Statistics
| Branch: | Revision:

root / target-arm / translate-a64.c @ 52a1f6a3

History | View | Annotate | Download (123.7 kB)

1
/*
2
 *  AArch64 translation
3
 *
4
 *  Copyright (c) 2013 Alexander Graf <agraf@suse.de>
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 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 <stdarg.h>
20
#include <stdlib.h>
21
#include <stdio.h>
22
#include <string.h>
23
#include <inttypes.h>
24

    
25
#include "cpu.h"
26
#include "tcg-op.h"
27
#include "qemu/log.h"
28
#include "translate.h"
29
#include "qemu/host-utils.h"
30

    
31
#include "exec/gen-icount.h"
32

    
33
#include "helper.h"
34
#define GEN_HELPER 1
35
#include "helper.h"
36

    
37
static TCGv_i64 cpu_X[32];
38
static TCGv_i64 cpu_pc;
39
static TCGv_i32 cpu_NF, cpu_ZF, cpu_CF, cpu_VF;
40

    
41
/* Load/store exclusive handling */
42
static TCGv_i64 cpu_exclusive_addr;
43
static TCGv_i64 cpu_exclusive_val;
44
static TCGv_i64 cpu_exclusive_high;
45
#ifdef CONFIG_USER_ONLY
46
static TCGv_i64 cpu_exclusive_test;
47
static TCGv_i32 cpu_exclusive_info;
48
#endif
49

    
50
static const char *regnames[] = {
51
    "x0", "x1", "x2", "x3", "x4", "x5", "x6", "x7",
52
    "x8", "x9", "x10", "x11", "x12", "x13", "x14", "x15",
53
    "x16", "x17", "x18", "x19", "x20", "x21", "x22", "x23",
54
    "x24", "x25", "x26", "x27", "x28", "x29", "lr", "sp"
55
};
56

    
57
enum a64_shift_type {
58
    A64_SHIFT_TYPE_LSL = 0,
59
    A64_SHIFT_TYPE_LSR = 1,
60
    A64_SHIFT_TYPE_ASR = 2,
61
    A64_SHIFT_TYPE_ROR = 3
62
};
63

    
64
/* initialize TCG globals.  */
65
void a64_translate_init(void)
66
{
67
    int i;
68

    
69
    cpu_pc = tcg_global_mem_new_i64(TCG_AREG0,
70
                                    offsetof(CPUARMState, pc),
71
                                    "pc");
72
    for (i = 0; i < 32; i++) {
73
        cpu_X[i] = tcg_global_mem_new_i64(TCG_AREG0,
74
                                          offsetof(CPUARMState, xregs[i]),
75
                                          regnames[i]);
76
    }
77

    
78
    cpu_NF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, NF), "NF");
79
    cpu_ZF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, ZF), "ZF");
80
    cpu_CF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, CF), "CF");
81
    cpu_VF = tcg_global_mem_new_i32(TCG_AREG0, offsetof(CPUARMState, VF), "VF");
82

    
83
    cpu_exclusive_addr = tcg_global_mem_new_i64(TCG_AREG0,
84
        offsetof(CPUARMState, exclusive_addr), "exclusive_addr");
85
    cpu_exclusive_val = tcg_global_mem_new_i64(TCG_AREG0,
86
        offsetof(CPUARMState, exclusive_val), "exclusive_val");
87
    cpu_exclusive_high = tcg_global_mem_new_i64(TCG_AREG0,
88
        offsetof(CPUARMState, exclusive_high), "exclusive_high");
89
#ifdef CONFIG_USER_ONLY
90
    cpu_exclusive_test = tcg_global_mem_new_i64(TCG_AREG0,
91
        offsetof(CPUARMState, exclusive_test), "exclusive_test");
92
    cpu_exclusive_info = tcg_global_mem_new_i32(TCG_AREG0,
93
        offsetof(CPUARMState, exclusive_info), "exclusive_info");
94
#endif
95
}
96

    
97
void aarch64_cpu_dump_state(CPUState *cs, FILE *f,
98
                            fprintf_function cpu_fprintf, int flags)
99
{
100
    ARMCPU *cpu = ARM_CPU(cs);
101
    CPUARMState *env = &cpu->env;
102
    uint32_t psr = pstate_read(env);
103
    int i;
104

    
105
    cpu_fprintf(f, "PC=%016"PRIx64"  SP=%016"PRIx64"\n",
106
            env->pc, env->xregs[31]);
107
    for (i = 0; i < 31; i++) {
108
        cpu_fprintf(f, "X%02d=%016"PRIx64, i, env->xregs[i]);
109
        if ((i % 4) == 3) {
110
            cpu_fprintf(f, "\n");
111
        } else {
112
            cpu_fprintf(f, " ");
113
        }
114
    }
115
    cpu_fprintf(f, "PSTATE=%08x (flags %c%c%c%c)\n",
116
                psr,
117
                psr & PSTATE_N ? 'N' : '-',
118
                psr & PSTATE_Z ? 'Z' : '-',
119
                psr & PSTATE_C ? 'C' : '-',
120
                psr & PSTATE_V ? 'V' : '-');
121
    cpu_fprintf(f, "\n");
122

    
123
    if (flags & CPU_DUMP_FPU) {
124
        int numvfpregs = 32;
125
        for (i = 0; i < numvfpregs; i += 2) {
126
            uint64_t vlo = float64_val(env->vfp.regs[i * 2]);
127
            uint64_t vhi = float64_val(env->vfp.regs[(i * 2) + 1]);
128
            cpu_fprintf(f, "q%02d=%016" PRIx64 ":%016" PRIx64 " ",
129
                        i, vhi, vlo);
130
            vlo = float64_val(env->vfp.regs[(i + 1) * 2]);
131
            vhi = float64_val(env->vfp.regs[((i + 1) * 2) + 1]);
132
            cpu_fprintf(f, "q%02d=%016" PRIx64 ":%016" PRIx64 "\n",
133
                        i + 1, vhi, vlo);
134
        }
135
        cpu_fprintf(f, "FPCR: %08x  FPSR: %08x\n",
136
                    vfp_get_fpcr(env), vfp_get_fpsr(env));
137
    }
138
}
139

    
140
static int get_mem_index(DisasContext *s)
141
{
142
#ifdef CONFIG_USER_ONLY
143
    return 1;
144
#else
145
    return s->user;
146
#endif
147
}
148

    
149
void gen_a64_set_pc_im(uint64_t val)
150
{
151
    tcg_gen_movi_i64(cpu_pc, val);
152
}
153

    
154
static void gen_exception(int excp)
155
{
156
    TCGv_i32 tmp = tcg_temp_new_i32();
157
    tcg_gen_movi_i32(tmp, excp);
158
    gen_helper_exception(cpu_env, tmp);
159
    tcg_temp_free_i32(tmp);
160
}
161

    
162
static void gen_exception_insn(DisasContext *s, int offset, int excp)
163
{
164
    gen_a64_set_pc_im(s->pc - offset);
165
    gen_exception(excp);
166
    s->is_jmp = DISAS_EXC;
167
}
168

    
169
static inline bool use_goto_tb(DisasContext *s, int n, uint64_t dest)
170
{
171
    /* No direct tb linking with singlestep or deterministic io */
172
    if (s->singlestep_enabled || (s->tb->cflags & CF_LAST_IO)) {
173
        return false;
174
    }
175

    
176
    /* Only link tbs from inside the same guest page */
177
    if ((s->tb->pc & TARGET_PAGE_MASK) != (dest & TARGET_PAGE_MASK)) {
178
        return false;
179
    }
180

    
181
    return true;
182
}
183

    
184
static inline void gen_goto_tb(DisasContext *s, int n, uint64_t dest)
185
{
186
    TranslationBlock *tb;
187

    
188
    tb = s->tb;
189
    if (use_goto_tb(s, n, dest)) {
190
        tcg_gen_goto_tb(n);
191
        gen_a64_set_pc_im(dest);
192
        tcg_gen_exit_tb((tcg_target_long)tb + n);
193
        s->is_jmp = DISAS_TB_JUMP;
194
    } else {
195
        gen_a64_set_pc_im(dest);
196
        if (s->singlestep_enabled) {
197
            gen_exception(EXCP_DEBUG);
198
        }
199
        tcg_gen_exit_tb(0);
200
        s->is_jmp = DISAS_JUMP;
201
    }
202
}
203

    
204
static void unallocated_encoding(DisasContext *s)
205
{
206
    gen_exception_insn(s, 4, EXCP_UDEF);
207
}
208

    
209
#define unsupported_encoding(s, insn)                                    \
210
    do {                                                                 \
211
        qemu_log_mask(LOG_UNIMP,                                         \
212
                      "%s:%d: unsupported instruction encoding 0x%08x "  \
213
                      "at pc=%016" PRIx64 "\n",                          \
214
                      __FILE__, __LINE__, insn, s->pc - 4);              \
215
        unallocated_encoding(s);                                         \
216
    } while (0);
217

    
218
static void init_tmp_a64_array(DisasContext *s)
219
{
220
#ifdef CONFIG_DEBUG_TCG
221
    int i;
222
    for (i = 0; i < ARRAY_SIZE(s->tmp_a64); i++) {
223
        TCGV_UNUSED_I64(s->tmp_a64[i]);
224
    }
225
#endif
226
    s->tmp_a64_count = 0;
227
}
228

    
229
static void free_tmp_a64(DisasContext *s)
230
{
231
    int i;
232
    for (i = 0; i < s->tmp_a64_count; i++) {
233
        tcg_temp_free_i64(s->tmp_a64[i]);
234
    }
235
    init_tmp_a64_array(s);
236
}
237

    
238
static TCGv_i64 new_tmp_a64(DisasContext *s)
239
{
240
    assert(s->tmp_a64_count < TMP_A64_MAX);
241
    return s->tmp_a64[s->tmp_a64_count++] = tcg_temp_new_i64();
242
}
243

    
244
static TCGv_i64 new_tmp_a64_zero(DisasContext *s)
245
{
246
    TCGv_i64 t = new_tmp_a64(s);
247
    tcg_gen_movi_i64(t, 0);
248
    return t;
249
}
250

    
251
/*
252
 * Register access functions
253
 *
254
 * These functions are used for directly accessing a register in where
255
 * changes to the final register value are likely to be made. If you
256
 * need to use a register for temporary calculation (e.g. index type
257
 * operations) use the read_* form.
258
 *
259
 * B1.2.1 Register mappings
260
 *
261
 * In instruction register encoding 31 can refer to ZR (zero register) or
262
 * the SP (stack pointer) depending on context. In QEMU's case we map SP
263
 * to cpu_X[31] and ZR accesses to a temporary which can be discarded.
264
 * This is the point of the _sp forms.
265
 */
266
static TCGv_i64 cpu_reg(DisasContext *s, int reg)
267
{
268
    if (reg == 31) {
269
        return new_tmp_a64_zero(s);
270
    } else {
271
        return cpu_X[reg];
272
    }
273
}
274

    
275
/* register access for when 31 == SP */
276
static TCGv_i64 cpu_reg_sp(DisasContext *s, int reg)
277
{
278
    return cpu_X[reg];
279
}
280

    
281
/* read a cpu register in 32bit/64bit mode. Returns a TCGv_i64
282
 * representing the register contents. This TCGv is an auto-freed
283
 * temporary so it need not be explicitly freed, and may be modified.
284
 */
285
static TCGv_i64 read_cpu_reg(DisasContext *s, int reg, int sf)
286
{
287
    TCGv_i64 v = new_tmp_a64(s);
288
    if (reg != 31) {
289
        if (sf) {
290
            tcg_gen_mov_i64(v, cpu_X[reg]);
291
        } else {
292
            tcg_gen_ext32u_i64(v, cpu_X[reg]);
293
        }
294
    } else {
295
        tcg_gen_movi_i64(v, 0);
296
    }
297
    return v;
298
}
299

    
300
static TCGv_i64 read_cpu_reg_sp(DisasContext *s, int reg, int sf)
301
{
302
    TCGv_i64 v = new_tmp_a64(s);
303
    if (sf) {
304
        tcg_gen_mov_i64(v, cpu_X[reg]);
305
    } else {
306
        tcg_gen_ext32u_i64(v, cpu_X[reg]);
307
    }
308
    return v;
309
}
310

    
311
/* Return the offset into CPUARMState of a slice (from
312
 * the least significant end) of FP register Qn (ie
313
 * Dn, Sn, Hn or Bn).
314
 * (Note that this is not the same mapping as for A32; see cpu.h)
315
 */
316
static inline int fp_reg_offset(int regno, TCGMemOp size)
317
{
318
    int offs = offsetof(CPUARMState, vfp.regs[regno * 2]);
319
#ifdef HOST_WORDS_BIGENDIAN
320
    offs += (8 - (1 << size));
321
#endif
322
    return offs;
323
}
324

    
325
/* Offset of the high half of the 128 bit vector Qn */
326
static inline int fp_reg_hi_offset(int regno)
327
{
328
    return offsetof(CPUARMState, vfp.regs[regno * 2 + 1]);
329
}
330

    
331
/* Convenience accessors for reading and writing single and double
332
 * FP registers. Writing clears the upper parts of the associated
333
 * 128 bit vector register, as required by the architecture.
334
 * Note that unlike the GP register accessors, the values returned
335
 * by the read functions must be manually freed.
336
 */
337
static TCGv_i64 read_fp_dreg(DisasContext *s, int reg)
338
{
339
    TCGv_i64 v = tcg_temp_new_i64();
340

    
341
    tcg_gen_ld_i64(v, cpu_env, fp_reg_offset(reg, MO_64));
342
    return v;
343
}
344

    
345
static TCGv_i32 read_fp_sreg(DisasContext *s, int reg)
346
{
347
    TCGv_i32 v = tcg_temp_new_i32();
348

    
349
    tcg_gen_ld_i32(v, cpu_env, fp_reg_offset(reg, MO_32));
350
    return v;
351
}
352

    
353
static void write_fp_dreg(DisasContext *s, int reg, TCGv_i64 v)
354
{
355
    TCGv_i64 tcg_zero = tcg_const_i64(0);
356

    
357
    tcg_gen_st_i64(v, cpu_env, fp_reg_offset(reg, MO_64));
358
    tcg_gen_st_i64(tcg_zero, cpu_env, fp_reg_hi_offset(reg));
359
    tcg_temp_free_i64(tcg_zero);
360
}
361

    
362
static void write_fp_sreg(DisasContext *s, int reg, TCGv_i32 v)
363
{
364
    TCGv_i64 tmp = tcg_temp_new_i64();
365

    
366
    tcg_gen_extu_i32_i64(tmp, v);
367
    write_fp_dreg(s, reg, tmp);
368
    tcg_temp_free_i64(tmp);
369
}
370

    
371
static TCGv_ptr get_fpstatus_ptr(void)
372
{
373
    TCGv_ptr statusptr = tcg_temp_new_ptr();
374
    int offset;
375

    
376
    /* In A64 all instructions (both FP and Neon) use the FPCR;
377
     * there is no equivalent of the A32 Neon "standard FPSCR value"
378
     * and all operations use vfp.fp_status.
379
     */
380
    offset = offsetof(CPUARMState, vfp.fp_status);
381
    tcg_gen_addi_ptr(statusptr, cpu_env, offset);
382
    return statusptr;
383
}
384

    
385
/* Set ZF and NF based on a 64 bit result. This is alas fiddlier
386
 * than the 32 bit equivalent.
387
 */
388
static inline void gen_set_NZ64(TCGv_i64 result)
389
{
390
    TCGv_i64 flag = tcg_temp_new_i64();
391

    
392
    tcg_gen_setcondi_i64(TCG_COND_NE, flag, result, 0);
393
    tcg_gen_trunc_i64_i32(cpu_ZF, flag);
394
    tcg_gen_shri_i64(flag, result, 32);
395
    tcg_gen_trunc_i64_i32(cpu_NF, flag);
396
    tcg_temp_free_i64(flag);
397
}
398

    
399
/* Set NZCV as for a logical operation: NZ as per result, CV cleared. */
400
static inline void gen_logic_CC(int sf, TCGv_i64 result)
401
{
402
    if (sf) {
403
        gen_set_NZ64(result);
404
    } else {
405
        tcg_gen_trunc_i64_i32(cpu_ZF, result);
406
        tcg_gen_trunc_i64_i32(cpu_NF, result);
407
    }
408
    tcg_gen_movi_i32(cpu_CF, 0);
409
    tcg_gen_movi_i32(cpu_VF, 0);
410
}
411

    
412
/* dest = T0 + T1; compute C, N, V and Z flags */
413
static void gen_add_CC(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
414
{
415
    if (sf) {
416
        TCGv_i64 result, flag, tmp;
417
        result = tcg_temp_new_i64();
418
        flag = tcg_temp_new_i64();
419
        tmp = tcg_temp_new_i64();
420

    
421
        tcg_gen_movi_i64(tmp, 0);
422
        tcg_gen_add2_i64(result, flag, t0, tmp, t1, tmp);
423

    
424
        tcg_gen_trunc_i64_i32(cpu_CF, flag);
425

    
426
        gen_set_NZ64(result);
427

    
428
        tcg_gen_xor_i64(flag, result, t0);
429
        tcg_gen_xor_i64(tmp, t0, t1);
430
        tcg_gen_andc_i64(flag, flag, tmp);
431
        tcg_temp_free_i64(tmp);
432
        tcg_gen_shri_i64(flag, flag, 32);
433
        tcg_gen_trunc_i64_i32(cpu_VF, flag);
434

    
435
        tcg_gen_mov_i64(dest, result);
436
        tcg_temp_free_i64(result);
437
        tcg_temp_free_i64(flag);
438
    } else {
439
        /* 32 bit arithmetic */
440
        TCGv_i32 t0_32 = tcg_temp_new_i32();
441
        TCGv_i32 t1_32 = tcg_temp_new_i32();
442
        TCGv_i32 tmp = tcg_temp_new_i32();
443

    
444
        tcg_gen_movi_i32(tmp, 0);
445
        tcg_gen_trunc_i64_i32(t0_32, t0);
446
        tcg_gen_trunc_i64_i32(t1_32, t1);
447
        tcg_gen_add2_i32(cpu_NF, cpu_CF, t0_32, tmp, t1_32, tmp);
448
        tcg_gen_mov_i32(cpu_ZF, cpu_NF);
449
        tcg_gen_xor_i32(cpu_VF, cpu_NF, t0_32);
450
        tcg_gen_xor_i32(tmp, t0_32, t1_32);
451
        tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
452
        tcg_gen_extu_i32_i64(dest, cpu_NF);
453

    
454
        tcg_temp_free_i32(tmp);
455
        tcg_temp_free_i32(t0_32);
456
        tcg_temp_free_i32(t1_32);
457
    }
458
}
459

    
460
/* dest = T0 - T1; compute C, N, V and Z flags */
461
static void gen_sub_CC(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
462
{
463
    if (sf) {
464
        /* 64 bit arithmetic */
465
        TCGv_i64 result, flag, tmp;
466

    
467
        result = tcg_temp_new_i64();
468
        flag = tcg_temp_new_i64();
469
        tcg_gen_sub_i64(result, t0, t1);
470

    
471
        gen_set_NZ64(result);
472

    
473
        tcg_gen_setcond_i64(TCG_COND_GEU, flag, t0, t1);
474
        tcg_gen_trunc_i64_i32(cpu_CF, flag);
475

    
476
        tcg_gen_xor_i64(flag, result, t0);
477
        tmp = tcg_temp_new_i64();
478
        tcg_gen_xor_i64(tmp, t0, t1);
479
        tcg_gen_and_i64(flag, flag, tmp);
480
        tcg_temp_free_i64(tmp);
481
        tcg_gen_shri_i64(flag, flag, 32);
482
        tcg_gen_trunc_i64_i32(cpu_VF, flag);
483
        tcg_gen_mov_i64(dest, result);
484
        tcg_temp_free_i64(flag);
485
        tcg_temp_free_i64(result);
486
    } else {
487
        /* 32 bit arithmetic */
488
        TCGv_i32 t0_32 = tcg_temp_new_i32();
489
        TCGv_i32 t1_32 = tcg_temp_new_i32();
490
        TCGv_i32 tmp;
491

    
492
        tcg_gen_trunc_i64_i32(t0_32, t0);
493
        tcg_gen_trunc_i64_i32(t1_32, t1);
494
        tcg_gen_sub_i32(cpu_NF, t0_32, t1_32);
495
        tcg_gen_mov_i32(cpu_ZF, cpu_NF);
496
        tcg_gen_setcond_i32(TCG_COND_GEU, cpu_CF, t0_32, t1_32);
497
        tcg_gen_xor_i32(cpu_VF, cpu_NF, t0_32);
498
        tmp = tcg_temp_new_i32();
499
        tcg_gen_xor_i32(tmp, t0_32, t1_32);
500
        tcg_temp_free_i32(t0_32);
501
        tcg_temp_free_i32(t1_32);
502
        tcg_gen_and_i32(cpu_VF, cpu_VF, tmp);
503
        tcg_temp_free_i32(tmp);
504
        tcg_gen_extu_i32_i64(dest, cpu_NF);
505
    }
506
}
507

    
508
/* dest = T0 + T1 + CF; do not compute flags. */
509
static void gen_adc(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
510
{
511
    TCGv_i64 flag = tcg_temp_new_i64();
512
    tcg_gen_extu_i32_i64(flag, cpu_CF);
513
    tcg_gen_add_i64(dest, t0, t1);
514
    tcg_gen_add_i64(dest, dest, flag);
515
    tcg_temp_free_i64(flag);
516

    
517
    if (!sf) {
518
        tcg_gen_ext32u_i64(dest, dest);
519
    }
520
}
521

    
522
/* dest = T0 + T1 + CF; compute C, N, V and Z flags. */
523
static void gen_adc_CC(int sf, TCGv_i64 dest, TCGv_i64 t0, TCGv_i64 t1)
524
{
525
    if (sf) {
526
        TCGv_i64 result, cf_64, vf_64, tmp;
527
        result = tcg_temp_new_i64();
528
        cf_64 = tcg_temp_new_i64();
529
        vf_64 = tcg_temp_new_i64();
530
        tmp = tcg_const_i64(0);
531

    
532
        tcg_gen_extu_i32_i64(cf_64, cpu_CF);
533
        tcg_gen_add2_i64(result, cf_64, t0, tmp, cf_64, tmp);
534
        tcg_gen_add2_i64(result, cf_64, result, cf_64, t1, tmp);
535
        tcg_gen_trunc_i64_i32(cpu_CF, cf_64);
536
        gen_set_NZ64(result);
537

