Statistics
| Branch: | Revision:

root / block.c @ a18f844f

History | View | Annotate | Download (61.7 kB)

1
/*
2
 * QEMU System Emulator block driver
3
 *
4
 * Copyright (c) 2003 Fabrice Bellard
5
 *
6
 * Permission is hereby granted, free of charge, to any person obtaining a copy
7
 * of this software and associated documentation files (the "Software"), to deal
8
 * in the Software without restriction, including without limitation the rights
9
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10
 * copies of the Software, and to permit persons to whom the Software is
11
 * furnished to do so, subject to the following conditions:
12
 *
13
 * The above copyright notice and this permission notice shall be included in
14
 * all copies or substantial portions of the Software.
15
 *
16
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22
 * THE SOFTWARE.
23
 */
24
#include "config-host.h"
25
#include "qemu-common.h"
26
#include "monitor.h"
27
#include "block_int.h"
28
#include "module.h"
29
#include "qemu-objects.h"
30

    
31
#ifdef CONFIG_BSD
32
#include <sys/types.h>
33
#include <sys/stat.h>
34
#include <sys/ioctl.h>
35
#include <sys/queue.h>
36
#ifndef __DragonFly__
37
#include <sys/disk.h>
38
#endif
39
#endif
40

    
41
#ifdef _WIN32
42
#include <windows.h>
43
#endif
44

    
45
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
46
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
47
        BlockDriverCompletionFunc *cb, void *opaque);
48
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
49
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
50
        BlockDriverCompletionFunc *cb, void *opaque);
51
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
52
        BlockDriverCompletionFunc *cb, void *opaque);
53
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
54
                        uint8_t *buf, int nb_sectors);
55
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
56
                         const uint8_t *buf, int nb_sectors);
57

    
58
BlockDriverState *bdrv_first;
59

    
60
static BlockDriver *first_drv;
61

    
62
/* If non-zero, use only whitelisted block drivers */
63
static int use_bdrv_whitelist;
64

    
65
int path_is_absolute(const char *path)
66
{
67
    const char *p;
68
#ifdef _WIN32
69
    /* specific case for names like: "\\.\d:" */
70
    if (*path == '/' || *path == '\\')
71
        return 1;
72
#endif
73
    p = strchr(path, ':');
74
    if (p)
75
        p++;
76
    else
77
        p = path;
78
#ifdef _WIN32
79
    return (*p == '/' || *p == '\\');
80
#else
81
    return (*p == '/');
82
#endif
83
}
84

    
85
/* if filename is absolute, just copy it to dest. Otherwise, build a
86
   path to it by considering it is relative to base_path. URL are
87
   supported. */
88
void path_combine(char *dest, int dest_size,
89
                  const char *base_path,
90
                  const char *filename)
91
{
92
    const char *p, *p1;
93
    int len;
94

    
95
    if (dest_size <= 0)
96
        return;
97
    if (path_is_absolute(filename)) {
98
        pstrcpy(dest, dest_size, filename);
99
    } else {
100
        p = strchr(base_path, ':');
101
        if (p)
102
            p++;
103
        else
104
            p = base_path;
105
        p1 = strrchr(base_path, '/');
106
#ifdef _WIN32
107
        {
108
            const char *p2;
109
            p2 = strrchr(base_path, '\\');
110
            if (!p1 || p2 > p1)
111
                p1 = p2;
112
        }
113
#endif
114
        if (p1)
115
            p1++;
116
        else
117
            p1 = base_path;
118
        if (p1 > p)
119
            p = p1;
120
        len = p - base_path;
121
        if (len > dest_size - 1)
122
            len = dest_size - 1;
123
        memcpy(dest, base_path, len);
124
        dest[len] = '\0';
125
        pstrcat(dest, dest_size, filename);
126
    }
127
}
128

    
129
void bdrv_register(BlockDriver *bdrv)
130
{
131
    if (!bdrv->bdrv_aio_readv) {
132
        /* add AIO emulation layer */
133
        bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
134
        bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
135
    } else if (!bdrv->bdrv_read) {
136
        /* add synchronous IO emulation layer */
137
        bdrv->bdrv_read = bdrv_read_em;
138
        bdrv->bdrv_write = bdrv_write_em;
139
    }
140

    
141
    if (!bdrv->bdrv_aio_flush)
142
        bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
143

