3_i32, s->T1);
5077         gen_helper_out_func(ot, s->tmp2_i32, s->tmp3_i32);
5078         gen_bpt_io(s, s->tmp2_i32, ot);
5079         break;
5080 
5081         /************************/
5082         /* control */
5083     case 0xc2: /* ret im */
5084         val = x86_ldsw_code(env, s);
5085         ot = gen_pop_T0(s);
5086         gen_stack_update(s, val + (1 << ot));
5087         /* Note that gen_pop_T0 uses a zero-extending load.  */
5088         gen_op_jmp_v(s, s->T0);
5089         gen_bnd_jmp(s);
5090         s->base.is_jmp = DISAS_JUMP;
5091         break;
5092     case 0xc3: /* ret */
5093         ot = gen_pop_T0(s);
5094         gen_pop_update(s, ot);
5095         /* Note that gen_pop_T0 uses a zero-extending load.  */
5096         gen_op_jmp_v(s, s->T0);
5097         gen_bnd_jmp(s);
5098         s->base.is_jmp = DISAS_JUMP;
5099         break;
5100     case 0xca: /* lret im */
5101         val = x86_ldsw_code(env, s);
5102     do_lret:
5103         if (PE(s) && !VM86(s)) {
5104             gen_update_cc_op(s);
5105             gen_update_eip_cur(s);
5106             gen_helper_lret_protected(tcg_env, tcg_constant_i32(dflag - 1),
5107                                       tcg_constant_i32(val));
5108         } else {
5109             gen_stack_A0(s);
5110             /* pop offset */
5111             gen_op_ld_v(s, dflag, s->T0, s->A0);
5112             /* NOTE: keeping EIP updated is not a problem in case of
5113                exception */
5114             gen_op_jmp_v(s, s->T0);
5115             /* pop selector */
5116             gen_add_A0_im(s, 1 << dflag);
5117             gen_op_ld_v(s, dflag, s->T0, s->A0);
5118             gen_op_movl_seg_T0_vm(s, R_CS);
5119             /* add stack offset */
5120             gen_stack_update(s, val + (2 << dflag));
5121         }
5122         s->base.is_jmp = DISAS_EOB_ONLY;
5123         break;
5124     case 0xcb: /* lret */
5125         val = 0;
5126         goto do_lret;
5127     case 0xcf: /* iret */
5128         gen_svm_check_intercept(s, SVM_EXIT_IRET);
5129         if (!PE(s) || VM86(s)) {
5130             /* real mode or vm86 mode */
5131             if (!check_vm86_iopl(s)) {
5132                 break;
5133             }
5134             gen_helper_iret_real(tcg_env, tcg_constant_i32(dflag - 1));
5135         } else {
5136             gen_helper_iret_protected(tcg_env, tcg_constant_i32(dflag - 1),
5137                                       eip_next_i32(s));
5138         }
5139         set_cc_op(s, CC_OP_EFLAGS);
5140         s->base.is_jmp = DISAS_EOB_ONLY;
5141         break;
5142     case 0xe8: /* call im */
5143         {
5144             int diff = (dflag != MO_16
5145                         ? (int32_t)insn_get(env, s, MO_32)
5146                         : (int16_t)insn_get(env, s, MO_16));
5147             gen_push_v(s, eip_next_tl(s));
5148             gen_bnd_jmp(s);
5149             gen_jmp_rel(s, dflag, diff, 0);
5150         }
5151         break;
5152     case 0x9a: /* lcall im */
5153         {
5154             unsigned int selector, offset;
5155 
5156             if (CODE64(s))
5157                 goto illegal_op;
5158             ot = dflag;
5159             offset = insn_get(env, s, ot);
5160             selector = insn_get(env, s, MO_16);
5161 
5162             tcg_gen_movi_tl(s->T0, selector);
5163             tcg_gen_movi_tl(s->T1, offset);
5164         }
5165         goto do_lcall;
5166     case 0xe9: /* jmp im */
5167         {
5168             int diff = (dflag != MO_16
5169                         ? (int32_t)insn_get(env, s, MO_32)
5170                         : (int16_t)insn_get(env, s, MO_16));
5171             gen_bnd_jmp(s);
5172             gen_jmp_rel(s, dflag, diff, 0);
5173         }
5174         break;
5175     case 0xea: /* ljmp im */
5176         {
5177             unsigned int selector, offset;
5178 
5179             if (CODE64(s))
5180                 goto illegal_op;
5181             ot = dflag;
5182             offset = insn_get(env, s, ot);
5183             selector = insn_get(env, s, MO_16);
5184 
5185             tcg_gen_movi_tl(s->T0, selector);
5186             tcg_gen_movi_tl(s->T1, offset);
5187         }
5188         goto do_ljmp;
5189     case 0xeb: /* jmp Jb */
5190         {
5191             int diff = (int8_t)insn_get(env, s, MO_8);
5192             gen_jmp_rel(s, dflag, diff, 0);
5193         }
5194         break;
5195     case 0x70 ... 0x7f: /* jcc Jb */
5196         {
5197             int diff = (int8_t)insn_get(env, s, MO_8);
5198             gen_bnd_jmp(s);
5199             gen_jcc(s, b, diff);
5200         }
5201         break;
5202     case 0x180 ... 0x18f: /* jcc Jv */
5203         {
5204             int diff = (dflag != MO_16
5205                         ? (int32_t)insn_get(env, s, MO_32)
5206                         : (int16_t)insn_get(env, s, MO_16));
5207             gen_bnd_jmp(s);
5208             gen_jcc(s, b, diff);
5209         }
5210         break;
5211 
5212     case 0x190 ... 0x19f: /* setcc Gv */
5213         modrm = x86_ldub_code(env, s);
5214         gen_setcc1(s, b, s->T0);
5215         gen_ldst_modrm(env, s, modrm, MO_8, OR_TMP0, 1);
5216         break;
5217     case 0x140 ... 0x14f: /* cmov Gv, Ev */
5218         if (!(s->cpuid_features & CPUID_CMOV)) {
5219             goto illegal_op;
5220         }
5221         ot = dflag;
5222         modrm = x86_ldub_code(env, s);
5223         reg = ((modrm >> 3) & 7) | REX_R(s);
5224         gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
5225         gen_cmovcc1(s, b ^ 1, s->T0, cpu_regs[reg]);
5226         gen_op_mov_reg_v(s, ot, reg, s->T0);
5227         break;
5228 
5229         /************************/
5230         /* flags */
5231     case 0x9c: /* pushf */
5232         gen_svm_check_intercept(s, SVM_EXIT_PUSHF);
5233         if (check_vm86_iopl(s)) {
5234             gen_update_cc_op(s);
5235             gen_helper_read_eflags(s->T0, tcg_env);
5236             gen_push_v(s, s->T0);
5237         }
5238         break;
5239     case 0x9d: /* popf */
5240         gen_svm_check_intercept(s, SVM_EXIT_POPF);
5241         if (check_vm86_iopl(s)) {
5242             int mask = TF_MASK | AC_MASK | ID_MASK | NT_MASK;
5243 
5244             if (CPL(s) == 0) {
5245                 mask |= IF_MASK | IOPL_MASK;
5246             } else if (CPL(s) <= IOPL(s)) {
5247                 mask |= IF_MASK;
5248             }
5249             if (dflag == MO_16) {
5250                 mask &= 0xffff;
5251             }
5252 
5253             ot = gen_pop_T0(s);
5254             gen_helper_write_eflags(tcg_env, s->T0, tcg_constant_i32(mask));
5255             gen_pop_update(s, ot);
5256             set_cc_op(s, CC_OP_EFLAGS);
5257             /* abort translation because TF/AC flag may change */
5258             s->base.is_jmp = DISAS_EOB_NEXT;
5259         }
5260         break;
5261     case 0x9e: /* sahf */
5262         if (CODE64(s) && !(s->cpuid_ext3_features & CPUID_EXT3_LAHF_LM))
5263             goto illegal_op;
5264         tcg_gen_shri_tl(s->T0, cpu_regs[R_EAX], 8);
5265         gen_compute_eflags(s);
5266         tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, CC_O);
5267         tcg_gen_andi_tl(s->T0, s->T0, CC_S | CC_Z | CC_A | CC_P | CC_C);
5268         tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, s->T0);
5269         break;
5270     case 0x9f: /* lahf */
5271         if (CODE64(s) && !(s->cpuid_ext3_features & CPUID_EXT3_LAHF_LM))
5272             goto illegal_op;
5273         gen_compute_eflags(s);
5274         /* Note: gen_compute_eflags() only gives the condition codes */
5275         tcg_gen_ori_tl(s->T0, cpu_cc_src, 0x02);
5276         tcg_gen_deposit_tl(cpu_regs[R_EAX], cpu_regs[R_EAX], s->T0, 8, 8);
5277         break;
5278     case 0xf5: /* cmc */
5279         gen_compute_eflags(s);
5280         tcg_gen_xori_tl(cpu_cc_src, cpu_cc_src, CC_C);
5281         break;
5282     case 0xf8: /* clc */
5283         gen_compute_eflags(s);
5284         tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, ~CC_C);
5285         break;
5286     case 0xf9: /* stc */
5287         gen_compute_eflags(s);
5288         tcg_gen_ori_tl(cpu_cc_src, cpu_cc_src, CC_C);
5289         break;
5290     case 0xfc: /* cld */
5291         tcg_gen_movi_i32(s->tmp2_i32, 1);
5292         tcg_gen_st_i32(s->tmp2_i32, tcg_env, offsetof(CPUX86State, df));
5293         break;
5294     case 0xfd: /* std */
5295         tcg_gen_movi_i32(s->tmp2_i32, -1);
5296         tcg_gen_st_i32(s->tmp2_i32, tcg_env, offsetof(CPUX86State, df));
5297         break;
5298 
5299         /************************/
5300         /* bit operations */
5301     case 0x1ba: /* bt/bts/btr/btc Gv, im */
5302         ot = dflag;
5303         modrm = x86_ldub_code(env, s);
5304         op = (modrm >> 3) & 7;
5305         mod = (modrm >> 6) & 3;
5306         rm = (modrm & 7) | REX_B(s);
5307         if (mod != 3) {
5308             s->rip_offset = 1;
5309             gen_lea_modrm(env, s, modrm);
5310             if (!(s->prefix & PREFIX_LOCK)) {
5311                 gen_op_ld_v(s, ot, s->T0, s->A0);
5312             }
5313         } else {
5314             gen_op_mov_v_reg(s, ot, s->T0, rm);
5315         }
5316         /* load shift */
5317         val = x86_ldub_code(env, s);
5318         tcg_gen_movi_tl(s->T1, val);
5319         if (op < 4)
5320             goto unknown_op;
5321         op -= 4;
5322         goto bt_op;
5323     case 0x1a3: /* bt Gv, Ev */
5324         op = 0;
5325         goto do_btx;
5326     case 0x1ab: /* bts */
5327         op = 1;
5328         goto do_btx;
5329     case 0x1b3: /* btr */
5330         op = 2;
5331         goto do_btx;
5332     case 0x1bb: /* btc */
5333         op = 3;
5334     do_btx:
5335         ot = dflag;
5336         modrm = x86_ldub_code(env, s);
5337         reg = ((modrm >> 3) & 7) | REX_R(s);
5338         mod = (modrm >> 6) & 3;
5339         rm = (modrm & 7) | REX_B(s);
5340         gen_op_mov_v_reg(s, MO_32, s->T1, reg);
5341         if (mod != 3) {
5342             AddressParts a = gen_lea_modrm_0(env, s, modrm);
5343             /* specific case: we need to add a displacement */
5344             gen_exts(ot, s->T1);
5345             tcg_gen_sari_tl(s->tmp0, s->T1, 3 + ot);
5346             tcg_gen_shli_tl(s->tmp0, s->tmp0, ot);
5347             tcg_gen_add_tl(s->A0, gen_lea_modrm_1(s, a, false), s->tmp0);