    
538
        tcg_gen_xor_i64(vf_64, result, t0);
539
        tcg_gen_xor_i64(tmp, t0, t1);
540
        tcg_gen_andc_i64(vf_64, vf_64, tmp);
541
        tcg_gen_shri_i64(vf_64, vf_64, 32);
542
        tcg_gen_trunc_i64_i32(cpu_VF, vf_64);
543

    
544
        tcg_gen_mov_i64(dest, result);
545

    
546
        tcg_temp_free_i64(tmp);
547
        tcg_temp_free_i64(vf_64);
548
        tcg_temp_free_i64(cf_64);
549
        tcg_temp_free_i64(result);
550
    } else {
551
        TCGv_i32 t0_32, t1_32, tmp;
552
        t0_32 = tcg_temp_new_i32();
553
        t1_32 = tcg_temp_new_i32();
554
        tmp = tcg_const_i32(0);
555

    
556
        tcg_gen_trunc_i64_i32(t0_32, t0);
557
        tcg_gen_trunc_i64_i32(t1_32, t1);
558
        tcg_gen_add2_i32(cpu_NF, cpu_CF, t0_32, tmp, cpu_CF, tmp);
559
        tcg_gen_add2_i32(cpu_NF, cpu_CF, cpu_NF, cpu_CF, t1_32, tmp);
560

    
561
        tcg_gen_mov_i32(cpu_ZF, cpu_NF);
562
        tcg_gen_xor_i32(cpu_VF, cpu_NF, t0_32);
563
        tcg_gen_xor_i32(tmp, t0_32, t1_32);
564
        tcg_gen_andc_i32(cpu_VF, cpu_VF, tmp);
565
        tcg_gen_extu_i32_i64(dest, cpu_NF);
566

    
567
        tcg_temp_free_i32(tmp);
568
        tcg_temp_free_i32(t1_32);
569
        tcg_temp_free_i32(t0_32);
570
    }
571
}
572

    
573
/*
574
 * Load/Store generators
575
 */
576

    
577
/*
578
 * Store from GPR register to memory
579
 */
580
static void do_gpr_st(DisasContext *s, TCGv_i64 source,
581
                      TCGv_i64 tcg_addr, int size)
582
{
583
    g_assert(size <= 3);
584
    tcg_gen_qemu_st_i64(source, tcg_addr, get_mem_index(s), MO_TE + size);
585
}
586

    
587
/*
588
 * Load from memory to GPR register
589
 */
590
static void do_gpr_ld(DisasContext *s, TCGv_i64 dest, TCGv_i64 tcg_addr,
591
                      int size, bool is_signed, bool extend)
592
{
593
    TCGMemOp memop = MO_TE + size;
594

    
595
    g_assert(size <= 3);
596

    
597
    if (is_signed) {
598
        memop += MO_SIGN;
599
    }
600

    
601
    tcg_gen_qemu_ld_i64(dest, tcg_addr, get_mem_index(s), memop);
602

    
603
    if (extend && is_signed) {
604
        g_assert(size < 3);
605
        tcg_gen_ext32u_i64(dest, dest);
606
    }
607
}
608

    
609
/*
610
 * Store from FP register to memory
611
 */
612
static void do_fp_st(DisasContext *s, int srcidx, TCGv_i64 tcg_addr, int size)
613
{
614
    /* This writes the bottom N bits of a 128 bit wide vector to memory */
615
    TCGv_i64 tmp = tcg_temp_new_i64();
616
    tcg_gen_ld_i64(tmp, cpu_env, fp_reg_offset(srcidx, MO_64));
617
    if (size < 4) {
618
        tcg_gen_qemu_st_i64(tmp, tcg_addr, get_mem_index(s), MO_TE + size);
619
    } else {
620
        TCGv_i64 tcg_hiaddr = tcg_temp_new_i64();
621
        tcg_gen_qemu_st_i64(tmp, tcg_addr, get_mem_index(s), MO_TEQ);
622
        tcg_gen_qemu_st64(tmp, tcg_addr, get_mem_index(s));
623
        tcg_gen_ld_i64(tmp, cpu_env, fp_reg_hi_offset(srcidx));
624
        tcg_gen_addi_i64(tcg_hiaddr, tcg_addr, 8);
625
        tcg_gen_qemu_st_i64(tmp, tcg_hiaddr, get_mem_index(s), MO_TEQ);
626
        tcg_temp_free_i64(tcg_hiaddr);
627
    }
628

    
629
    tcg_temp_free_i64(tmp);
630
}
631

    
632
/*
633
 * Load from memory to FP register
634
 */
635
static void do_fp_ld(DisasContext *s, int destidx, TCGv_i64 tcg_addr, int size)
636
{
637
    /* This always zero-extends and writes to a full 128 bit wide vector */
638
    TCGv_i64 tmplo = tcg_temp_new_i64();
639
    TCGv_i64 tmphi;
640

    
641
    if (size < 4) {
642
        TCGMemOp memop = MO_TE + size;
643
        tmphi = tcg_const_i64(0);
644
        tcg_gen_qemu_ld_i64(tmplo, tcg_addr, get_mem_index(s), memop);
645
    } else {
646
        TCGv_i64 tcg_hiaddr;
647
        tmphi = tcg_temp_new_i64();
648
        tcg_hiaddr = tcg_temp_new_i64();
649

    
650
        tcg_gen_qemu_ld_i64(tmplo, tcg_addr, get_mem_index(s), MO_TEQ);
651
        tcg_gen_addi_i64(tcg_hiaddr, tcg_addr, 8);
652
        tcg_gen_qemu_ld_i64(tmphi, tcg_hiaddr, get_mem_index(s), MO_TEQ);
653
        tcg_temp_free_i64(tcg_hiaddr);
654
    }
655

    
656
    tcg_gen_st_i64(tmplo, cpu_env, fp_reg_offset(destidx, MO_64));
657
    tcg_gen_st_i64(tmphi, cpu_env, fp_reg_hi_offset(destidx));
658

    
659
    tcg_temp_free_i64(tmplo);
660
    tcg_temp_free_i64(tmphi);
661
}
662

    
663
/*
664
 * This utility function is for doing register extension with an
665
 * optional shift. You will likely want to pass a temporary for the
666
 * destination register. See DecodeRegExtend() in the ARM ARM.
667
 */
668
static void ext_and_shift_reg(TCGv_i64 tcg_out, TCGv_i64 tcg_in,
669
                              int option, unsigned int shift)
670
{
671
    int extsize = extract32(option, 0, 2);
672
    bool is_signed = extract32(option, 2, 1);
673

    
674
    if (is_signed) {
675
        switch (extsize) {
676
        case 0:
677
            tcg_gen_ext8s_i64(tcg_out, tcg_in);
678
            break;
679
        case 1:
680
            tcg_gen_ext16s_i64(tcg_out, tcg_in);
681
            break;
682
        case 2:
683
            tcg_gen_ext32s_i64(tcg_out, tcg_in);
684
            break;
685
        case 3:
686
            tcg_gen_mov_i64(tcg_out, tcg_in);
687
            break;
688
        }
689
    } else {
690
        switch (extsize) {
691
        case 0:
692
            tcg_gen_ext8u_i64(tcg_out, tcg_in);
693
            break;
694
        case 1:
695
            tcg_gen_ext16u_i64(tcg_out, tcg_in);
696
            break;
697
        case 2:
698
            tcg_gen_ext32u_i64(tcg_out, tcg_in);
699
            break;
700
        case 3:
701
            tcg_gen_mov_i64(tcg_out, tcg_in);
702
            break;
703
        }
704
    }
705

    
706
    if (shift) {
707
        tcg_gen_shli_i64(tcg_out, tcg_out, shift);
708
    }
709
}
710

    
711
static inline void gen_check_sp_alignment(DisasContext *s)
712
{
713
    /* The AArch64 architecture mandates that (if enabled via PSTATE
714
     * or SCTLR bits) there is a check that SP is 16-aligned on every
715
     * SP-relative load or store (with an exception generated if it is not).
716
     * In line with general QEMU practice regarding misaligned accesses,
717
     * we omit these checks for the sake of guest program performance.
718
     * This function is provided as a hook so we can more easily add these
719
     * checks in future (possibly as a "favour catching guest program bugs
720
     * over speed" user selectable option).
721
     */
722
}
723

    
724
/*
725
 * the instruction disassembly implemented here matches
726
 * the instruction encoding classifications in chapter 3 (C3)
727
 * of the ARM Architecture Reference Manual (DDI0487A_a)
728
 */
729

    
730
/* C3.2.7 Unconditional branch (immediate)
731
 *   31  30       26 25                                  0
732
 * +----+-----------+-------------------------------------+
733
 * | op | 0 0 1 0 1 |                 imm26               |
734
 * +----+-----------+-------------------------------------+
735
 */
736
static void disas_uncond_b_imm(DisasContext *s, uint32_t insn)
737
{
738
    uint64_t addr = s->pc + sextract32(insn, 0, 26) * 4 - 4;
739

    
740
    if (insn & (1 << 31)) {
741
        /* C5.6.26 BL Branch with link */
742
        tcg_gen_movi_i64(cpu_reg(s, 30), s->pc);
743
    }
744

    
745
    /* C5.6.20 B Branch / C5.6.26 BL Branch with link */
746
    gen_goto_tb(s, 0, addr);
747
}
748

    
749
/* C3.2.1 Compare & branch (immediate)
750
 *   31  30         25  24  23                  5 4      0
751
 * +----+-------------+----+---------------------+--------+
752
 * | sf | 0 1 1 0 1 0 | op |         imm19       |   Rt   |
753
 * +----+-------------+----+---------------------+--------+
754
 */
755
static void disas_comp_b_imm(DisasContext *s, uint32_t insn)
756
{
757
    unsigned int sf, op, rt;
758
    uint64_t addr;
759
    int label_match;
760
    TCGv_i64 tcg_cmp;
761

    
762
    sf = extract32(insn, 31, 1);
763
    op = extract32(insn, 24, 1); /* 0: CBZ; 1: CBNZ */
764
    rt = extract32(insn, 0, 5);
765
    addr = s->pc + sextract32(insn, 5, 19) * 4 - 4;
766

    
767
    tcg_cmp = read_cpu_reg(s, rt, sf);
768
    label_match = gen_new_label();
769

    
770
    tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
771
                        tcg_cmp, 0, label_match);
772

    
773
    gen_goto_tb(s, 0, s->pc);
774
    gen_set_label(label_match);
775
    gen_goto_tb(s, 1, addr);
776
}
777

    
778
/* C3.2.5 Test & branch (immediate)
779
 *   31  30         25  24  23   19 18          5 4    0
780
 * +----+-------------+----+-------+-------------+------+
781
 * | b5 | 0 1 1 0 1 1 | op |  b40  |    imm14    |  Rt  |
782
 * +----+-------------+----+-------+-------------+------+
783
 */
784
static void disas_test_b_imm(DisasContext *s, uint32_t insn)
785
{
786
    unsigned int bit_pos, op, rt;
787
    uint64_t addr;
788
    int label_match;
789
    TCGv_i64 tcg_cmp;
790

    
791
    bit_pos = (extract32(insn, 31, 1) << 5) | extract32(insn, 19, 5);
792
    op = extract32(insn, 24, 1); /* 0: TBZ; 1: TBNZ */
793
    addr = s->pc + sextract32(insn, 5, 14) * 4 - 4;
794
    rt = extract32(insn, 0, 5);
795

    
796
    tcg_cmp = tcg_temp_new_i64();
797
    tcg_gen_andi_i64(tcg_cmp, cpu_reg(s, rt), (1ULL << bit_pos));
798
    label_match = gen_new_label();
799
    tcg_gen_brcondi_i64(op ? TCG_COND_NE : TCG_COND_EQ,
800
                        tcg_cmp, 0, label_match);
801
    tcg_temp_free_i64(tcg_cmp);
802
    gen_goto_tb(s, 0, s->pc);
803
    gen_set_label(label_match);
804
    gen_goto_tb(s, 1, addr);
805
}
806

    
807
/* C3.2.2 / C5.6.19 Conditional branch (immediate)
808
 *  31           25  24  23                  5   4  3    0
809
 * +---------------+----+---------------------+----+------+
810
 * | 0 1 0 1 0 1 0 | o1 |         imm19       | o0 | cond |
811
 * +---------------+----+---------------------+----+------+
812
 */
813
static void disas_cond_b_imm(DisasContext *s, uint32_t insn)
814
{
815
    unsigned int cond;
816
    uint64_t addr;
817

    
818
    if ((insn & (1 << 4)) || (insn & (1 << 24))) {
819
        unallocated_encoding(s);
820
        return;
821
    }
822
    addr = s->pc + sextract32(insn, 5, 19) * 4 - 4;
823
    cond = extract32(insn, 0, 4);
824

    
825
    if (cond < 0x0e) {
826
        /* genuinely conditional branches */
827
        int label_match = gen_new_label();
828
        arm_gen_test_cc(cond, label_match);
829
        gen_goto_tb(s, 0, s->pc);
830
        gen_set_label(label_match);
831
        gen_goto_tb(s, 1, addr);
832
    } else {
833
        /* 0xe and 0xf are both "always" conditions */
834
        gen_goto_tb(s, 0, addr);
835
    }
836
}
837

    
838
/* C5.6.68 HINT */
839
static void handle_hint(DisasContext *s, uint32_t insn,
840
                        unsigned int op1, unsigned int op2, unsigned int crm)
841
{
842
    unsigned int selector = crm << 3 | op2;
843

    
844
    if (op1 != 3) {
845
        unallocated_encoding(s);
846
        return;
847
    }
848

    
849
    switch (selector) {
850
    case 0: /* NOP */
851
        return;
852
    case 1: /* YIELD */
853
    case 2: /* WFE */
854
    case 3: /* WFI */
855
    case 4: /* SEV */
856
    case 5: /* SEVL */
857
        /* we treat all as NOP at least for now */
858
        return;
859
    default:
860
        /* default specified as NOP equivalent */
861
        return;
862
    }
863
}
864

    
865
static void gen_clrex(DisasContext *s, uint32_t insn)
866
{
867
    tcg_gen_movi_i64(cpu_exclusive_addr, -1);
868
}
869

    
870
/* CLREX, DSB, DMB, ISB */
871
static void handle_sync(DisasContext *s, uint32_t insn,
872
                        unsigned int op1, unsigned int op2, unsigned int crm)
873
{
874
    if (op1 != 3) {
875
        unallocated_encoding(s);
876
        return;
877
    }
878

    
879
    switch (op2) {
880
    case 2: /* CLREX */
881
        gen_clrex(s, insn);
882
        return;
883
    case 4: /* DSB */
884
    case 5: /* DMB */
885
    case 6: /* ISB */
886
        /* We don't emulate caches so barriers are no-ops */
887
        return;
888
    default:
889
        unallocated_encoding(s);
890
        return;
891
    }
892
}
893

    
894
/* C5.6.130 MSR (immediate) - move immediate to processor state field */
895
static void handle_msr_i(DisasContext *s, uint32_t insn,
896
                         unsigned int op1, unsigned int op2, unsigned int crm)
897
{
898
    unsupported_encoding(s, insn);
899
}
900

    
901
static void gen_get_nzcv(TCGv_i64 tcg_rt)
902
{
903
    TCGv_i32 tmp = tcg_temp_new_i32();
904
    TCGv_i32 nzcv = tcg_temp_new_i32();
905

    
906
    /* build bit 31, N */
907
    tcg_gen_andi_i32(nzcv, cpu_NF, (1 << 31));
908
    /* build bit 30, Z */
909
    tcg_gen_setcondi_i32(TCG_COND_EQ, tmp, cpu_ZF, 0);
910
    tcg_gen_deposit_i32(nzcv, nzcv, tmp, 30, 1);
911
    /* build bit 29, C */
912
    tcg_gen_deposit_i32(nzcv, nzcv, cpu_CF, 29, 1);
913
    /* build bit 28, V */
914
    tcg_gen_shri_i32(tmp, cpu_VF, 31);
915
    tcg_gen_deposit_i32(nzcv, nzcv, tmp, 28, 1);
916
    /* generate result */
917
    tcg_gen_extu_i32_i64(tcg_rt, nzcv);
918

    
919
    tcg_temp_free_i32(nzcv);
920
    tcg_temp_free_i32(tmp);
921
}
922

    
923
static void gen_set_nzcv(TCGv_i64 tcg_rt)
924

    
925
{
926
    TCGv_i32 nzcv = tcg_temp_new_i32();
927

    
928
    /* take NZCV from R[t] */
929
    tcg_gen_trunc_i64_i32(nzcv, tcg_rt);
930

    
931
    /* bit 31, N */
932
    tcg_gen_andi_i32(cpu_NF, nzcv, (1 << 31));
933
    /* bit 30, Z */
934
    tcg_gen_andi_i32(cpu_ZF, nzcv, (1 << 30));
935
    tcg_gen_setcondi_i32(TCG_COND_EQ, cpu_ZF, cpu_ZF, 0);
936
    /* bit 29, C */
937
    tcg_gen_andi_i32(cpu_CF, nzcv, (1 << 29));
938
    tcg_gen_shri_i32(cpu_CF, cpu_CF, 29);
939
    /* bit 28, V */
940
    tcg_gen_andi_i32(cpu_VF, nzcv, (1 << 28));
941
    tcg_gen_shli_i32(cpu_VF, cpu_VF, 3);
942
    tcg_temp_free_i32(nzcv);
943
}
944

    
945
/* C5.6.129 MRS - move from system register
946
 * C5.6.131 MSR (register) - move to system register
947
 * C5.6.204 SYS
948
 * C5.6.205 SYSL
949
 * These are all essentially the same insn in 'read' and 'write'
950
 * versions, with varying op0 fields.
951
 */
952
static void handle_sys(DisasContext *s, uint32_t insn, bool isread,
953
                       unsigned int op0, unsigned int op1, unsigned int op2,
954
                       unsigned int crn, unsigned int crm, unsigned int rt)
955
{
956
    const ARMCPRegInfo *ri;
957
    TCGv_i64 tcg_rt;
958

    
959
    ri = get_arm_cp_reginfo(s->cp_regs,
960
                            ENCODE_AA64_CP_REG(CP_REG_ARM64_SYSREG_CP,
961
                                               crn, crm, op0, op1, op2));
962

    
963
    if (!ri) {
964
        /* Unknown register */
965
        unallocated_encoding(s);
966
        return;
967
    }
968

    
969
    /* Check access permissions */
970
    if (!cp_access_ok(s->current_pl, ri, isread)) {
971
        unallocated_encoding(s);
972
        return;
973
    }
974

    
975
    /* Handle special cases first */
976
    switch (ri->type & ~(ARM_CP_FLAG_MASK & ~ARM_CP_SPECIAL)) {
977
    case ARM_CP_NOP:
978
        return;
979
    case ARM_CP_NZCV:
980
        tcg_rt = cpu_reg(s, rt);
981
        if (isread) {
982
            gen_get_nzcv(tcg_rt);
983
        } else {
984
            gen_set_nzcv(tcg_rt);
985
        }
986
        return;
987
    default:
988
        break;
989
    }
990

    
991
    if (use_icount && (ri->type & ARM_CP_IO)) {
992
        gen_io_start();
993
    }
994

    
995
    tcg_rt = cpu_reg(s, rt);
996

    
997
    if (isread) {
998
        if (ri->type & ARM_CP_CONST) {
999
            tcg_gen_movi_i64(tcg_rt, ri->resetvalue);
1000
        } else if (ri->readfn) {
1001
            TCGv_ptr tmpptr;
1002
            gen_a64_set_pc_im(s->pc - 4);
1003
            tmpptr = tcg_const_ptr(ri);
1004
            gen_helper_get_cp_reg64(tcg_rt, cpu_env, tmpptr);
1005
            tcg_temp_free_ptr(tmpptr);
1006
        } else {
1007
            tcg_gen_ld_i64(tcg_rt, cpu_env, ri->fieldoffset);
1008
        }
1009
    } else {
1010
        if (ri->type & ARM_CP_CONST) {
1011
            /* If not forbidden by access permissions, treat as WI */
1012
            return;
1013
        } else if (ri->writefn) {
1014
            TCGv_ptr tmpptr;
1015
            gen_a64_set_pc_im(s->pc - 4);
1016
            tmpptr = tcg_const_ptr(ri);
1017
            gen_helper_set_cp_reg64(cpu_env, tmpptr, tcg_rt);
1018
            tcg_temp_free_ptr(tmpptr);
1019
        } else {
1020
            tcg_gen_st_i64(tcg_rt, cpu_env, ri->fieldoffset);
1021
        }
1022
    }
1023

    
1024
    if (use_icount && (ri->type & ARM_CP_IO)) {
1025
        /* I/O operations must end the TB here (whether read or write) */
1026
        gen_io_end();
1027
        s->is_jmp = DISAS_UPDATE;
1028
    } else if (!isread && !(ri->type & ARM_CP_SUPPRESS_TB_END)) {
1029
        /* We default to ending the TB on a coprocessor register write,
1030
         * but allow this to be suppressed by the register definition
1031
         * (usually only necessary to work around guest bugs).
1032
         */
1033
        s->is_jmp = DISAS_UPDATE;
1034
    }
1035
}
1036

    
1037
/* C3.2.4 System
1038
 *  31                 22 21  20 19 18 16 15   12 11    8 7   5 4    0
1039
 * +---------------------+---+-----+-----+-------+-------+-----+------+
1040
 * | 1 1 0 1 0 1 0 1 0 0 | L | op0 | op1 |  CRn  |  CRm  | op2 |  Rt  |
1041
 * +---------------------+---+-----+-----+-------+-------+-----+------+
1042
 */
1043
static void disas_system(DisasContext *s, uint32_t insn)
1044
{
1045
    unsigned int l, op0, op1, crn, crm, op2, rt;
1046
    l = extract32(insn, 21, 1);
1047
    op0 = extract32(insn, 19, 2);
1048
    op1 = extract32(insn, 16, 3);
1049
    crn = extract32(insn, 12, 4);
1050
    crm = extract32(insn, 8, 4);
1051
    op2 = extract32(insn, 5, 3);
1052
    rt = extract32(insn, 0, 5);
1053

    
1054
    if (op0 == 0) {
1055
        if (l || rt != 31) {
1056
            unallocated_encoding(s);
1057
            return;
1058
        }
1059
        switch (crn) {
1060
        case 2: /* C5.6.68 HINT */
1061
            handle_hint(s, insn, op1, op2, crm);
1062
            break;
1063
        case 3: /* CLREX, DSB, DMB, ISB */
1064
            handle_sync(s, insn, op1, op2, crm);
1065
            break;
1066
        case 4: /* C5.6.130 MSR (immediate) */
1067
            handle_msr_i(s, insn, op1, op2, crm);
1068
            break;
1069
        default:
1070
            unallocated_encoding(s);
1071
            break;
1072
        }
1073
        return;
1074
    }
1075
    handle_sys(s, insn, l, op0, op1, op2, crn, crm, rt);
1076
}
1077