    
144
    bdrv->next = first_drv;
145
    first_drv = bdrv;
146
}
147

    
148
/* create a new block device (by default it is empty) */
149
BlockDriverState *bdrv_new(const char *device_name)
150
{
151
    BlockDriverState **pbs, *bs;
152

    
153
    bs = qemu_mallocz(sizeof(BlockDriverState));
154
    pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
155
    if (device_name[0] != '\0') {
156
        /* insert at the end */
157
        pbs = &bdrv_first;
158
        while (*pbs != NULL)
159
            pbs = &(*pbs)->next;
160
        *pbs = bs;
161
    }
162
    return bs;
163
}
164

    
165
BlockDriver *bdrv_find_format(const char *format_name)
166
{
167
    BlockDriver *drv1;
168
    for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
169
        if (!strcmp(drv1->format_name, format_name))
170
            return drv1;
171
    }
172
    return NULL;
173
}
174

    
175
static int bdrv_is_whitelisted(BlockDriver *drv)
176
{
177
    static const char *whitelist[] = {
178
        CONFIG_BDRV_WHITELIST
179
    };
180
    const char **p;
181

    
182
    if (!whitelist[0])
183
        return 1;               /* no whitelist, anything goes */
184

    
185
    for (p = whitelist; *p; p++) {
186
        if (!strcmp(drv->format_name, *p)) {
187
            return 1;
188
        }
189
    }
190
    return 0;
191
}
192

    
193
BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
194
{
195
    BlockDriver *drv = bdrv_find_format(format_name);
196
    return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
197
}
198

    
199
int bdrv_create(BlockDriver *drv, const char* filename,
200
    QEMUOptionParameter *options)
201
{
202
    if (!drv->bdrv_create)
203
        return -ENOTSUP;
204

    
205
    return drv->bdrv_create(filename, options);
206
}
207

    
208
#ifdef _WIN32
209
void get_tmp_filename(char *filename, int size)
210
{
211
    char temp_dir[MAX_PATH];
212

    
213
    GetTempPath(MAX_PATH, temp_dir);
214
    GetTempFileName(temp_dir, "qem", 0, filename);
215
}
216
#else
217
void get_tmp_filename(char *filename, int size)
218
{
219
    int fd;
220
    const char *tmpdir;
221
    /* XXX: race condition possible */
222
    tmpdir = getenv("TMPDIR");
223
    if (!tmpdir)
224
        tmpdir = "/tmp";
225
    snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
226
    fd = mkstemp(filename);
227
    close(fd);
228
}
229
#endif
230

    
231
#ifdef _WIN32
232
static int is_windows_drive_prefix(const char *filename)
233
{
234
    return (((filename[0] >= 'a' && filename[0] <= 'z') ||
235
             (filename[0] >= 'A' && filename[0] <= 'Z')) &&
236
            filename[1] == ':');
237
}
238

    
239
int is_windows_drive(const char *filename)
240
{
241
    if (is_windows_drive_prefix(filename) &&
242
        filename[2] == '\0')
243
        return 1;
244
    if (strstart(filename, "\\\\.\\", NULL) ||
245
        strstart(filename, "//./", NULL))
246
        return 1;
247
    return 0;
248
}
249
#endif
250

    
251
static BlockDriver *find_protocol(const char *filename)
252
{
253
    BlockDriver *drv1;
254
    char protocol[128];
255
    int len;
256
    const char *p;
257

    
258
#ifdef _WIN32
259
    if (is_windows_drive(filename) ||
260
        is_windows_drive_prefix(filename))
261
        return bdrv_find_format("raw");
262
#endif
263
    p = strchr(filename, ':');
264
    if (!p)
265
        return bdrv_find_format("raw");
266
    len = p - filename;
267
    if (len > sizeof(protocol) - 1)
268
        len = sizeof(protocol) - 1;
269
    memcpy(protocol, filename, len);
270
    protocol[len] = '\0';
271
    for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
272
        if (drv1->protocol_name &&
273
            !strcmp(drv1->protocol_name, protocol))
274
            return drv1;
275
    }
276
    return NULL;
277
}
278