5348             gen_lea_v_seg(s, s->aflag, s->A0, a.def_seg, s->override);
5349             if (!(s->prefix & PREFIX_LOCK)) {
5350                 gen_op_ld_v(s, ot, s->T0, s->A0);
5351             }
5352         } else {
5353             gen_op_mov_v_reg(s, ot, s->T0, rm);
5354         }
5355     bt_op:
5356         tcg_gen_andi_tl(s->T1, s->T1, (1 << (3 + ot)) - 1);
5357         tcg_gen_movi_tl(s->tmp0, 1);
5358         tcg_gen_shl_tl(s->tmp0, s->tmp0, s->T1);
5359         if (s->prefix & PREFIX_LOCK) {
5360             switch (op) {
5361             case 0: /* bt */
5362                 /* Needs no atomic ops; we suppressed the normal
5363                    memory load for LOCK above so do it now.  */
5364                 gen_op_ld_v(s, ot, s->T0, s->A0);
5365                 break;
5366             case 1: /* bts */
5367                 tcg_gen_atomic_fetch_or_tl(s->T0, s->A0, s->tmp0,
5368                                            s->mem_index, ot | MO_LE);
5369                 break;
5370             case 2: /* btr */
5371                 tcg_gen_not_tl(s->tmp0, s->tmp0);
5372                 tcg_gen_atomic_fetch_and_tl(s->T0, s->A0, s->tmp0,
5373                                             s->mem_index, ot | MO_LE);
5374                 break;
5375             default:
5376             case 3: /* btc */
5377                 tcg_gen_atomic_fetch_xor_tl(s->T0, s->A0, s->tmp0,
5378                                             s->mem_index, ot | MO_LE);
5379                 break;
5380             }
5381             tcg_gen_shr_tl(s->tmp4, s->T0, s->T1);
5382         } else {
5383             tcg_gen_shr_tl(s->tmp4, s->T0, s->T1);
5384             switch (op) {
5385             case 0: /* bt */
5386                 /* Data already loaded; nothing to do.  */
5387                 break;
5388             case 1: /* bts */
5389                 tcg_gen_or_tl(s->T0, s->T0, s->tmp0);
5390                 break;
5391             case 2: /* btr */
5392                 tcg_gen_andc_tl(s->T0, s->T0, s->tmp0);
5393                 break;
5394             default:
5395             case 3: /* btc */
5396                 tcg_gen_xor_tl(s->T0, s->T0, s->tmp0);
5397                 break;
5398             }
5399             if (op != 0) {
5400                 if (mod != 3) {
5401                     gen_op_st_v(s, ot, s->T0, s->A0);
5402                 } else {
5403                     gen_op_mov_reg_v(s, ot, rm, s->T0);
5404                 }
5405             }
5406         }
5407 
5408         /* Delay all CC updates until after the store above.  Note that
5409            C is the result of the test, Z is unchanged, and the others
5410            are all undefined.  */
5411         switch (s->cc_op) {
5412         case CC_OP_MULB ... CC_OP_MULQ:
5413         case CC_OP_ADDB ... CC_OP_ADDQ:
5414         case CC_OP_ADCB ... CC_OP_ADCQ:
5415         case CC_OP_SUBB ... CC_OP_SUBQ:
5416         case CC_OP_SBBB ... CC_OP_SBBQ:
5417         case CC_OP_LOGICB ... CC_OP_LOGICQ:
5418         case CC_OP_INCB ... CC_OP_INCQ:
5419         case CC_OP_DECB ... CC_OP_DECQ:
5420         case CC_OP_SHLB ... CC_OP_SHLQ:
5421         case CC_OP_SARB ... CC_OP_SARQ:
5422         case CC_OP_BMILGB ... CC_OP_BMILGQ:
5423             /* Z was going to be computed from the non-zero status of CC_DST.
5424                We can get that same Z value (and the new C value) by leaving
5425                CC_DST alone, setting CC_SRC, and using a CC_OP_SAR of the
5426                same width.  */
5427             tcg_gen_mov_tl(cpu_cc_src, s->tmp4);
5428             set_cc_op(s, ((s->cc_op - CC_OP_MULB) & 3) + CC_OP_SARB);
5429             break;
5430         default:
5431             /* Otherwise, generate EFLAGS and replace the C bit.  */
5432             gen_compute_eflags(s);
5433             tcg_gen_deposit_tl(cpu_cc_src, cpu_cc_src, s->tmp4,
5434                                ctz32(CC_C), 1);
5435             break;
5436         }
5437         break;
5438     case 0x1bc: /* bsf / tzcnt */
5439     case 0x1bd: /* bsr / lzcnt */
5440         ot = dflag;
5441         modrm = x86_ldub_code(env, s);
5442         reg = ((modrm >> 3) & 7) | REX_R(s);
5443         gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
5444         gen_extu(ot, s->T0);
5445 
5446         /* Note that lzcnt and tzcnt are in different extensions.  */
5447         if ((prefixes & PREFIX_REPZ)
5448             && (b & 1
5449                 ? s->cpuid_ext3_features & CPUID_EXT3_ABM
5450                 : s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_BMI1)) {
5451             int size = 8 << ot;
5452             /* For lzcnt/tzcnt, C bit is defined related to the input. */
5453             tcg_gen_mov_tl(cpu_cc_src, s->T0);
5454             if (b & 1) {
5455                 /* For lzcnt, reduce the target_ulong result by the
5456                    number of zeros that we expect to find at the top.  */
5457                 tcg_gen_clzi_tl(s->T0, s->T0, TARGET_LONG_BITS);
5458                 tcg_gen_subi_tl(s->T0, s->T0, TARGET_LONG_BITS - size);
5459             } else {
5460                 /* For tzcnt, a zero input must return the operand size.  */
5461                 tcg_gen_ctzi_tl(s->T0, s->T0, size);
5462             }
5463             /* For lzcnt/tzcnt, Z bit is defined related to the result.  */
5464             gen_op_update1_cc(s);
5465             set_cc_op(s, CC_OP_BMILGB + ot);
5466         } else {
5467             /* For bsr/bsf, only the Z bit is defined and it is related
5468                to the input and not the result.  */
5469             tcg_gen_mov_tl(cpu_cc_dst, s->T0);
5470             set_cc_op(s, CC_OP_LOGICB + ot);
5471 
5472             /* ??? The manual says that the output is undefined when the
5473                input is zero, but real hardware leaves it unchanged, and
5474                real programs appear to depend on that.  Accomplish this
5475                by passing the output as the value to return upon zero.  */
5476             if (b & 1) {
5477                 /* For bsr, return the bit index of the first 1 bit,
5478                    not the count of leading zeros.  */
5479                 tcg_gen_xori_tl(s->T1, cpu_regs[reg], TARGET_LONG_BITS - 1);
5480                 tcg_gen_clz_tl(s->T0, s->T0, s->T1);
5481                 tcg_gen_xori_tl(s->T0, s->T0, TARGET_LONG_BITS - 1);
5482             } else {
5483                 tcg_gen_ctz_tl(s->T0, s->T0, cpu_regs[reg]);
5484             }
5485         }
5486         gen_op_mov_reg_v(s, ot, reg, s->T0);
5487         break;
5488         /************************/
5489         /* bcd */
5490     case 0x27: /* daa */
5491         if (CODE64(s))
5492             goto illegal_op;
5493         gen_update_cc_op(s);
5494         gen_helper_daa(tcg_env);
5495         set_cc_op(s, CC_OP_EFLAGS);
5496         break;
5497     case 0x2f: /* das */
5498         if (CODE64(s))
5499             goto illegal_op;
5500         gen_update_cc_op(s);
5501         gen_helper_das(tcg_env);
5502         set_cc_op(s, CC_OP_EFLAGS);
5503         break;
5504     case 0x37: /* aaa */
5505         if (CODE64(s))
5506             goto illegal_op;
5507         gen_update_cc_op(s);
5508         gen_helper_aaa(tcg_env);
5509         set_cc_op(s, CC_OP_EFLAGS);
5510         break;
5511     case 0x3f: /* aas */
5512         if (CODE64(s))
5513             goto illegal_op;
5514         gen_update_cc_op(s);
5515         gen_helper_aas(tcg_env);
5516         set_cc_op(s, CC_OP_EFLAGS);
5517         break;
5518     case 0xd4: /* aam */
5519         if (CODE64(s))
5520             goto illegal_op;
5521         val = x86_ldub_code(env, s);
5522         if (val == 0) {
5523             gen_exception(s, EXCP00_DIVZ);
5524         } else {
5525             gen_helper_aam(tcg_env, tcg_constant_i32(val));
5526             set_cc_op(s, CC_OP_LOGICB);
5527         }
5528         break;
5529     case 0xd5: /* aad */
5530         if (CODE64(s))
5531             goto illegal_op;
5532         val = x86_ldub_code(env, s);
5533         gen_helper_aad(tcg_env, tcg_constant_i32(val));
5534         set_cc_op(s, CC_OP_LOGICB);
5535         break;
5536         /************************/
5537         /* misc */
5538     case 0x90: /* nop */
5539         /* XXX: correct lock test for all insn */
5540         if (prefixes & PREFIX_LOCK) {
5541             goto illegal_op;
5542         }
5543         /* If REX_B is set, then this is xchg eax, r8d, not a nop.  */
5544         if (REX_B(s)) {
5545             goto do_xchg_reg_eax;
5546         }
5547         if (prefixes & PREFIX_REPZ) {
5548             gen_update_cc_op(s);
5549             gen_update_eip_cur(s);
5550             gen_helper_pause(tcg_env, cur_insn_len_i32(s));
5551             s->base.is_jmp = DISAS_NORETURN;
5552         }
5553         break;
5554     case 0x9b: /* fwait */
5555         if ((s->flags & (HF_MP_MASK | HF_TS_MASK)) ==
5556             (HF_MP_MASK | HF_TS_MASK)) {
5557             gen_exception(s, EXCP07_PREX);
5558         } else {
5559             /* needs to be treated as I/O because of ferr_irq */
5560             translator_io_start(&s->base);
5561             gen_helper_fwait(tcg_env);
5562         }
5563         break;
5564     case 0xcc: /* int3 */
5565         gen_interrupt(s, EXCP03_INT3);
5566         break;
5567     case 0xcd: /* int N */
5568         val = x86_ldub_code(env, s);
5569         if (check_vm86_iopl(s)) {
5570             gen_interrupt(s, val);
5571         }
5572         break;
5573     case 0xce: /* into */
5574         if (CODE64(s))
5575             goto illegal_op;
5576         gen_update_cc_op(s);
5577         gen_update_eip_cur(s);
5578         gen_helper_into(tcg_env, cur_insn_len_i32(s));
5579         break;
5580 #ifdef WANT_ICEBP
5581     case 0xf1: /* icebp (undocumented, exits to external debugger) */
5582         gen_svm_check_intercept(s, SVM_EXIT_ICEBP);
5583         gen_debug(s);
5584         break;
5585 #endif
5586     case 0xfa: /* cli */
5587         if (check_iopl(s)) {
5588             gen_reset_eflags(s, IF_MASK);
5589         }
5590         break;
5591     case 0xfb: /* sti */
5592         if (check_iopl(s)) {
5593             gen_set_eflags(s, IF_MASK);
5594             /* interruptions are enabled only the first insn after sti */