    
1078
/* C3.2.3 Exception generation
1079
 *
1080
 *  31             24 23 21 20                     5 4   2 1  0
1081
 * +-----------------+-----+------------------------+-----+----+
1082
 * | 1 1 0 1 0 1 0 0 | opc |          imm16         | op2 | LL |
1083
 * +-----------------------+------------------------+----------+
1084
 */
1085
static void disas_exc(DisasContext *s, uint32_t insn)
1086
{
1087
    int opc = extract32(insn, 21, 3);
1088
    int op2_ll = extract32(insn, 0, 5);
1089

    
1090
    switch (opc) {
1091
    case 0:
1092
        /* SVC, HVC, SMC; since we don't support the Virtualization
1093
         * or TrustZone extensions these all UNDEF except SVC.
1094
         */
1095
        if (op2_ll != 1) {
1096
            unallocated_encoding(s);
1097
            break;
1098
        }
1099
        gen_exception_insn(s, 0, EXCP_SWI);
1100
        break;
1101
    case 1:
1102
        if (op2_ll != 0) {
1103
            unallocated_encoding(s);
1104
            break;
1105
        }
1106
        /* BRK */
1107
        gen_exception_insn(s, 0, EXCP_BKPT);
1108
        break;
1109
    case 2:
1110
        if (op2_ll != 0) {
1111
            unallocated_encoding(s);
1112
            break;
1113
        }
1114
        /* HLT */
1115
        unsupported_encoding(s, insn);
1116
        break;
1117
    case 5:
1118
        if (op2_ll < 1 || op2_ll > 3) {
1119
            unallocated_encoding(s);
1120
            break;
1121
        }
1122
        /* DCPS1, DCPS2, DCPS3 */
1123
        unsupported_encoding(s, insn);
1124
        break;
1125
    default:
1126
        unallocated_encoding(s);
1127
        break;
1128
    }
1129
}
1130

    
1131
/* C3.2.7 Unconditional branch (register)
1132
 *  31           25 24   21 20   16 15   10 9    5 4     0
1133
 * +---------------+-------+-------+-------+------+-------+
1134
 * | 1 1 0 1 0 1 1 |  opc  |  op2  |  op3  |  Rn  |  op4  |
1135
 * +---------------+-------+-------+-------+------+-------+
1136
 */
1137
static void disas_uncond_b_reg(DisasContext *s, uint32_t insn)
1138
{
1139
    unsigned int opc, op2, op3, rn, op4;
1140

    
1141
    opc = extract32(insn, 21, 4);
1142
    op2 = extract32(insn, 16, 5);
1143
    op3 = extract32(insn, 10, 6);
1144
    rn = extract32(insn, 5, 5);
1145
    op4 = extract32(insn, 0, 5);
1146

    
1147
    if (op4 != 0x0 || op3 != 0x0 || op2 != 0x1f) {
1148
        unallocated_encoding(s);
1149
        return;
1150
    }
1151

    
1152
    switch (opc) {
1153
    case 0: /* BR */
1154
    case 2: /* RET */
1155
        break;
1156
    case 1: /* BLR */
1157
        tcg_gen_movi_i64(cpu_reg(s, 30), s->pc);
1158
        break;
1159
    case 4: /* ERET */
1160
    case 5: /* DRPS */
1161
        if (rn != 0x1f) {
1162
            unallocated_encoding(s);
1163
        } else {
1164
            unsupported_encoding(s, insn);
1165
        }
1166
        return;
1167
    default:
1168
        unallocated_encoding(s);
1169
        return;
1170
    }
1171

    
1172
    tcg_gen_mov_i64(cpu_pc, cpu_reg(s, rn));
1173
    s->is_jmp = DISAS_JUMP;
1174
}
1175

    
1176
/* C3.2 Branches, exception generating and system instructions */
1177
static void disas_b_exc_sys(DisasContext *s, uint32_t insn)
1178
{
1179
    switch (extract32(insn, 25, 7)) {
1180
    case 0x0a: case 0x0b:
1181
    case 0x4a: case 0x4b: /* Unconditional branch (immediate) */
1182
        disas_uncond_b_imm(s, insn);
1183
        break;
1184
    case 0x1a: case 0x5a: /* Compare & branch (immediate) */
1185
        disas_comp_b_imm(s, insn);
1186
        break;
1187
    case 0x1b: case 0x5b: /* Test & branch (immediate) */
1188
        disas_test_b_imm(s, insn);
1189
        break;
1190
    case 0x2a: /* Conditional branch (immediate) */
1191
        disas_cond_b_imm(s, insn);
1192
        break;
1193
    case 0x6a: /* Exception generation / System */
1194
        if (insn & (1 << 24)) {
1195
            disas_system(s, insn);
1196
        } else {
1197
            disas_exc(s, insn);
1198
        }
1199
        break;
1200
    case 0x6b: /* Unconditional branch (register) */
1201
        disas_uncond_b_reg(s, insn);
1202
        break;
1203
    default:
1204
        unallocated_encoding(s);
1205
        break;
1206
    }
1207
}
1208

    
1209
/*
1210
 * Load/Store exclusive instructions are implemented by remembering
1211
 * the value/address loaded, and seeing if these are the same
1212
 * when the store is performed. This is not actually the architecturally
1213
 * mandated semantics, but it works for typical guest code sequences
1214
 * and avoids having to monitor regular stores.
1215
 *
1216
 * In system emulation mode only one CPU will be running at once, so
1217
 * this sequence is effectively atomic.  In user emulation mode we
1218
 * throw an exception and handle the atomic operation elsewhere.
1219
 */
1220
static void gen_load_exclusive(DisasContext *s, int rt, int rt2,
1221
                               TCGv_i64 addr, int size, bool is_pair)
1222
{
1223
    TCGv_i64 tmp = tcg_temp_new_i64();
1224
    TCGMemOp memop = MO_TE + size;
1225

    
1226
    g_assert(size <= 3);
1227
    tcg_gen_qemu_ld_i64(tmp, addr, get_mem_index(s), memop);
1228

    
1229
    if (is_pair) {
1230
        TCGv_i64 addr2 = tcg_temp_new_i64();
1231
        TCGv_i64 hitmp = tcg_temp_new_i64();
1232

    
1233
        g_assert(size >= 2);
1234
        tcg_gen_addi_i64(addr2, addr, 1 << size);
1235
        tcg_gen_qemu_ld_i64(hitmp, addr2, get_mem_index(s), memop);
1236
        tcg_temp_free_i64(addr2);
1237
        tcg_gen_mov_i64(cpu_exclusive_high, hitmp);
1238
        tcg_gen_mov_i64(cpu_reg(s, rt2), hitmp);
1239
        tcg_temp_free_i64(hitmp);
1240
    }
1241

    
1242
    tcg_gen_mov_i64(cpu_exclusive_val, tmp);
1243
    tcg_gen_mov_i64(cpu_reg(s, rt), tmp);
1244

    
1245
    tcg_temp_free_i64(tmp);
1246
    tcg_gen_mov_i64(cpu_exclusive_addr, addr);
1247
}
1248

    
1249
#ifdef CONFIG_USER_ONLY
1250
static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
1251
                                TCGv_i64 addr, int size, int is_pair)
1252
{
1253
    tcg_gen_mov_i64(cpu_exclusive_test, addr);
1254
    tcg_gen_movi_i32(cpu_exclusive_info,
1255
                     size | is_pair << 2 | (rd << 4) | (rt << 9) | (rt2 << 14));
1256
    gen_exception_insn(s, 4, EXCP_STREX);
1257
}
1258
#else
1259
static void gen_store_exclusive(DisasContext *s, int rd, int rt, int rt2,
1260
                                TCGv_i64 addr, int size, int is_pair)
1261
{
1262
    qemu_log_mask(LOG_UNIMP,
1263
                  "%s:%d: system mode store_exclusive unsupported "
1264
                  "at pc=%016" PRIx64 "\n",
1265
                  __FILE__, __LINE__, s->pc - 4);
1266
}
1267
#endif
1268

    
1269
/* C3.3.6 Load/store exclusive
1270
 *
1271
 *  31 30 29         24  23  22   21  20  16  15  14   10 9    5 4    0
1272
 * +-----+-------------+----+---+----+------+----+-------+------+------+
1273
 * | sz  | 0 0 1 0 0 0 | o2 | L | o1 |  Rs  | o0 |  Rt2  |  Rn  | Rt   |
1274
 * +-----+-------------+----+---+----+------+----+-------+------+------+
1275
 *
1276
 *  sz: 00 -> 8 bit, 01 -> 16 bit, 10 -> 32 bit, 11 -> 64 bit
1277
 *   L: 0 -> store, 1 -> load
1278
 *  o2: 0 -> exclusive, 1 -> not
1279
 *  o1: 0 -> single register, 1 -> register pair
1280
 *  o0: 1 -> load-acquire/store-release, 0 -> not
1281
 *
1282
 *  o0 == 0 AND o2 == 1 is un-allocated
1283
 *  o1 == 1 is un-allocated except for 32 and 64 bit sizes
1284
 */
1285
static void disas_ldst_excl(DisasContext *s, uint32_t insn)
1286
{
1287
    int rt = extract32(insn, 0, 5);
1288
    int rn = extract32(insn, 5, 5);
1289
    int rt2 = extract32(insn, 10, 5);
1290
    int is_lasr = extract32(insn, 15, 1);
1291
    int rs = extract32(insn, 16, 5);
1292
    int is_pair = extract32(insn, 21, 1);
1293
    int is_store = !extract32(insn, 22, 1);
1294
    int is_excl = !extract32(insn, 23, 1);
1295
    int size = extract32(insn, 30, 2);
1296
    TCGv_i64 tcg_addr;
1297

    
1298
    if ((!is_excl && !is_lasr) ||
1299
        (is_pair && size < 2)) {
1300
        unallocated_encoding(s);
1301
        return;
1302
    }
1303

    
1304
    if (rn == 31) {
1305
        gen_check_sp_alignment(s);
1306
    }
1307
    tcg_addr = read_cpu_reg_sp(s, rn, 1);
1308

    
1309
    /* Note that since TCG is single threaded load-acquire/store-release
1310
     * semantics require no extra if (is_lasr) { ... } handling.
1311
     */
1312

    
1313
    if (is_excl) {
1314
        if (!is_store) {
1315
            gen_load_exclusive(s, rt, rt2, tcg_addr, size, is_pair);
1316
        } else {
1317
            gen_store_exclusive(s, rs, rt, rt2, tcg_addr, size, is_pair);
1318
        }
1319
    } else {
1320
        TCGv_i64 tcg_rt = cpu_reg(s, rt);
1321
        if (is_store) {
1322
            do_gpr_st(s, tcg_rt, tcg_addr, size);
1323
        } else {
1324
            do_gpr_ld(s, tcg_rt, tcg_addr, size, false, false);
1325
        }
1326
        if (is_pair) {
1327
            TCGv_i64 tcg_rt2 = cpu_reg(s, rt);
1328
            tcg_gen_addi_i64(tcg_addr, tcg_addr, 1 << size);
1329
            if (is_store) {
1330
                do_gpr_st(s, tcg_rt2, tcg_addr, size);
1331
            } else {
1332
                do_gpr_ld(s, tcg_rt2, tcg_addr, size, false, false);
1333
            }
1334
        }
1335
    }
1336
}
1337

    
1338
/*
1339
 * C3.3.5 Load register (literal)
1340
 *
1341
 *  31 30 29   27  26 25 24 23                5 4     0
1342
 * +-----+-------+---+-----+-------------------+-------+
1343
 * | opc | 0 1 1 | V | 0 0 |     imm19         |  Rt   |
1344
 * +-----+-------+---+-----+-------------------+-------+
1345
 *
1346
 * V: 1 -> vector (simd/fp)
1347
 * opc (non-vector): 00 -> 32 bit, 01 -> 64 bit,
1348
 *                   10-> 32 bit signed, 11 -> prefetch
1349
 * opc (vector): 00 -> 32 bit, 01 -> 64 bit, 10 -> 128 bit (11 unallocated)
1350
 */
1351
static void disas_ld_lit(DisasContext *s, uint32_t insn)
1352
{
1353
    int rt = extract32(insn, 0, 5);
1354
    int64_t imm = sextract32(insn, 5, 19) << 2;
1355
    bool is_vector = extract32(insn, 26, 1);
1356
    int opc = extract32(insn, 30, 2);
1357
    bool is_signed = false;
1358
    int size = 2;
1359
    TCGv_i64 tcg_rt, tcg_addr;
1360

    
1361
    if (is_vector) {
1362
        if (opc == 3) {
1363
            unallocated_encoding(s);
1364
            return;
1365
        }
1366
        size = 2 + opc;
1367
    } else {
1368
        if (opc == 3) {
1369
            /* PRFM (literal) : prefetch */
1370
            return;
1371
        }
1372
        size = 2 + extract32(opc, 0, 1);
1373
        is_signed = extract32(opc, 1, 1);
1374
    }
1375

    
1376
    tcg_rt = cpu_reg(s, rt);
1377

    
1378
    tcg_addr = tcg_const_i64((s->pc - 4) + imm);
1379
    if (is_vector) {
1380
        do_fp_ld(s, rt, tcg_addr, size);
1381
    } else {
1382
        do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, false);
1383
    }
1384
    tcg_temp_free_i64(tcg_addr);
1385
}
1386

    
1387
/*
1388
 * C5.6.80 LDNP (Load Pair - non-temporal hint)
1389
 * C5.6.81 LDP (Load Pair - non vector)
1390
 * C5.6.82 LDPSW (Load Pair Signed Word - non vector)
1391
 * C5.6.176 STNP (Store Pair - non-temporal hint)
1392
 * C5.6.177 STP (Store Pair - non vector)
1393
 * C6.3.165 LDNP (Load Pair of SIMD&FP - non-temporal hint)
1394
 * C6.3.165 LDP (Load Pair of SIMD&FP)
1395
 * C6.3.284 STNP (Store Pair of SIMD&FP - non-temporal hint)
1396
 * C6.3.284 STP (Store Pair of SIMD&FP)
1397
 *
1398
 *  31 30 29   27  26  25 24   23  22 21   15 14   10 9    5 4    0
1399
 * +-----+-------+---+---+-------+---+-----------------------------+
1400
 * | opc | 1 0 1 | V | 0 | index | L |  imm7 |  Rt2  |  Rn  | Rt   |
1401
 * +-----+-------+---+---+-------+---+-------+-------+------+------+
1402
 *
1403
 * opc: LDP/STP/LDNP/STNP        00 -> 32 bit, 10 -> 64 bit
1404
 *      LDPSW                    01
1405
 *      LDP/STP/LDNP/STNP (SIMD) 00 -> 32 bit, 01 -> 64 bit, 10 -> 128 bit
1406
 *   V: 0 -> GPR, 1 -> Vector
1407
 * idx: 00 -> signed offset with non-temporal hint, 01 -> post-index,
1408
 *      10 -> signed offset, 11 -> pre-index
1409
 *   L: 0 -> Store 1 -> Load
1410
 *
1411
 * Rt, Rt2 = GPR or SIMD registers to be stored
1412
 * Rn = general purpose register containing address
1413
 * imm7 = signed offset (multiple of 4 or 8 depending on size)
1414
 */
1415
static void disas_ldst_pair(DisasContext *s, uint32_t insn)
1416
{
1417
    int rt = extract32(insn, 0, 5);
1418
    int rn = extract32(insn, 5, 5);
1419
    int rt2 = extract32(insn, 10, 5);
1420
    int64_t offset = sextract32(insn, 15, 7);
1421
    int index = extract32(insn, 23, 2);
1422
    bool is_vector = extract32(insn, 26, 1);
1423
    bool is_load = extract32(insn, 22, 1);
1424
    int opc = extract32(insn, 30, 2);
1425

    
1426
    bool is_signed = false;
1427
    bool postindex = false;
1428
    bool wback = false;
1429

    
1430
    TCGv_i64 tcg_addr; /* calculated address */
1431
    int size;
1432

    
1433
    if (opc == 3) {
1434
        unallocated_encoding(s);
1435
        return;
1436
    }
1437

    
1438
    if (is_vector) {
1439
        size = 2 + opc;
1440
    } else {
1441
        size = 2 + extract32(opc, 1, 1);
1442
        is_signed = extract32(opc, 0, 1);
1443
        if (!is_load && is_signed) {
1444
            unallocated_encoding(s);
1445
            return;
1446
        }
1447
    }
1448

    
1449
    switch (index) {
1450
    case 1: /* post-index */
1451
        postindex = true;
1452
        wback = true;
1453
        break;
1454
    case 0:
1455
        /* signed offset with "non-temporal" hint. Since we don't emulate
1456
         * caches we don't care about hints to the cache system about
1457
         * data access patterns, and handle this identically to plain
1458
         * signed offset.
1459
         */
1460
        if (is_signed) {
1461
            /* There is no non-temporal-hint version of LDPSW */
1462
            unallocated_encoding(s);
1463
            return;
1464
        }
1465
        postindex = false;
1466
        break;
1467
    case 2: /* signed offset, rn not updated */
1468
        postindex = false;
1469
        break;
1470
    case 3: /* pre-index */
1471
        postindex = false;
1472
        wback = true;
1473
        break;
1474
    }
1475

    
1476
    offset <<= size;
1477

    
1478
    if (rn == 31) {
1479
        gen_check_sp_alignment(s);
1480
    }
1481

    
1482
    tcg_addr = read_cpu_reg_sp(s, rn, 1);
1483

    
1484
    if (!postindex) {
1485
        tcg_gen_addi_i64(tcg_addr, tcg_addr, offset);
1486
    }
1487

    
1488
    if (is_vector) {
1489
        if (is_load) {
1490
            do_fp_ld(s, rt, tcg_addr, size);
1491
        } else {
1492
            do_fp_st(s, rt, tcg_addr, size);
1493
        }
1494
    } else {
1495
        TCGv_i64 tcg_rt = cpu_reg(s, rt);
1496
        if (is_load) {
1497
            do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, false);
1498
        } else {
1499
            do_gpr_st(s, tcg_rt, tcg_addr, size);
1500
        }
1501
    }
1502
    tcg_gen_addi_i64(tcg_addr, tcg_addr, 1 << size);
1503
    if (is_vector) {
1504
        if (is_load) {
1505
            do_fp_ld(s, rt2, tcg_addr, size);
1506
        } else {
1507
            do_fp_st(s, rt2, tcg_addr, size);
1508
        }
1509
    } else {
1510
        TCGv_i64 tcg_rt2 = cpu_reg(s, rt2);
1511
        if (is_load) {
1512
            do_gpr_ld(s, tcg_rt2, tcg_addr, size, is_signed, false);
1513
        } else {
1514
            do_gpr_st(s, tcg_rt2, tcg_addr, size);
1515
        }
1516
    }
1517

    
1518
    if (wback) {
1519
        if (postindex) {
1520
            tcg_gen_addi_i64(tcg_addr, tcg_addr, offset - (1 << size));
1521
        } else {
1522
            tcg_gen_subi_i64(tcg_addr, tcg_addr, 1 << size);
1523
        }
1524
        tcg_gen_mov_i64(cpu_reg_sp(s, rn), tcg_addr);
1525
    }
1526
}
1527

    
1528
/*
1529
 * C3.3.8 Load/store (immediate post-indexed)
1530
 * C3.3.9 Load/store (immediate pre-indexed)
1531
 * C3.3.12 Load/store (unscaled immediate)
1532
 *
1533
 * 31 30 29   27  26 25 24 23 22 21  20    12 11 10 9    5 4    0
1534
 * +----+-------+---+-----+-----+---+--------+-----+------+------+
1535
 * |size| 1 1 1 | V | 0 0 | opc | 0 |  imm9  | idx |  Rn  |  Rt  |
1536
 * +----+-------+---+-----+-----+---+--------+-----+------+------+
1537
 *
1538
 * idx = 01 -> post-indexed, 11 pre-indexed, 00 unscaled imm. (no writeback)
1539
 * V = 0 -> non-vector
1540
 * size: 00 -> 8 bit, 01 -> 16 bit, 10 -> 32 bit, 11 -> 64bit
1541
 * opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
1542
 */
1543
static void disas_ldst_reg_imm9(DisasContext *s, uint32_t insn)
1544
{
1545
    int rt = extract32(insn, 0, 5);
1546
    int rn = extract32(insn, 5, 5);
1547
    int imm9 = sextract32(insn, 12, 9);
1548
    int opc = extract32(insn, 22, 2);
1549
    int size = extract32(insn, 30, 2);
1550
    int idx = extract32(insn, 10, 2);
1551
    bool is_signed = false;
1552
    bool is_store = false;
1553
    bool is_extended = false;
1554
    bool is_vector = extract32(insn, 26, 1);
1555
    bool post_index;
1556
    bool writeback;
1557

    
1558
    TCGv_i64 tcg_addr;
1559

    
1560
    if (is_vector) {
1561
        size |= (opc & 2) << 1;
1562
        if (size > 4) {
1563
            unallocated_encoding(s);
1564
            return;
1565
        }
1566
        is_store = ((opc & 1) == 0);
1567
    } else {
1568
        if (size == 3 && opc == 2) {
1569
            /* PRFM - prefetch */
1570
            return;
1571
        }
1572
        if (opc == 3 && size > 1) {
1573
            unallocated_encoding(s);
1574
            return;
1575
        }
1576
        is_store = (opc == 0);
1577
        is_signed = opc & (1<<1);
1578
        is_extended = (size < 3) && (opc & 1);
1579
    }
1580

    
1581
    switch (idx) {
1582
    case 0:
1583
        post_index = false;
1584
        writeback = false;
1585
        break;
1586
    case 1:
1587
        post_index = true;
1588
        writeback = true;
1589
        break;
1590
    case 3:
1591
        post_index = false;
1592
        writeback = true;
1593
        break;
1594
    case 2:
1595
        g_assert(false);
1596
        break;
1597
    }
1598

    
1599
    if (rn == 31) {
1600
        gen_check_sp_alignment(s);
1601
    }
1602
    tcg_addr = read_cpu_reg_sp(s, rn, 1);
1603

    
1604
    if (!post_index) {
1605
        tcg_gen_addi_i64(tcg_addr, tcg_addr, imm9);
1606
    }
1607

    
1608
    if (is_vector) {
1609
        if (is_store) {
1610
            do_fp_st(s, rt, tcg_addr, size);
1611
        } else {
1612
            do_fp_ld(s, rt, tcg_addr, size);
1613
        }
1614
    } else {
1615
        TCGv_i64 tcg_rt = cpu_reg(s, rt);
1616
        if (is_store) {
1617
            do_gpr_st(s, tcg_rt, tcg_addr, size);
1618
        } else {
1619
            do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended);
1620
        }
1621
    }
1622