    
279
/*
280
 * Detect host devices. By convention, /dev/cdrom[N] is always
281
 * recognized as a host CDROM.
282
 */
283
static BlockDriver *find_hdev_driver(const char *filename)
284
{
285
    int score_max = 0, score;
286
    BlockDriver *drv = NULL, *d;
287

    
288
    for (d = first_drv; d; d = d->next) {
289
        if (d->bdrv_probe_device) {
290
            score = d->bdrv_probe_device(filename);
291
            if (score > score_max) {
292
                score_max = score;
293
                drv = d;
294
            }
295
        }
296
    }
297

    
298
    return drv;
299
}
300

    
301
static BlockDriver *find_image_format(const char *filename)
302
{
303
    int ret, score, score_max;
304
    BlockDriver *drv1, *drv;
305
    uint8_t buf[2048];
306
    BlockDriverState *bs;
307

    
308
    drv = find_protocol(filename);
309
    /* no need to test disk image formats for vvfat */
310
    if (drv && strcmp(drv->format_name, "vvfat") == 0)
311
        return drv;
312

    
313
    ret = bdrv_file_open(&bs, filename, 0);
314
    if (ret < 0)
315
        return NULL;
316
    ret = bdrv_pread(bs, 0, buf, sizeof(buf));
317
    bdrv_delete(bs);
318
    if (ret < 0) {
319
        return NULL;
320
    }
321

    
322
    score_max = 0;
323
    for(drv1 = first_drv; drv1 != NULL; drv1 = drv1->next) {
324
        if (drv1->bdrv_probe) {
325
            score = drv1->bdrv_probe(buf, ret, filename);
326
            if (score > score_max) {
327
                score_max = score;
328
                drv = drv1;
329
            }
330
        }
331
    }
332
    return drv;
333
}
334

    
335
int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
336
{
337
    BlockDriverState *bs;
338
    int ret;
339

    
340
    bs = bdrv_new("");
341
    ret = bdrv_open2(bs, filename, flags | BDRV_O_FILE, NULL);
342
    if (ret < 0) {
343
        bdrv_delete(bs);
344
        return ret;
345
    }
346
    bs->growable = 1;
347
    *pbs = bs;
348
    return 0;
349
}
350

    
351
int bdrv_open(BlockDriverState *bs, const char *filename, int flags)
352
{
353
    return bdrv_open2(bs, filename, flags, NULL);
354
}
355

    
356
int bdrv_open2(BlockDriverState *bs, const char *filename, int flags,
357
               BlockDriver *drv)
358
{
359
    int ret, open_flags;
360
    char tmp_filename[PATH_MAX];
361
    char backing_filename[PATH_MAX];
362

    
363
    bs->is_temporary = 0;
364
    bs->encrypted = 0;
365
    bs->valid_key = 0;
366
    bs->open_flags = flags;
367
    /* buffer_alignment defaulted to 512, drivers can change this value */
368
    bs->buffer_alignment = 512;
369

    
370
    if (flags & BDRV_O_SNAPSHOT) {
371
        BlockDriverState *bs1;
372
        int64_t total_size;
373
        int is_protocol = 0;
374
        BlockDriver *bdrv_qcow2;
375
        QEMUOptionParameter *options;
376

    
377
        /* if snapshot, we create a temporary backing file and open it
378
           instead of opening 'filename' directly */
379

    
380
        /* if there is a backing file, use it */
381
        bs1 = bdrv_new("");
382
        ret = bdrv_open2(bs1, filename, 0, drv);
383
        if (ret < 0) {
384
            bdrv_delete(bs1);
385
            return ret;
386
        }
387
        total_size = bdrv_getlength(bs1) >> BDRV_SECTOR_BITS;
388

    
389
        if (bs1->drv && bs1->drv->protocol_name)
390
            is_protocol = 1;
391

    
392
        bdrv_delete(bs1);
393

    
394
        get_tmp_filename(tmp_filename, sizeof(tmp_filename));
395

    
396
        /* Real path is meaningless for protocols */
397
        if (is_protocol)
398
            snprintf(backing_filename, sizeof(backing_filename),
399
                     "%s", filename);
400
        else if (!realpath(filename, backing_filename))
401
            return -errno;
402

    
403
        bdrv_qcow2 = bdrv_find_format("qcow2");
404
        options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
405

    
406
        set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size * 512);
407
        set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
408
        if (drv) {
409
            set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
410
                drv->format_name);
411
        }
412