5595             gen_update_eip_next(s);
5596             gen_eob_inhibit_irq(s, true);
5597         }
5598         break;
5599     case 0x62: /* bound */
5600         if (CODE64(s))
5601             goto illegal_op;
5602         ot = dflag;
5603         modrm = x86_ldub_code(env, s);
5604         reg = (modrm >> 3) & 7;
5605         mod = (modrm >> 6) & 3;
5606         if (mod == 3)
5607             goto illegal_op;
5608         gen_op_mov_v_reg(s, ot, s->T0, reg);
5609         gen_lea_modrm(env, s, modrm);
5610         tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
5611         if (ot == MO_16) {
5612             gen_helper_boundw(tcg_env, s->A0, s->tmp2_i32);
5613         } else {
5614             gen_helper_boundl(tcg_env, s->A0, s->tmp2_i32);
5615         }
5616         break;
5617     case 0x1c8 ... 0x1cf: /* bswap reg */
5618         reg = (b & 7) | REX_B(s);
5619 #ifdef TARGET_X86_64
5620         if (dflag == MO_64) {
5621             tcg_gen_bswap64_i64(cpu_regs[reg], cpu_regs[reg]);
5622             break;
5623         }
5624 #endif
5625         tcg_gen_bswap32_tl(cpu_regs[reg], cpu_regs[reg], TCG_BSWAP_OZ);
5626         break;
5627     case 0xd6: /* salc */
5628         if (CODE64(s))
5629             goto illegal_op;
5630         gen_compute_eflags_c(s, s->T0);
5631         tcg_gen_neg_tl(s->T0, s->T0);
5632         gen_op_mov_reg_v(s, MO_8, R_EAX, s->T0);
5633         break;
5634     case 0xe0: /* loopnz */
5635     case 0xe1: /* loopz */
5636     case 0xe2: /* loop */
5637     case 0xe3: /* jecxz */
5638         {
5639             TCGLabel *l1, *l2;
5640             int diff = (int8_t)insn_get(env, s, MO_8);
5641 
5642             l1 = gen_new_label();
5643             l2 = gen_new_label();
5644             gen_update_cc_op(s);
5645             b &= 3;
5646             switch(b) {
5647             case 0: /* loopnz */
5648             case 1: /* loopz */
5649                 gen_op_add_reg_im(s, s->aflag, R_ECX, -1);
5650                 gen_op_jz_ecx(s, l2);
5651                 gen_jcc1(s, (JCC_Z << 1) | (b ^ 1), l1);
5652                 break;
5653             case 2: /* loop */
5654                 gen_op_add_reg_im(s, s->aflag, R_ECX, -1);
5655                 gen_op_jnz_ecx(s, l1);
5656                 break;
5657             default:
5658             case 3: /* jcxz */
5659                 gen_op_jz_ecx(s, l1);
5660                 break;
5661             }
5662 
5663             gen_set_label(l2);
5664             gen_jmp_rel_csize(s, 0, 1);
5665 
5666             gen_set_label(l1);
5667             gen_jmp_rel(s, dflag, diff, 0);
5668         }
5669         break;
5670     case 0x130: /* wrmsr */
5671     case 0x132: /* rdmsr */
5672         if (check_cpl0(s)) {
5673             gen_update_cc_op(s);
5674             gen_update_eip_cur(s);
5675             if (b & 2) {
5676                 gen_helper_rdmsr(tcg_env);
5677             } else {
5678                 gen_helper_wrmsr(tcg_env);
5679                 s->base.is_jmp = DISAS_EOB_NEXT;
5680             }
5681         }
5682         break;
5683     case 0x131: /* rdtsc */
5684         gen_update_cc_op(s);
5685         gen_update_eip_cur(s);
5686         translator_io_start(&s->base);
5687         gen_helper_rdtsc(tcg_env);
5688         break;
5689     case 0x133: /* rdpmc */
5690         gen_update_cc_op(s);
5691         gen_update_eip_cur(s);
5692         gen_helper_rdpmc(tcg_env);
5693         s->base.is_jmp = DISAS_NORETURN;
5694         break;
5695     case 0x134: /* sysenter */
5696         /* For AMD SYSENTER is not valid in long mode */
5697         if (LMA(s) && env->cpuid_vendor1 != CPUID_VENDOR_INTEL_1) {
5698             goto illegal_op;
5699         }
5700         if (!PE(s)) {
5701             gen_exception_gpf(s);
5702         } else {
5703             gen_helper_sysenter(tcg_env);
5704             s->base.is_jmp = DISAS_EOB_ONLY;
5705         }
5706         break;
5707     case 0x135: /* sysexit */
5708         /* For AMD SYSEXIT is not valid in long mode */
5709         if (LMA(s) && env->cpuid_vendor1 != CPUID_VENDOR_INTEL_1) {
5710             goto illegal_op;
5711         }
5712         if (!PE(s) || CPL(s) != 0) {
5713             gen_exception_gpf(s);
5714         } else {
5715             gen_helper_sysexit(tcg_env, tcg_constant_i32(dflag - 1));
5716             s->base.is_jmp = DISAS_EOB_ONLY;
5717         }
5718         break;
5719     case 0x105: /* syscall */
5720         /* For Intel SYSCALL is only valid in long mode */
5721         if (!LMA(s) && env->cpuid_vendor1 == CPUID_VENDOR_INTEL_1) {
5722             goto illegal_op;
5723         }
5724         gen_update_cc_op(s);
5725         gen_update_eip_cur(s);
5726         gen_helper_syscall(tcg_env, cur_insn_len_i32(s));
5727         /* TF handling for the syscall insn is different. The TF bit is  checked
5728            after the syscall insn completes. This allows #DB to not be
5729            generated after one has entered CPL0 if TF is set in FMASK.  */
5730         gen_eob_worker(s, false, true);
5731         break;
5732     case 0x107: /* sysret */
5733         /* For Intel SYSRET is only valid in long mode */
5734         if (!LMA(s) && env->cpuid_vendor1 == CPUID_VENDOR_INTEL_1) {
5735             goto illegal_op;
5736         }
5737         if (!PE(s) || CPL(s) != 0) {
5738             gen_exception_gpf(s);
5739         } else {
5740             gen_helper_sysret(tcg_env, tcg_constant_i32(dflag - 1));
5741             /* condition codes are modified only in long mode */
5742             if (LMA(s)) {
5743                 set_cc_op(s, CC_OP_EFLAGS);
5744             }
5745             /* TF handling for the sysret insn is different. The TF bit is
5746                checked after the sysret insn completes. This allows #DB to be
5747                generated "as if" the syscall insn in userspace has just
5748                completed.  */
5749             gen_eob_worker(s, false, true);
5750         }
5751         break;
5752     case 0x1a2: /* cpuid */
5753         gen_update_cc_op(s);
5754         gen_update_eip_cur(s);
5755         gen_helper_cpuid(tcg_env);
5756         break;
5757     case 0xf4: /* hlt */
5758         if (check_cpl0(s)) {
5759             gen_update_cc_op(s);
5760             gen_update_eip_cur(s);
5761             gen_helper_hlt(tcg_env, cur_insn_len_i32(s));
5762             s->base.is_jmp = DISAS_NORETURN;
5763         }
5764         break;
5765     case 0x100:
5766         modrm = x86_ldub_code(env, s);
5767         mod = (modrm >> 6) & 3;
5768         op = (modrm >> 3) & 7;
5769         switch(op) {
5770         case 0: /* sldt */
5771             if (!PE(s) || VM86(s))
5772                 goto illegal_op;
5773             if (s->flags & HF_UMIP_MASK && !check_cpl0(s)) {
5774                 break;
5775             }
5776             gen_svm_check_intercept(s, SVM_EXIT_LDTR_READ);
5777             tcg_gen_ld32u_tl(s->T0, tcg_env,
5778                              offsetof(CPUX86State, ldt.selector));
5779             ot = mod == 3 ? dflag : MO_16;
5780             gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
5781             break;
5782         case 2: /* lldt */
5783             if (!PE(s) || VM86(s))
5784                 goto illegal_op;
5785             if (check_cpl0(s)) {
5786                 gen_svm_check_intercept(s, SVM_EXIT_LDTR_WRITE);
5787                 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
5788                 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
5789                 gen_helper_lldt(tcg_env, s->tmp2_i32);
5790             }
5791             break;
5792         case 1: /* str */
5793             if (!PE(s) || VM86(s))
5794                 goto illegal_op;
5795             if (s->flags & HF_UMIP_MASK && !check_cpl0(s)) {
5796                 break;
5797             }
5798             gen_svm_check_intercept(s, SVM_EXIT_TR_READ);
5799             tcg_gen_ld32u_tl(s->T0, tcg_env,
5800                              offsetof(CPUX86State, tr.selector));
5801             ot = mod == 3 ? dflag : MO_16;
5802             gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
5803             break;
5804         case 3: /* ltr */
5805             if (!PE(s) || VM86(s))
5806                 goto illegal_op;
5807             if (check_cpl0(s)) {
5808                 gen_svm_check_intercept(s, SVM_EXIT_TR_WRITE);
5809                 gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
5810                 tcg_gen_trunc_tl_i32(s->tmp2_i32, s->T0);
5811                 gen_helper_ltr(tcg_env, s->tmp2_i32);
5812             }
5813             break;
5814         case 4: /* verr */
5815         case 5: /* verw */
5816             if (!PE(s) || VM86(s))
5817                 goto illegal_op;
5818             gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
5819             gen_update_cc_op(s);
5820             if (op == 4) {
5821                 gen_helper_verr(tcg_env, s->T0);
5822             } else {
5823                 gen_helper_verw(tcg_env, s->T0);
5824             }
5825             set_cc_op(s, CC_OP_EFLAGS);
5826             break;
5827         default:
5828             goto unknown_op;
5829         }
5830         break;
5831 
5832     case 0x101:
5833         modrm = x86_ldub_code(env, s);
5834         switch (modrm) {
5835         CASE_MODRM_MEM_OP(0): /* sgdt */
5836             if (s->flags & HF_UMIP_MASK && !check_cpl0(s)) {
5837                 break;
5838             }
5839             gen_svm_check_intercept(s, SVM_EXIT_GDTR_READ);
5840             gen_lea_modrm(env, s, modrm);
5841             tcg_gen_ld32u_tl(s->T0,
5842                              tcg_env, offsetof(CPUX86State, gdt.limit));
5843             gen_op_st_v(s, MO_16, s->T0, s->A0);
5844             gen_add_A0_im(s, 2);
5845             tcg_gen_ld_tl(s->T0, tcg_env, offsetof(CPUX86State, gdt.base));
5846             if (dflag == MO_16) {
5847                 tcg_gen_andi_tl(s->T0, s->T0, 0xffffff);
5848             }
5849             gen_op_st_v(s, CODE64(s) + MO_32, s->T0, s->A0);
5850             break;
5851 
5852         case 0xc8: /* monitor */
5853             if (!(s->cpuid_ext_features & CPUID_EXT_MONITOR) || CPL(s) != 0) {
5854                 goto illegal_op;
5855             }
5856             gen_update_cc_op(s);
