Blame view

kqemu.c 25.3 KB
bellard authored
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
/*
 *  KQEMU support
 * 
 *  Copyright (c) 2005 Fabrice Bellard
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */
#include "config.h"
#ifdef _WIN32
#include <windows.h>
23
#include <winioctl.h>
bellard authored
24
25
26
#else
#include <sys/types.h>
#include <sys/mman.h>
27
#include <sys/ioctl.h>
bellard authored
28
#endif
29
#ifdef HOST_SOLARIS
30
#include <sys/ioccom.h>
31
#endif
bellard authored
32
33
34
35
36
37
38
39
40
41
42
43
44
45
#include <stdlib.h>
#include <stdio.h>
#include <stdarg.h>
#include <string.h>
#include <errno.h>
#include <unistd.h>
#include <inttypes.h>

#include "cpu.h"
#include "exec-all.h"

#ifdef USE_KQEMU

#define DEBUG
46
//#define PROFILE
bellard authored
47
48
49

#include <unistd.h>
#include <fcntl.h>
bellard authored
50
#include "kqemu.h"
bellard authored
51
bellard authored
52
53
54
55
/* compatibility stuff */
#ifndef KQEMU_RET_SYSCALL
#define KQEMU_RET_SYSCALL   0x0300 /* syscall insn */
#endif
56
57
58
59
#ifndef KQEMU_MAX_RAM_PAGES_TO_UPDATE
#define KQEMU_MAX_RAM_PAGES_TO_UPDATE 512
#define KQEMU_RAM_PAGES_UPDATE_ALL (KQEMU_MAX_RAM_PAGES_TO_UPDATE + 1)
#endif
60
61
62
#ifndef KQEMU_MAX_MODIFIED_RAM_PAGES
#define KQEMU_MAX_MODIFIED_RAM_PAGES 512
#endif
bellard authored
63
64
65
66
#ifdef _WIN32
#define KQEMU_DEVICE "\\\\.\\kqemu"
#else
bellard authored
67
#define KQEMU_DEVICE "/dev/kqemu"
68
69
70
71
72
73
74
75
76
77
78
#endif

#ifdef _WIN32
#define KQEMU_INVALID_FD INVALID_HANDLE_VALUE
HANDLE kqemu_fd = KQEMU_INVALID_FD;
#define kqemu_closefd(x) CloseHandle(x)
#else
#define KQEMU_INVALID_FD -1
int kqemu_fd = KQEMU_INVALID_FD;
#define kqemu_closefd(x) close(x)
#endif
bellard authored
79
80
81
82
83
/* 0 = not allowed
   1 = user kqemu
   2 = kernel kqemu
*/
bellard authored
84
85
86
int kqemu_allowed = 1;
unsigned long *pages_to_flush;
unsigned int nb_pages_to_flush;
87
88
unsigned long *ram_pages_to_update;
unsigned int nb_ram_pages_to_update;
89
90
91
unsigned long *modified_ram_pages;
unsigned int nb_modified_ram_pages;
uint8_t *modified_ram_pages_table;
bellard authored
92
93
94
95
96
97
98
extern uint32_t **l1_phys_map;

#define cpuid(index, eax, ebx, ecx, edx) \
  asm volatile ("cpuid" \
                : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx) \
                : "0" (index))
bellard authored
99
100
101
102
103
104
#ifdef __x86_64__
static int is_cpuid_supported(void)
{
    return 1;
}
#else
bellard authored
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
static int is_cpuid_supported(void)
{
    int v0, v1;
    asm volatile ("pushf\n"
                  "popl %0\n"
                  "movl %0, %1\n"
                  "xorl $0x00200000, %0\n"
                  "pushl %0\n"
                  "popf\n"
                  "pushf\n"
                  "popl %0\n"
                  : "=a" (v0), "=d" (v1)
                  :
                  : "cc");
    return (v0 != v1);
}
bellard authored
121
#endif
bellard authored
122
123
124

static void kqemu_update_cpuid(CPUState *env)
{
bellard authored
125
    int critical_features_mask, features, ext_features, ext_features_mask;
bellard authored
126
127
128
129
130
131
132
133
134
    uint32_t eax, ebx, ecx, edx;