    
413
        ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
414
        if (ret < 0) {
415
            return ret;
416
        }
417

    
418
        filename = tmp_filename;
419
        drv = bdrv_qcow2;
420
        bs->is_temporary = 1;
421
    }
422

    
423
    pstrcpy(bs->filename, sizeof(bs->filename), filename);
424
    if (flags & BDRV_O_FILE) {
425
        drv = find_protocol(filename);
426
    } else if (!drv) {
427
        drv = find_hdev_driver(filename);
428
        if (!drv) {
429
            drv = find_image_format(filename);
430
        }
431
    }
432

    
433
    if (!drv) {
434
        ret = -ENOENT;
435
        goto unlink_and_fail;
436
    }
437
    if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
438
        ret = -ENOTSUP;
439
        goto unlink_and_fail;
440
    }
441

    
442
    bs->drv = drv;
443
    bs->opaque = qemu_mallocz(drv->instance_size);
444

    
445
    /*
446
     * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
447
     * write cache to the guest.  We do need the fdatasync to flush
448
     * out transactions for block allocations, and we maybe have a
449
     * volatile write cache in our backing device to deal with.
450
     */
451
    if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
452
        bs->enable_write_cache = 1;
453

    
454
    /*
455
     * Clear flags that are internal to the block layer before opening the
456
     * image.
457
     */
458
    open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
459

    
460
    /*
461
     * Snapshots should be writeable.
462
     *
463
     * XXX(hch): and what is the point of a snapshot during a read-only open?
464
     */
465
    if (!(flags & BDRV_O_FILE) && bs->is_temporary) {
466
        open_flags |= BDRV_O_RDWR;
467
    }
468

    
469
    ret = drv->bdrv_open(bs, filename, open_flags);
470
    if (ret < 0) {
471
        goto free_and_fail;
472
    }
473

    
474
    bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
475
    if (drv->bdrv_getlength) {
476
        bs->total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
477
    }
478
#ifndef _WIN32
479
    if (bs->is_temporary) {
480
        unlink(filename);
481
    }
482
#endif
483
    if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
484
        /* if there is a backing file, use it */
485
        BlockDriver *back_drv = NULL;
486
        bs->backing_hd = bdrv_new("");
487
        path_combine(backing_filename, sizeof(backing_filename),
488
                     filename, bs->backing_file);
489
        if (bs->backing_format[0] != '\0')
490
            back_drv = bdrv_find_format(bs->backing_format);
491

    
492
        /* backing files always opened read-only */
493
        open_flags &= ~BDRV_O_RDWR;
494
        
495
        ret = bdrv_open2(bs->backing_hd, backing_filename, open_flags,
496
                         back_drv);
497
        if (ret < 0) {
498
            bdrv_close(bs);
499
            return ret;
500
        }
501
        if (bs->is_temporary) {
502
            bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
503
        } else {
504
            /* base image inherits from "parent" */
505
            bs->backing_hd->keep_read_only = bs->keep_read_only;
506
        }
507
    }
508

    
509
    if (!bdrv_key_required(bs)) {
510
        /* call the change callback */
511
        bs->media_changed = 1;
512
        if (bs->change_cb)
513
            bs->change_cb(bs->change_opaque);
514
    }
515
    return 0;
516

    
517
free_and_fail:
518
    qemu_free(bs->opaque);
519
    bs->opaque = NULL;
520
    bs->drv = NULL;
521
unlink_and_fail:
522
    if (bs->is_temporary)
523
        unlink(filename);
524
    return ret;
525
}
526

    
527
void bdrv_close(BlockDriverState *bs)
528
{
529
    if (bs->drv) {
530
        if (bs->backing_hd)
531
            bdrv_delete(bs->backing_hd);
532
        bs->drv->bdrv_close(bs);
533
        qemu_free(bs->opaque);
534
#ifdef _WIN32
535
        if (bs->is_temporary) {
536
            unlink(bs->filename);
537
        }
538
#endif
539
        bs->opaque = NULL;
540
        bs->drv = NULL;
541

    
542
        /* call the change callback */
543
        bs->media_changed = 1;
544
        if (bs->change_cb)
545
            bs->change_cb(bs->change_opaque);
546
    }
547
}
548