5857             gen_update_eip_cur(s);
5858             tcg_gen_mov_tl(s->A0, cpu_regs[R_EAX]);
5859             gen_add_A0_ds_seg(s);
5860             gen_helper_monitor(tcg_env, s->A0);
5861             break;
5862 
5863         case 0xc9: /* mwait */
5864             if (!(s->cpuid_ext_features & CPUID_EXT_MONITOR) || CPL(s) != 0) {
5865                 goto illegal_op;
5866             }
5867             gen_update_cc_op(s);
5868             gen_update_eip_cur(s);
5869             gen_helper_mwait(tcg_env, cur_insn_len_i32(s));
5870             s->base.is_jmp = DISAS_NORETURN;
5871             break;
5872 
5873         case 0xca: /* clac */
5874             if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_SMAP)
5875                 || CPL(s) != 0) {
5876                 goto illegal_op;
5877             }
5878             gen_reset_eflags(s, AC_MASK);
5879             s->base.is_jmp = DISAS_EOB_NEXT;
5880             break;
5881 
5882         case 0xcb: /* stac */
5883             if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_SMAP)
5884                 || CPL(s) != 0) {
5885                 goto illegal_op;
5886             }
5887             gen_set_eflags(s, AC_MASK);
5888             s->base.is_jmp = DISAS_EOB_NEXT;
5889             break;
5890 
5891         CASE_MODRM_MEM_OP(1): /* sidt */
5892             if (s->flags & HF_UMIP_MASK && !check_cpl0(s)) {
5893                 break;
5894             }
5895             gen_svm_check_intercept(s, SVM_EXIT_IDTR_READ);
5896             gen_lea_modrm(env, s, modrm);
5897             tcg_gen_ld32u_tl(s->T0, tcg_env, offsetof(CPUX86State, idt.limit));
5898             gen_op_st_v(s, MO_16, s->T0, s->A0);
5899             gen_add_A0_im(s, 2);
5900             tcg_gen_ld_tl(s->T0, tcg_env, offsetof(CPUX86State, idt.base));
5901             if (dflag == MO_16) {
5902                 tcg_gen_andi_tl(s->T0, s->T0, 0xffffff);
5903             }
5904             gen_op_st_v(s, CODE64(s) + MO_32, s->T0, s->A0);
5905             break;
5906 
5907         case 0xd0: /* xgetbv */
5908             if ((s->cpuid_ext_features & CPUID_EXT_XSAVE) == 0
5909                 || (s->prefix & (PREFIX_LOCK | PREFIX_DATA
5910                                  | PREFIX_REPZ | PREFIX_REPNZ))) {
5911                 goto illegal_op;
5912             }
5913             tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_ECX]);
5914             gen_helper_xgetbv(s->tmp1_i64, tcg_env, s->tmp2_i32);
5915             tcg_gen_extr_i64_tl(cpu_regs[R_EAX], cpu_regs[R_EDX], s->tmp1_i64);
5916             break;
5917 
5918         case 0xd1: /* xsetbv */
5919             if ((s->cpuid_ext_features & CPUID_EXT_XSAVE) == 0
5920                 || (s->prefix & (PREFIX_LOCK | PREFIX_DATA
5921                                  | PREFIX_REPZ | PREFIX_REPNZ))) {
5922                 goto illegal_op;
5923             }
5924             gen_svm_check_intercept(s, SVM_EXIT_XSETBV);
5925             if (!check_cpl0(s)) {
5926                 break;
5927             }
5928             tcg_gen_concat_tl_i64(s->tmp1_i64, cpu_regs[R_EAX],
5929                                   cpu_regs[R_EDX]);
5930             tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_ECX]);
5931             gen_helper_xsetbv(tcg_env, s->tmp2_i32, s->tmp1_i64);
5932             /* End TB because translation flags may change.  */
5933             s->base.is_jmp = DISAS_EOB_NEXT;
5934             break;
5935 
5936         case 0xd8: /* VMRUN */
5937             if (!SVME(s) || !PE(s)) {
5938                 goto illegal_op;
5939             }
5940             if (!check_cpl0(s)) {
5941                 break;
5942             }
5943             gen_update_cc_op(s);
5944             gen_update_eip_cur(s);
5945             gen_helper_vmrun(tcg_env, tcg_constant_i32(s->aflag - 1),
5946                              cur_insn_len_i32(s));
5947             tcg_gen_exit_tb(NULL, 0);
5948             s->base.is_jmp = DISAS_NORETURN;
5949             break;
5950 
5951         case 0xd9: /* VMMCALL */
5952             if (!SVME(s)) {
5953                 goto illegal_op;
5954             }
5955             gen_update_cc_op(s);
5956             gen_update_eip_cur(s);
5957             gen_helper_vmmcall(tcg_env);
5958             break;
5959 
5960         case 0xda: /* VMLOAD */
5961             if (!SVME(s) || !PE(s)) {
5962                 goto illegal_op;
5963             }
5964             if (!check_cpl0(s)) {
5965                 break;
5966             }
5967             gen_update_cc_op(s);
5968             gen_update_eip_cur(s);
5969             gen_helper_vmload(tcg_env, tcg_constant_i32(s->aflag - 1));
5970             break;
5971 
5972         case 0xdb: /* VMSAVE */
5973             if (!SVME(s) || !PE(s)) {
5974                 goto illegal_op;
5975             }
5976             if (!check_cpl0(s)) {
5977                 break;
5978             }
5979             gen_update_cc_op(s);
5980             gen_update_eip_cur(s);
5981             gen_helper_vmsave(tcg_env, tcg_constant_i32(s->aflag - 1));
5982             break;
5983 
5984         case 0xdc: /* STGI */
5985             if ((!SVME(s) && !(s->cpuid_ext3_features & CPUID_EXT3_SKINIT))
5986                 || !PE(s)) {
5987                 goto illegal_op;
5988             }
5989             if (!check_cpl0(s)) {
5990                 break;
5991             }
5992             gen_update_cc_op(s);
5993             gen_helper_stgi(tcg_env);
5994             s->base.is_jmp = DISAS_EOB_NEXT;
5995             break;
5996 
5997         case 0xdd: /* CLGI */
5998             if (!SVME(s) || !PE(s)) {
5999                 goto illegal_op;
6000             }
6001             if (!check_cpl0(s)) {
6002                 break;
6003             }
6004             gen_update_cc_op(s);
6005             gen_update_eip_cur(s);
6006             gen_helper_clgi(tcg_env);
6007             break;
6008 
6009         case 0xde: /* SKINIT */
6010             if ((!SVME(s) && !(s->cpuid_ext3_features & CPUID_EXT3_SKINIT))
6011                 || !PE(s)) {
6012                 goto illegal_op;
6013             }
6014             gen_svm_check_intercept(s, SVM_EXIT_SKINIT);
6015             /* If not intercepted, not implemented -- raise #UD. */
6016             goto illegal_op;
6017 
6018         case 0xdf: /* INVLPGA */
6019             if (!SVME(s) || !PE(s)) {
6020                 goto illegal_op;
6021             }
6022             if (!check_cpl0(s)) {
6023                 break;
6024             }
6025             gen_svm_check_intercept(s, SVM_EXIT_INVLPGA);
6026             if (s->aflag == MO_64) {
6027                 tcg_gen_mov_tl(s->A0, cpu_regs[R_EAX]);
6028             } else {
6029                 tcg_gen_ext32u_tl(s->A0, cpu_regs[R_EAX]);
6030             }
6031             gen_helper_flush_page(tcg_env, s->A0);
6032             s->base.is_jmp = DISAS_EOB_NEXT;
6033             break;
6034 
6035         CASE_MODRM_MEM_OP(2): /* lgdt */
6036             if (!check_cpl0(s)) {
6037                 break;
6038             }
6039             gen_svm_check_intercept(s, SVM_EXIT_GDTR_WRITE);
6040             gen_lea_modrm(env, s, modrm);
6041             gen_op_ld_v(s, MO_16, s->T1, s->A0);
6042             gen_add_A0_im(s, 2);
6043             gen_op_ld_v(s, CODE64(s) + MO_32, s->T0, s->A0);
6044             if (dflag == MO_16) {
6045                 tcg_gen_andi_tl(s->T0, s->T0, 0xffffff);
6046             }
6047             tcg_gen_st_tl(s->T0, tcg_env, offsetof(CPUX86State, gdt.base));
6048             tcg_gen_st32_tl(s->T1, tcg_env, offsetof(CPUX86State, gdt.limit));
6049             break;
6050 
6051         CASE_MODRM_MEM_OP(3): /* lidt */
6052             if (!check_cpl0(s)) {
6053                 break;
6054             }
6055             gen_svm_check_intercept(s, SVM_EXIT_IDTR_WRITE);
6056             gen_lea_modrm(env, s, modrm);
6057             gen_op_ld_v(s, MO_16, s->T1, s->A0);
6058             gen_add_A0_im(s, 2);
6059             gen_op_ld_v(s, CODE64(s) + MO_32, s->T0, s->A0);
6060             if (dflag == MO_16) {
6061                 tcg_gen_andi_tl(s->T0, s->T0, 0xffffff);
6062             }
6063             tcg_gen_st_tl(s->T0, tcg_env, offsetof(CPUX86State, idt.base));
6064             tcg_gen_st32_tl(s->T1, tcg_env, offsetof(CPUX86State, idt.limit));
6065             break;
6066 
6067         CASE_MODRM_OP(4): /* smsw */
6068             if (s->flags & HF_UMIP_MASK && !check_cpl0(s)) {
6069                 break;
6070             }
6071             gen_svm_check_intercept(s, SVM_EXIT_READ_CR0);
6072             tcg_gen_ld_tl(s->T0, tcg_env, offsetof(CPUX86State, cr[0]));
6073             /*
6074              * In 32-bit mode, the higher 16 bits of the destination
6075              * register are undefined.  In practice CR0[31:0] is stored
6076              * just like in 64-bit mode.
6077              */
6078             mod = (modrm >> 6) & 3;
6079             ot = (mod != 3 ? MO_16 : s->dflag);
6080             gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 1);
6081             break;
6082         case 0xee: /* rdpkru */
6083             if (prefixes & PREFIX_LOCK) {
6084                 goto illegal_op;
6085             }
6086             tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_ECX]);
6087             gen_helper_rdpkru(s->tmp1_i64, tcg_env, s->tmp2_i32);
6088             tcg_gen_extr_i64_tl(cpu_regs[R_EAX], cpu_regs[R_EDX], s->tmp1_i64);
6089             break;
6090         case 0xef: /* wrpkru */
6091             if (prefixes & PREFIX_LOCK) {
6092                 goto illegal_op;
6093             }
6094             tcg_gen_concat_tl_i64(s->tmp1_i64, cpu_regs[R_EAX],
6095                                   cpu_regs[R_EDX]);
6096             tcg_gen_trunc_tl_i32(s->tmp2_i32, cpu_regs[R_ECX]);
6097             gen_helper_wrpkru(tcg_env, s->tmp2_i32, s->tmp1_i64);
6098             break;
6099 
6100         CASE_MODRM_OP(6): /* lmsw */
6101             if (!check_cpl0(s)) {
6102                 break;
6103             }
6104             gen_svm_check_intercept(s, SVM_EXIT_WRITE_CR0);
6105             gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
6106             /*
6107              * Only the 4 lower bits of CR0 are modified.
6108              * PE cannot be set to zero if already set to one.