    /* the following features are kept identical on the host and
       target cpus because they are important for user code. Strictly
       speaking, only SSE really matters because the OS must support
       it if the user code uses it. */
    critical_features_mask = 
        CPUID_CMOV | CPUID_CX8 | 
        CPUID_FXSR | CPUID_MMX | CPUID_SSE | 
bellard authored
135
        CPUID_SSE2 | CPUID_SEP;
bellard authored
136
    ext_features_mask = CPUID_EXT_SSE3 | CPUID_EXT_MONITOR;
bellard authored
137
138
    if (!is_cpuid_supported()) {
        features = 0;
bellard authored
139
        ext_features = 0;
bellard authored
140
141
142
    } else {
        cpuid(1, eax, ebx, ecx, edx);
        features = edx;
bellard authored
143
        ext_features = ecx;
bellard authored
144
    }
bellard authored
145
146
147
148
149
150
#ifdef __x86_64__
    /* NOTE: on x86_64 CPUs, SYSENTER is not supported in
       compatibility mode, so in order to have the best performances
       it is better not to use it */
    features &= ~CPUID_SEP;
#endif
bellard authored
151
152
    env->cpuid_features = (env->cpuid_features & ~critical_features_mask) |
        (features & critical_features_mask);
bellard authored
153
154
    env->cpuid_ext_features = (env->cpuid_ext_features & ~ext_features_mask) |
        (ext_features & ext_features_mask);
bellard authored
155
156
157
158
159
160
161
162
163
    /* XXX: we could update more of the target CPUID state so that the
       non accelerated code sees exactly the same CPU features as the
       accelerated code */
}

int kqemu_init(CPUState *env)
{
    struct kqemu_init init;
    int ret, version;
164
165
166
#ifdef _WIN32
    DWORD temp;
#endif
bellard authored
167
168
169
170

    if (!kqemu_allowed)
        return -1;
171
172
173
174
175
176
#ifdef _WIN32
    kqemu_fd = CreateFile(KQEMU_DEVICE, GENERIC_WRITE | GENERIC_READ,
                          FILE_SHARE_READ | FILE_SHARE_WRITE,
                          NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL,
                          NULL);
#else
bellard authored
177
    kqemu_fd = open(KQEMU_DEVICE, O_RDWR);
178
179
#endif
    if (kqemu_fd == KQEMU_INVALID_FD) {
180
181
        fprintf(stderr, "Could not open '%s' - QEMU acceleration layer not activated: %s\n",
                KQEMU_DEVICE, strerror(errno));
bellard authored
182
183
184
        return -1;
    }
    version = 0;
185
186
187
188
#ifdef _WIN32
    DeviceIoControl(kqemu_fd, KQEMU_GET_VERSION, NULL, 0,
                    &version, sizeof(version), &temp, NULL);
#else
bellard authored
189
    ioctl(kqemu_fd, KQEMU_GET_VERSION, &version);
190
#endif
bellard authored
191
192
193
194
195
196
197
198
199
200
201
    if (version != KQEMU_VERSION) {
        fprintf(stderr, "Version mismatch between kqemu module and qemu (%08x %08x) - disabling kqemu use\n",
                version, KQEMU_VERSION);
        goto fail;
    }

    pages_to_flush = qemu_vmalloc(KQEMU_MAX_PAGES_TO_FLUSH * 
                                  sizeof(unsigned long));
    if (!pages_to_flush)
        goto fail;
202
203
204
205
206
    ram_pages_to_update = qemu_vmalloc(KQEMU_MAX_RAM_PAGES_TO_UPDATE * 
                                       sizeof(unsigned long));
    if (!ram_pages_to_update)
        goto fail;
207
208
209
210
211
212
213
214
    modified_ram_pages = qemu_vmalloc(KQEMU_MAX_MODIFIED_RAM_PAGES * 
                                      sizeof(unsigned long));
    if (!modified_ram_pages)
        goto fail;
    modified_ram_pages_table = qemu_mallocz(phys_ram_size >> TARGET_PAGE_BITS);
    if (!modified_ram_pages_table)
        goto fail;
bellard authored
215
216
217
218
219
    init.ram_base = phys_ram_base;
    init.ram_size = phys_ram_size;
    init.ram_dirty = phys_ram_dirty;
    init.phys_to_ram_map = l1_phys_map;
    init.pages_to_flush = pages_to_flush;
220
221
222
#if KQEMU_VERSION >= 0x010200
    init.ram_pages_to_update = ram_pages_to_update;
#endif
223
224
225
#if KQEMU_VERSION >= 0x010300
    init.modified_ram_pages = modified_ram_pages;
#endif
226
227
228
229
#ifdef _WIN32
    ret = DeviceIoControl(kqemu_fd, KQEMU_INIT, &init, sizeof(init),
                          NULL, 0, &temp, NULL) == TRUE ? 0 : -1;
#else
bellard authored
230
    ret = ioctl(kqemu_fd, KQEMU_INIT, &init);
231
#endif
bellard authored
232
233
234
    if (ret < 0) {
        fprintf(stderr, "Error %d while initializing QEMU acceleration layer - disabling it for now\n", ret);
    fail:
235
236
        kqemu_closefd(kqemu_fd);
        kqemu_fd = KQEMU_INVALID_FD;
bellard authored
237
238
239
        return -1;
    }
    kqemu_update_cpuid(env);
240
    env->kqemu_enabled = kqemu_allowed;
bellard authored
241
    nb_pages_to_flush = 0;
242
    nb_ram_pages_to_update = 0;
bellard authored
243
244
245
246
247
    return 0;
}