    
549
void bdrv_delete(BlockDriverState *bs)
550
{
551
    BlockDriverState **pbs;
552

    
553
    pbs = &bdrv_first;
554
    while (*pbs != bs && *pbs != NULL)
555
        pbs = &(*pbs)->next;
556
    if (*pbs == bs)
557
        *pbs = bs->next;
558

    
559
    bdrv_close(bs);
560
    qemu_free(bs);
561
}
562

    
563
/*
564
 * Run consistency checks on an image
565
 *
566
 * Returns the number of errors or -errno when an internal error occurs
567
 */
568
int bdrv_check(BlockDriverState *bs)
569
{
570
    if (bs->drv->bdrv_check == NULL) {
571
        return -ENOTSUP;
572
    }
573

    
574
    return bs->drv->bdrv_check(bs);
575
}
576

    
577
/* commit COW file into the raw image */
578
int bdrv_commit(BlockDriverState *bs)
579
{
580
    BlockDriver *drv = bs->drv;
581
    int64_t i, total_sectors;
582
    int n, j, ro, open_flags;
583
    int ret = 0, rw_ret = 0;
584
    unsigned char sector[512];
585
    char filename[1024];
586
    BlockDriverState *bs_rw, *bs_ro;
587

    
588
    if (!drv)
589
        return -ENOMEDIUM;
590
    
591
    if (!bs->backing_hd) {
592
        return -ENOTSUP;
593
    }
594

    
595
    if (bs->backing_hd->keep_read_only) {
596
        return -EACCES;
597
    }
598
    
599
    ro = bs->backing_hd->read_only;
600
    strncpy(filename, bs->backing_hd->filename, sizeof(filename));
601
    open_flags =  bs->backing_hd->open_flags;
602

    
603
    if (ro) {
604
        /* re-open as RW */
605
        bdrv_delete(bs->backing_hd);
606
        bs->backing_hd = NULL;
607
        bs_rw = bdrv_new("");
608
        rw_ret = bdrv_open2(bs_rw, filename, open_flags | BDRV_O_RDWR, NULL);
609
        if (rw_ret < 0) {
610
            bdrv_delete(bs_rw);
611
            /* try to re-open read-only */
612
            bs_ro = bdrv_new("");
613
            ret = bdrv_open2(bs_ro, filename, open_flags & ~BDRV_O_RDWR, NULL);
614
            if (ret < 0) {
615
                bdrv_delete(bs_ro);
616
                /* drive not functional anymore */
617
                bs->drv = NULL;
618
                return ret;
619
            }
620
            bs->backing_hd = bs_ro;
621
            return rw_ret;
622
        }
623
        bs->backing_hd = bs_rw;
624
    }
625

    
626
    total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
627
    for (i = 0; i < total_sectors;) {
628
        if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
629
            for(j = 0; j < n; j++) {
630
                if (bdrv_read(bs, i, sector, 1) != 0) {
631
                    ret = -EIO;
632
                    goto ro_cleanup;
633
                }
634

    
635
                if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
636
                    ret = -EIO;
637
                    goto ro_cleanup;
638
                }
639
                i++;
640
            }
641
        } else {
642
            i += n;
643
        }
644
    }
645

    
646
    if (drv->bdrv_make_empty) {
647
        ret = drv->bdrv_make_empty(bs);
648
        bdrv_flush(bs);
649
    }
650

    
651
    /*
652
     * Make sure all data we wrote to the backing device is actually
653
     * stable on disk.
654
     */
655
    if (bs->backing_hd)
656
        bdrv_flush(bs->backing_hd);
657

    
658
ro_cleanup:
659

    
660
    if (ro) {
661
        /* re-open as RO */
662
        bdrv_delete(bs->backing_hd);
663
        bs->backing_hd = NULL;
664
        bs_ro = bdrv_new("");
665
        ret = bdrv_open2(bs_ro, filename, open_flags & ~BDRV_O_RDWR, NULL);
666
        if (ret < 0) {
667
            bdrv_delete(bs_ro);
668
            /* drive not functional anymore */
669
            bs->drv = NULL;
670
            return ret;
671
        }
672
        bs->backing_hd = bs_ro;
673
        bs->backing_hd->keep_read_only = 0;
674
    }
675