6109              */
6110             tcg_gen_ld_tl(s->T1, tcg_env, offsetof(CPUX86State, cr[0]));
6111             tcg_gen_andi_tl(s->T0, s->T0, 0xf);
6112             tcg_gen_andi_tl(s->T1, s->T1, ~0xe);
6113             tcg_gen_or_tl(s->T0, s->T0, s->T1);
6114             gen_helper_write_crN(tcg_env, tcg_constant_i32(0), s->T0);
6115             s->base.is_jmp = DISAS_EOB_NEXT;
6116             break;
6117 
6118         CASE_MODRM_MEM_OP(7): /* invlpg */
6119             if (!check_cpl0(s)) {
6120                 break;
6121             }
6122             gen_svm_check_intercept(s, SVM_EXIT_INVLPG);
6123             gen_lea_modrm(env, s, modrm);
6124             gen_helper_flush_page(tcg_env, s->A0);
6125             s->base.is_jmp = DISAS_EOB_NEXT;
6126             break;
6127 
6128         case 0xf8: /* swapgs */
6129 #ifdef TARGET_X86_64
6130             if (CODE64(s)) {
6131                 if (check_cpl0(s)) {
6132                     tcg_gen_mov_tl(s->T0, cpu_seg_base[R_GS]);
6133                     tcg_gen_ld_tl(cpu_seg_base[R_GS], tcg_env,
6134                                   offsetof(CPUX86State, kernelgsbase));
6135                     tcg_gen_st_tl(s->T0, tcg_env,
6136                                   offsetof(CPUX86State, kernelgsbase));
6137                 }
6138                 break;
6139             }
6140 #endif
6141             goto illegal_op;
6142 
6143         case 0xf9: /* rdtscp */
6144             if (!(s->cpuid_ext2_features & CPUID_EXT2_RDTSCP)) {
6145                 goto illegal_op;
6146             }
6147             gen_update_cc_op(s);
6148             gen_update_eip_cur(s);
6149             translator_io_start(&s->base);
6150             gen_helper_rdtsc(tcg_env);
6151             gen_helper_rdpid(s->T0, tcg_env);
6152             gen_op_mov_reg_v(s, dflag, R_ECX, s->T0);
6153             break;
6154 
6155         default:
6156             goto unknown_op;
6157         }
6158         break;
6159 
6160     case 0x108: /* invd */
6161     case 0x109: /* wbinvd; wbnoinvd with REPZ prefix */
6162         if (check_cpl0(s)) {
6163             gen_svm_check_intercept(s, (b & 1) ? SVM_EXIT_WBINVD : SVM_EXIT_INVD);
6164             /* nothing to do */
6165         }
6166         break;
6167     case 0x63: /* arpl or movslS (x86_64) */
6168 #ifdef TARGET_X86_64
6169         if (CODE64(s)) {
6170             int d_ot;
6171             /* d_ot is the size of destination */
6172             d_ot = dflag;
6173 
6174             modrm = x86_ldub_code(env, s);
6175             reg = ((modrm >> 3) & 7) | REX_R(s);
6176             mod = (modrm >> 6) & 3;
6177             rm = (modrm & 7) | REX_B(s);
6178 
6179             if (mod == 3) {
6180                 gen_op_mov_v_reg(s, MO_32, s->T0, rm);
6181                 /* sign extend */
6182                 if (d_ot == MO_64) {
6183                     tcg_gen_ext32s_tl(s->T0, s->T0);
6184                 }
6185                 gen_op_mov_reg_v(s, d_ot, reg, s->T0);
6186             } else {
6187                 gen_lea_modrm(env, s, modrm);
6188                 gen_op_ld_v(s, MO_32 | MO_SIGN, s->T0, s->A0);
6189                 gen_op_mov_reg_v(s, d_ot, reg, s->T0);
6190             }
6191         } else
6192 #endif
6193         {
6194             TCGLabel *label1;
6195             TCGv t0, t1, t2;
6196 
6197             if (!PE(s) || VM86(s))
6198                 goto illegal_op;
6199             t0 = tcg_temp_new();
6200             t1 = tcg_temp_new();
6201             t2 = tcg_temp_new();
6202             ot = MO_16;
6203             modrm = x86_ldub_code(env, s);
6204             reg = (modrm >> 3) & 7;
6205             mod = (modrm >> 6) & 3;
6206             rm = modrm & 7;
6207             if (mod != 3) {
6208                 gen_lea_modrm(env, s, modrm);
6209                 gen_op_ld_v(s, ot, t0, s->A0);
6210             } else {
6211                 gen_op_mov_v_reg(s, ot, t0, rm);
6212             }
6213             gen_op_mov_v_reg(s, ot, t1, reg);
6214             tcg_gen_andi_tl(s->tmp0, t0, 3);
6215             tcg_gen_andi_tl(t1, t1, 3);
6216             tcg_gen_movi_tl(t2, 0);
6217             label1 = gen_new_label();
6218             tcg_gen_brcond_tl(TCG_COND_GE, s->tmp0, t1, label1);
6219             tcg_gen_andi_tl(t0, t0, ~3);
6220             tcg_gen_or_tl(t0, t0, t1);
6221             tcg_gen_movi_tl(t2, CC_Z);
6222             gen_set_label(label1);
6223             if (mod != 3) {
6224                 gen_op_st_v(s, ot, t0, s->A0);
6225            } else {
6226                 gen_op_mov_reg_v(s, ot, rm, t0);
6227             }
6228             gen_compute_eflags(s);
6229             tcg_gen_andi_tl(cpu_cc_src, cpu_cc_src, ~CC_Z);
6230             tcg_gen_or_tl(cpu_cc_src, cpu_cc_src, t2);
6231         }
6232         break;
6233     case 0x102: /* lar */
6234     case 0x103: /* lsl */
6235         {
6236             TCGLabel *label1;
6237             TCGv t0;
6238             if (!PE(s) || VM86(s))
6239                 goto illegal_op;
6240             ot = dflag != MO_16 ? MO_32 : MO_16;
6241             modrm = x86_ldub_code(env, s);
6242             reg = ((modrm >> 3) & 7) | REX_R(s);
6243             gen_ldst_modrm(env, s, modrm, MO_16, OR_TMP0, 0);
6244             t0 = tcg_temp_new();
6245             gen_update_cc_op(s);
6246             if (b == 0x102) {
6247                 gen_helper_lar(t0, tcg_env, s->T0);
6248             } else {
6249                 gen_helper_lsl(t0, tcg_env, s->T0);
6250             }
6251             tcg_gen_andi_tl(s->tmp0, cpu_cc_src, CC_Z);
6252             label1 = gen_new_label();
6253             tcg_gen_brcondi_tl(TCG_COND_EQ, s->tmp0, 0, label1);
6254             gen_op_mov_reg_v(s, ot, reg, t0);
6255             gen_set_label(label1);
6256             set_cc_op(s, CC_OP_EFLAGS);
6257         }
6258         break;
6259     case 0x118:
6260         modrm = x86_ldub_code(env, s);
6261         mod = (modrm >> 6) & 3;
6262         op = (modrm >> 3) & 7;
6263         switch(op) {
6264         case 0: /* prefetchnta */
6265         case 1: /* prefetchnt0 */
6266         case 2: /* prefetchnt0 */
6267         case 3: /* prefetchnt0 */
6268             if (mod == 3)
6269                 goto illegal_op;
6270             gen_nop_modrm(env, s, modrm);
6271             /* nothing more to do */
6272             break;
6273         default: /* nop (multi byte) */
6274             gen_nop_modrm(env, s, modrm);
6275             break;
6276         }
6277         break;
6278     case 0x11a:
6279         modrm = x86_ldub_code(env, s);
6280         if (s->flags & HF_MPX_EN_MASK) {
6281             mod = (modrm >> 6) & 3;
6282             reg = ((modrm >> 3) & 7) | REX_R(s);
6283             if (prefixes & PREFIX_REPZ) {
6284                 /* bndcl */
6285                 if (reg >= 4
6286                     || (prefixes & PREFIX_LOCK)
6287                     || s->aflag == MO_16) {
6288                     goto illegal_op;
6289                 }
6290                 gen_bndck(env, s, modrm, TCG_COND_LTU, cpu_bndl[reg]);
6291             } else if (prefixes & PREFIX_REPNZ) {
6292                 /* bndcu */
6293                 if (reg >= 4
6294                     || (prefixes & PREFIX_LOCK)
6295                     || s->aflag == MO_16) {
6296                     goto illegal_op;
6297                 }
6298                 TCGv_i64 notu = tcg_temp_new_i64();
6299                 tcg_gen_not_i64(notu, cpu_bndu[reg]);
6300                 gen_bndck(env, s, modrm, TCG_COND_GTU, notu);
6301             } else if (prefixes & PREFIX_DATA) {
6302                 /* bndmov -- from reg/mem */
6303                 if (reg >= 4 || s->aflag == MO_16) {
6304                     goto illegal_op;
6305                 }
6306                 if (mod == 3) {
6307                     int reg2 = (modrm & 7) | REX_B(s);
6308                     if (reg2 >= 4 || (prefixes & PREFIX_LOCK)) {
6309                         goto illegal_op;
6310                     }
6311                     if (s->flags & HF_MPX_IU_MASK) {
6312                         tcg_gen_mov_i64(cpu_bndl[reg], cpu_bndl[reg2]);
6313                         tcg_gen_mov_i64(cpu_bndu[reg], cpu_bndu[reg2]);
6314                     }
6315                 } else {
6316                     gen_lea_modrm(env, s, modrm);
6317                     if (CODE64(s)) {
6318                         tcg_gen_qemu_ld_i64(cpu_bndl[reg], s->A0,
6319                                             s->mem_index, MO_LEUQ);
6320                         tcg_gen_addi_tl(s->A0, s->A0, 8);
6321                         tcg_gen_qemu_ld_i64(cpu_bndu[reg], s->A0,
6322                                             s->mem_index, MO_LEUQ);
6323                     } else {
6324                         tcg_gen_qemu_ld_i64(cpu_bndl[reg], s->A0,
6325                                             s->mem_index, MO_LEUL);
6326                         tcg_gen_addi_tl(s->A0, s->A0, 4);
6327                         tcg_gen_qemu_ld_i64(cpu_bndu[reg], s->A0,