void kqemu_flush_page(CPUState *env, target_ulong addr)
{
248
#if defined(DEBUG)
bellard authored
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
    if (loglevel & CPU_LOG_INT) {
        fprintf(logfile, "kqemu_flush_page: addr=" TARGET_FMT_lx "\n", addr);
    }
#endif
    if (nb_pages_to_flush >= KQEMU_MAX_PAGES_TO_FLUSH)
        nb_pages_to_flush = KQEMU_FLUSH_ALL;
    else
        pages_to_flush[nb_pages_to_flush++] = addr;
}

void kqemu_flush(CPUState *env, int global)
{
#ifdef DEBUG
    if (loglevel & CPU_LOG_INT) {
        fprintf(logfile, "kqemu_flush:\n");
    }
#endif
    nb_pages_to_flush = KQEMU_FLUSH_ALL;
}
269
270
271
272
273
274
275
void kqemu_set_notdirty(CPUState *env, ram_addr_t ram_addr)
{
#ifdef DEBUG
    if (loglevel & CPU_LOG_INT) {
        fprintf(logfile, "kqemu_set_notdirty: addr=%08lx\n", ram_addr);
    }
#endif
bellard authored
276
277
278
    /* we only track transitions to dirty state */
    if (phys_ram_dirty[ram_addr >> TARGET_PAGE_BITS] != 0xff)
        return;
279
280
281
282
283
284
    if (nb_ram_pages_to_update >= KQEMU_MAX_RAM_PAGES_TO_UPDATE)
        nb_ram_pages_to_update = KQEMU_RAM_PAGES_UPDATE_ALL;
    else
        ram_pages_to_update[nb_ram_pages_to_update++] = ram_addr;
}
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
static void kqemu_reset_modified_ram_pages(void)
{
    int i;
    unsigned long page_index;

    for(i = 0; i < nb_modified_ram_pages; i++) {
        page_index = modified_ram_pages[i] >> TARGET_PAGE_BITS;
        modified_ram_pages_table[page_index] = 0;
    }
    nb_modified_ram_pages = 0;
}

void kqemu_modify_page(CPUState *env, ram_addr_t ram_addr)
{
    unsigned long page_index;
    int ret;
#ifdef _WIN32
    DWORD temp;
#endif

    page_index = ram_addr >> TARGET_PAGE_BITS;
    if (!modified_ram_pages_table[page_index]) {
#if 0
        printf("%d: modify_page=%08lx\n", nb_modified_ram_pages, ram_addr);
#endif
        modified_ram_pages_table[page_index] = 1;
        modified_ram_pages[nb_modified_ram_pages++] = ram_addr;
        if (nb_modified_ram_pages >= KQEMU_MAX_MODIFIED_RAM_PAGES) {
            /* flush */
#ifdef _WIN32
            ret = DeviceIoControl(kqemu_fd, KQEMU_MODIFY_RAM_PAGES, 
                                  &nb_modified_ram_pages, 
                                  sizeof(nb_modified_ram_pages),
                                  NULL, 0, &temp, NULL);
#else
            ret = ioctl(kqemu_fd, KQEMU_MODIFY_RAM_PAGES, 
                        &nb_modified_ram_pages);
#endif
            kqemu_reset_modified_ram_pages();
        }
    }
}
bellard authored
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
struct fpstate {
    uint16_t fpuc;
    uint16_t dummy1;
    uint16_t fpus;
    uint16_t dummy2;
    uint16_t fptag;
    uint16_t dummy3;

    uint32_t fpip;
    uint32_t fpcs;
    uint32_t fpoo;
    uint32_t fpos;
    uint8_t fpregs1[8 * 10];
};