    
676
    return ret;
677
}
678

    
679
/*
680
 * Return values:
681
 * 0        - success
682
 * -EINVAL  - backing format specified, but no file
683
 * -ENOSPC  - can't update the backing file because no space is left in the
684
 *            image file header
685
 * -ENOTSUP - format driver doesn't support changing the backing file
686
 */
687
int bdrv_change_backing_file(BlockDriverState *bs,
688
    const char *backing_file, const char *backing_fmt)
689
{
690
    BlockDriver *drv = bs->drv;
691

    
692
    if (drv->bdrv_change_backing_file != NULL) {
693
        return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
694
    } else {
695
        return -ENOTSUP;
696
    }
697
}
698

    
699
static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
700
                                   size_t size)
701
{
702
    int64_t len;
703

    
704
    if (!bdrv_is_inserted(bs))
705
        return -ENOMEDIUM;
706

    
707
    if (bs->growable)
708
        return 0;
709

    
710
    len = bdrv_getlength(bs);
711

    
712
    if (offset < 0)
713
        return -EIO;
714

    
715
    if ((offset > len) || (len - offset < size))
716
        return -EIO;
717

    
718
    return 0;
719
}
720

    
721
static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
722
                              int nb_sectors)
723
{
724
    return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);
725
}
726

    
727
/* return < 0 if error. See bdrv_write() for the return codes */
728
int bdrv_read(BlockDriverState *bs, int64_t sector_num,
729
              uint8_t *buf, int nb_sectors)
730
{
731
    BlockDriver *drv = bs->drv;
732

    
733
    if (!drv)
734
        return -ENOMEDIUM;
735
    if (bdrv_check_request(bs, sector_num, nb_sectors))
736
        return -EIO;
737

    
738
    return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
739
}
740

    
741
static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
742
                             int nb_sectors, int dirty)
743
{
744
    int64_t start, end;
745
    unsigned long val, idx, bit;
746

    
747
    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
748
    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
749

    
750
    for (; start <= end; start++) {
751
        idx = start / (sizeof(unsigned long) * 8);
752
        bit = start % (sizeof(unsigned long) * 8);
753
        val = bs->dirty_bitmap[idx];
754
        if (dirty) {
755
            if (!(val & (1 << bit))) {
756
                bs->dirty_count++;
757
                val |= 1 << bit;
758
            }
759
        } else {
760
            if (val & (1 << bit)) {
761
                bs->dirty_count--;
762
                val &= ~(1 << bit);
763
            }
764
        }
765
        bs->dirty_bitmap[idx] = val;
766
    }
767
}
768

    
769
/* Return < 0 if error. Important errors are:
770
  -EIO         generic I/O error (may happen for all errors)
771
  -ENOMEDIUM   No media inserted.
772
  -EINVAL      Invalid sector number or nb_sectors
773
  -EACCES      Trying to write a read-only device
774
*/
775
int bdrv_write(BlockDriverState *bs, int64_t sector_num,
776
               const uint8_t *buf, int nb_sectors)
777
{
778
    BlockDriver *drv = bs->drv;
779
    if (!bs->drv)
780
        return -ENOMEDIUM;
781
    if (bs->read_only)
782
        return -EACCES;
783
    if (bdrv_check_request(bs, sector_num, nb_sectors))
784
        return -EIO;
785

    
786
    if (bs->dirty_bitmap) {
787
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
788
    }
789

    
790
    return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
791
}
792

    
793
int bdrv_pread(BlockDriverState *bs, int64_t offset,
794
               void *buf, int count1)
795
{
796
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
797
    int len, nb_sectors, count;
798
    int64_t sector_num;
799
    int ret;
800

    
801
    count = count1;
802
    /* first read to align to sector start */
803
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
804
    if (len > count)
805
        len = count;
806
    sector_num = offset >> BDRV_SECTOR_BITS;
807
    if (len > 0) {
808
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
809
            return ret;
810
        memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
811
        count -= len;
812
        if (count == 0)
813
            return count1;
814
        sector_num++;
815
        buf += len;
816
    }
817

    
818
    /* read the sectors "in place" */
819
    nb_sectors = count >> BDRV_SECTOR_BITS;
820
    if (nb_sectors > 0) {
821
        if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
822
            return ret;
823
        sector_num += nb_sectors;
824
        len = nb_sectors << BDRV_SECTOR_BITS;
825
        buf += len;
826
        count -= len;
827
    }
828