    
1623
    if (writeback) {
1624
        TCGv_i64 tcg_rn = cpu_reg_sp(s, rn);
1625
        if (post_index) {
1626
            tcg_gen_addi_i64(tcg_addr, tcg_addr, imm9);
1627
        }
1628
        tcg_gen_mov_i64(tcg_rn, tcg_addr);
1629
    }
1630
}
1631

    
1632
/*
1633
 * C3.3.10 Load/store (register offset)
1634
 *
1635
 * 31 30 29   27  26 25 24 23 22 21  20  16 15 13 12 11 10 9  5 4  0
1636
 * +----+-------+---+-----+-----+---+------+-----+--+-----+----+----+
1637
 * |size| 1 1 1 | V | 0 0 | opc | 1 |  Rm  | opt | S| 1 0 | Rn | Rt |
1638
 * +----+-------+---+-----+-----+---+------+-----+--+-----+----+----+
1639
 *
1640
 * For non-vector:
1641
 *   size: 00-> byte, 01 -> 16 bit, 10 -> 32bit, 11 -> 64bit
1642
 *   opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
1643
 * For vector:
1644
 *   size is opc<1>:size<1:0> so 100 -> 128 bit; 110 and 111 unallocated
1645
 *   opc<0>: 0 -> store, 1 -> load
1646
 * V: 1 -> vector/simd
1647
 * opt: extend encoding (see DecodeRegExtend)
1648
 * S: if S=1 then scale (essentially index by sizeof(size))
1649
 * Rt: register to transfer into/out of
1650
 * Rn: address register or SP for base
1651
 * Rm: offset register or ZR for offset
1652
 */
1653
static void disas_ldst_reg_roffset(DisasContext *s, uint32_t insn)
1654
{
1655
    int rt = extract32(insn, 0, 5);
1656
    int rn = extract32(insn, 5, 5);
1657
    int shift = extract32(insn, 12, 1);
1658
    int rm = extract32(insn, 16, 5);
1659
    int opc = extract32(insn, 22, 2);
1660
    int opt = extract32(insn, 13, 3);
1661
    int size = extract32(insn, 30, 2);
1662
    bool is_signed = false;
1663
    bool is_store = false;
1664
    bool is_extended = false;
1665
    bool is_vector = extract32(insn, 26, 1);
1666

    
1667
    TCGv_i64 tcg_rm;
1668
    TCGv_i64 tcg_addr;
1669

    
1670
    if (extract32(opt, 1, 1) == 0) {
1671
        unallocated_encoding(s);
1672
        return;
1673
    }
1674

    
1675
    if (is_vector) {
1676
        size |= (opc & 2) << 1;
1677
        if (size > 4) {
1678
            unallocated_encoding(s);
1679
            return;
1680
        }
1681
        is_store = !extract32(opc, 0, 1);
1682
    } else {
1683
        if (size == 3 && opc == 2) {
1684
            /* PRFM - prefetch */
1685
            return;
1686
        }
1687
        if (opc == 3 && size > 1) {
1688
            unallocated_encoding(s);
1689
            return;
1690
        }
1691
        is_store = (opc == 0);
1692
        is_signed = extract32(opc, 1, 1);
1693
        is_extended = (size < 3) && extract32(opc, 0, 1);
1694
    }
1695

    
1696
    if (rn == 31) {
1697
        gen_check_sp_alignment(s);
1698
    }
1699
    tcg_addr = read_cpu_reg_sp(s, rn, 1);
1700

    
1701
    tcg_rm = read_cpu_reg(s, rm, 1);
1702
    ext_and_shift_reg(tcg_rm, tcg_rm, opt, shift ? size : 0);
1703

    
1704
    tcg_gen_add_i64(tcg_addr, tcg_addr, tcg_rm);
1705

    
1706
    if (is_vector) {
1707
        if (is_store) {
1708
            do_fp_st(s, rt, tcg_addr, size);
1709
        } else {
1710
            do_fp_ld(s, rt, tcg_addr, size);
1711
        }
1712
    } else {
1713
        TCGv_i64 tcg_rt = cpu_reg(s, rt);
1714
        if (is_store) {
1715
            do_gpr_st(s, tcg_rt, tcg_addr, size);
1716
        } else {
1717
            do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended);
1718
        }
1719
    }
1720
}
1721

    
1722
/*
1723
 * C3.3.13 Load/store (unsigned immediate)
1724
 *
1725
 * 31 30 29   27  26 25 24 23 22 21        10 9     5
1726
 * +----+-------+---+-----+-----+------------+-------+------+
1727
 * |size| 1 1 1 | V | 0 1 | opc |   imm12    |  Rn   |  Rt  |
1728
 * +----+-------+---+-----+-----+------------+-------+------+
1729
 *
1730
 * For non-vector:
1731
 *   size: 00-> byte, 01 -> 16 bit, 10 -> 32bit, 11 -> 64bit
1732
 *   opc: 00 -> store, 01 -> loadu, 10 -> loads 64, 11 -> loads 32
1733
 * For vector:
1734
 *   size is opc<1>:size<1:0> so 100 -> 128 bit; 110 and 111 unallocated
1735
 *   opc<0>: 0 -> store, 1 -> load
1736
 * Rn: base address register (inc SP)
1737
 * Rt: target register
1738
 */
1739
static void disas_ldst_reg_unsigned_imm(DisasContext *s, uint32_t insn)
1740
{
1741
    int rt = extract32(insn, 0, 5);
1742
    int rn = extract32(insn, 5, 5);
1743
    unsigned int imm12 = extract32(insn, 10, 12);
1744
    bool is_vector = extract32(insn, 26, 1);
1745
    int size = extract32(insn, 30, 2);
1746
    int opc = extract32(insn, 22, 2);
1747
    unsigned int offset;
1748

    
1749
    TCGv_i64 tcg_addr;
1750

    
1751
    bool is_store;
1752
    bool is_signed = false;
1753
    bool is_extended = false;
1754

    
1755
    if (is_vector) {
1756
        size |= (opc & 2) << 1;
1757
        if (size > 4) {
1758
            unallocated_encoding(s);
1759
            return;
1760
        }
1761
        is_store = !extract32(opc, 0, 1);
1762
    } else {
1763
        if (size == 3 && opc == 2) {
1764
            /* PRFM - prefetch */
1765
            return;
1766
        }
1767
        if (opc == 3 && size > 1) {
1768
            unallocated_encoding(s);
1769
            return;
1770
        }
1771
        is_store = (opc == 0);
1772
        is_signed = extract32(opc, 1, 1);
1773
        is_extended = (size < 3) && extract32(opc, 0, 1);
1774
    }
1775

    
1776
    if (rn == 31) {
1777
        gen_check_sp_alignment(s);
1778
    }
1779
    tcg_addr = read_cpu_reg_sp(s, rn, 1);
1780
    offset = imm12 << size;
1781
    tcg_gen_addi_i64(tcg_addr, tcg_addr, offset);
1782

    
1783
    if (is_vector) {
1784
        if (is_store) {
1785
            do_fp_st(s, rt, tcg_addr, size);
1786
        } else {
1787
            do_fp_ld(s, rt, tcg_addr, size);
1788
        }
1789
    } else {
1790
        TCGv_i64 tcg_rt = cpu_reg(s, rt);
1791
        if (is_store) {
1792
            do_gpr_st(s, tcg_rt, tcg_addr, size);
1793
        } else {
1794
            do_gpr_ld(s, tcg_rt, tcg_addr, size, is_signed, is_extended);
1795
        }
1796
    }
1797
}
1798

    
1799
/* Load/store register (immediate forms) */
1800
static void disas_ldst_reg_imm(DisasContext *s, uint32_t insn)
1801
{
1802
    switch (extract32(insn, 10, 2)) {
1803
    case 0: case 1: case 3:
1804
        /* Load/store register (unscaled immediate) */
1805
        /* Load/store immediate pre/post-indexed */
1806
        disas_ldst_reg_imm9(s, insn);
1807
        break;
1808
    case 2:
1809
        /* Load/store register unprivileged */
1810
        unsupported_encoding(s, insn);
1811
        break;
1812
    default:
1813
        unallocated_encoding(s);
1814
        break;
1815
    }
1816
}
1817

    
1818
/* Load/store register (all forms) */
1819
static void disas_ldst_reg(DisasContext *s, uint32_t insn)
1820
{
1821
    switch (extract32(insn, 24, 2)) {
1822
    case 0:
1823
        if (extract32(insn, 21, 1) == 1 && extract32(insn, 10, 2) == 2) {
1824
            disas_ldst_reg_roffset(s, insn);
1825
        } else {
1826
            disas_ldst_reg_imm(s, insn);
1827
        }
1828
        break;
1829
    case 1:
1830
        disas_ldst_reg_unsigned_imm(s, insn);
1831
        break;
1832
    default:
1833
        unallocated_encoding(s);
1834
        break;
1835
    }
1836
}
1837

    
1838
/* AdvSIMD load/store multiple structures */
1839
static void disas_ldst_multiple_struct(DisasContext *s, uint32_t insn)
1840
{
1841
    unsupported_encoding(s, insn);
1842
}
1843

    
1844
/* AdvSIMD load/store single structure */
1845
static void disas_ldst_single_struct(DisasContext *s, uint32_t insn)
1846
{
1847
    unsupported_encoding(s, insn);
1848
}
1849

    
1850
/* C3.3 Loads and stores */
1851
static void disas_ldst(DisasContext *s, uint32_t insn)
1852
{
1853
    switch (extract32(insn, 24, 6)) {
1854
    case 0x08: /* Load/store exclusive */
1855
        disas_ldst_excl(s, insn);
1856
        break;
1857
    case 0x18: case 0x1c: /* Load register (literal) */
1858
        disas_ld_lit(s, insn);
1859
        break;
1860
    case 0x28: case 0x29:
1861
    case 0x2c: case 0x2d: /* Load/store pair (all forms) */
1862
        disas_ldst_pair(s, insn);
1863
        break;
1864
    case 0x38: case 0x39:
1865
    case 0x3c: case 0x3d: /* Load/store register (all forms) */
1866
        disas_ldst_reg(s, insn);
1867
        break;
1868
    case 0x0c: /* AdvSIMD load/store multiple structures */
1869
        disas_ldst_multiple_struct(s, insn);
1870
        break;
1871
    case 0x0d: /* AdvSIMD load/store single structure */
1872
        disas_ldst_single_struct(s, insn);
1873
        break;
1874
    default:
1875
        unallocated_encoding(s);
1876
        break;
1877
    }
1878
}
1879

    
1880
/* C3.4.6 PC-rel. addressing
1881
 *   31  30   29 28       24 23                5 4    0
1882
 * +----+-------+-----------+-------------------+------+
1883
 * | op | immlo | 1 0 0 0 0 |       immhi       |  Rd  |
1884
 * +----+-------+-----------+-------------------+------+
1885
 */
1886
static void disas_pc_rel_adr(DisasContext *s, uint32_t insn)
1887
{
1888
    unsigned int page, rd;
1889
    uint64_t base;
1890
    int64_t offset;
1891

    
1892
    page = extract32(insn, 31, 1);
1893
    /* SignExtend(immhi:immlo) -> offset */
1894
    offset = ((int64_t)sextract32(insn, 5, 19) << 2) | extract32(insn, 29, 2);
1895
    rd = extract32(insn, 0, 5);
1896
    base = s->pc - 4;
1897

    
1898
    if (page) {
1899
        /* ADRP (page based) */
1900
        base &= ~0xfff;
1901
        offset <<= 12;
1902
    }
1903

    
1904
    tcg_gen_movi_i64(cpu_reg(s, rd), base + offset);
1905
}
1906

    
1907
/*
1908
 * C3.4.1 Add/subtract (immediate)
1909
 *
1910
 *  31 30 29 28       24 23 22 21         10 9   5 4   0
1911
 * +--+--+--+-----------+-----+-------------+-----+-----+
1912
 * |sf|op| S| 1 0 0 0 1 |shift|    imm12    |  Rn | Rd  |
1913
 * +--+--+--+-----------+-----+-------------+-----+-----+
1914
 *
1915
 *    sf: 0 -> 32bit, 1 -> 64bit
1916
 *    op: 0 -> add  , 1 -> sub
1917
 *     S: 1 -> set flags
1918
 * shift: 00 -> LSL imm by 0, 01 -> LSL imm by 12
1919
 */
1920
static void disas_add_sub_imm(DisasContext *s, uint32_t insn)
1921
{
1922
    int rd = extract32(insn, 0, 5);
1923
    int rn = extract32(insn, 5, 5);
1924
    uint64_t imm = extract32(insn, 10, 12);
1925
    int shift = extract32(insn, 22, 2);
1926
    bool setflags = extract32(insn, 29, 1);
1927
    bool sub_op = extract32(insn, 30, 1);
1928
    bool is_64bit = extract32(insn, 31, 1);
1929

    
1930
    TCGv_i64 tcg_rn = cpu_reg_sp(s, rn);
1931
    TCGv_i64 tcg_rd = setflags ? cpu_reg(s, rd) : cpu_reg_sp(s, rd);
1932
    TCGv_i64 tcg_result;
1933

    
1934
    switch (shift) {
1935
    case 0x0:
1936
        break;
1937
    case 0x1:
1938
        imm <<= 12;
1939
        break;
1940
    default:
1941
        unallocated_encoding(s);
1942
        return;
1943
    }
1944

    
1945
    tcg_result = tcg_temp_new_i64();
1946
    if (!setflags) {
1947
        if (sub_op) {
1948
            tcg_gen_subi_i64(tcg_result, tcg_rn, imm);
1949
        } else {
1950
            tcg_gen_addi_i64(tcg_result, tcg_rn, imm);
1951
        }
1952
    } else {
1953
        TCGv_i64 tcg_imm = tcg_const_i64(imm);
1954
        if (sub_op) {
1955
            gen_sub_CC(is_64bit, tcg_result, tcg_rn, tcg_imm);
1956
        } else {
1957
            gen_add_CC(is_64bit, tcg_result, tcg_rn, tcg_imm);
1958
        }
1959
        tcg_temp_free_i64(tcg_imm);
1960
    }
1961

    
1962
    if (is_64bit) {
1963
        tcg_gen_mov_i64(tcg_rd, tcg_result);
1964
    } else {
1965
        tcg_gen_ext32u_i64(tcg_rd, tcg_result);
1966
    }
1967

    
1968
    tcg_temp_free_i64(tcg_result);
1969
}
1970

    
1971
/* The input should be a value in the bottom e bits (with higher
1972
 * bits zero); returns that value replicated into every element
1973
 * of size e in a 64 bit integer.
1974
 */
1975
static uint64_t bitfield_replicate(uint64_t mask, unsigned int e)
1976
{
1977
    assert(e != 0);
1978
    while (e < 64) {
1979
        mask |= mask << e;
1980
        e *= 2;
1981
    }
1982
    return mask;
1983
}
1984

    
1985
/* Return a value with the bottom len bits set (where 0 < len <= 64) */
1986
static inline uint64_t bitmask64(unsigned int length)
1987
{
1988
    assert(length > 0 && length <= 64);
1989
    return ~0ULL >> (64 - length);
1990
}
1991

    
1992
/* Simplified variant of pseudocode DecodeBitMasks() for the case where we
1993
 * only require the wmask. Returns false if the imms/immr/immn are a reserved
1994
 * value (ie should cause a guest UNDEF exception), and true if they are
1995
 * valid, in which case the decoded bit pattern is written to result.
1996
 */
1997
static bool logic_imm_decode_wmask(uint64_t *result, unsigned int immn,
1998
                                   unsigned int imms, unsigned int immr)
1999
{
2000
    uint64_t mask;
2001
    unsigned e, levels, s, r;
2002
    int len;
2003

    
2004
    assert(immn < 2 && imms < 64 && immr < 64);
2005

    
2006
    /* The bit patterns we create here are 64 bit patterns which
2007
     * are vectors of identical elements of size e = 2, 4, 8, 16, 32 or
2008
     * 64 bits each. Each element contains the same value: a run
2009
     * of between 1 and e-1 non-zero bits, rotated within the
2010
     * element by between 0 and e-1 bits.
2011
     *
2012
     * The element size and run length are encoded into immn (1 bit)
2013
     * and imms (6 bits) as follows:
2014
     * 64 bit elements: immn = 1, imms = <length of run - 1>
2015
     * 32 bit elements: immn = 0, imms = 0 : <length of run - 1>
2016
     * 16 bit elements: immn = 0, imms = 10 : <length of run - 1>
2017
     *  8 bit elements: immn = 0, imms = 110 : <length of run - 1>
2018
     *  4 bit elements: immn = 0, imms = 1110 : <length of run - 1>
2019
     *  2 bit elements: immn = 0, imms = 11110 : <length of run - 1>
2020
     * Notice that immn = 0, imms = 11111x is the only combination
2021
     * not covered by one of the above options; this is reserved.
2022
     * Further, <length of run - 1> all-ones is a reserved pattern.
2023
     *
2024
     * In all cases the rotation is by immr % e (and immr is 6 bits).
2025
     */
2026

    
2027
    /* First determine the element size */
2028
    len = 31 - clz32((immn << 6) | (~imms & 0x3f));
2029
    if (len < 1) {
2030
        /* This is the immn == 0, imms == 0x11111x case */
2031
        return false;
2032
    }
2033
    e = 1 << len;
2034

    
2035
    levels = e - 1;
2036
    s = imms & levels;
2037
    r = immr & levels;
2038

    
2039
    if (s == levels) {
2040
        /* <length of run - 1> mustn't be all-ones. */
2041
        return false;
2042
    }
2043

    
2044
    /* Create the value of one element: s+1 set bits rotated
2045
     * by r within the element (which is e bits wide)...
2046
     */
2047
    mask = bitmask64(s + 1);
2048
    mask = (mask >> r) | (mask << (e - r));
2049
    /* ...then replicate the element over the whole 64 bit value */
2050
    mask = bitfield_replicate(mask, e);
2051
    *result = mask;
2052
    return true;
2053
}
2054

    
2055
/* C3.4.4 Logical (immediate)
2056
 *   31  30 29 28         23 22  21  16 15  10 9    5 4    0
2057
 * +----+-----+-------------+---+------+------+------+------+
2058
 * | sf | opc | 1 0 0 1 0 0 | N | immr | imms |  Rn  |  Rd  |
2059
 * +----+-----+-------------+---+------+------+------+------+
2060
 */
2061
static void disas_logic_imm(DisasContext *s, uint32_t insn)
2062
{
2063
    unsigned int sf, opc, is_n, immr, imms, rn, rd;
2064
    TCGv_i64 tcg_rd, tcg_rn;
2065
    uint64_t wmask;
2066
    bool is_and = false;
2067

    
2068
    sf = extract32(insn, 31, 1);
2069
    opc = extract32(insn, 29, 2);
2070
    is_n = extract32(insn, 22, 1);
2071
    immr = extract32(insn, 16, 6);
2072
    imms = extract32(insn, 10, 6);
2073
    rn = extract32(insn, 5, 5);
2074
    rd = extract32(insn, 0, 5);
2075

    
2076
    if (!sf && is_n) {
2077
        unallocated_encoding(s);
2078
        return;
2079
    }
2080

    
2081
    if (opc == 0x3) { /* ANDS */
2082
        tcg_rd = cpu_reg(s, rd);
2083
    } else {
2084
        tcg_rd = cpu_reg_sp(s, rd);
2085
    }
2086
    tcg_rn = cpu_reg(s, rn);
2087

    
2088
    if (!logic_imm_decode_wmask(&wmask, is_n, imms, immr)) {
2089
        /* some immediate field values are reserved */
2090
        unallocated_encoding(s);
2091
        return;
2092
    }
2093

    
2094
    if (!sf) {
2095
        wmask &= 0xffffffff;
2096
    }
2097

    
2098
    switch (opc) {
2099
    case 0x3: /* ANDS */
2100
    case 0x0: /* AND */
2101
        tcg_gen_andi_i64(tcg_rd, tcg_rn, wmask);
2102
        is_and = true;
2103
        break;
2104
    case 0x1: /* ORR */
2105
        tcg_gen_ori_i64(tcg_rd, tcg_rn, wmask);
2106
        break;
2107
    case 0x2: /* EOR */
2108
        tcg_gen_xori_i64(tcg_rd, tcg_rn, wmask);
2109
        break;
2110
    default:
2111
        assert(FALSE); /* must handle all above */
2112
        break;
2113
    }
2114

    
2115
    if (!sf && !is_and) {
2116
        /* zero extend final result; we know we can skip this for AND
2117
         * since the immediate had the high 32 bits clear.
2118
         */
2119
        tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2120
    }
2121

    
2122
    if (opc == 3) { /* ANDS */
2123
        gen_logic_CC(sf, tcg_rd);
2124
    }
2125
}
2126

    
2127
/*
2128
 * C3.4.5 Move wide (immediate)
2129
 *
2130
 *  31 30 29 28         23 22 21 20             5 4    0
2131
 * +--+-----+-------------+-----+----------------+------+
2132
 * |sf| opc | 1 0 0 1 0 1 |  hw |  imm16         |  Rd  |
2133
 * +--+-----+-------------+-----+----------------+------+
2134
 *
2135
 * sf: 0 -> 32 bit, 1 -> 64 bit
2136
 * opc: 00 -> N, 10 -> Z, 11 -> K
2137
 * hw: shift/16 (0,16, and sf only 32, 48)
2138
 */
2139
static void disas_movw_imm(DisasContext *s, uint32_t insn)
2140
{
2141
    int rd = extract32(insn, 0, 5);
2142
    uint64_t imm = extract32(insn, 5, 16);
2143
    int sf = extract32(insn, 31, 1);
2144
    int opc = extract32(insn, 29, 2);
2145
    int pos = extract32(insn, 21, 2) << 4;
2146
    TCGv_i64 tcg_rd = cpu_reg(s, rd);
2147
    TCGv_i64 tcg_imm;
2148

    
2149
    if (!sf && (pos >= 32)) {
2150
        unallocated_encoding(s);
2151
        return;
2152
    }
2153

    
2154
    switch (opc) {
2155
    case 0: /* MOVN */
2156
    case 2: /* MOVZ */
2157
        imm <<= pos;
2158
        if (opc == 0) {
2159
            imm = ~imm;
2160
        }
2161
        if (!sf) {
2162
            imm &= 0xffffffffu;
2163
        }
2164
        tcg_gen_movi_i64(tcg_rd, imm);
2165
        break;
2166
    case 3: /* MOVK */
2167
        tcg_imm = tcg_const_i64(imm);
2168
        tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_imm, pos, 16);
2169
        tcg_temp_free_i64(tcg_imm);
2170
        if (!sf) {
2171
            tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2172
        }
2173
        break;
2174
    default:
2175
        unallocated_encoding(s);
2176
        break;
2177
    }
2178
}
2179