struct fpxstate {
    uint16_t fpuc;
    uint16_t fpus;
    uint16_t fptag;
    uint16_t fop;
    uint32_t fpuip;
    uint16_t cs_sel;
    uint16_t dummy0;
    uint32_t fpudp;
    uint16_t ds_sel;
    uint16_t dummy1;
    uint32_t mxcsr;
    uint32_t mxcsr_mask;
    uint8_t fpregs1[8 * 16];
bellard authored
357
358
    uint8_t xmm_regs[16 * 16];
    uint8_t dummy2[96];
bellard authored
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
};

static struct fpxstate fpx1 __attribute__((aligned(16)));

static void restore_native_fp_frstor(CPUState *env)
{
    int fptag, i, j;
    struct fpstate fp1, *fp = &fp1;

    fp->fpuc = env->fpuc;
    fp->fpus = (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11;
    fptag = 0;
    for (i=7; i>=0; i--) {
	fptag <<= 2;
	if (env->fptags[i]) {
            fptag |= 3;
        } else {
            /* the FPU automatically computes it */
        }
    }
    fp->fptag = fptag;
    j = env->fpstt;
    for(i = 0;i < 8; i++) {
        memcpy(&fp->fpregs1[i * 10], &env->fpregs[j].d, 10);
        j = (j + 1) & 7;
    }
    asm volatile ("frstor %0" : "=m" (*fp));
}

static void save_native_fp_fsave(CPUState *env)
{
    int fptag, i, j;
    uint16_t fpuc;
    struct fpstate fp1, *fp = &fp1;

    asm volatile ("fsave %0" : : "m" (*fp));
    env->fpuc = fp->fpuc;
    env->fpstt = (fp->fpus >> 11) & 7;
    env->fpus = fp->fpus & ~0x3800;
    fptag = fp->fptag;
    for(i = 0;i < 8; i++) {
        env->fptags[i] = ((fptag & 3) == 3);
        fptag >>= 2;
    }
    j = env->fpstt;
    for(i = 0;i < 8; i++) {
        memcpy(&env->fpregs[j].d, &fp->fpregs1[i * 10], 10);
        j = (j + 1) & 7;
    }
    /* we must restore the default rounding state */
    fpuc = 0x037f | (env->fpuc & (3 << 10));
    asm volatile("fldcw %0" : : "m" (fpuc));
}

static void restore_native_fp_fxrstor(CPUState *env)
{
    struct fpxstate *fp = &fpx1;
    int i, j, fptag;

    fp->fpuc = env->fpuc;
    fp->fpus = (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11;
    fptag = 0;
    for(i = 0; i < 8; i++)
        fptag |= (env->fptags[i] << i);
    fp->fptag = fptag ^ 0xff;

    j = env->fpstt;
    for(i = 0;i < 8; i++) {
        memcpy(&fp->fpregs1[i * 16], &env->fpregs[j].d, 10);
        j = (j + 1) & 7;
    }
    if (env->cpuid_features & CPUID_SSE) {
        fp->mxcsr = env->mxcsr;
        /* XXX: check if DAZ is not available */
        fp->mxcsr_mask = 0xffff;
bellard authored
434
        memcpy(fp->xmm_regs, env->xmm_regs, CPU_NB_REGS * 16);
bellard authored
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
    }
    asm volatile ("fxrstor %0" : "=m" (*fp));
}

static void save_native_fp_fxsave(CPUState *env)
{
    struct fpxstate *fp = &fpx1;
    int fptag, i, j;
    uint16_t fpuc;

    asm volatile ("fxsave %0" : : "m" (*fp));
    env->fpuc = fp->fpuc;
    env->fpstt = (fp->fpus >> 11) & 7;
    env->fpus = fp->fpus & ~0x3800;
    fptag = fp->fptag ^ 0xff;
    for(i = 0;i < 8; i++) {
        env->fptags[i] = (fptag >> i) & 1;
    }
    j = env->fpstt;
    for(i = 0;i < 8; i++) {
        memcpy(&env->fpregs[j].d, &fp->fpregs1[i * 16], 10);
        j = (j + 1) & 7;
    }
    if (env->cpuid_features & CPUID_SSE) {
        env->mxcsr = fp->mxcsr;
bellard authored
460
        memcpy(env->xmm_regs, fp->xmm_regs, CPU_NB_REGS * 16);
bellard authored
461
462
463
464
465
466
467
468
    }