    
829
    /* add data from the last sector */
830
    if (count > 0) {
831
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
832
            return ret;
833
        memcpy(buf, tmp_buf, count);
834
    }
835
    return count1;
836
}
837

    
838
int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
839
                const void *buf, int count1)
840
{
841
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
842
    int len, nb_sectors, count;
843
    int64_t sector_num;
844
    int ret;
845

    
846
    count = count1;
847
    /* first write to align to sector start */
848
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
849
    if (len > count)
850
        len = count;
851
    sector_num = offset >> BDRV_SECTOR_BITS;
852
    if (len > 0) {
853
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
854
            return ret;
855
        memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
856
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
857
            return ret;
858
        count -= len;
859
        if (count == 0)
860
            return count1;
861
        sector_num++;
862
        buf += len;
863
    }
864

    
865
    /* write the sectors "in place" */
866
    nb_sectors = count >> BDRV_SECTOR_BITS;
867
    if (nb_sectors > 0) {
868
        if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
869
            return ret;
870
        sector_num += nb_sectors;
871
        len = nb_sectors << BDRV_SECTOR_BITS;
872
        buf += len;
873
        count -= len;
874
    }
875

    
876
    /* add data from the last sector */
877
    if (count > 0) {
878
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
879
            return ret;
880
        memcpy(tmp_buf, buf, count);
881
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
882
            return ret;
883
    }
884
    return count1;
885
}
886

    
887
/**
888
 * Truncate file to 'offset' bytes (needed only for file protocols)
889
 */
890
int bdrv_truncate(BlockDriverState *bs, int64_t offset)
891
{
892
    BlockDriver *drv = bs->drv;
893
    if (!drv)
894
        return -ENOMEDIUM;
895
    if (!drv->bdrv_truncate)
896
        return -ENOTSUP;
897
    if (bs->read_only)
898
        return -EACCES;
899
    return drv->bdrv_truncate(bs, offset);
900
}
901

    
902
/**
903
 * Length of a file in bytes. Return < 0 if error or unknown.
904
 */
905
int64_t bdrv_getlength(BlockDriverState *bs)
906
{
907
    BlockDriver *drv = bs->drv;
908
    if (!drv)
909
        return -ENOMEDIUM;
910
    if (!drv->bdrv_getlength) {
911
        /* legacy mode */
912
        return bs->total_sectors * BDRV_SECTOR_SIZE;
913
    }
914
    return drv->bdrv_getlength(bs);
915
}
916

    
917
/* return 0 as number of sectors if no device present or error */
918
void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
919
{
920
    int64_t length;
921
    length = bdrv_getlength(bs);
922
    if (length < 0)
923
        length = 0;
924
    else
925
        length = length >> BDRV_SECTOR_BITS;
926
    *nb_sectors_ptr = length;
927
}
928

    
929
struct partition {
930
        uint8_t boot_ind;           /* 0x80 - active */
931
        uint8_t head;               /* starting head */
932
        uint8_t sector;             /* starting sector */
933
        uint8_t cyl;                /* starting cylinder */
934
        uint8_t sys_ind;            /* What partition type */
935
        uint8_t end_head;           /* end head */
936
        uint8_t end_sector;         /* end sector */
937
        uint8_t end_cyl;            /* end cylinder */
938
        uint32_t start_sect;        /* starting sector counting from 0 */
939
        uint32_t nr_sects;          /* nr of sectors in partition */
940
} __attribute__((packed));
941

    
942
/* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
943
static int guess_disk_lchs(BlockDriverState *bs,
944
                           int *pcylinders, int *pheads, int *psectors)
945
{
946
    uint8_t buf[512];
947
    int ret, i, heads, sectors, cylinders;
948
    struct partition *p;
949
    uint32_t nr_sects;
950
    uint64_t nb_sectors;
951