    
2180
/* C3.4.2 Bitfield
2181
 *   31  30 29 28         23 22  21  16 15  10 9    5 4    0
2182
 * +----+-----+-------------+---+------+------+------+------+
2183
 * | sf | opc | 1 0 0 1 1 0 | N | immr | imms |  Rn  |  Rd  |
2184
 * +----+-----+-------------+---+------+------+------+------+
2185
 */
2186
static void disas_bitfield(DisasContext *s, uint32_t insn)
2187
{
2188
    unsigned int sf, n, opc, ri, si, rn, rd, bitsize, pos, len;
2189
    TCGv_i64 tcg_rd, tcg_tmp;
2190

    
2191
    sf = extract32(insn, 31, 1);
2192
    opc = extract32(insn, 29, 2);
2193
    n = extract32(insn, 22, 1);
2194
    ri = extract32(insn, 16, 6);
2195
    si = extract32(insn, 10, 6);
2196
    rn = extract32(insn, 5, 5);
2197
    rd = extract32(insn, 0, 5);
2198
    bitsize = sf ? 64 : 32;
2199

    
2200
    if (sf != n || ri >= bitsize || si >= bitsize || opc > 2) {
2201
        unallocated_encoding(s);
2202
        return;
2203
    }
2204

    
2205
    tcg_rd = cpu_reg(s, rd);
2206
    tcg_tmp = read_cpu_reg(s, rn, sf);
2207

    
2208
    /* OPTME: probably worth recognizing common cases of ext{8,16,32}{u,s} */
2209

    
2210
    if (opc != 1) { /* SBFM or UBFM */
2211
        tcg_gen_movi_i64(tcg_rd, 0);
2212
    }
2213

    
2214
    /* do the bit move operation */
2215
    if (si >= ri) {
2216
        /* Wd<s-r:0> = Wn<s:r> */
2217
        tcg_gen_shri_i64(tcg_tmp, tcg_tmp, ri);
2218
        pos = 0;
2219
        len = (si - ri) + 1;
2220
    } else {
2221
        /* Wd<32+s-r,32-r> = Wn<s:0> */
2222
        pos = bitsize - ri;
2223
        len = si + 1;
2224
    }
2225

    
2226
    tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, pos, len);
2227

    
2228
    if (opc == 0) { /* SBFM - sign extend the destination field */
2229
        tcg_gen_shli_i64(tcg_rd, tcg_rd, 64 - (pos + len));
2230
        tcg_gen_sari_i64(tcg_rd, tcg_rd, 64 - (pos + len));
2231
    }
2232

    
2233
    if (!sf) { /* zero extend final result */
2234
        tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2235
    }
2236
}
2237

    
2238
/* C3.4.3 Extract
2239
 *   31  30  29 28         23 22   21  20  16 15    10 9    5 4    0
2240
 * +----+------+-------------+---+----+------+--------+------+------+
2241
 * | sf | op21 | 1 0 0 1 1 1 | N | o0 |  Rm  |  imms  |  Rn  |  Rd  |
2242
 * +----+------+-------------+---+----+------+--------+------+------+
2243
 */
2244
static void disas_extract(DisasContext *s, uint32_t insn)
2245
{
2246
    unsigned int sf, n, rm, imm, rn, rd, bitsize, op21, op0;
2247

    
2248
    sf = extract32(insn, 31, 1);
2249
    n = extract32(insn, 22, 1);
2250
    rm = extract32(insn, 16, 5);
2251
    imm = extract32(insn, 10, 6);
2252
    rn = extract32(insn, 5, 5);
2253
    rd = extract32(insn, 0, 5);
2254
    op21 = extract32(insn, 29, 2);
2255
    op0 = extract32(insn, 21, 1);
2256
    bitsize = sf ? 64 : 32;
2257

    
2258
    if (sf != n || op21 || op0 || imm >= bitsize) {
2259
        unallocated_encoding(s);
2260
    } else {
2261
        TCGv_i64 tcg_rd, tcg_rm, tcg_rn;
2262

    
2263
        tcg_rd = cpu_reg(s, rd);
2264

    
2265
        if (imm) {
2266
            /* OPTME: we can special case rm==rn as a rotate */
2267
            tcg_rm = read_cpu_reg(s, rm, sf);
2268
            tcg_rn = read_cpu_reg(s, rn, sf);
2269
            tcg_gen_shri_i64(tcg_rm, tcg_rm, imm);
2270
            tcg_gen_shli_i64(tcg_rn, tcg_rn, bitsize - imm);
2271
            tcg_gen_or_i64(tcg_rd, tcg_rm, tcg_rn);
2272
            if (!sf) {
2273
                tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2274
            }
2275
        } else {
2276
            /* tcg shl_i32/shl_i64 is undefined for 32/64 bit shifts,
2277
             * so an extract from bit 0 is a special case.
2278
             */
2279
            if (sf) {
2280
                tcg_gen_mov_i64(tcg_rd, cpu_reg(s, rm));
2281
            } else {
2282
                tcg_gen_ext32u_i64(tcg_rd, cpu_reg(s, rm));
2283
            }
2284
        }
2285

    
2286
    }
2287
}
2288

    
2289
/* C3.4 Data processing - immediate */
2290
static void disas_data_proc_imm(DisasContext *s, uint32_t insn)
2291
{
2292
    switch (extract32(insn, 23, 6)) {
2293
    case 0x20: case 0x21: /* PC-rel. addressing */
2294
        disas_pc_rel_adr(s, insn);
2295
        break;
2296
    case 0x22: case 0x23: /* Add/subtract (immediate) */
2297
        disas_add_sub_imm(s, insn);
2298
        break;
2299
    case 0x24: /* Logical (immediate) */
2300
        disas_logic_imm(s, insn);
2301
        break;
2302
    case 0x25: /* Move wide (immediate) */
2303
        disas_movw_imm(s, insn);
2304
        break;
2305
    case 0x26: /* Bitfield */
2306
        disas_bitfield(s, insn);
2307
        break;
2308
    case 0x27: /* Extract */
2309
        disas_extract(s, insn);
2310
        break;
2311
    default:
2312
        unallocated_encoding(s);
2313
        break;
2314
    }
2315
}
2316

    
2317
/* Shift a TCGv src by TCGv shift_amount, put result in dst.
2318
 * Note that it is the caller's responsibility to ensure that the
2319
 * shift amount is in range (ie 0..31 or 0..63) and provide the ARM
2320
 * mandated semantics for out of range shifts.
2321
 */
2322
static void shift_reg(TCGv_i64 dst, TCGv_i64 src, int sf,
2323
                      enum a64_shift_type shift_type, TCGv_i64 shift_amount)
2324
{
2325
    switch (shift_type) {
2326
    case A64_SHIFT_TYPE_LSL:
2327
        tcg_gen_shl_i64(dst, src, shift_amount);
2328
        break;
2329
    case A64_SHIFT_TYPE_LSR:
2330
        tcg_gen_shr_i64(dst, src, shift_amount);
2331
        break;
2332
    case A64_SHIFT_TYPE_ASR:
2333
        if (!sf) {
2334
            tcg_gen_ext32s_i64(dst, src);
2335
        }
2336
        tcg_gen_sar_i64(dst, sf ? src : dst, shift_amount);
2337
        break;
2338
    case A64_SHIFT_TYPE_ROR:
2339
        if (sf) {
2340
            tcg_gen_rotr_i64(dst, src, shift_amount);
2341
        } else {
2342
            TCGv_i32 t0, t1;
2343
            t0 = tcg_temp_new_i32();
2344
            t1 = tcg_temp_new_i32();
2345
            tcg_gen_trunc_i64_i32(t0, src);
2346
            tcg_gen_trunc_i64_i32(t1, shift_amount);
2347
            tcg_gen_rotr_i32(t0, t0, t1);
2348
            tcg_gen_extu_i32_i64(dst, t0);
2349
            tcg_temp_free_i32(t0);
2350
            tcg_temp_free_i32(t1);
2351
        }
2352
        break;
2353
    default:
2354
        assert(FALSE); /* all shift types should be handled */
2355
        break;
2356
    }
2357

    
2358
    if (!sf) { /* zero extend final result */
2359
        tcg_gen_ext32u_i64(dst, dst);
2360
    }
2361
}
2362

    
2363
/* Shift a TCGv src by immediate, put result in dst.
2364
 * The shift amount must be in range (this should always be true as the
2365
 * relevant instructions will UNDEF on bad shift immediates).
2366
 */
2367
static void shift_reg_imm(TCGv_i64 dst, TCGv_i64 src, int sf,
2368
                          enum a64_shift_type shift_type, unsigned int shift_i)
2369
{
2370
    assert(shift_i < (sf ? 64 : 32));
2371

    
2372
    if (shift_i == 0) {
2373
        tcg_gen_mov_i64(dst, src);
2374
    } else {
2375
        TCGv_i64 shift_const;
2376

    
2377
        shift_const = tcg_const_i64(shift_i);
2378
        shift_reg(dst, src, sf, shift_type, shift_const);
2379
        tcg_temp_free_i64(shift_const);
2380
    }
2381
}
2382

    
2383
/* C3.5.10 Logical (shifted register)
2384
 *   31  30 29 28       24 23   22 21  20  16 15    10 9    5 4    0
2385
 * +----+-----+-----------+-------+---+------+--------+------+------+
2386
 * | sf | opc | 0 1 0 1 0 | shift | N |  Rm  |  imm6  |  Rn  |  Rd  |
2387
 * +----+-----+-----------+-------+---+------+--------+------+------+
2388
 */
2389
static void disas_logic_reg(DisasContext *s, uint32_t insn)
2390
{
2391
    TCGv_i64 tcg_rd, tcg_rn, tcg_rm;
2392
    unsigned int sf, opc, shift_type, invert, rm, shift_amount, rn, rd;
2393

    
2394
    sf = extract32(insn, 31, 1);
2395
    opc = extract32(insn, 29, 2);
2396
    shift_type = extract32(insn, 22, 2);
2397
    invert = extract32(insn, 21, 1);
2398
    rm = extract32(insn, 16, 5);
2399
    shift_amount = extract32(insn, 10, 6);
2400
    rn = extract32(insn, 5, 5);
2401
    rd = extract32(insn, 0, 5);
2402

    
2403
    if (!sf && (shift_amount & (1 << 5))) {
2404
        unallocated_encoding(s);
2405
        return;
2406
    }
2407

    
2408
    tcg_rd = cpu_reg(s, rd);
2409

    
2410
    if (opc == 1 && shift_amount == 0 && shift_type == 0 && rn == 31) {
2411
        /* Unshifted ORR and ORN with WZR/XZR is the standard encoding for
2412
         * register-register MOV and MVN, so it is worth special casing.
2413
         */
2414
        tcg_rm = cpu_reg(s, rm);
2415
        if (invert) {
2416
            tcg_gen_not_i64(tcg_rd, tcg_rm);
2417
            if (!sf) {
2418
                tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2419
            }
2420
        } else {
2421
            if (sf) {
2422
                tcg_gen_mov_i64(tcg_rd, tcg_rm);
2423
            } else {
2424
                tcg_gen_ext32u_i64(tcg_rd, tcg_rm);
2425
            }
2426
        }
2427
        return;
2428
    }
2429

    
2430
    tcg_rm = read_cpu_reg(s, rm, sf);
2431

    
2432
    if (shift_amount) {
2433
        shift_reg_imm(tcg_rm, tcg_rm, sf, shift_type, shift_amount);
2434
    }
2435

    
2436
    tcg_rn = cpu_reg(s, rn);
2437

    
2438
    switch (opc | (invert << 2)) {
2439
    case 0: /* AND */
2440
    case 3: /* ANDS */
2441
        tcg_gen_and_i64(tcg_rd, tcg_rn, tcg_rm);
2442
        break;
2443
    case 1: /* ORR */
2444
        tcg_gen_or_i64(tcg_rd, tcg_rn, tcg_rm);
2445
        break;
2446
    case 2: /* EOR */
2447
        tcg_gen_xor_i64(tcg_rd, tcg_rn, tcg_rm);
2448
        break;
2449
    case 4: /* BIC */
2450
    case 7: /* BICS */
2451
        tcg_gen_andc_i64(tcg_rd, tcg_rn, tcg_rm);
2452
        break;
2453
    case 5: /* ORN */
2454
        tcg_gen_orc_i64(tcg_rd, tcg_rn, tcg_rm);
2455
        break;
2456
    case 6: /* EON */
2457
        tcg_gen_eqv_i64(tcg_rd, tcg_rn, tcg_rm);
2458
        break;
2459
    default:
2460
        assert(FALSE);
2461
        break;
2462
    }
2463

    
2464
    if (!sf) {
2465
        tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2466
    }
2467

    
2468
    if (opc == 3) {
2469
        gen_logic_CC(sf, tcg_rd);
2470
    }
2471
}
2472

    
2473
/*
2474
 * C3.5.1 Add/subtract (extended register)
2475
 *
2476
 *  31|30|29|28       24|23 22|21|20   16|15  13|12  10|9  5|4  0|
2477
 * +--+--+--+-----------+-----+--+-------+------+------+----+----+
2478
 * |sf|op| S| 0 1 0 1 1 | opt | 1|  Rm   |option| imm3 | Rn | Rd |
2479
 * +--+--+--+-----------+-----+--+-------+------+------+----+----+
2480
 *
2481
 *  sf: 0 -> 32bit, 1 -> 64bit
2482
 *  op: 0 -> add  , 1 -> sub
2483
 *   S: 1 -> set flags
2484
 * opt: 00
2485
 * option: extension type (see DecodeRegExtend)
2486
 * imm3: optional shift to Rm
2487
 *
2488
 * Rd = Rn + LSL(extend(Rm), amount)
2489
 */
2490
static void disas_add_sub_ext_reg(DisasContext *s, uint32_t insn)
2491
{
2492
    int rd = extract32(insn, 0, 5);
2493
    int rn = extract32(insn, 5, 5);
2494
    int imm3 = extract32(insn, 10, 3);
2495
    int option = extract32(insn, 13, 3);
2496
    int rm = extract32(insn, 16, 5);
2497
    bool setflags = extract32(insn, 29, 1);
2498
    bool sub_op = extract32(insn, 30, 1);
2499
    bool sf = extract32(insn, 31, 1);
2500

    
2501
    TCGv_i64 tcg_rm, tcg_rn; /* temps */
2502
    TCGv_i64 tcg_rd;
2503
    TCGv_i64 tcg_result;
2504

    
2505
    if (imm3 > 4) {
2506
        unallocated_encoding(s);
2507
        return;
2508
    }
2509

    
2510
    /* non-flag setting ops may use SP */
2511
    if (!setflags) {
2512
        tcg_rn = read_cpu_reg_sp(s, rn, sf);
2513
        tcg_rd = cpu_reg_sp(s, rd);
2514
    } else {
2515
        tcg_rn = read_cpu_reg(s, rn, sf);
2516
        tcg_rd = cpu_reg(s, rd);
2517
    }
2518

    
2519
    tcg_rm = read_cpu_reg(s, rm, sf);
2520
    ext_and_shift_reg(tcg_rm, tcg_rm, option, imm3);
2521

    
2522
    tcg_result = tcg_temp_new_i64();
2523

    
2524
    if (!setflags) {
2525
        if (sub_op) {
2526
            tcg_gen_sub_i64(tcg_result, tcg_rn, tcg_rm);
2527
        } else {
2528
            tcg_gen_add_i64(tcg_result, tcg_rn, tcg_rm);
2529
        }
2530
    } else {
2531
        if (sub_op) {
2532
            gen_sub_CC(sf, tcg_result, tcg_rn, tcg_rm);
2533
        } else {
2534
            gen_add_CC(sf, tcg_result, tcg_rn, tcg_rm);
2535
        }
2536
    }
2537

    
2538
    if (sf) {
2539
        tcg_gen_mov_i64(tcg_rd, tcg_result);
2540
    } else {
2541
        tcg_gen_ext32u_i64(tcg_rd, tcg_result);
2542
    }
2543

    
2544
    tcg_temp_free_i64(tcg_result);
2545
}
2546

    
2547
/*
2548
 * C3.5.2 Add/subtract (shifted register)
2549
 *
2550
 *  31 30 29 28       24 23 22 21 20   16 15     10 9    5 4    0
2551
 * +--+--+--+-----------+-----+--+-------+---------+------+------+
2552
 * |sf|op| S| 0 1 0 1 1 |shift| 0|  Rm   |  imm6   |  Rn  |  Rd  |
2553
 * +--+--+--+-----------+-----+--+-------+---------+------+------+
2554
 *
2555
 *    sf: 0 -> 32bit, 1 -> 64bit
2556
 *    op: 0 -> add  , 1 -> sub
2557
 *     S: 1 -> set flags
2558
 * shift: 00 -> LSL, 01 -> LSR, 10 -> ASR, 11 -> RESERVED
2559
 *  imm6: Shift amount to apply to Rm before the add/sub
2560
 */
2561
static void disas_add_sub_reg(DisasContext *s, uint32_t insn)
2562
{
2563
    int rd = extract32(insn, 0, 5);
2564
    int rn = extract32(insn, 5, 5);
2565
    int imm6 = extract32(insn, 10, 6);
2566
    int rm = extract32(insn, 16, 5);
2567
    int shift_type = extract32(insn, 22, 2);
2568
    bool setflags = extract32(insn, 29, 1);
2569
    bool sub_op = extract32(insn, 30, 1);
2570
    bool sf = extract32(insn, 31, 1);
2571

    
2572
    TCGv_i64 tcg_rd = cpu_reg(s, rd);
2573
    TCGv_i64 tcg_rn, tcg_rm;
2574
    TCGv_i64 tcg_result;
2575

    
2576
    if ((shift_type == 3) || (!sf && (imm6 > 31))) {
2577
        unallocated_encoding(s);
2578
        return;
2579
    }
2580

    
2581
    tcg_rn = read_cpu_reg(s, rn, sf);
2582
    tcg_rm = read_cpu_reg(s, rm, sf);
2583

    
2584
    shift_reg_imm(tcg_rm, tcg_rm, sf, shift_type, imm6);
2585

    
2586
    tcg_result = tcg_temp_new_i64();
2587

    
2588
    if (!setflags) {
2589
        if (sub_op) {
2590
            tcg_gen_sub_i64(tcg_result, tcg_rn, tcg_rm);
2591
        } else {
2592
            tcg_gen_add_i64(tcg_result, tcg_rn, tcg_rm);
2593
        }
2594
    } else {
2595
        if (sub_op) {
2596
            gen_sub_CC(sf, tcg_result, tcg_rn, tcg_rm);
2597
        } else {
2598
            gen_add_CC(sf, tcg_result, tcg_rn, tcg_rm);
2599
        }
2600
    }
2601

    
2602
    if (sf) {
2603
        tcg_gen_mov_i64(tcg_rd, tcg_result);
2604
    } else {
2605
        tcg_gen_ext32u_i64(tcg_rd, tcg_result);
2606
    }
2607

    
2608
    tcg_temp_free_i64(tcg_result);
2609
}
2610

    
2611
/* C3.5.9 Data-processing (3 source)
2612

2613
   31 30  29 28       24 23 21  20  16  15  14  10 9    5 4    0
2614
  +--+------+-----------+------+------+----+------+------+------+
2615
  |sf| op54 | 1 1 0 1 1 | op31 |  Rm  | o0 |  Ra  |  Rn  |  Rd  |
2616
  +--+------+-----------+------+------+----+------+------+------+
2617

2618
 */
2619
static void disas_data_proc_3src(DisasContext *s, uint32_t insn)
2620
{
2621
    int rd = extract32(insn, 0, 5);
2622
    int rn = extract32(insn, 5, 5);
2623
    int ra = extract32(insn, 10, 5);
2624
    int rm = extract32(insn, 16, 5);
2625
    int op_id = (extract32(insn, 29, 3) << 4) |
2626
        (extract32(insn, 21, 3) << 1) |
2627
        extract32(insn, 15, 1);
2628
    bool sf = extract32(insn, 31, 1);
2629
    bool is_sub = extract32(op_id, 0, 1);
2630
    bool is_high = extract32(op_id, 2, 1);
2631
    bool is_signed = false;
2632
    TCGv_i64 tcg_op1;
2633
    TCGv_i64 tcg_op2;
2634
    TCGv_i64 tcg_tmp;
2635

    
2636
    /* Note that op_id is sf:op54:op31:o0 so it includes the 32/64 size flag */
2637
    switch (op_id) {
2638
    case 0x42: /* SMADDL */
2639
    case 0x43: /* SMSUBL */
2640
    case 0x44: /* SMULH */
2641
        is_signed = true;
2642
        break;
2643
    case 0x0: /* MADD (32bit) */
2644
    case 0x1: /* MSUB (32bit) */
2645
    case 0x40: /* MADD (64bit) */
2646
    case 0x41: /* MSUB (64bit) */
2647
    case 0x4a: /* UMADDL */
2648
    case 0x4b: /* UMSUBL */
2649
    case 0x4c: /* UMULH */
2650
        break;
2651
    default:
2652
        unallocated_encoding(s);
2653
        return;
2654
    }
2655

    
2656
    if (is_high) {
2657
        TCGv_i64 low_bits = tcg_temp_new_i64(); /* low bits discarded */
2658
        TCGv_i64 tcg_rd = cpu_reg(s, rd);
2659
        TCGv_i64 tcg_rn = cpu_reg(s, rn);
2660
        TCGv_i64 tcg_rm = cpu_reg(s, rm);
2661

    
2662
        if (is_signed) {
2663
            tcg_gen_muls2_i64(low_bits, tcg_rd, tcg_rn, tcg_rm);
2664
        } else {
2665
            tcg_gen_mulu2_i64(low_bits, tcg_rd, tcg_rn, tcg_rm);
2666
        }
2667

    
2668
        tcg_temp_free_i64(low_bits);
2669
        return;
2670
    }
2671

    
2672
    tcg_op1 = tcg_temp_new_i64();
2673
    tcg_op2 = tcg_temp_new_i64();
2674
    tcg_tmp = tcg_temp_new_i64();
2675

    
2676
    if (op_id < 0x42) {
2677
        tcg_gen_mov_i64(tcg_op1, cpu_reg(s, rn));
2678
        tcg_gen_mov_i64(tcg_op2, cpu_reg(s, rm));
2679
    } else {
2680
        if (is_signed) {
2681
            tcg_gen_ext32s_i64(tcg_op1, cpu_reg(s, rn));
2682
            tcg_gen_ext32s_i64(tcg_op2, cpu_reg(s, rm));
2683
        } else {
2684
            tcg_gen_ext32u_i64(tcg_op1, cpu_reg(s, rn));
2685
            tcg_gen_ext32u_i64(tcg_op2, cpu_reg(s, rm));
2686
        }
2687
    }
2688