    /* we must restore the default rounding state */
    asm volatile ("fninit");
    fpuc = 0x037f | (env->fpuc & (3 << 10));
    asm volatile("fldcw %0" : : "m" (fpuc));
}
bellard authored
469
470
471
472
473
474
475
476
static int do_syscall(CPUState *env,
                      struct kqemu_cpu_state *kenv)
{
    int selector;

    selector = (env->star >> 32) & 0xffff;
#ifdef __x86_64__
    if (env->hflags & HF_LMA_MASK) {
477
478
        int code64;
bellard authored
479
480
481
        env->regs[R_ECX] = kenv->next_eip;
        env->regs[11] = env->eflags;
482
483
        code64 = env->hflags & HF_CS64_MASK;
bellard authored
484
485
486
        cpu_x86_set_cpl(env, 0);
        cpu_x86_load_seg_cache(env, R_CS, selector & 0xfffc, 
                               0, 0xffffffff, 
bellard authored
487
                               DESC_G_MASK | DESC_P_MASK |
bellard authored
488
489
490
491
492
493
494
495
                               DESC_S_MASK |
                               DESC_CS_MASK | DESC_R_MASK | DESC_A_MASK | DESC_L_MASK);
        cpu_x86_load_seg_cache(env, R_SS, (selector + 8) & 0xfffc, 
                               0, 0xffffffff,
                               DESC_G_MASK | DESC_B_MASK | DESC_P_MASK |
                               DESC_S_MASK |
                               DESC_W_MASK | DESC_A_MASK);
        env->eflags &= ~env->fmask;
496
        if (code64)
bellard authored
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
            env->eip = env->lstar;
        else
            env->eip = env->cstar;
    } else 
#endif
    {
        env->regs[R_ECX] = (uint32_t)kenv->next_eip;

        cpu_x86_set_cpl(env, 0);
        cpu_x86_load_seg_cache(env, R_CS, selector & 0xfffc, 
                           0, 0xffffffff, 
                               DESC_G_MASK | DESC_B_MASK | DESC_P_MASK |
                               DESC_S_MASK |
                               DESC_CS_MASK | DESC_R_MASK | DESC_A_MASK);
        cpu_x86_load_seg_cache(env, R_SS, (selector + 8) & 0xfffc, 
                               0, 0xffffffff,
                               DESC_G_MASK | DESC_B_MASK | DESC_P_MASK |
                               DESC_S_MASK |
                               DESC_W_MASK | DESC_A_MASK);
        env->eflags &= ~(IF_MASK | RF_MASK | VM_MASK);
        env->eip = (uint32_t)env->star;
    }
    return 2;
}
522
#ifdef CONFIG_PROFILER
523
524
525
526
527
528
529
530
531
532
533

#define PC_REC_SIZE 1
#define PC_REC_HASH_BITS 16
#define PC_REC_HASH_SIZE (1 << PC_REC_HASH_BITS)

typedef struct PCRecord {
    unsigned long pc;
    int64_t count;
    struct PCRecord *next;
} PCRecord;
534
535
static PCRecord *pc_rec_hash[PC_REC_HASH_SIZE];
static int nb_pc_records;
536
537
static void kqemu_record_pc(unsigned long pc)
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
{
    unsigned long h;
    PCRecord **pr, *r;

    h = pc / PC_REC_SIZE;
    h = h ^ (h >> PC_REC_HASH_BITS);
    h &= (PC_REC_HASH_SIZE - 1);
    pr = &pc_rec_hash[h];
    for(;;) {
        r = *pr;
        if (r == NULL)
            break;
        if (r->pc == pc) {
            r->count++;
            return;
        }
        pr = &r->next;
    }
    r = malloc(sizeof(PCRecord));
    r->count = 1;
    r->pc = pc;
    r->next = NULL;
    *pr = r;
    nb_pc_records++;
}
564
static int pc_rec_cmp(const void *p1, const void *p2)
565
566
567
568
569
570
571
572
573
574
575
{
    PCRecord *r1 = *(PCRecord **)p1;
    PCRecord *r2 = *(PCRecord **)p2;
    if (r1->count < r2->count)
        return 1;
    else if (r1->count == r2->count)
        return 0;
    else
        return -1;
}
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
static void kqemu_record_flush(void)
{
    PCRecord *r, *r_next;
    int h;

    for(h = 0; h < PC_REC_HASH_SIZE; h++) {
        for(r = pc_rec_hash[h]; r != NULL; r = r_next) {
            r_next = r->next;
            free(r);
        }
        pc_rec_hash[h] = NULL;
    }
    nb_pc_records = 0;
}
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
void kqemu_record_dump(void)
{
    PCRecord **pr, *r;
    int i, h;
    FILE *f;
    int64_t total, sum;