6328                                             s->mem_index, MO_LEUL);
6329                     }
6330                     /* bnd registers are now in-use */
6331                     gen_set_hflag(s, HF_MPX_IU_MASK);
6332                 }
6333             } else if (mod != 3) {
6334                 /* bndldx */
6335                 AddressParts a = gen_lea_modrm_0(env, s, modrm);
6336                 if (reg >= 4
6337                     || (prefixes & PREFIX_LOCK)
6338                     || s->aflag == MO_16
6339                     || a.base < -1) {
6340                     goto illegal_op;
6341                 }
6342                 if (a.base >= 0) {
6343                     tcg_gen_addi_tl(s->A0, cpu_regs[a.base], a.disp);
6344                 } else {
6345                     tcg_gen_movi_tl(s->A0, 0);
6346                 }
6347                 gen_lea_v_seg(s, s->aflag, s->A0, a.def_seg, s->override);
6348                 if (a.index >= 0) {
6349                     tcg_gen_mov_tl(s->T0, cpu_regs[a.index]);
6350                 } else {
6351                     tcg_gen_movi_tl(s->T0, 0);
6352                 }
6353                 if (CODE64(s)) {
6354                     gen_helper_bndldx64(cpu_bndl[reg], tcg_env, s->A0, s->T0);
6355                     tcg_gen_ld_i64(cpu_bndu[reg], tcg_env,
6356                                    offsetof(CPUX86State, mmx_t0.MMX_Q(0)));
6357                 } else {
6358                     gen_helper_bndldx32(cpu_bndu[reg], tcg_env, s->A0, s->T0);
6359                     tcg_gen_ext32u_i64(cpu_bndl[reg], cpu_bndu[reg]);
6360                     tcg_gen_shri_i64(cpu_bndu[reg], cpu_bndu[reg], 32);
6361                 }
6362                 gen_set_hflag(s, HF_MPX_IU_MASK);
6363             }
6364         }
6365         gen_nop_modrm(env, s, modrm);
6366         break;
6367     case 0x11b:
6368         modrm = x86_ldub_code(env, s);
6369         if (s->flags & HF_MPX_EN_MASK) {
6370             mod = (modrm >> 6) & 3;
6371             reg = ((modrm >> 3) & 7) | REX_R(s);
6372             if (mod != 3 && (prefixes & PREFIX_REPZ)) {
6373                 /* bndmk */
6374                 if (reg >= 4
6375                     || (prefixes & PREFIX_LOCK)
6376                     || s->aflag == MO_16) {
6377                     goto illegal_op;
6378                 }
6379                 AddressParts a = gen_lea_modrm_0(env, s, modrm);
6380                 if (a.base >= 0) {
6381                     tcg_gen_extu_tl_i64(cpu_bndl[reg], cpu_regs[a.base]);
6382                     if (!CODE64(s)) {
6383                         tcg_gen_ext32u_i64(cpu_bndl[reg], cpu_bndl[reg]);
6384                     }
6385                 } else if (a.base == -1) {
6386                     /* no base register has lower bound of 0 */
6387                     tcg_gen_movi_i64(cpu_bndl[reg], 0);
6388                 } else {
6389                     /* rip-relative generates #ud */
6390                     goto illegal_op;
6391                 }
6392                 tcg_gen_not_tl(s->A0, gen_lea_modrm_1(s, a, false));
6393                 if (!CODE64(s)) {
6394                     tcg_gen_ext32u_tl(s->A0, s->A0);
6395                 }
6396                 tcg_gen_extu_tl_i64(cpu_bndu[reg], s->A0);
6397                 /* bnd registers are now in-use */
6398                 gen_set_hflag(s, HF_MPX_IU_MASK);
6399                 break;
6400             } else if (prefixes & PREFIX_REPNZ) {
6401                 /* bndcn */
6402                 if (reg >= 4
6403                     || (prefixes & PREFIX_LOCK)
6404                     || s->aflag == MO_16) {
6405                     goto illegal_op;
6406                 }
6407                 gen_bndck(env, s, modrm, TCG_COND_GTU, cpu_bndu[reg]);
6408             } else if (prefixes & PREFIX_DATA) {
6409                 /* bndmov -- to reg/mem */
6410                 if (reg >= 4 || s->aflag == MO_16) {
6411                     goto illegal_op;
6412                 }
6413                 if (mod == 3) {
6414                     int reg2 = (modrm & 7) | REX_B(s);
6415                     if (reg2 >= 4 || (prefixes & PREFIX_LOCK)) {
6416                         goto illegal_op;
6417                     }
6418                     if (s->flags & HF_MPX_IU_MASK) {
6419                         tcg_gen_mov_i64(cpu_bndl[reg2], cpu_bndl[reg]);
6420                         tcg_gen_mov_i64(cpu_bndu[reg2], cpu_bndu[reg]);
6421                     }
6422                 } else {
6423                     gen_lea_modrm(env, s, modrm);
6424                     if (CODE64(s)) {
6425                         tcg_gen_qemu_st_i64(cpu_bndl[reg], s->A0,
6426                                             s->mem_index, MO_LEUQ);
6427                         tcg_gen_addi_tl(s->A0, s->A0, 8);
6428                         tcg_gen_qemu_st_i64(cpu_bndu[reg], s->A0,
6429                                             s->mem_index, MO_LEUQ);
6430                     } else {
6431                         tcg_gen_qemu_st_i64(cpu_bndl[reg], s->A0,
6432                                             s->mem_index, MO_LEUL);
6433                         tcg_gen_addi_tl(s->A0, s->A0, 4);
6434                         tcg_gen_qemu_st_i64(cpu_bndu[reg], s->A0,
6435                                             s->mem_index, MO_LEUL);
6436                     }
6437                 }
6438             } else if (mod != 3) {
6439                 /* bndstx */
6440                 AddressParts a = gen_lea_modrm_0(env, s, modrm);
6441                 if (reg >= 4
6442                     || (prefixes & PREFIX_LOCK)
6443                     || s->aflag == MO_16
6444                     || a.base < -1) {
6445                     goto illegal_op;
6446                 }
6447                 if (a.base >= 0) {
6448                     tcg_gen_addi_tl(s->A0, cpu_regs[a.base], a.disp);
6449                 } else {
6450                     tcg_gen_movi_tl(s->A0, 0);
6451                 }
6452                 gen_lea_v_seg(s, s->aflag, s->A0, a.def_seg, s->override);
6453                 if (a.index >= 0) {
6454                     tcg_gen_mov_tl(s->T0, cpu_regs[a.index]);
6455                 } else {
6456                     tcg_gen_movi_tl(s->T0, 0);
6457                 }
6458                 if (CODE64(s)) {
6459                     gen_helper_bndstx64(tcg_env, s->A0, s->T0,
6460                                         cpu_bndl[reg], cpu_bndu[reg]);
6461                 } else {
6462                     gen_helper_bndstx32(tcg_env, s->A0, s->T0,
6463                                         cpu_bndl[reg], cpu_bndu[reg]);
6464                 }
6465             }
6466         }
6467         gen_nop_modrm(env, s, modrm);
6468         break;
6469     case 0x119: case 0x11c ... 0x11f: /* nop (multi byte) */
6470         modrm = x86_ldub_code(env, s);
6471         gen_nop_modrm(env, s, modrm);
6472         break;
6473 
6474     case 0x120: /* mov reg, crN */
6475     case 0x122: /* mov crN, reg */
6476         if (!check_cpl0(s)) {
6477             break;
6478         }
6479         modrm = x86_ldub_code(env, s);
6480         /*
6481          * Ignore the mod bits (assume (modrm&0xc0)==0xc0).
6482          * AMD documentation (24594.pdf) and testing of Intel 386 and 486
6483          * processors all show that the mod bits are assumed to be 1's,
6484          * regardless of actual values.
6485          */
6486         rm = (modrm & 7) | REX_B(s);
6487         reg = ((modrm >> 3) & 7) | REX_R(s);
6488         switch (reg) {
6489         case 0:
6490             if ((prefixes & PREFIX_LOCK) &&
6491                 (s->cpuid_ext3_features & CPUID_EXT3_CR8LEG)) {
6492                 reg = 8;
6493             }
6494             break;
6495         case 2:
6496         case 3:
6497         case 4:
6498         case 8:
6499             break;
6500         default:
6501             goto unknown_op;
6502         }
6503         ot  = (CODE64(s) ? MO_64 : MO_32);
6504 
6505         translator_io_start(&s->base);
6506         if (b & 2) {
6507             gen_svm_check_intercept(s, SVM_EXIT_WRITE_CR0 + reg);
6508             gen_op_mov_v_reg(s, ot, s->T0, rm);
6509             gen_helper_write_crN(tcg_env, tcg_constant_i32(reg), s->T0);
6510             s->base.is_jmp = DISAS_EOB_NEXT;
6511         } else {
6512             gen_svm_check_intercept(s, SVM_EXIT_READ_CR0 + reg);
6513             gen_helper_read_crN(s->T0, tcg_env, tcg_constant_i32(reg));
6514             gen_op_mov_reg_v(s, ot, rm, s->T0);
6515         }
6516         break;
6517 
6518     case 0x121: /* mov reg, drN */
6519     case 0x123: /* mov drN, reg */
6520         if (check_cpl0(s)) {
6521             modrm = x86_ldub_code(env, s);
6522             /* Ignore the mod bits (assume (modrm&0xc0)==0xc0).
6523              * AMD documentation (24594.pdf) and testing of
6524              * intel 386 and 486 processors all show that the mod bits
6525              * are assumed to be 1's, regardless of actual values.