    
2689
    if (ra == 31 && !is_sub) {
2690
        /* Special-case MADD with rA == XZR; it is the standard MUL alias */
2691
        tcg_gen_mul_i64(cpu_reg(s, rd), tcg_op1, tcg_op2);
2692
    } else {
2693
        tcg_gen_mul_i64(tcg_tmp, tcg_op1, tcg_op2);
2694
        if (is_sub) {
2695
            tcg_gen_sub_i64(cpu_reg(s, rd), cpu_reg(s, ra), tcg_tmp);
2696
        } else {
2697
            tcg_gen_add_i64(cpu_reg(s, rd), cpu_reg(s, ra), tcg_tmp);
2698
        }
2699
    }
2700

    
2701
    if (!sf) {
2702
        tcg_gen_ext32u_i64(cpu_reg(s, rd), cpu_reg(s, rd));
2703
    }
2704

    
2705
    tcg_temp_free_i64(tcg_op1);
2706
    tcg_temp_free_i64(tcg_op2);
2707
    tcg_temp_free_i64(tcg_tmp);
2708
}
2709

    
2710
/* C3.5.3 - Add/subtract (with carry)
2711
 *  31 30 29 28 27 26 25 24 23 22 21  20  16  15   10  9    5 4   0
2712
 * +--+--+--+------------------------+------+---------+------+-----+
2713
 * |sf|op| S| 1  1  0  1  0  0  0  0 |  rm  | opcode2 |  Rn  |  Rd |
2714
 * +--+--+--+------------------------+------+---------+------+-----+
2715
 *                                            [000000]
2716
 */
2717

    
2718
static void disas_adc_sbc(DisasContext *s, uint32_t insn)
2719
{
2720
    unsigned int sf, op, setflags, rm, rn, rd;
2721
    TCGv_i64 tcg_y, tcg_rn, tcg_rd;
2722

    
2723
    if (extract32(insn, 10, 6) != 0) {
2724
        unallocated_encoding(s);
2725
        return;
2726
    }
2727

    
2728
    sf = extract32(insn, 31, 1);
2729
    op = extract32(insn, 30, 1);
2730
    setflags = extract32(insn, 29, 1);
2731
    rm = extract32(insn, 16, 5);
2732
    rn = extract32(insn, 5, 5);
2733
    rd = extract32(insn, 0, 5);
2734

    
2735
    tcg_rd = cpu_reg(s, rd);
2736
    tcg_rn = cpu_reg(s, rn);
2737

    
2738
    if (op) {
2739
        tcg_y = new_tmp_a64(s);
2740
        tcg_gen_not_i64(tcg_y, cpu_reg(s, rm));
2741
    } else {
2742
        tcg_y = cpu_reg(s, rm);
2743
    }
2744

    
2745
    if (setflags) {
2746
        gen_adc_CC(sf, tcg_rd, tcg_rn, tcg_y);
2747
    } else {
2748
        gen_adc(sf, tcg_rd, tcg_rn, tcg_y);
2749
    }
2750
}
2751

    
2752
/* C3.5.4 - C3.5.5 Conditional compare (immediate / register)
2753
 *  31 30 29 28 27 26 25 24 23 22 21  20    16 15  12  11  10  9   5  4 3   0
2754
 * +--+--+--+------------------------+--------+------+----+--+------+--+-----+
2755
 * |sf|op| S| 1  1  0  1  0  0  1  0 |imm5/rm | cond |i/r |o2|  Rn  |o3|nzcv |
2756
 * +--+--+--+------------------------+--------+------+----+--+------+--+-----+
2757
 *        [1]                             y                [0]       [0]
2758
 */
2759
static void disas_cc(DisasContext *s, uint32_t insn)
2760
{
2761
    unsigned int sf, op, y, cond, rn, nzcv, is_imm;
2762
    int label_continue = -1;
2763
    TCGv_i64 tcg_tmp, tcg_y, tcg_rn;
2764

    
2765
    if (!extract32(insn, 29, 1)) {
2766
        unallocated_encoding(s);
2767
        return;
2768
    }
2769
    if (insn & (1 << 10 | 1 << 4)) {
2770
        unallocated_encoding(s);
2771
        return;
2772
    }
2773
    sf = extract32(insn, 31, 1);
2774
    op = extract32(insn, 30, 1);
2775
    is_imm = extract32(insn, 11, 1);
2776
    y = extract32(insn, 16, 5); /* y = rm (reg) or imm5 (imm) */
2777
    cond = extract32(insn, 12, 4);
2778
    rn = extract32(insn, 5, 5);
2779
    nzcv = extract32(insn, 0, 4);
2780

    
2781
    if (cond < 0x0e) { /* not always */
2782
        int label_match = gen_new_label();
2783
        label_continue = gen_new_label();
2784
        arm_gen_test_cc(cond, label_match);
2785
        /* nomatch: */
2786
        tcg_tmp = tcg_temp_new_i64();
2787
        tcg_gen_movi_i64(tcg_tmp, nzcv << 28);
2788
        gen_set_nzcv(tcg_tmp);
2789
        tcg_temp_free_i64(tcg_tmp);
2790
        tcg_gen_br(label_continue);
2791
        gen_set_label(label_match);
2792
    }
2793
    /* match, or condition is always */
2794
    if (is_imm) {
2795
        tcg_y = new_tmp_a64(s);
2796
        tcg_gen_movi_i64(tcg_y, y);
2797
    } else {
2798
        tcg_y = cpu_reg(s, y);
2799
    }
2800
    tcg_rn = cpu_reg(s, rn);
2801

    
2802
    tcg_tmp = tcg_temp_new_i64();
2803
    if (op) {
2804
        gen_sub_CC(sf, tcg_tmp, tcg_rn, tcg_y);
2805
    } else {
2806
        gen_add_CC(sf, tcg_tmp, tcg_rn, tcg_y);
2807
    }
2808
    tcg_temp_free_i64(tcg_tmp);
2809

    
2810
    if (cond < 0x0e) { /* continue */
2811
        gen_set_label(label_continue);
2812
    }
2813
}
2814

    
2815
/* C3.5.6 Conditional select
2816
 *   31   30  29  28             21 20  16 15  12 11 10 9    5 4    0
2817
 * +----+----+---+-----------------+------+------+-----+------+------+
2818
 * | sf | op | S | 1 1 0 1 0 1 0 0 |  Rm  | cond | op2 |  Rn  |  Rd  |
2819
 * +----+----+---+-----------------+------+------+-----+------+------+
2820
 */
2821
static void disas_cond_select(DisasContext *s, uint32_t insn)
2822
{
2823
    unsigned int sf, else_inv, rm, cond, else_inc, rn, rd;
2824
    TCGv_i64 tcg_rd, tcg_src;
2825

    
2826
    if (extract32(insn, 29, 1) || extract32(insn, 11, 1)) {
2827
        /* S == 1 or op2<1> == 1 */
2828
        unallocated_encoding(s);
2829
        return;
2830
    }
2831
    sf = extract32(insn, 31, 1);
2832
    else_inv = extract32(insn, 30, 1);
2833
    rm = extract32(insn, 16, 5);
2834
    cond = extract32(insn, 12, 4);
2835
    else_inc = extract32(insn, 10, 1);
2836
    rn = extract32(insn, 5, 5);
2837
    rd = extract32(insn, 0, 5);
2838

    
2839
    if (rd == 31) {
2840
        /* silly no-op write; until we use movcond we must special-case
2841
         * this to avoid a dead temporary across basic blocks.
2842
         */
2843
        return;
2844
    }
2845

    
2846
    tcg_rd = cpu_reg(s, rd);
2847

    
2848
    if (cond >= 0x0e) { /* condition "always" */
2849
        tcg_src = read_cpu_reg(s, rn, sf);
2850
        tcg_gen_mov_i64(tcg_rd, tcg_src);
2851
    } else {
2852
        /* OPTME: we could use movcond here, at the cost of duplicating
2853
         * a lot of the arm_gen_test_cc() logic.
2854
         */
2855
        int label_match = gen_new_label();
2856
        int label_continue = gen_new_label();
2857

    
2858
        arm_gen_test_cc(cond, label_match);
2859
        /* nomatch: */
2860
        tcg_src = cpu_reg(s, rm);
2861

    
2862
        if (else_inv && else_inc) {
2863
            tcg_gen_neg_i64(tcg_rd, tcg_src);
2864
        } else if (else_inv) {
2865
            tcg_gen_not_i64(tcg_rd, tcg_src);
2866
        } else if (else_inc) {
2867
            tcg_gen_addi_i64(tcg_rd, tcg_src, 1);
2868
        } else {
2869
            tcg_gen_mov_i64(tcg_rd, tcg_src);
2870
        }
2871
        if (!sf) {
2872
            tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
2873
        }
2874
        tcg_gen_br(label_continue);
2875
        /* match: */
2876
        gen_set_label(label_match);
2877
        tcg_src = read_cpu_reg(s, rn, sf);
2878
        tcg_gen_mov_i64(tcg_rd, tcg_src);
2879
        /* continue: */
2880
        gen_set_label(label_continue);
2881
    }
2882
}
2883

    
2884
static void handle_clz(DisasContext *s, unsigned int sf,
2885
                       unsigned int rn, unsigned int rd)
2886
{
2887
    TCGv_i64 tcg_rd, tcg_rn;
2888
    tcg_rd = cpu_reg(s, rd);
2889
    tcg_rn = cpu_reg(s, rn);
2890

    
2891
    if (sf) {
2892
        gen_helper_clz64(tcg_rd, tcg_rn);
2893
    } else {
2894
        TCGv_i32 tcg_tmp32 = tcg_temp_new_i32();
2895
        tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn);
2896
        gen_helper_clz(tcg_tmp32, tcg_tmp32);
2897
        tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32);
2898
        tcg_temp_free_i32(tcg_tmp32);
2899
    }
2900
}
2901

    
2902
static void handle_cls(DisasContext *s, unsigned int sf,
2903
                       unsigned int rn, unsigned int rd)
2904
{
2905
    TCGv_i64 tcg_rd, tcg_rn;
2906
    tcg_rd = cpu_reg(s, rd);
2907
    tcg_rn = cpu_reg(s, rn);
2908

    
2909
    if (sf) {
2910
        gen_helper_cls64(tcg_rd, tcg_rn);
2911
    } else {
2912
        TCGv_i32 tcg_tmp32 = tcg_temp_new_i32();
2913
        tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn);
2914
        gen_helper_cls32(tcg_tmp32, tcg_tmp32);
2915
        tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32);
2916
        tcg_temp_free_i32(tcg_tmp32);
2917
    }
2918
}
2919

    
2920
static void handle_rbit(DisasContext *s, unsigned int sf,
2921
                        unsigned int rn, unsigned int rd)
2922
{
2923
    TCGv_i64 tcg_rd, tcg_rn;
2924
    tcg_rd = cpu_reg(s, rd);
2925
    tcg_rn = cpu_reg(s, rn);
2926

    
2927
    if (sf) {
2928
        gen_helper_rbit64(tcg_rd, tcg_rn);
2929
    } else {
2930
        TCGv_i32 tcg_tmp32 = tcg_temp_new_i32();
2931
        tcg_gen_trunc_i64_i32(tcg_tmp32, tcg_rn);
2932
        gen_helper_rbit(tcg_tmp32, tcg_tmp32);
2933
        tcg_gen_extu_i32_i64(tcg_rd, tcg_tmp32);
2934
        tcg_temp_free_i32(tcg_tmp32);
2935
    }
2936
}
2937

    
2938
/* C5.6.149 REV with sf==1, opcode==3 ("REV64") */
2939
static void handle_rev64(DisasContext *s, unsigned int sf,
2940
                         unsigned int rn, unsigned int rd)
2941
{
2942
    if (!sf) {
2943
        unallocated_encoding(s);
2944
        return;
2945
    }
2946
    tcg_gen_bswap64_i64(cpu_reg(s, rd), cpu_reg(s, rn));
2947
}
2948

    
2949
/* C5.6.149 REV with sf==0, opcode==2
2950
 * C5.6.151 REV32 (sf==1, opcode==2)
2951
 */
2952
static void handle_rev32(DisasContext *s, unsigned int sf,
2953
                         unsigned int rn, unsigned int rd)
2954
{
2955
    TCGv_i64 tcg_rd = cpu_reg(s, rd);
2956

    
2957
    if (sf) {
2958
        TCGv_i64 tcg_tmp = tcg_temp_new_i64();
2959
        TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf);
2960

    
2961
        /* bswap32_i64 requires zero high word */
2962
        tcg_gen_ext32u_i64(tcg_tmp, tcg_rn);
2963
        tcg_gen_bswap32_i64(tcg_rd, tcg_tmp);
2964
        tcg_gen_shri_i64(tcg_tmp, tcg_rn, 32);
2965
        tcg_gen_bswap32_i64(tcg_tmp, tcg_tmp);
2966
        tcg_gen_concat32_i64(tcg_rd, tcg_rd, tcg_tmp);
2967

    
2968
        tcg_temp_free_i64(tcg_tmp);
2969
    } else {
2970
        tcg_gen_ext32u_i64(tcg_rd, cpu_reg(s, rn));
2971
        tcg_gen_bswap32_i64(tcg_rd, tcg_rd);
2972
    }
2973
}
2974

    
2975
/* C5.6.150 REV16 (opcode==1) */
2976
static void handle_rev16(DisasContext *s, unsigned int sf,
2977
                         unsigned int rn, unsigned int rd)
2978
{
2979
    TCGv_i64 tcg_rd = cpu_reg(s, rd);
2980
    TCGv_i64 tcg_tmp = tcg_temp_new_i64();
2981
    TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf);
2982

    
2983
    tcg_gen_andi_i64(tcg_tmp, tcg_rn, 0xffff);
2984
    tcg_gen_bswap16_i64(tcg_rd, tcg_tmp);
2985

    
2986
    tcg_gen_shri_i64(tcg_tmp, tcg_rn, 16);
2987
    tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
2988
    tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
2989
    tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 16, 16);
2990

    
2991
    if (sf) {
2992
        tcg_gen_shri_i64(tcg_tmp, tcg_rn, 32);
2993
        tcg_gen_andi_i64(tcg_tmp, tcg_tmp, 0xffff);
2994
        tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
2995
        tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 32, 16);
2996

    
2997
        tcg_gen_shri_i64(tcg_tmp, tcg_rn, 48);
2998
        tcg_gen_bswap16_i64(tcg_tmp, tcg_tmp);
2999
        tcg_gen_deposit_i64(tcg_rd, tcg_rd, tcg_tmp, 48, 16);
3000
    }
3001

    
3002
    tcg_temp_free_i64(tcg_tmp);
3003
}
3004

    
3005
/* C3.5.7 Data-processing (1 source)
3006
 *   31  30  29  28             21 20     16 15    10 9    5 4    0
3007
 * +----+---+---+-----------------+---------+--------+------+------+
3008
 * | sf | 1 | S | 1 1 0 1 0 1 1 0 | opcode2 | opcode |  Rn  |  Rd  |
3009
 * +----+---+---+-----------------+---------+--------+------+------+
3010
 */
3011
static void disas_data_proc_1src(DisasContext *s, uint32_t insn)
3012
{
3013
    unsigned int sf, opcode, rn, rd;
3014

    
3015
    if (extract32(insn, 29, 1) || extract32(insn, 16, 5)) {
3016
        unallocated_encoding(s);
3017
        return;
3018
    }
3019

    
3020
    sf = extract32(insn, 31, 1);
3021
    opcode = extract32(insn, 10, 6);
3022
    rn = extract32(insn, 5, 5);
3023
    rd = extract32(insn, 0, 5);
3024

    
3025
    switch (opcode) {
3026
    case 0: /* RBIT */
3027
        handle_rbit(s, sf, rn, rd);
3028
        break;
3029
    case 1: /* REV16 */
3030
        handle_rev16(s, sf, rn, rd);
3031
        break;
3032
    case 2: /* REV32 */
3033
        handle_rev32(s, sf, rn, rd);
3034
        break;
3035
    case 3: /* REV64 */
3036
        handle_rev64(s, sf, rn, rd);
3037
        break;
3038
    case 4: /* CLZ */
3039
        handle_clz(s, sf, rn, rd);
3040
        break;
3041
    case 5: /* CLS */
3042
        handle_cls(s, sf, rn, rd);
3043
        break;
3044
    }
3045
}
3046

    
3047
static void handle_div(DisasContext *s, bool is_signed, unsigned int sf,
3048
                       unsigned int rm, unsigned int rn, unsigned int rd)
3049
{
3050
    TCGv_i64 tcg_n, tcg_m, tcg_rd;
3051
    tcg_rd = cpu_reg(s, rd);
3052

    
3053
    if (!sf && is_signed) {
3054
        tcg_n = new_tmp_a64(s);
3055
        tcg_m = new_tmp_a64(s);
3056
        tcg_gen_ext32s_i64(tcg_n, cpu_reg(s, rn));
3057
        tcg_gen_ext32s_i64(tcg_m, cpu_reg(s, rm));
3058
    } else {
3059
        tcg_n = read_cpu_reg(s, rn, sf);
3060
        tcg_m = read_cpu_reg(s, rm, sf);
3061
    }
3062

    
3063
    if (is_signed) {
3064
        gen_helper_sdiv64(tcg_rd, tcg_n, tcg_m);
3065
    } else {
3066
        gen_helper_udiv64(tcg_rd, tcg_n, tcg_m);
3067
    }
3068

    
3069
    if (!sf) { /* zero extend final result */
3070
        tcg_gen_ext32u_i64(tcg_rd, tcg_rd);
3071
    }
3072
}
3073

    
3074
/* C5.6.115 LSLV, C5.6.118 LSRV, C5.6.17 ASRV, C5.6.154 RORV */
3075
static void handle_shift_reg(DisasContext *s,
3076
                             enum a64_shift_type shift_type, unsigned int sf,
3077
                             unsigned int rm, unsigned int rn, unsigned int rd)
3078
{
3079
    TCGv_i64 tcg_shift = tcg_temp_new_i64();
3080
    TCGv_i64 tcg_rd = cpu_reg(s, rd);
3081
    TCGv_i64 tcg_rn = read_cpu_reg(s, rn, sf);
3082

    
3083
    tcg_gen_andi_i64(tcg_shift, cpu_reg(s, rm), sf ? 63 : 31);
3084
    shift_reg(tcg_rd, tcg_rn, sf, shift_type, tcg_shift);
3085
    tcg_temp_free_i64(tcg_shift);
3086
}
3087

    
3088
/* C3.5.8 Data-processing (2 source)
3089
 *   31   30  29 28             21 20  16 15    10 9    5 4    0
3090
 * +----+---+---+-----------------+------+--------+------+------+
3091
 * | sf | 0 | S | 1 1 0 1 0 1 1 0 |  Rm  | opcode |  Rn  |  Rd  |
3092
 * +----+---+---+-----------------+------+--------+------+------+
3093
 */
3094
static void disas_data_proc_2src(DisasContext *s, uint32_t insn)
3095
{
3096
    unsigned int sf, rm, opcode, rn, rd;
3097
    sf = extract32(insn, 31, 1);
3098
    rm = extract32(insn, 16, 5);
3099
    opcode = extract32(insn, 10, 6);
3100
    rn = extract32(insn, 5, 5);
3101
    rd = extract32(insn, 0, 5);
3102

    
3103
    if (extract32(insn, 29, 1)) {
3104
        unallocated_encoding(s);
3105
        return;
3106
    }
3107

    
3108
    switch (opcode) {
3109
    case 2: /* UDIV */
3110
        handle_div(s, false, sf, rm, rn, rd);
3111
        break;
3112
    case 3: /* SDIV */
3113
        handle_div(s, true, sf, rm, rn, rd);
3114
        break;
3115
    case 8: /* LSLV */
3116
        handle_shift_reg(s, A64_SHIFT_TYPE_LSL, sf, rm, rn, rd);
3117
        break;
3118
    case 9: /* LSRV */
3119
        handle_shift_reg(s, A64_SHIFT_TYPE_LSR, sf, rm, rn, rd);
3120
        break;
3121
    case 10: /* ASRV */
3122
        handle_shift_reg(s, A64_SHIFT_TYPE_ASR, sf, rm, rn, rd);
3123
        break;
3124
    case 11: /* RORV */
3125
        handle_shift_reg(s, A64_SHIFT_TYPE_ROR, sf, rm, rn, rd);
3126
        break;
3127
    case 16:
3128
    case 17:
3129
    case 18:
3130
    case 19:
3131
    case 20:
3132
    case 21:
3133
    case 22:
3134
    case 23: /* CRC32 */
3135
        unsupported_encoding(s, insn);
3136
        break;
3137
    default:
3138
        unallocated_encoding(s);
3139
        break;
3140
    }
3141
}
3142

    
3143
/* C3.5 Data processing - register */
3144
static void disas_data_proc_reg(DisasContext *s, uint32_t insn)
3145
{
3146
    switch (extract32(insn, 24, 5)) {
3147
    case 0x0a: /* Logical (shifted register) */
3148
        disas_logic_reg(s, insn);
3149
        break;
3150
    case 0x0b: /* Add/subtract */
3151
        if (insn & (1 << 21)) { /* (extended register) */
3152
            disas_add_sub_ext_reg(s, insn);
3153
        } else {
3154
            disas_add_sub_reg(s, insn);
3155
        }
3156
        break;
3157
    case 0x1b: /* Data-processing (3 source) */
3158
        disas_data_proc_3src(s, insn);
3159
        break;
3160
    case 0x1a:
3161
        switch (extract32(insn, 21, 3)) {
3162
        case 0x0: /* Add/subtract (with carry) */
3163
            disas_adc_sbc(s, insn);
3164
            break;
3165
        case 0x2: /* Conditional compare */
3166
            disas_cc(s, insn); /* both imm and reg forms */
3167
            break;
3168
        case 0x4: /* Conditional select */
3169
            disas_cond_select(s, insn);
3170
            break;
3171
        case 0x6: /* Data-processing */
3172
            if (insn & (1 << 30)) { /* (1 source) */
3173
                disas_data_proc_1src(s, insn);
3174
            } else {            /* (2 source) */
3175
                disas_data_proc_2src(s, insn);
3176
            }
3177
            break;
3178
        default:
3179
            unallocated_encoding(s);
3180
            break;
3181
        }
3182
        break;
3183
    default:
3184
        unallocated_encoding(s);
3185
        break;
3186
    }
3187
}
3188