    pr = malloc(sizeof(PCRecord *) * nb_pc_records);
    i = 0;
    total = 0;
    for(h = 0; h < PC_REC_HASH_SIZE; h++) {
        for(r = pc_rec_hash[h]; r != NULL; r = r->next) {
            pr[i++] = r;
            total += r->count;
        }
    }
    qsort(pr, nb_pc_records, sizeof(PCRecord *), pc_rec_cmp);

    f = fopen("/tmp/kqemu.stats", "w");
    if (!f) {
        perror("/tmp/kqemu.stats");
        exit(1);
    }
bellard authored
614
    fprintf(f, "total: %" PRId64 "\n", total);
615
616
617
618
    sum = 0;
    for(i = 0; i < nb_pc_records; i++) {
        r = pr[i];
        sum += r->count;
bellard authored
619
        fprintf(f, "%08lx: %" PRId64 " %0.2f%% %0.2f%%\n", 
620
621
622
623
624
625
626
                r->pc, 
                r->count, 
                (double)r->count / (double)total * 100.0,
                (double)sum / (double)total * 100.0);
    }
    fclose(f);
    free(pr);
627
628

    kqemu_record_flush();
629
630
631
}
#endif
bellard authored
632
633
634
int kqemu_cpu_exec(CPUState *env)
{
    struct kqemu_cpu_state kcpu_state, *kenv = &kcpu_state;
635
636
637
638
639
    int ret, cpl, i;
#ifdef CONFIG_PROFILER
    int64_t ti;
#endif
640
641
642
#ifdef _WIN32
    DWORD temp;
#endif
bellard authored
643
644
645
646
#ifdef CONFIG_PROFILER
    ti = profile_getclock();
#endif
bellard authored
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
#ifdef DEBUG
    if (loglevel & CPU_LOG_INT) {
        fprintf(logfile, "kqemu: cpu_exec: enter\n");
        cpu_dump_state(env, logfile, fprintf, 0);
    }
#endif
    memcpy(kenv->regs, env->regs, sizeof(kenv->regs));
    kenv->eip = env->eip;
    kenv->eflags = env->eflags;
    memcpy(&kenv->segs, &env->segs, sizeof(env->segs));
    memcpy(&kenv->ldt, &env->ldt, sizeof(env->ldt));
    memcpy(&kenv->tr, &env->tr, sizeof(env->tr));
    memcpy(&kenv->gdt, &env->gdt, sizeof(env->gdt));
    memcpy(&kenv->idt, &env->idt, sizeof(env->idt));
    kenv->cr0 = env->cr[0];
    kenv->cr2 = env->cr[2];
    kenv->cr3 = env->cr[3];
    kenv->cr4 = env->cr[4];
    kenv->a20_mask = env->a20_mask;
666
#if KQEMU_VERSION >= 0x010100
bellard authored
667
668
    kenv->efer = env->efer;
#endif
669
670
671
672
673
674
675
676
677
678
679
680
681
#if KQEMU_VERSION >= 0x010300
    kenv->tsc_offset = 0;
    kenv->star = env->star;
    kenv->sysenter_cs = env->sysenter_cs;
    kenv->sysenter_esp = env->sysenter_esp;
    kenv->sysenter_eip = env->sysenter_eip;
#ifdef __x86_64__
    kenv->lstar = env->lstar;
    kenv->cstar = env->cstar;
    kenv->fmask = env->fmask;
    kenv->kernelgsbase = env->kernelgsbase;
#endif
#endif
bellard authored
682
683
684
685
686
687
688
689
690
691
    if (env->dr[7] & 0xff) {
        kenv->dr7 = env->dr[7];
        kenv->dr0 = env->dr[0];
        kenv->dr1 = env->dr[1];
        kenv->dr2 = env->dr[2];
        kenv->dr3 = env->dr[3];
    } else {
        kenv->dr7 = 0;
    }
    kenv->dr6 = env->dr[6];
692
693
    cpl = (env->hflags & HF_CPL_MASK);
    kenv->cpl = cpl;
bellard authored
694
    kenv->nb_pages_to_flush = nb_pages_to_flush;
695
#if KQEMU_VERSION >= 0x010200
696
    kenv->user_only = (env->kqemu_enabled == 1);
697
698
699
    kenv->nb_ram_pages_to_update = nb_ram_pages_to_update;
#endif
    nb_ram_pages_to_update = 0;
bellard authored
700
701
702
703
704
705
706
707
708
709
#if KQEMU_VERSION >= 0x010300
    kenv->nb_modified_ram_pages = nb_modified_ram_pages;
#endif
    kqemu_reset_modified_ram_pages();