6526              */
6527             rm = (modrm & 7) | REX_B(s);
6528             reg = ((modrm >> 3) & 7) | REX_R(s);
6529             if (CODE64(s))
6530                 ot = MO_64;
6531             else
6532                 ot = MO_32;
6533             if (reg >= 8) {
6534                 goto illegal_op;
6535             }
6536             if (b & 2) {
6537                 gen_svm_check_intercept(s, SVM_EXIT_WRITE_DR0 + reg);
6538                 gen_op_mov_v_reg(s, ot, s->T0, rm);
6539                 tcg_gen_movi_i32(s->tmp2_i32, reg);
6540                 gen_helper_set_dr(tcg_env, s->tmp2_i32, s->T0);
6541                 s->base.is_jmp = DISAS_EOB_NEXT;
6542             } else {
6543                 gen_svm_check_intercept(s, SVM_EXIT_READ_DR0 + reg);
6544                 tcg_gen_movi_i32(s->tmp2_i32, reg);
6545                 gen_helper_get_dr(s->T0, tcg_env, s->tmp2_i32);
6546                 gen_op_mov_reg_v(s, ot, rm, s->T0);
6547             }
6548         }
6549         break;
6550     case 0x106: /* clts */
6551         if (check_cpl0(s)) {
6552             gen_svm_check_intercept(s, SVM_EXIT_WRITE_CR0);
6553             gen_helper_clts(tcg_env);
6554             /* abort block because static cpu state changed */
6555             s->base.is_jmp = DISAS_EOB_NEXT;
6556         }
6557         break;
6558     /* MMX/3DNow!/SSE/SSE2/SSE3/SSSE3/SSE4 support */
6559     case 0x1c3: /* MOVNTI reg, mem */
6560         if (!(s->cpuid_features & CPUID_SSE2))
6561             goto illegal_op;
6562         ot = mo_64_32(dflag);
6563         modrm = x86_ldub_code(env, s);
6564         mod = (modrm >> 6) & 3;
6565         if (mod == 3)
6566             goto illegal_op;
6567         reg = ((modrm >> 3) & 7) | REX_R(s);
6568         /* generate a generic store */
6569         gen_ldst_modrm(env, s, modrm, ot, reg, 1);
6570         break;
6571     case 0x1ae:
6572         modrm = x86_ldub_code(env, s);
6573         switch (modrm) {
6574         CASE_MODRM_MEM_OP(0): /* fxsave */
6575             if (!(s->cpuid_features & CPUID_FXSR)
6576                 || (prefixes & PREFIX_LOCK)) {
6577                 goto illegal_op;
6578             }
6579             if ((s->flags & HF_EM_MASK) || (s->flags & HF_TS_MASK)) {
6580                 gen_exception(s, EXCP07_PREX);
6581                 break;
6582             }
6583             gen_lea_modrm(env, s, modrm);
6584             gen_helper_fxsave(tcg_env, s->A0);
6585             break;
6586 
6587         CASE_MODRM_MEM_OP(1): /* fxrstor */
6588             if (!(s->cpuid_features & CPUID_FXSR)
6589                 || (prefixes & PREFIX_LOCK)) {
6590                 goto illegal_op;
6591             }
6592             if ((s->flags & HF_EM_MASK) || (s->flags & HF_TS_MASK)) {
6593                 gen_exception(s, EXCP07_PREX);
6594                 break;
6595             }
6596             gen_lea_modrm(env, s, modrm);
6597             gen_helper_fxrstor(tcg_env, s->A0);
6598             break;
6599 
6600         CASE_MODRM_MEM_OP(2): /* ldmxcsr */
6601             if ((s->flags & HF_EM_MASK) || !(s->flags & HF_OSFXSR_MASK)) {
6602                 goto illegal_op;
6603             }
6604             if (s->flags & HF_TS_MASK) {
6605                 gen_exception(s, EXCP07_PREX);
6606                 break;
6607             }
6608             gen_lea_modrm(env, s, modrm);
6609             tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0, s->mem_index, MO_LEUL);
6610             gen_helper_ldmxcsr(tcg_env, s->tmp2_i32);
6611             break;
6612 
6613         CASE_MODRM_MEM_OP(3): /* stmxcsr */
6614             if ((s->flags & HF_EM_MASK) || !(s->flags & HF_OSFXSR_MASK)) {
6615                 goto illegal_op;
6616             }
6617             if (s->flags & HF_TS_MASK) {
6618                 gen_exception(s, EXCP07_PREX);
6619                 break;
6620             }
6621             gen_helper_update_mxcsr(tcg_env);
6622             gen_lea_modrm(env, s, modrm);
6623             tcg_gen_ld32u_tl(s->T0, tcg_env, offsetof(CPUX86State, mxcsr));
6624             gen_op_st_v(s, MO_32, s->T0, s->A0);
6625             break;
6626 
6627         CASE_MODRM_MEM_OP(4): /* xsave */
6628             if ((s->cpuid_ext_features & CPUID_EXT_XSAVE) == 0
6629                 || (prefixes & (PREFIX_LOCK | PREFIX_DATA
6630                                 | PREFIX_REPZ | PREFIX_REPNZ))) {
6631                 goto illegal_op;
6632             }
6633             gen_lea_modrm(env, s, modrm);
6634             tcg_gen_concat_tl_i64(s->tmp1_i64, cpu_regs[R_EAX],
6635                                   cpu_regs[R_EDX]);
6636             gen_helper_xsave(tcg_env, s->A0, s->tmp1_i64);
6637             break;
6638 
6639         CASE_MODRM_MEM_OP(5): /* xrstor */
6640             if ((s->cpuid_ext_features & CPUID_EXT_XSAVE) == 0
6641                 || (prefixes & (PREFIX_LOCK | PREFIX_DATA
6642                                 | PREFIX_REPZ | PREFIX_REPNZ))) {
6643                 goto illegal_op;
6644             }
6645             gen_lea_modrm(env, s, modrm);
6646             tcg_gen_concat_tl_i64(s->tmp1_i64, cpu_regs[R_EAX],
6647                                   cpu_regs[R_EDX]);
6648             gen_helper_xrstor(tcg_env, s->A0, s->tmp1_i64);
6649             /* XRSTOR is how MPX is enabled, which changes how
6650                we translate.  Thus we need to end the TB.  */
6651             s->base.is_jmp = DISAS_EOB_NEXT;
6652             break;
6653 
6654         CASE_MODRM_MEM_OP(6): /* xsaveopt / clwb */
6655             if (prefixes & PREFIX_LOCK) {
6656                 goto illegal_op;
6657             }
6658             if (prefixes & PREFIX_DATA) {
6659                 /* clwb */
6660                 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_CLWB)) {
6661                     goto illegal_op;
6662                 }
6663                 gen_nop_modrm(env, s, modrm);
6664             } else {
6665                 /* xsaveopt */
6666                 if ((s->cpuid_ext_features & CPUID_EXT_XSAVE) == 0
6667                     || (s->cpuid_xsave_features & CPUID_XSAVE_XSAVEOPT) == 0
6668                     || (prefixes & (PREFIX_REPZ | PREFIX_REPNZ))) {
6669                     goto illegal_op;
6670                 }
6671                 gen_lea_modrm(env, s, modrm);
6672                 tcg_gen_concat_tl_i64(s->tmp1_i64, cpu_regs[R_EAX],
6673                                       cpu_regs[R_EDX]);
6674                 gen_helper_xsaveopt(tcg_env, s->A0, s->tmp1_i64);
6675             }
6676             break;
6677 
6678         CASE_MODRM_MEM_OP(7): /* clflush / clflushopt */
6679             if (prefixes & PREFIX_LOCK) {
6680                 goto illegal_op;
6681             }
6682             if (prefixes & PREFIX_DATA) {
6683                 /* clflushopt */
6684                 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_CLFLUSHOPT)) {
6685                     goto illegal_op;
6686                 }
6687             } else {
6688                 /* clflush */
6689                 if ((s->prefix & (PREFIX_REPZ | PREFIX_REPNZ))
6690                     || !(s->cpuid_features & CPUID_CLFLUSH)) {
6691                     goto illegal_op;
6692                 }
6693             }
6694             gen_nop_modrm(env, s, modrm);
6695             break;
6696 
6697         case 0xc0 ... 0xc7: /* rdfsbase (f3 0f ae /0) */
6698         case 0xc8 ... 0xcf: /* rdgsbase (f3 0f ae /1) */
6699         case 0xd0 ... 0xd7: /* wrfsbase (f3 0f ae /2) */
6700         case 0xd8 ... 0xdf: /* wrgsbase (f3 0f ae /3) */
6701             if (CODE64(s)
6702                 && (prefixes & PREFIX_REPZ)
6703                 && !(prefixes & PREFIX_LOCK)
6704                 && (s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_FSGSBASE)) {
6705                 TCGv base, treg, src, dst;
6706 
6707                 /* Preserve hflags bits by testing CR4 at runtime.  */
6708                 tcg_gen_movi_i32(s->tmp2_i32, CR4_FSGSBASE_MASK);
6709                 gen_helper_cr4_testbit(tcg_env, s->tmp2_i32);
6710 
6711                 base = cpu_seg_base[modrm & 8 ? R_GS : R_FS];
6712                 treg = cpu_regs[(modrm & 7) | REX_B(s)];
6713 
6714                 if (modrm & 0x10) {
6715                     /* wr*base */
6716                     dst = base, src = treg;
6717                 } else {
6718                     /* rd*base */
6719                     dst = treg, src = base;
6720                 }
6721 
6722                 if (s->dflag == MO_32) {
6723                     tcg_gen_ext32u_tl(dst, src);
6724                 } else {
6725                     tcg_gen_mov_tl(dst, src);
6726                 }
6727                 break;
6728             }
6729             goto unknown_op;
6730 
6731         case 0xf8: /* sfence / pcommit */
6732             if (prefixes & PREFIX_DATA) {
6733                 /* pcommit */
6734                 if (!(s->cpuid_7_0_ebx_features & CPUID_7_0_EBX_PCOMMIT)
6735                     || (prefixes & PREFIX_LOCK)) {
6736                     goto illegal_op;
6737                 }
6738                 break;
6739             }
6740             /* fallthru */
6741         case 0xf9 ... 0xff: /* sfence */
6742             if (!(s->cpuid_features & CPUID_SSE)
6743                 || (prefixes & PREFIX_LOCK)) {
6744                 goto illegal_op;
6745             }
6746             tcg_gen_mb(TCG_MO_ST_ST | TCG_BAR_SC);
6747             break;
6748         case 0xe8 ... 0xef: /* lfence */
6749             if (!(s->cpuid_features & CPUID_SSE)
6750                 || (prefixes & PREFIX_LOCK)) {
6751                 goto illegal_op;
6752             }
6753             tcg_gen_mb(TCG_MO_LD_LD | TCG_BAR_SC);
6754             break;
6755         case 0xf0 ... 0xf7: /* mfence */
6756             if (!(s->cpuid_features & CPUID_SSE2)
6757                 || (prefixes & PREFIX_LOCK)) {
6758                 goto illegal_op;
6759             }
6760             tcg_gen_mb(TCG_MO_ALL | TCG_BAR_SC);
6761             break;
6762 
6763         default:
6764             goto unknown_op;
6765         }
6766         break;
6767 
6768     case 0x10d: /* 3DNow! prefetch(w) */
6769         modrm = x86_ldub_code(env, s);
6770         mod = (modrm >> 6) & 3;
6771         if (mod == 3)
6772             goto illegal_op;
6773         gen_nop_modrm(env, s, modrm);
6774         break;
6775     case 0x1aa: /* rsm */
6776         gen_svm_check_intercept(s, SVM_EXIT_RSM);
6777         if (!(s->flags & HF_SMM_MASK))
6778             goto illegal_op;
6779 #ifdef CONFIG_USER_ONLY
6780         /* we should not be in SMM mode */
6781         g_assert_not_reached();
6782 #else
6783         gen_update_cc_op(s);
6784         gen_update_eip_next(s);
6785         gen_helper_rsm(tcg_env);
6786 #endif /* CONFIG_USER_ONLY */
6787         s->base.is_jmp = DISAS_EOB_ONLY;
6788         break;
6789     case 0x1b8: /* SSE4.2 popcnt */
6790         if ((prefixes & (PREFIX_REPZ | PREFIX_LOCK | PREFIX_REPNZ)) !=
6791              PREFIX_REPZ)