    
3189
/* Convert ARM rounding mode to softfloat */
3190
static inline int arm_rmode_to_sf(int rmode)
3191
{
3192
    switch (rmode) {
3193
    case FPROUNDING_TIEAWAY:
3194
        rmode = float_round_ties_away;
3195
        break;
3196
    case FPROUNDING_ODD:
3197
        /* FIXME: add support for TIEAWAY and ODD */
3198
        qemu_log_mask(LOG_UNIMP, "arm: unimplemented rounding mode: %d\n",
3199
                      rmode);
3200
    case FPROUNDING_TIEEVEN:
3201
    default:
3202
        rmode = float_round_nearest_even;
3203
        break;
3204
    case FPROUNDING_POSINF:
3205
        rmode = float_round_up;
3206
        break;
3207
    case FPROUNDING_NEGINF:
3208
        rmode = float_round_down;
3209
        break;
3210
    case FPROUNDING_ZERO:
3211
        rmode = float_round_to_zero;
3212
        break;
3213
    }
3214
    return rmode;
3215
}
3216

    
3217
static void handle_fp_compare(DisasContext *s, bool is_double,
3218
                              unsigned int rn, unsigned int rm,
3219
                              bool cmp_with_zero, bool signal_all_nans)
3220
{
3221
    TCGv_i64 tcg_flags = tcg_temp_new_i64();
3222
    TCGv_ptr fpst = get_fpstatus_ptr();
3223

    
3224
    if (is_double) {
3225
        TCGv_i64 tcg_vn, tcg_vm;
3226

    
3227
        tcg_vn = read_fp_dreg(s, rn);
3228
        if (cmp_with_zero) {
3229
            tcg_vm = tcg_const_i64(0);
3230
        } else {
3231
            tcg_vm = read_fp_dreg(s, rm);
3232
        }
3233
        if (signal_all_nans) {
3234
            gen_helper_vfp_cmped_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3235
        } else {
3236
            gen_helper_vfp_cmpd_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3237
        }
3238
        tcg_temp_free_i64(tcg_vn);
3239
        tcg_temp_free_i64(tcg_vm);
3240
    } else {
3241
        TCGv_i32 tcg_vn, tcg_vm;
3242

    
3243
        tcg_vn = read_fp_sreg(s, rn);
3244
        if (cmp_with_zero) {
3245
            tcg_vm = tcg_const_i32(0);
3246
        } else {
3247
            tcg_vm = read_fp_sreg(s, rm);
3248
        }
3249
        if (signal_all_nans) {
3250
            gen_helper_vfp_cmpes_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3251
        } else {
3252
            gen_helper_vfp_cmps_a64(tcg_flags, tcg_vn, tcg_vm, fpst);
3253
        }
3254
        tcg_temp_free_i32(tcg_vn);
3255
        tcg_temp_free_i32(tcg_vm);
3256
    }
3257

    
3258
    tcg_temp_free_ptr(fpst);
3259

    
3260
    gen_set_nzcv(tcg_flags);
3261

    
3262
    tcg_temp_free_i64(tcg_flags);
3263
}
3264

    
3265
/* C3.6.22 Floating point compare
3266
 *   31  30  29 28       24 23  22  21 20  16 15 14 13  10    9    5 4     0
3267
 * +---+---+---+-----------+------+---+------+-----+---------+------+-------+
3268
 * | M | 0 | S | 1 1 1 1 0 | type | 1 |  Rm  | op  | 1 0 0 0 |  Rn  |  op2  |
3269
 * +---+---+---+-----------+------+---+------+-----+---------+------+-------+
3270
 */
3271
static void disas_fp_compare(DisasContext *s, uint32_t insn)
3272
{
3273
    unsigned int mos, type, rm, op, rn, opc, op2r;
3274

    
3275
    mos = extract32(insn, 29, 3);
3276
    type = extract32(insn, 22, 2); /* 0 = single, 1 = double */
3277
    rm = extract32(insn, 16, 5);
3278
    op = extract32(insn, 14, 2);
3279
    rn = extract32(insn, 5, 5);
3280
    opc = extract32(insn, 3, 2);
3281
    op2r = extract32(insn, 0, 3);
3282

    
3283
    if (mos || op || op2r || type > 1) {
3284
        unallocated_encoding(s);
3285
        return;
3286
    }
3287

    
3288
    handle_fp_compare(s, type, rn, rm, opc & 1, opc & 2);
3289
}
3290

    
3291
/* C3.6.23 Floating point conditional compare
3292
 *   31  30  29 28       24 23  22  21 20  16 15  12 11 10 9    5  4   3    0
3293
 * +---+---+---+-----------+------+---+------+------+-----+------+----+------+
3294
 * | M | 0 | S | 1 1 1 1 0 | type | 1 |  Rm  | cond | 0 1 |  Rn  | op | nzcv |
3295
 * +---+---+---+-----------+------+---+------+------+-----+------+----+------+
3296
 */
3297
static void disas_fp_ccomp(DisasContext *s, uint32_t insn)
3298
{
3299
    unsigned int mos, type, rm, cond, rn, op, nzcv;
3300
    TCGv_i64 tcg_flags;
3301
    int label_continue = -1;
3302

    
3303
    mos = extract32(insn, 29, 3);
3304
    type = extract32(insn, 22, 2); /* 0 = single, 1 = double */
3305
    rm = extract32(insn, 16, 5);
3306
    cond = extract32(insn, 12, 4);
3307
    rn = extract32(insn, 5, 5);
3308
    op = extract32(insn, 4, 1);
3309
    nzcv = extract32(insn, 0, 4);
3310

    
3311
    if (mos || type > 1) {
3312
        unallocated_encoding(s);
3313
        return;
3314
    }
3315

    
3316
    if (cond < 0x0e) { /* not always */
3317
        int label_match = gen_new_label();
3318
        label_continue = gen_new_label();
3319
        arm_gen_test_cc(cond, label_match);
3320
        /* nomatch: */
3321
        tcg_flags = tcg_const_i64(nzcv << 28);
3322
        gen_set_nzcv(tcg_flags);
3323
        tcg_temp_free_i64(tcg_flags);
3324
        tcg_gen_br(label_continue);
3325
        gen_set_label(label_match);
3326
    }
3327

    
3328
    handle_fp_compare(s, type, rn, rm, false, op);
3329

    
3330
    if (cond < 0x0e) {
3331
        gen_set_label(label_continue);
3332
    }
3333
}
3334

    
3335
/* copy src FP register to dst FP register; type specifies single or double */
3336
static void gen_mov_fp2fp(DisasContext *s, int type, int dst, int src)
3337
{
3338
    if (type) {
3339
        TCGv_i64 v = read_fp_dreg(s, src);
3340
        write_fp_dreg(s, dst, v);
3341
        tcg_temp_free_i64(v);
3342
    } else {
3343
        TCGv_i32 v = read_fp_sreg(s, src);
3344
        write_fp_sreg(s, dst, v);
3345
        tcg_temp_free_i32(v);
3346
    }
3347
}
3348

    
3349
/* C3.6.24 Floating point conditional select
3350
 *   31  30  29 28       24 23  22  21 20  16 15  12 11 10 9    5 4    0
3351
 * +---+---+---+-----------+------+---+------+------+-----+------+------+
3352
 * | M | 0 | S | 1 1 1 1 0 | type | 1 |  Rm  | cond | 1 1 |  Rn  |  Rd  |
3353
 * +---+---+---+-----------+------+---+------+------+-----+------+------+
3354
 */
3355
static void disas_fp_csel(DisasContext *s, uint32_t insn)
3356
{
3357
    unsigned int mos, type, rm, cond, rn, rd;
3358
    int label_continue = -1;
3359

    
3360
    mos = extract32(insn, 29, 3);
3361
    type = extract32(insn, 22, 2); /* 0 = single, 1 = double */
3362
    rm = extract32(insn, 16, 5);
3363
    cond = extract32(insn, 12, 4);
3364
    rn = extract32(insn, 5, 5);
3365
    rd = extract32(insn, 0, 5);
3366

    
3367
    if (mos || type > 1) {
3368
        unallocated_encoding(s);
3369
        return;
3370
    }
3371

    
3372
    if (cond < 0x0e) { /* not always */
3373
        int label_match = gen_new_label();
3374
        label_continue = gen_new_label();
3375
        arm_gen_test_cc(cond, label_match);
3376
        /* nomatch: */
3377
        gen_mov_fp2fp(s, type, rd, rm);
3378
        tcg_gen_br(label_continue);
3379
        gen_set_label(label_match);
3380
    }
3381

    
3382
    gen_mov_fp2fp(s, type, rd, rn);
3383

    
3384
    if (cond < 0x0e) { /* continue */
3385
        gen_set_label(label_continue);
3386
    }
3387
}
3388

    
3389
/* C3.6.25 Floating point data-processing (1 source)
3390
 *   31  30  29 28       24 23  22  21 20    15 14       10 9    5 4    0
3391
 * +---+---+---+-----------+------+---+--------+-----------+------+------+
3392
 * | M | 0 | S | 1 1 1 1 0 | type | 1 | opcode | 1 0 0 0 0 |  Rn  |  Rd  |
3393
 * +---+---+---+-----------+------+---+--------+-----------+------+------+
3394
 */
3395
static void disas_fp_1src(DisasContext *s, uint32_t insn)
3396
{
3397
    unsupported_encoding(s, insn);
3398
}
3399

    
3400
/* C3.6.26 Floating-point data-processing (2 source) - single precision */
3401
static void handle_fp_2src_single(DisasContext *s, int opcode,
3402
                                  int rd, int rn, int rm)
3403
{
3404
    TCGv_i32 tcg_op1;
3405
    TCGv_i32 tcg_op2;
3406
    TCGv_i32 tcg_res;
3407
    TCGv_ptr fpst;
3408

    
3409
    tcg_res = tcg_temp_new_i32();
3410
    fpst = get_fpstatus_ptr();
3411
    tcg_op1 = read_fp_sreg(s, rn);
3412
    tcg_op2 = read_fp_sreg(s, rm);
3413

    
3414
    switch (opcode) {
3415
    case 0x0: /* FMUL */
3416
        gen_helper_vfp_muls(tcg_res, tcg_op1, tcg_op2, fpst);
3417
        break;
3418
    case 0x1: /* FDIV */
3419
        gen_helper_vfp_divs(tcg_res, tcg_op1, tcg_op2, fpst);
3420
        break;
3421
    case 0x2: /* FADD */
3422
        gen_helper_vfp_adds(tcg_res, tcg_op1, tcg_op2, fpst);
3423
        break;
3424
    case 0x3: /* FSUB */
3425
        gen_helper_vfp_subs(tcg_res, tcg_op1, tcg_op2, fpst);
3426
        break;
3427
    case 0x4: /* FMAX */
3428
        gen_helper_vfp_maxs(tcg_res, tcg_op1, tcg_op2, fpst);
3429
        break;
3430
    case 0x5: /* FMIN */
3431
        gen_helper_vfp_mins(tcg_res, tcg_op1, tcg_op2, fpst);
3432
        break;
3433
    case 0x6: /* FMAXNM */
3434
        gen_helper_vfp_maxnums(tcg_res, tcg_op1, tcg_op2, fpst);
3435
        break;
3436
    case 0x7: /* FMINNM */
3437
        gen_helper_vfp_minnums(tcg_res, tcg_op1, tcg_op2, fpst);
3438
        break;
3439
    case 0x8: /* FNMUL */
3440
        gen_helper_vfp_muls(tcg_res, tcg_op1, tcg_op2, fpst);
3441
        gen_helper_vfp_negs(tcg_res, tcg_res);
3442
        break;
3443
    }
3444

    
3445
    write_fp_sreg(s, rd, tcg_res);
3446

    
3447
    tcg_temp_free_ptr(fpst);
3448
    tcg_temp_free_i32(tcg_op1);
3449
    tcg_temp_free_i32(tcg_op2);
3450
    tcg_temp_free_i32(tcg_res);
3451
}
3452

    
3453
/* C3.6.26 Floating-point data-processing (2 source) - double precision */
3454
static void handle_fp_2src_double(DisasContext *s, int opcode,
3455
                                  int rd, int rn, int rm)
3456
{
3457
    TCGv_i64 tcg_op1;
3458
    TCGv_i64 tcg_op2;
3459
    TCGv_i64 tcg_res;
3460
    TCGv_ptr fpst;
3461

    
3462
    tcg_res = tcg_temp_new_i64();
3463
    fpst = get_fpstatus_ptr();
3464
    tcg_op1 = read_fp_dreg(s, rn);
3465
    tcg_op2 = read_fp_dreg(s, rm);
3466

    
3467
    switch (opcode) {
3468
    case 0x0: /* FMUL */
3469
        gen_helper_vfp_muld(tcg_res, tcg_op1, tcg_op2, fpst);
3470
        break;
3471
    case 0x1: /* FDIV */
3472
        gen_helper_vfp_divd(tcg_res, tcg_op1, tcg_op2, fpst);
3473
        break;
3474
    case 0x2: /* FADD */
3475
        gen_helper_vfp_addd(tcg_res, tcg_op1, tcg_op2, fpst);
3476
        break;
3477
    case 0x3: /* FSUB */
3478
        gen_helper_vfp_subd(tcg_res, tcg_op1, tcg_op2, fpst);
3479
        break;
3480
    case 0x4: /* FMAX */
3481
        gen_helper_vfp_maxd(tcg_res, tcg_op1, tcg_op2, fpst);
3482
        break;
3483
    case 0x5: /* FMIN */
3484
        gen_helper_vfp_mind(tcg_res, tcg_op1, tcg_op2, fpst);
3485
        break;
3486
    case 0x6: /* FMAXNM */
3487
        gen_helper_vfp_maxnumd(tcg_res, tcg_op1, tcg_op2, fpst);
3488
        break;
3489
    case 0x7: /* FMINNM */
3490
        gen_helper_vfp_minnumd(tcg_res, tcg_op1, tcg_op2, fpst);
3491
        break;
3492
    case 0x8: /* FNMUL */
3493
        gen_helper_vfp_muld(tcg_res, tcg_op1, tcg_op2, fpst);
3494
        gen_helper_vfp_negd(tcg_res, tcg_res);
3495
        break;
3496
    }
3497

    
3498
    write_fp_dreg(s, rd, tcg_res);
3499

    
3500
    tcg_temp_free_ptr(fpst);
3501
    tcg_temp_free_i64(tcg_op1);
3502
    tcg_temp_free_i64(tcg_op2);
3503
    tcg_temp_free_i64(tcg_res);
3504
}
3505

    
3506
/* C3.6.26 Floating point data-processing (2 source)
3507
 *   31  30  29 28       24 23  22  21 20  16 15    12 11 10 9    5 4    0
3508
 * +---+---+---+-----------+------+---+------+--------+-----+------+------+
3509
 * | M | 0 | S | 1 1 1 1 0 | type | 1 |  Rm  | opcode | 1 0 |  Rn  |  Rd  |
3510
 * +---+---+---+-----------+------+---+------+--------+-----+------+------+
3511
 */
3512
static void disas_fp_2src(DisasContext *s, uint32_t insn)
3513
{
3514
    int type = extract32(insn, 22, 2);
3515
    int rd = extract32(insn, 0, 5);
3516
    int rn = extract32(insn, 5, 5);
3517
    int rm = extract32(insn, 16, 5);
3518
    int opcode = extract32(insn, 12, 4);
3519

    
3520
    if (opcode > 8) {
3521
        unallocated_encoding(s);
3522
        return;
3523
    }
3524

    
3525
    switch (type) {
3526
    case 0:
3527
        handle_fp_2src_single(s, opcode, rd, rn, rm);
3528
        break;
3529
    case 1:
3530
        handle_fp_2src_double(s, opcode, rd, rn, rm);
3531
        break;
3532
    default:
3533
        unallocated_encoding(s);
3534
    }
3535
}
3536

    
3537
/* C3.6.27 Floating-point data-processing (3 source) - single precision */
3538
static void handle_fp_3src_single(DisasContext *s, bool o0, bool o1,
3539
                                  int rd, int rn, int rm, int ra)
3540
{
3541
    TCGv_i32 tcg_op1, tcg_op2, tcg_op3;
3542
    TCGv_i32 tcg_res = tcg_temp_new_i32();
3543
    TCGv_ptr fpst = get_fpstatus_ptr();
3544

    
3545
    tcg_op1 = read_fp_sreg(s, rn);
3546
    tcg_op2 = read_fp_sreg(s, rm);
3547
    tcg_op3 = read_fp_sreg(s, ra);
3548

    
3549
    /* These are fused multiply-add, and must be done as one
3550
     * floating point operation with no rounding between the
3551
     * multiplication and addition steps.
3552
     * NB that doing the negations here as separate steps is
3553
     * correct : an input NaN should come out with its sign bit
3554
     * flipped if it is a negated-input.
3555
     */
3556
    if (o1 == true) {
3557
        gen_helper_vfp_negs(tcg_op3, tcg_op3);
3558
    }
3559

    
3560
    if (o0 != o1) {
3561
        gen_helper_vfp_negs(tcg_op1, tcg_op1);
3562
    }
3563

    
3564
    gen_helper_vfp_muladds(tcg_res, tcg_op1, tcg_op2, tcg_op3, fpst);
3565

    
3566
    write_fp_sreg(s, rd, tcg_res);
3567

    
3568
    tcg_temp_free_ptr(fpst);
3569
    tcg_temp_free_i32(tcg_op1);
3570
    tcg_temp_free_i32(tcg_op2);
3571
    tcg_temp_free_i32(tcg_op3);
3572
    tcg_temp_free_i32(tcg_res);
3573
}
3574

    
3575
/* C3.6.27 Floating-point data-processing (3 source) - double precision */
3576
static void handle_fp_3src_double(DisasContext *s, bool o0, bool o1,
3577
                                  int rd, int rn, int rm, int ra)
3578
{
3579
    TCGv_i64 tcg_op1, tcg_op2, tcg_op3;
3580
    TCGv_i64 tcg_res = tcg_temp_new_i64();
3581
    TCGv_ptr fpst = get_fpstatus_ptr();
3582

    
3583
    tcg_op1 = read_fp_dreg(s, rn);
3584
    tcg_op2 = read_fp_dreg(s, rm);
3585
    tcg_op3 = read_fp_dreg(s, ra);
3586

    
3587
    /* These are fused multiply-add, and must be done as one
3588
     * floating point operation with no rounding between the
3589
     * multiplication and addition steps.
3590
     * NB that doing the negations here as separate steps is
3591
     * correct : an input NaN should come out with its sign bit
3592
     * flipped if it is a negated-input.
3593
     */
3594
    if (o1 == true) {
3595
        gen_helper_vfp_negd(tcg_op3, tcg_op3);
3596
    }
3597

    
3598
    if (o0 != o1) {
3599
        gen_helper_vfp_negd(tcg_op1, tcg_op1);
3600
    }
3601

    
3602
    gen_helper_vfp_muladdd(tcg_res, tcg_op1, tcg_op2, tcg_op3, fpst);
3603

    
3604
    write_fp_dreg(s, rd, tcg_res);
3605

    
3606
    tcg_temp_free_ptr(fpst);
3607
    tcg_temp_free_i64(tcg_op1);
3608
    tcg_temp_free_i64(tcg_op2);
3609
    tcg_temp_free_i64(tcg_op3);
3610
    tcg_temp_free_i64(tcg_res);
3611
}
3612

    
3613
/* C3.6.27 Floating point data-processing (3 source)
3614
 *   31  30  29 28       24 23  22  21  20  16  15  14  10 9    5 4    0
3615
 * +---+---+---+-----------+------+----+------+----+------+------+------+
3616
 * | M | 0 | S | 1 1 1 1 1 | type | o1 |  Rm  | o0 |  Ra  |  Rn  |  Rd  |
3617
 * +---+---+---+-----------+------+----+------+----+------+------+------+
3618
 */
3619
static void disas_fp_3src(DisasContext *s, uint32_t insn)
3620
{
3621
    int type = extract32(insn, 22, 2);
3622
    int rd = extract32(insn, 0, 5);
3623
    int rn = extract32(insn, 5, 5);
3624
    int ra = extract32(insn, 10, 5);
3625
    int rm = extract32(insn, 16, 5);
3626
    bool o0 = extract32(insn, 15, 1);
3627
    bool o1 = extract32(insn, 21, 1);
3628

    
3629
    switch (type) {
3630
    case 0:
3631
        handle_fp_3src_single(s, o0, o1, rd, rn, rm, ra);
3632
        break;
3633
    case 1:
3634
        handle_fp_3src_double(s, o0, o1, rd, rn, rm, ra);
3635
        break;
3636
    default:
3637
        unallocated_encoding(s);
3638
    }
3639
}
3640

    
3641
/* C3.6.28 Floating point immediate
3642
 *   31  30  29 28       24 23  22  21 20        13 12   10 9    5 4    0
3643
 * +---+---+---+-----------+------+---+------------+-------+------+------+
3644
 * | M | 0 | S | 1 1 1 1 0 | type | 1 |    imm8    | 1 0 0 | imm5 |  Rd  |
3645
 * +---+---+---+-----------+------+---+------------+-------+------+------+
3646
 */
3647
static void disas_fp_imm(DisasContext *s, uint32_t insn)
3648
{
3649
    int rd = extract32(insn, 0, 5);
3650
    int imm8 = extract32(insn, 13, 8);
3651
    int is_double = extract32(insn, 22, 2);
3652
    uint64_t imm;
3653
    TCGv_i64 tcg_res;
3654

    
3655
    if (is_double > 1) {
3656
        unallocated_encoding(s);
3657
        return;
3658
    }
3659

    
3660
    /* The imm8 encodes the sign bit, enough bits to represent
3661
     * an exponent in the range 01....1xx to 10....0xx,
3662
     * and the most significant 4 bits of the mantissa; see
3663
     * VFPExpandImm() in the v8 ARM ARM.
3664
     */
3665
    if (is_double) {
3666
        imm = (extract32(imm8, 7, 1) ? 0x8000 : 0) |
3667
            (extract32(imm8, 6, 1) ? 0x3fc0 : 0x4000) |
3668
            extract32(imm8, 0, 6);
3669
        imm <<= 48;
3670
    } else {
3671
        imm = (extract32(imm8, 7, 1) ? 0x8000 : 0) |
3672
            (extract32(imm8, 6, 1) ? 0x3e00 : 0x4000) |
3673
            (extract32(imm8, 0, 6) << 3);
3674
        imm <<= 16;
3675
    }
3676

    
3677
    tcg_res = tcg_const_i64(imm);
3678
    write_fp_dreg(s, rd, tcg_res);
3679
    tcg_temp_free_i64(tcg_res);
3680
}
3681