    if (env->cpuid_features & CPUID_FXSR)
        restore_native_fp_fxrstor(env);
    else
        restore_native_fp_frstor(env);
bellard authored
710
711
#ifdef _WIN32
712
713
714
715
716
717
718
719
    if (DeviceIoControl(kqemu_fd, KQEMU_EXEC,
                        kenv, sizeof(struct kqemu_cpu_state),
                        kenv, sizeof(struct kqemu_cpu_state),
                        &temp, NULL)) {
        ret = kenv->retval;
    } else {
        ret = -1;
    }
720
721
722
723
724
#else
#if KQEMU_VERSION >= 0x010100
    ioctl(kqemu_fd, KQEMU_EXEC, kenv);
    ret = kenv->retval;
#else
bellard authored
725
    ret = ioctl(kqemu_fd, KQEMU_EXEC, kenv);
726
727
#endif
#endif
728
729
730
731
    if (env->cpuid_features & CPUID_FXSR)
        save_native_fp_fxsave(env);
    else
        save_native_fp_fsave(env);
bellard authored
732
733
734
735
736

    memcpy(env->regs, kenv->regs, sizeof(env->regs));
    env->eip = kenv->eip;
    env->eflags = kenv->eflags;
    memcpy(env->segs, kenv->segs, sizeof(env->segs));
737
738
    cpu_x86_set_cpl(env, kenv->cpl);
    memcpy(&env->ldt, &kenv->ldt, sizeof(env->ldt));
bellard authored
739
740
741
742
743
744
745
#if 0
    /* no need to restore that */
    memcpy(env->tr, kenv->tr, sizeof(env->tr));
    memcpy(env->gdt, kenv->gdt, sizeof(env->gdt));
    memcpy(env->idt, kenv->idt, sizeof(env->idt));
    env->a20_mask = kenv->a20_mask;
#endif
746
747
748
    env->cr[0] = kenv->cr0;
    env->cr[4] = kenv->cr4;
    env->cr[3] = kenv->cr3;
bellard authored
749
750
    env->cr[2] = kenv->cr2;
    env->dr[6] = kenv->dr6;
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
#if KQEMU_VERSION >= 0x010300
#ifdef __x86_64__
    env->kernelgsbase = kenv->kernelgsbase;
#endif
#endif

    /* flush pages as indicated by kqemu */
    if (kenv->nb_pages_to_flush >= KQEMU_FLUSH_ALL) {
        tlb_flush(env, 1);
    } else {
        for(i = 0; i < kenv->nb_pages_to_flush; i++) {
            tlb_flush_page(env, pages_to_flush[i]);
        }
    }
    nb_pages_to_flush = 0;