6792             goto illegal_op;
6793         if (!(s->cpuid_ext_features & CPUID_EXT_POPCNT))
6794             goto illegal_op;
6795 
6796         modrm = x86_ldub_code(env, s);
6797         reg = ((modrm >> 3) & 7) | REX_R(s);
6798 
6799         if (s->prefix & PREFIX_DATA) {
6800             ot = MO_16;
6801         } else {
6802             ot = mo_64_32(dflag);
6803         }
6804 
6805         gen_ldst_modrm(env, s, modrm, ot, OR_TMP0, 0);
6806         gen_extu(ot, s->T0);
6807         tcg_gen_mov_tl(cpu_cc_src, s->T0);
6808         tcg_gen_ctpop_tl(s->T0, s->T0);
6809         gen_op_mov_reg_v(s, ot, reg, s->T0);
6810 
6811         set_cc_op(s, CC_OP_POPCNT);
6812         break;
6813     case 0x10e ... 0x117:
6814     case 0x128 ... 0x12f:
6815     case 0x138 ... 0x13a:
6816     case 0x150 ... 0x179:
6817     case 0x17c ... 0x17f:
6818     case 0x1c2:
6819     case 0x1c4 ... 0x1c6:
6820     case 0x1d0 ... 0x1fe:
6821         disas_insn_new(s, cpu, b);
6822         break;
6823     default:
6824         goto unknown_op;
6825     }
6826     return true;
6827  illegal_op:
6828     gen_illegal_opcode(s);
6829     return true;
6830  unknown_op:
6831     gen_unknown_opcode(env, s);
6832     return true;
6833 }
6834 
6835 void tcg_x86_init(void)
6836 {
6837     static const char reg_names[CPU_NB_REGS][4] = {
6838 #ifdef TARGET_X86_64
6839         [R_EAX] = "rax",
6840         [R_EBX] = "rbx",
6841         [R_ECX] = "rcx",
6842         [R_EDX] = "rdx",
6843         [R_ESI] = "rsi",
6844         [R_EDI] = "rdi",
6845         [R_EBP] = "rbp",
6846         [R_ESP] = "rsp",
6847         [8]  = "r8",
6848         [9]  = "r9",
6849         [10] = "r10",
6850         [11] = "r11",
6851         [12] = "r12",
6852         [13] = "r13",
6853         [14] = "r14",
6854         [15] = "r15",
6855 #else
6856         [R_EAX] = "eax",
6857         [R_EBX] = "ebx",
6858         [R_ECX] = "ecx",
6859         [R_EDX] = "edx",
6860         [R_ESI] = "esi",
6861         [R_EDI] = "edi",
6862         [R_EBP] = "ebp",
6863         [R_ESP] = "esp",
6864 #endif
6865     };
6866     static const char eip_name[] = {
6867 #ifdef TARGET_X86_64
6868         "rip"
6869 #else
6870         "eip"
6871 #endif
6872     };
6873     static const char seg_base_names[6][8] = {
6874         [R_CS] = "cs_base",
6875         [R_DS] = "ds_base",
6876         [R_ES] = "es_base",
6877         [R_FS] = "fs_base",
6878         [R_GS] = "gs_base",
6879         [R_SS] = "ss_base",
6880     };
6881     static const char bnd_regl_names[4][8] = {
6882         "bnd0_lb", "bnd1_lb", "bnd2_lb", "bnd3_lb"
6883     };
6884     static const char bnd_regu_names[4][8] = {
6885         "bnd0_ub", "bnd1_ub", "bnd2_ub", "bnd3_ub"
6886     };
6887     int i;
6888 
6889     cpu_cc_op = tcg_global_mem_new_i32(tcg_env,
6890                                        offsetof(CPUX86State, cc_op), "cc_op");
6891     cpu_cc_dst = tcg_global_mem_new(tcg_env, offsetof(CPUX86State, cc_dst),
6892                                     "cc_dst");
6893     cpu_cc_src = tcg_global_mem_new(tcg_env, offsetof(CPUX86State, cc_src),
6894                                     "cc_src");
6895     cpu_cc_src2 = tcg_global_mem_new(tcg_env, offsetof(CPUX86State, cc_src2),
6896                                      "cc_src2");
6897     cpu_eip = tcg_global_mem_new(tcg_env, offsetof(CPUX86State, eip), eip_name);
6898 
6899     for (i = 0; i < CPU_NB_REGS; ++i) {
6900         cpu_regs[i] = tcg_global_mem_new(tcg_env,
6901                                          offsetof(CPUX86State, regs[i]),
6902                                          reg_names[i]);
6903     }
6904 
6905     for (i = 0; i < 6; ++i) {
6906         cpu_seg_base[i]
6907             = tcg_global_mem_new(tcg_env,
6908                                  offsetof(CPUX86State, segs[i].base),
6909                                  seg_base_names[i]);
6910     }
6911 
6912     for (i = 0; i < 4; ++i) {
6913         cpu_bndl[i]
6914             = tcg_global_mem_new_i64(tcg_env,
6915                                      offsetof(CPUX86State, bnd_regs[i].lb),
6916                                      bnd_regl_names[i]);
6917         cpu_bndu[i]
6918             = tcg_global_mem_new_i64(tcg_env,
6919                                      offsetof(CPUX86State, bnd_regs[i].ub),
6920                                      bnd_regu_names[i]);
6921     }
6922 }
6923 
6924 static void i386_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cpu)
6925 {
6926     DisasContext *dc = container_of(dcbase, DisasContext, base);
6927     CPUX86State *env = cpu_env(cpu);
6928     uint32_t flags = dc->base.tb->flags;
6929     uint32_t cflags = tb_cflags(dc->base.tb);
6930     int cpl = (flags >> HF_CPL_SHIFT) & 3;
6931     int iopl = (flags >> IOPL_SHIFT) & 3;
6932 
6933     dc->cs_base = dc->base.tb->cs_base;
6934     dc->pc_save = dc->base.pc_next;
6935     dc->flags = flags;
6936 #ifndef CONFIG_USER_ONLY
6937     dc->cpl = cpl;
6938     dc->iopl = iopl;
6939 #endif
6940 
6941     /* We make some simplifying assumptions; validate they're correct. */
6942     g_assert(PE(dc) == ((flags & HF_PE_MASK) != 0));
6943     g_assert(CPL(dc) == cpl);
6944     g_assert(IOPL(dc) == iopl);
6945     g_assert(VM86(dc) == ((flags & HF_VM_MASK) != 0));
6946     g_assert(CODE32(dc) == ((flags & HF_CS32_MASK) != 0));
6947     g_assert(CODE64(dc) == ((flags & HF_CS64_MASK) != 0));
6948     g_assert(SS32(dc) == ((flags & HF_SS32_MASK) != 0));
6949     g_assert(LMA(dc) == ((flags & HF_LMA_MASK) != 0));
6950     g_assert(ADDSEG(dc) == ((flags & HF_ADDSEG_MASK) != 0));
6951     g_assert(SVME(dc) == ((flags & HF_SVME_MASK) != 0));
6952     g_assert(GUEST(dc) == ((flags & HF_GUEST_MASK) != 0));
6953 
6954     dc->cc_op = CC_OP_DYNAMIC;
6955     dc->cc_op_dirty = false;
6956     dc->popl_esp_hack = 0;
6957     /* select memory access functions */
6958     dc->mem_index = cpu_mmu_index(env, false);
6959     dc->cpuid_features = env->features[FEAT_1_EDX];
6960     dc->cpuid_ext_features = env->features[FEAT_1_ECX];
6961     dc->cpuid_ext2_features = env->features[FEAT_8000_0001_EDX];
6962     dc->cpuid_ext3_features = env->features[FEAT_8000_0001_ECX];
6963     dc->cpuid_7_0_ebx_features = env->features[FEAT_7_0_EBX];
6964     dc->cpuid_7_0_ecx_features = env->features[FEAT_7_0_ECX];
6965     dc->cpuid_7_1_eax_features = env->features[FEAT_7_1_EAX];
6966     dc->cpuid_xsave_features = env->features[FEAT_XSAVE];
6967     dc->jmp_opt = !((cflags & CF_NO_GOTO_TB) ||
6968                     (flags & (HF_TF_MASK | HF_INHIBIT_IRQ_MASK)));
6969     /*
6970      * If jmp_opt, we want to handle each string instruction individually.
6971      * For icount also disable repz optimization so that each iteration
6972      * is accounted separately.
6973      */
6974     dc->repz_opt = !dc->jmp_opt && !(cflags & CF_USE_ICOUNT);
6975 
6976     dc->T0 = tcg_temp_new();
6977     dc->T1 = tcg_temp_new();
6978     dc->A0 = tcg_temp_new();
6979 
6980     dc->tmp0 = tcg_temp_new();
6981     dc->tmp1_i64 = tcg_temp_new_i64();
6982     dc->tmp2_i32 = tcg_temp_new_i32();
6983     dc->tmp3_i32 = tcg_temp_new_i32();
6984     dc->tmp4 = tcg_temp_new();
6985     dc->cc_srcT = tcg_temp_new();
6986 }
6987 
6988 static void i386_tr_tb_start(DisasContextBase *db, CPUState *cpu)
6989 {
6990 }
6991 
6992 static void i386_tr_insn_start(DisasContextBase *dcbase, CPUState *cpu)
6993 {
6994     DisasContext *dc = container_of(dcbase, DisasContext, base);
6995     target_ulong pc_arg = dc->base.pc_next;
6996 
6997     dc->prev_insn_end = tcg_last_op();
6998     if (tb_cflags(dcbase->tb) & CF_PCREL) {
6999         pc_arg &= ~TARGET_PAGE_MASK;
7000     }
7001     tcg_gen_insn_start(pc_arg, dc->cc_op);
7002 }
7003 
7004 static void i386_tr_translate_insn(DisasContextBase *dcbase, CPUState *cpu)
7005 {
7006     DisasContext *dc = container_of(dcbase, DisasContext, base);
7007 
7008 #ifdef TARGET_VSYSCALL_PAGE
7009     /*
7010      * Detect entry into the vsyscall page and invoke the syscall.
7011      */
7012     if ((dc->base.pc_next & TARGET_PAGE_MASK) == TARGET_VSYSCALL_PAGE) {
7013         gen_exception(dc, EXCP_VSYSCALL);
7014         dc->base.pc_next = dc->pc + 1;
7015         return;
7016     }
7017 #endif
7018 
7019     if (disas_insn(dc, cpu)) {
7020         target_ulong pc_next = dc->pc;
7021         dc->base.pc_next = pc_next;
7022 
7023         if (dc->base.is_jmp == DISAS_NEXT) {
7024             if (dc->flags & (HF_TF_MASK | HF_INHIBIT_IRQ_MASK)) {
7025                 /*
7026                  * If single step mode, we generate only one instruction and
7027                  * generate an exception.
7028                  * If irq were inhibited with HF_INHIBIT_IRQ_MASK, we clear
7029                  * the flag and abort the translation to give the irqs a
7030                  * chance to happen.
7031                  */
7032                 dc->base.is_jmp = DISAS_EOB_NEXT;
7033             } else if (!is_same_page(&dc->base, pc_next)) {
7034                 dc->base.is_jmp = DISAS_TOO_MANY;
7035             }
7036         }
7037     }
7038 }
7039 
7040 static void i386_tr_tb_stop(DisasContextBase *dcbase, CPUState *cpu)
7041 {
7042     DisasContext *dc = container_of(dcbase, DisasContext, base);
7043 
7044     switch (dc->base.is_jmp) {
7045     case DISAS_NORETURN:
7046         break;
7047     case DISAS_TOO_MANY:
7048         gen_update_cc_op(dc);
7049         gen_jmp_rel_csize(dc, 0, 0);
7050         break;
7051     case DISAS_EOB_NEXT:
7052         gen_update_cc_op(dc);
7053         gen_update_eip_cur(dc);
7054         /* fall through */
7055     case DISAS_EOB_ONLY:
7056         gen_eob(dc);
7057         break;
7058     case DISAS_EOB_INHIBIT_IRQ:
7059         gen_update_cc_op(dc);
7060         gen_update_eip_cur(dc);
7061         gen_eob_inhibit_irq(dc, true);
7062         break;
7063     case DISAS_JUMP:
7064         gen_jr(dc);
7065         break;
7066     default:
7067         g_assert_not_reached();
7068     }
7069 }
7070 
7071 static void i386_tr_disas_log(const DisasContextBase *dcbase,
7072                               CPUState *cpu, FILE *logfile)
7073 {
7074     DisasContext *dc = container_of(dcbase, DisasContext, base);
7075 
7076     fprintf(logfile, "IN: %s\n", lookup_symbol(dc->base.pc_first));
7077     target_disas(logfile, cpu, dc->base.pc_first, dc->base.tb->size);
7078 }
7079 
7080 static const TranslatorOps i386_tr_ops = {
7081     .init_disas_context = i386_tr_init_disas_context,
7082     .tb_start           = i386_tr_tb_start,
7083     .insn_start         = i386_tr_insn_start,
7084     .translate_insn     = i386_tr_translate_insn,
7085     .tb_stop            = i386_tr_tb_stop,
7086     .disas_log          = i386_tr_disas_log,
7087 };
7088 
7089 /* generate intermediate code for basic block 'tb'.  */
7090 void gen_intermediate_code(CPUState *cpu, TranslationBlock *tb, int *max_insns,
7091                            target_ulong pc, void *host_pc)
7092 {
7093     DisasContext dc;
7094 
7095     translator_loop(cpu, tb, max_insns, pc, host_pc, &i386_tr_ops, &dc.base);
7096 }
7097