    
3682
/* Handle floating point <=> fixed point conversions. Note that we can
3683
 * also deal with fp <=> integer conversions as a special case (scale == 64)
3684
 * OPTME: consider handling that special case specially or at least skipping
3685
 * the call to scalbn in the helpers for zero shifts.
3686
 */
3687
static void handle_fpfpcvt(DisasContext *s, int rd, int rn, int opcode,
3688
                           bool itof, int rmode, int scale, int sf, int type)
3689
{
3690
    bool is_signed = !(opcode & 1);
3691
    bool is_double = type;
3692
    TCGv_ptr tcg_fpstatus;
3693
    TCGv_i32 tcg_shift;
3694

    
3695
    tcg_fpstatus = get_fpstatus_ptr();
3696

    
3697
    tcg_shift = tcg_const_i32(64 - scale);
3698

    
3699
    if (itof) {
3700
        TCGv_i64 tcg_int = cpu_reg(s, rn);
3701
        if (!sf) {
3702
            TCGv_i64 tcg_extend = new_tmp_a64(s);
3703

    
3704
            if (is_signed) {
3705
                tcg_gen_ext32s_i64(tcg_extend, tcg_int);
3706
            } else {
3707
                tcg_gen_ext32u_i64(tcg_extend, tcg_int);
3708
            }
3709

    
3710
            tcg_int = tcg_extend;
3711
        }
3712

    
3713
        if (is_double) {
3714
            TCGv_i64 tcg_double = tcg_temp_new_i64();
3715
            if (is_signed) {
3716
                gen_helper_vfp_sqtod(tcg_double, tcg_int,
3717
                                     tcg_shift, tcg_fpstatus);
3718
            } else {
3719
                gen_helper_vfp_uqtod(tcg_double, tcg_int,
3720
                                     tcg_shift, tcg_fpstatus);
3721
            }
3722
            write_fp_dreg(s, rd, tcg_double);
3723
            tcg_temp_free_i64(tcg_double);
3724
        } else {
3725
            TCGv_i32 tcg_single = tcg_temp_new_i32();
3726
            if (is_signed) {
3727
                gen_helper_vfp_sqtos(tcg_single, tcg_int,
3728
                                     tcg_shift, tcg_fpstatus);
3729
            } else {
3730
                gen_helper_vfp_uqtos(tcg_single, tcg_int,
3731
                                     tcg_shift, tcg_fpstatus);
3732
            }
3733
            write_fp_sreg(s, rd, tcg_single);
3734
            tcg_temp_free_i32(tcg_single);
3735
        }
3736
    } else {
3737
        TCGv_i64 tcg_int = cpu_reg(s, rd);
3738
        TCGv_i32 tcg_rmode;
3739

    
3740
        if (extract32(opcode, 2, 1)) {
3741
            /* There are too many rounding modes to all fit into rmode,
3742
             * so FCVTA[US] is a special case.
3743
             */
3744
            rmode = FPROUNDING_TIEAWAY;
3745
        }
3746

    
3747
        tcg_rmode = tcg_const_i32(arm_rmode_to_sf(rmode));
3748

    
3749
        gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3750

    
3751
        if (is_double) {
3752
            TCGv_i64 tcg_double = read_fp_dreg(s, rn);
3753
            if (is_signed) {
3754
                if (!sf) {
3755
                    gen_helper_vfp_tosld(tcg_int, tcg_double,
3756
                                         tcg_shift, tcg_fpstatus);
3757
                } else {
3758
                    gen_helper_vfp_tosqd(tcg_int, tcg_double,
3759
                                         tcg_shift, tcg_fpstatus);
3760
                }
3761
            } else {
3762
                if (!sf) {
3763
                    gen_helper_vfp_tould(tcg_int, tcg_double,
3764
                                         tcg_shift, tcg_fpstatus);
3765
                } else {
3766
                    gen_helper_vfp_touqd(tcg_int, tcg_double,
3767
                                         tcg_shift, tcg_fpstatus);
3768
                }
3769
            }
3770
            tcg_temp_free_i64(tcg_double);
3771
        } else {
3772
            TCGv_i32 tcg_single = read_fp_sreg(s, rn);
3773
            if (sf) {
3774
                if (is_signed) {
3775
                    gen_helper_vfp_tosqs(tcg_int, tcg_single,
3776
                                         tcg_shift, tcg_fpstatus);
3777
                } else {
3778
                    gen_helper_vfp_touqs(tcg_int, tcg_single,
3779
                                         tcg_shift, tcg_fpstatus);
3780
                }
3781
            } else {
3782
                TCGv_i32 tcg_dest = tcg_temp_new_i32();
3783
                if (is_signed) {
3784
                    gen_helper_vfp_tosls(tcg_dest, tcg_single,
3785
                                         tcg_shift, tcg_fpstatus);
3786
                } else {
3787
                    gen_helper_vfp_touls(tcg_dest, tcg_single,
3788
                                         tcg_shift, tcg_fpstatus);
3789
                }
3790
                tcg_gen_extu_i32_i64(tcg_int, tcg_dest);
3791
                tcg_temp_free_i32(tcg_dest);
3792
            }
3793
            tcg_temp_free_i32(tcg_single);
3794
        }
3795

    
3796
        gen_helper_set_rmode(tcg_rmode, tcg_rmode, cpu_env);
3797
        tcg_temp_free_i32(tcg_rmode);
3798

    
3799
        if (!sf) {
3800
            tcg_gen_ext32u_i64(tcg_int, tcg_int);
3801
        }
3802
    }
3803

    
3804
    tcg_temp_free_ptr(tcg_fpstatus);
3805
    tcg_temp_free_i32(tcg_shift);
3806
}
3807

    
3808
/* C3.6.29 Floating point <-> fixed point conversions
3809
 *   31   30  29 28       24 23  22  21 20   19 18    16 15   10 9    5 4    0
3810
 * +----+---+---+-----------+------+---+-------+--------+-------+------+------+
3811
 * | sf | 0 | S | 1 1 1 1 0 | type | 0 | rmode | opcode | scale |  Rn  |  Rd  |
3812
 * +----+---+---+-----------+------+---+-------+--------+-------+------+------+
3813
 */
3814
static void disas_fp_fixed_conv(DisasContext *s, uint32_t insn)
3815
{
3816
    int rd = extract32(insn, 0, 5);
3817
    int rn = extract32(insn, 5, 5);
3818
    int scale = extract32(insn, 10, 6);
3819
    int opcode = extract32(insn, 16, 3);
3820
    int rmode = extract32(insn, 19, 2);
3821
    int type = extract32(insn, 22, 2);
3822
    bool sbit = extract32(insn, 29, 1);
3823
    bool sf = extract32(insn, 31, 1);
3824
    bool itof;
3825

    
3826
    if (sbit || (type > 1)
3827
        || (!sf && scale < 32)) {
3828
        unallocated_encoding(s);
3829
        return;
3830
    }
3831

    
3832
    switch ((rmode << 3) | opcode) {
3833
    case 0x2: /* SCVTF */
3834
    case 0x3: /* UCVTF */
3835
        itof = true;
3836
        break;
3837
    case 0x18: /* FCVTZS */
3838
    case 0x19: /* FCVTZU */
3839
        itof = false;
3840
        break;
3841
    default:
3842
        unallocated_encoding(s);
3843
        return;
3844
    }
3845

    
3846
    handle_fpfpcvt(s, rd, rn, opcode, itof, FPROUNDING_ZERO, scale, sf, type);
3847
}
3848

    
3849
static void handle_fmov(DisasContext *s, int rd, int rn, int type, bool itof)
3850
{
3851
    /* FMOV: gpr to or from float, double, or top half of quad fp reg,
3852
     * without conversion.
3853
     */
3854

    
3855
    if (itof) {
3856
        TCGv_i64 tcg_rn = cpu_reg(s, rn);
3857

    
3858
        switch (type) {
3859
        case 0:
3860
        {
3861
            /* 32 bit */
3862
            TCGv_i64 tmp = tcg_temp_new_i64();
3863
            tcg_gen_ext32u_i64(tmp, tcg_rn);
3864
            tcg_gen_st_i64(tmp, cpu_env, fp_reg_offset(rd, MO_64));
3865
            tcg_gen_movi_i64(tmp, 0);
3866
            tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(rd));
3867
            tcg_temp_free_i64(tmp);
3868
            break;
3869
        }
3870
        case 1:
3871
        {
3872
            /* 64 bit */
3873
            TCGv_i64 tmp = tcg_const_i64(0);
3874
            tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_offset(rd, MO_64));
3875
            tcg_gen_st_i64(tmp, cpu_env, fp_reg_hi_offset(rd));
3876
            tcg_temp_free_i64(tmp);
3877
            break;
3878
        }
3879
        case 2:
3880
            /* 64 bit to top half. */
3881
            tcg_gen_st_i64(tcg_rn, cpu_env, fp_reg_hi_offset(rd));
3882
            break;
3883
        }
3884
    } else {
3885
        TCGv_i64 tcg_rd = cpu_reg(s, rd);
3886

    
3887
        switch (type) {
3888
        case 0:
3889
            /* 32 bit */
3890
            tcg_gen_ld32u_i64(tcg_rd, cpu_env, fp_reg_offset(rn, MO_32));
3891
            break;
3892
        case 1:
3893
            /* 64 bit */
3894
            tcg_gen_ld_i64(tcg_rd, cpu_env, fp_reg_offset(rn, MO_64));
3895
            break;
3896
        case 2:
3897
            /* 64 bits from top half */
3898
            tcg_gen_ld_i64(tcg_rd, cpu_env, fp_reg_hi_offset(rn));
3899
            break;
3900
        }
3901
    }
3902
}
3903

    
3904
/* C3.6.30 Floating point <-> integer conversions
3905
 *   31   30  29 28       24 23  22  21 20   19 18 16 15         10 9  5 4  0
3906
 * +----+---+---+-----------+------+---+-------+-----+-------------+----+----+
3907
 * | sf | 0 | S | 1 1 1 1 0 | type | 0 | rmode | opc | 0 0 0 0 0 0 | Rn | Rd |
3908
 * +----+---+---+-----------+------+---+-------+-----+-------------+----+----+
3909
 */
3910
static void disas_fp_int_conv(DisasContext *s, uint32_t insn)
3911
{
3912
    int rd = extract32(insn, 0, 5);
3913
    int rn = extract32(insn, 5, 5);
3914
    int opcode = extract32(insn, 16, 3);
3915
    int rmode = extract32(insn, 19, 2);
3916
    int type = extract32(insn, 22, 2);
3917
    bool sbit = extract32(insn, 29, 1);
3918
    bool sf = extract32(insn, 31, 1);
3919

    
3920
    if (!sbit && (rmode < 2) && (opcode > 5)) {
3921
        /* FMOV */
3922
        bool itof = opcode & 1;
3923

    
3924
        switch (sf << 3 | type << 1 | rmode) {
3925
        case 0x0: /* 32 bit */
3926
        case 0xa: /* 64 bit */
3927
        case 0xd: /* 64 bit to top half of quad */
3928
            break;
3929
        default:
3930
            /* all other sf/type/rmode combinations are invalid */
3931
            unallocated_encoding(s);
3932
            break;
3933
        }
3934

    
3935
        handle_fmov(s, rd, rn, type, itof);
3936
    } else {
3937
        /* actual FP conversions */
3938
        unsupported_encoding(s, insn);
3939
    }
3940
}
3941

    
3942
/* FP-specific subcases of table C3-6 (SIMD and FP data processing)
3943
 *   31  30  29 28     25 24                          0
3944
 * +---+---+---+---------+-----------------------------+
3945
 * |   | 0 |   | 1 1 1 1 |                             |
3946
 * +---+---+---+---------+-----------------------------+
3947
 */
3948
static void disas_data_proc_fp(DisasContext *s, uint32_t insn)
3949
{
3950
    if (extract32(insn, 24, 1)) {
3951
        /* Floating point data-processing (3 source) */
3952
        disas_fp_3src(s, insn);
3953
    } else if (extract32(insn, 21, 1) == 0) {
3954
        /* Floating point to fixed point conversions */
3955
        disas_fp_fixed_conv(s, insn);
3956
    } else {
3957
        switch (extract32(insn, 10, 2)) {
3958
        case 1:
3959
            /* Floating point conditional compare */
3960
            disas_fp_ccomp(s, insn);
3961
            break;
3962
        case 2:
3963
            /* Floating point data-processing (2 source) */
3964
            disas_fp_2src(s, insn);
3965
            break;
3966
        case 3:
3967
            /* Floating point conditional select */
3968
            disas_fp_csel(s, insn);
3969
            break;
3970
        case 0:
3971
            switch (ctz32(extract32(insn, 12, 4))) {
3972
            case 0: /* [15:12] == xxx1 */
3973
                /* Floating point immediate */
3974
                disas_fp_imm(s, insn);
3975
                break;
3976
            case 1: /* [15:12] == xx10 */
3977
                /* Floating point compare */
3978
                disas_fp_compare(s, insn);
3979
                break;
3980
            case 2: /* [15:12] == x100 */
3981
                /* Floating point data-processing (1 source) */
3982
                disas_fp_1src(s, insn);
3983
                break;
3984
            case 3: /* [15:12] == 1000 */
3985
                unallocated_encoding(s);
3986
                break;
3987
            default: /* [15:12] == 0000 */
3988
                /* Floating point <-> integer conversions */
3989
                disas_fp_int_conv(s, insn);
3990
                break;
3991
            }
3992
            break;
3993
        }
3994
    }
3995
}
3996

    
3997
static void disas_data_proc_simd(DisasContext *s, uint32_t insn)
3998
{
3999
    /* Note that this is called with all non-FP cases from
4000
     * table C3-6 so it must UNDEF for entries not specifically
4001
     * allocated to instructions in that table.
4002
     */
4003
    unsupported_encoding(s, insn);
4004
}
4005

    
4006
/* C3.6 Data processing - SIMD and floating point */
4007
static void disas_data_proc_simd_fp(DisasContext *s, uint32_t insn)
4008
{
4009
    if (extract32(insn, 28, 1) == 1 && extract32(insn, 30, 1) == 0) {
4010
        disas_data_proc_fp(s, insn);
4011
    } else {
4012
        /* SIMD, including crypto */
4013
        disas_data_proc_simd(s, insn);
4014
    }
4015
}
4016

    
4017
/* C3.1 A64 instruction index by encoding */
4018
static void disas_a64_insn(CPUARMState *env, DisasContext *s)
4019
{
4020
    uint32_t insn;
4021

    
4022
    insn = arm_ldl_code(env, s->pc, s->bswap_code);
4023
    s->insn = insn;
4024
    s->pc += 4;
4025

    
4026
    switch (extract32(insn, 25, 4)) {
4027
    case 0x0: case 0x1: case 0x2: case 0x3: /* UNALLOCATED */
4028
        unallocated_encoding(s);
4029
        break;
4030
    case 0x8: case 0x9: /* Data processing - immediate */
4031
        disas_data_proc_imm(s, insn);
4032
        break;
4033
    case 0xa: case 0xb: /* Branch, exception generation and system insns */
4034
        disas_b_exc_sys(s, insn);
4035
        break;
4036
    case 0x4:
4037
    case 0x6:
4038
    case 0xc:
4039
    case 0xe:      /* Loads and stores */
4040
        disas_ldst(s, insn);
4041
        break;
4042
    case 0x5:
4043
    case 0xd:      /* Data processing - register */
4044
        disas_data_proc_reg(s, insn);
4045
        break;
4046
    case 0x7:
4047
    case 0xf:      /* Data processing - SIMD and floating point */
4048
        disas_data_proc_simd_fp(s, insn);
4049
        break;
4050
    default:
4051
        assert(FALSE); /* all 15 cases should be handled above */
4052
        break;
4053
    }
4054

    
4055
    /* if we allocated any temporaries, free them here */
4056
    free_tmp_a64(s);
4057
}
4058

    
4059
void gen_intermediate_code_internal_a64(ARMCPU *cpu,
4060
                                        TranslationBlock *tb,
4061
                                        bool search_pc)
4062
{
4063
    CPUState *cs = CPU(cpu);
4064
    CPUARMState *env = &cpu->env;
4065
    DisasContext dc1, *dc = &dc1;
4066
    CPUBreakpoint *bp;
4067
    uint16_t *gen_opc_end;
4068
    int j, lj;
4069
    target_ulong pc_start;
4070
    target_ulong next_page_start;
4071
    int num_insns;
4072
    int max_insns;
4073

    
4074
    pc_start = tb->pc;
4075

    
4076
    dc->tb = tb;
4077

    
4078
    gen_opc_end = tcg_ctx.gen_opc_buf + OPC_MAX_SIZE;
4079

    
4080
    dc->is_jmp = DISAS_NEXT;
4081
    dc->pc = pc_start;
4082
    dc->singlestep_enabled = cs->singlestep_enabled;
4083
    dc->condjmp = 0;
4084

    
4085
    dc->aarch64 = 1;
4086
    dc->thumb = 0;
4087
    dc->bswap_code = 0;
4088
    dc->condexec_mask = 0;
4089
    dc->condexec_cond = 0;
4090
#if !defined(CONFIG_USER_ONLY)
4091
    dc->user = 0;
4092
#endif
4093
    dc->vfp_enabled = 0;
4094
    dc->vec_len = 0;
4095
    dc->vec_stride = 0;
4096
    dc->cp_regs = cpu->cp_regs;
4097
    dc->current_pl = arm_current_pl(env);
4098

    
4099
    init_tmp_a64_array(dc);
4100

    
4101
    next_page_start = (pc_start & TARGET_PAGE_MASK) + TARGET_PAGE_SIZE;
4102
    lj = -1;
4103
    num_insns = 0;
4104
    max_insns = tb->cflags & CF_COUNT_MASK;
4105
    if (max_insns == 0) {
4106
        max_insns = CF_COUNT_MASK;
4107
    }
4108

    
4109
    gen_tb_start();
4110

    
4111
    tcg_clear_temp_count();
4112

    
4113
    do {
4114
        if (unlikely(!QTAILQ_EMPTY(&env->breakpoints))) {
4115
            QTAILQ_FOREACH(bp, &env->breakpoints, entry) {
4116
                if (bp->pc == dc->pc) {
4117
                    gen_exception_insn(dc, 0, EXCP_DEBUG);
4118
                    /* Advance PC so that clearing the breakpoint will
4119
                       invalidate this TB.  */
4120
                    dc->pc += 2;
4121
                    goto done_generating;
4122
                }
4123
            }
4124
        }
4125

    
4126
        if (search_pc) {
4127
            j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
4128
            if (lj < j) {
4129
                lj++;
4130
                while (lj < j) {
4131
                    tcg_ctx.gen_opc_instr_start[lj++] = 0;
4132
                }
4133
            }
4134
            tcg_ctx.gen_opc_pc[lj] = dc->pc;
4135
            tcg_ctx.gen_opc_instr_start[lj] = 1;
4136
            tcg_ctx.gen_opc_icount[lj] = num_insns;
4137
        }
4138

    
4139
        if (num_insns + 1 == max_insns && (tb->cflags & CF_LAST_IO)) {
4140
            gen_io_start();
4141
        }
4142

    
4143
        if (unlikely(qemu_loglevel_mask(CPU_LOG_TB_OP | CPU_LOG_TB_OP_OPT))) {
4144
            tcg_gen_debug_insn_start(dc->pc);
4145
        }
4146

    
4147
        disas_a64_insn(env, dc);
4148

    
4149
        if (tcg_check_temp_count()) {
4150
            fprintf(stderr, "TCG temporary leak before "TARGET_FMT_lx"\n",
4151
                    dc->pc);
4152
        }
4153

    
4154
        /* Translation stops when a conditional branch is encountered.
4155
         * Otherwise the subsequent code could get translated several times.
4156
         * Also stop translation when a page boundary is reached.  This
4157
         * ensures prefetch aborts occur at the right place.
4158
         */
4159
        num_insns++;
4160
    } while (!dc->is_jmp && tcg_ctx.gen_opc_ptr < gen_opc_end &&
4161
             !cs->singlestep_enabled &&
4162
             !singlestep &&
4163
             dc->pc < next_page_start &&
4164
             num_insns < max_insns);
4165

    
4166
    if (tb->cflags & CF_LAST_IO) {
4167
        gen_io_end();
4168
    }
4169

    
4170
    if (unlikely(cs->singlestep_enabled) && dc->is_jmp != DISAS_EXC) {
4171
        /* Note that this means single stepping WFI doesn't halt the CPU.
4172
         * For conditional branch insns this is harmless unreachable code as
4173
         * gen_goto_tb() has already handled emitting the debug exception
4174
         * (and thus a tb-jump is not possible when singlestepping).
4175
         */
4176
        assert(dc->is_jmp != DISAS_TB_JUMP);
4177
        if (dc->is_jmp != DISAS_JUMP) {
4178
            gen_a64_set_pc_im(dc->pc);
4179
        }
4180
        gen_exception(EXCP_DEBUG);
4181
    } else {
4182
        switch (dc->is_jmp) {
4183
        case DISAS_NEXT:
4184
            gen_goto_tb(dc, 1, dc->pc);
4185
            break;
4186
        default:
4187
        case DISAS_UPDATE:
4188
            gen_a64_set_pc_im(dc->pc);
4189
            /* fall through */
4190
        case DISAS_JUMP:
4191
            /* indicate that the hash table must be used to find the next TB */
4192
            tcg_gen_exit_tb(0);
4193
            break;
4194
        case DISAS_TB_JUMP:
4195
        case DISAS_EXC:
4196
        case DISAS_SWI:
4197
            break;
4198
        case DISAS_WFI:
4199
            /* This is a special case because we don't want to just halt the CPU
4200
             * if trying to debug across a WFI.
4201
             */
4202
            gen_helper_wfi(cpu_env);
4203
            break;
4204
        }
4205
    }
4206

    
4207
done_generating:
4208
    gen_tb_end(tb, num_insns);
4209
    *tcg_ctx.gen_opc_ptr = INDEX_op_end;
4210

    
4211
#ifdef DEBUG_DISAS
4212
    if (qemu_loglevel_mask(CPU_LOG_TB_IN_ASM)) {
4213
        qemu_log("----------------\n");
4214
        qemu_log("IN: %s\n", lookup_symbol(pc_start));
4215
        log_target_disas(env, pc_start, dc->pc - pc_start,
4216
                         dc->thumb | (dc->bswap_code << 1));
4217
        qemu_log("\n");
4218
    }
4219
#endif
4220
    if (search_pc) {
4221
        j = tcg_ctx.gen_opc_ptr - tcg_ctx.gen_opc_buf;
4222
        lj++;
4223
        while (lj <= j) {
4224
            tcg_ctx.gen_opc_instr_start[lj++] = 0;
4225
        }
4226
    } else {
4227
        tb->size = dc->pc - pc_start;
4228
        tb->icount = num_insns;
4229
    }
4230
}