#ifdef CONFIG_PROFILER
    kqemu_time += profile_getclock() - ti;
    kqemu_exec_count++;
#endif
bellard authored
771
772
773
774
775
776
777
#if KQEMU_VERSION >= 0x010200
    if (kenv->nb_ram_pages_to_update > 0) {
        cpu_tlb_update_dirty(env);
    }
#endif
778
779
780
781
782
783
784
785
786
787
#if KQEMU_VERSION >= 0x010300
    if (kenv->nb_modified_ram_pages > 0) {
        for(i = 0; i < kenv->nb_modified_ram_pages; i++) {
            unsigned long addr;
            addr = modified_ram_pages[i];
            tb_invalidate_phys_page_range(addr, addr + TARGET_PAGE_SIZE, 0);
        }
    }
#endif
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
    /* restore the hidden flags */
    {
        unsigned int new_hflags;
#ifdef TARGET_X86_64
        if ((env->hflags & HF_LMA_MASK) && 
            (env->segs[R_CS].flags & DESC_L_MASK)) {
            /* long mode */
            new_hflags = HF_CS32_MASK | HF_SS32_MASK | HF_CS64_MASK;
        } else
#endif
        {
            /* legacy / compatibility case */
            new_hflags = (env->segs[R_CS].flags & DESC_B_MASK)
                >> (DESC_B_SHIFT - HF_CS32_SHIFT);
            new_hflags |= (env->segs[R_SS].flags & DESC_B_MASK)
                >> (DESC_B_SHIFT - HF_SS32_SHIFT);
            if (!(env->cr[0] & CR0_PE_MASK) || 
                   (env->eflags & VM_MASK) ||
                   !(env->hflags & HF_CS32_MASK)) {
                /* XXX: try to avoid this test. The problem comes from the
                   fact that is real mode or vm86 mode we only modify the
                   'base' and 'selector' fields of the segment cache to go
                   faster. A solution may be to force addseg to one in
                   translate-i386.c. */
                new_hflags |= HF_ADDSEG_MASK;
            } else {
                new_hflags |= ((env->segs[R_DS].base | 
                                env->segs[R_ES].base |
                                env->segs[R_SS].base) != 0) << 
                    HF_ADDSEG_SHIFT;
            }
        }
        env->hflags = (env->hflags & 
           ~(HF_CS32_MASK | HF_SS32_MASK | HF_CS64_MASK | HF_ADDSEG_MASK)) |
            new_hflags;
    }
824
825
826
827
828
829
830
831
    /* update FPU flags */
    env->hflags = (env->hflags & ~(HF_MP_MASK | HF_EM_MASK | HF_TS_MASK)) |
        ((env->cr[0] << (HF_MP_SHIFT - 1)) & (HF_MP_MASK | HF_EM_MASK | HF_TS_MASK));
    if (env->cr[4] & CR4_OSFXSR_MASK)
        env->hflags |= HF_OSFXSR_MASK;
    else
        env->hflags &= ~HF_OSFXSR_MASK;
bellard authored
832
833
834
835
836
#ifdef DEBUG
    if (loglevel & CPU_LOG_INT) {
        fprintf(logfile, "kqemu: kqemu_cpu_exec: ret=0x%x\n", ret);
    }
#endif
bellard authored
837
838
839
840
    if (ret == KQEMU_RET_SYSCALL) {
        /* syscall instruction */
        return do_syscall(env, kenv);
    } else 
bellard authored
841
842
843
844
845
    if ((ret & 0xff00) == KQEMU_RET_INT) {
        env->exception_index = ret & 0xff;
        env->error_code = 0;
        env->exception_is_int = 1;
        env->exception_next_eip = kenv->next_eip;
846
847
848
#ifdef CONFIG_PROFILER
        kqemu_ret_int_count++;
#endif
bellard authored
849
#ifdef DEBUG
bellard authored
850
851
852
853
854
        if (loglevel & CPU_LOG_INT) {
            fprintf(logfile, "kqemu: interrupt v=%02x:\n", 
                    env->exception_index);
            cpu_dump_state(env, logfile, fprintf, 0);
        }
bellard authored
855
856
857
858
859
860
861
#endif
        return 1;
    } else if ((ret & 0xff00) == KQEMU_RET_EXCEPTION) {
        env->exception_index = ret & 0xff;
        env->error_code = kenv->error_code;
        env->exception_is_int = 0;
        env->exception_next_eip = 0;
862
863
864
#ifdef CONFIG_PROFILER
        kqemu_ret_excp_count++;
#endif
bellard authored
865
866
867
868
869
870
871
872
873
#ifdef DEBUG
        if (loglevel & CPU_LOG_INT) {
            fprintf(logfile, "kqemu: exception v=%02x e=%04x:\n",
                    env->exception_index, env->error_code);
            cpu_dump_state(env, logfile, fprintf, 0);
        }
#endif
        return 1;
    } else if (ret == KQEMU_RET_INTR) {
874
875
876
#ifdef CONFIG_PROFILER
        kqemu_ret_intr_count++;
#endif
877
878
879
880
881
#ifdef DEBUG
        if (loglevel & CPU_LOG_INT) {
            cpu_dump_state(env, logfile, fprintf, 0);
        }
#endif
bellard authored
882
883
        return 0;
    } else if (ret == KQEMU_RET_SOFTMMU) { 
884
885
886
887
888
#ifdef CONFIG_PROFILER
        {
            unsigned long pc = env->eip + env->segs[R_CS].base;
            kqemu_record_pc(pc);
        }
889
890
891
892
893
894
#endif
#ifdef DEBUG
        if (loglevel & CPU_LOG_INT) {
            cpu_dump_state(env, logfile, fprintf, 0);
        }
#endif
bellard authored
895
896
897
898
899
900
901
902
903
        return 2;
    } else {
        cpu_dump_state(env, stderr, fprintf, 0);
        fprintf(stderr, "Unsupported return value: 0x%x\n", ret);
        exit(1);
    }
    return 0;
}
904
905
906
907
908
909
910
911
912
void kqemu_cpu_interrupt(CPUState *env)
{
#if defined(_WIN32) && KQEMU_VERSION >= 0x010101
    /* cancelling the I/O request causes KQEMU to finish executing the 
       current block and successfully returning. */
    CancelIo(kqemu_fd);
#endif
}
bellard authored
913
#endif