    
952
    bdrv_get_geometry(bs, &nb_sectors);
953

    
954
    ret = bdrv_read(bs, 0, buf, 1);
955
    if (ret < 0)
956
        return -1;
957
    /* test msdos magic */
958
    if (buf[510] != 0x55 || buf[511] != 0xaa)
959
        return -1;
960
    for(i = 0; i < 4; i++) {
961
        p = ((struct partition *)(buf + 0x1be)) + i;
962
        nr_sects = le32_to_cpu(p->nr_sects);
963
        if (nr_sects && p->end_head) {
964
            /* We make the assumption that the partition terminates on
965
               a cylinder boundary */
966
            heads = p->end_head + 1;
967
            sectors = p->end_sector & 63;
968
            if (sectors == 0)
969
                continue;
970
            cylinders = nb_sectors / (heads * sectors);
971
            if (cylinders < 1 || cylinders > 16383)
972
                continue;
973
            *pheads = heads;
974
            *psectors = sectors;
975
            *pcylinders = cylinders;
976
#if 0
977
            printf("guessed geometry: LCHS=%d %d %d\n",
978
                   cylinders, heads, sectors);
979
#endif
980
            return 0;
981
        }
982
    }
983
    return -1;
984
}
985

    
986
void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
987
{
988
    int translation, lba_detected = 0;
989
    int cylinders, heads, secs;
990
    uint64_t nb_sectors;
991

    
992
    /* if a geometry hint is available, use it */
993
    bdrv_get_geometry(bs, &nb_sectors);
994
    bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
995
    translation = bdrv_get_translation_hint(bs);
996
    if (cylinders != 0) {
997
        *pcyls = cylinders;
998
        *pheads = heads;
999
        *psecs = secs;
1000
    } else {
1001
        if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1002
            if (heads > 16) {
1003
                /* if heads > 16, it means that a BIOS LBA
1004
                   translation was active, so the default
1005
                   hardware geometry is OK */
1006
                lba_detected = 1;
1007
                goto default_geometry;
1008
            } else {
1009
                *pcyls = cylinders;
1010
                *pheads = heads;
1011
                *psecs = secs;
1012
                /* disable any translation to be in sync with
1013
                   the logical geometry */
1014
                if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1015
                    bdrv_set_translation_hint(bs,
1016
                                              BIOS_ATA_TRANSLATION_NONE);
1017
                }
1018
            }
1019
        } else {
1020
        default_geometry:
1021
            /* if no geometry, use a standard physical disk geometry */
1022
            cylinders = nb_sectors / (16 * 63);
1023

    
1024
            if (cylinders > 16383)
1025
                cylinders = 16383;
1026
            else if (cylinders < 2)
1027
                cylinders = 2;
1028
            *pcyls = cylinders;
1029
            *pheads = 16;
1030
            *psecs = 63;
1031
            if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1032
                if ((*pcyls * *pheads) <= 131072) {
1033
                    bdrv_set_translation_hint(bs,
1034
                                              BIOS_ATA_TRANSLATION_LARGE);
1035
                } else {
1036
                    bdrv_set_translation_hint(bs,
1037
                                              BIOS_ATA_TRANSLATION_LBA);
1038
                }
1039
            }
1040
        }
1041
        bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1042
    }
1043
}
1044

    
1045
void bdrv_set_geometry_hint(BlockDriverState *bs,
1046
                            int cyls, int heads, int secs)
1047
{
1048
    bs->cyls = cyls;
1049
    bs->heads = heads;
1050
    bs->secs = secs;
1051
}
1052

    
1053
void bdrv_set_type_hint(BlockDriverState *bs, int type)
1054
{
1055
    bs->type = type;
1056
    bs->removable = ((type == BDRV_TYPE_CDROM ||
1057
                      type == BDRV_TYPE_FLOPPY));
1058
}
1059

    
1060
void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1061
{
1062
    bs->translation = translation;
1063
}
1064

    
1065
void bdrv_get_geometry_hint(BlockDriverState *bs,
1066
                            int *pcyls, int *pheads, int *psecs)
1067
{
1068
    *pcyls = bs->cyls;
1069
    *pheads = bs->heads;
1070
    *psecs = bs->secs;
1071
}
1072

    
1073
int bdrv_get_type_hint(BlockDriverState *bs)
1074
{
1075
    return bs->type;
1076
}
1077

    
1078
int bdrv_get_translation_hint(BlockDriverState *bs)
1079
{
1080
    return bs->translation;
1081
}
1082

    
1083
int bdrv_is_removable(BlockDriverState *bs)
1084
{
1085
    return bs->removable;
1086
}
1087

    
1088
int bdrv_is_read_only(BlockDriverState *bs)
1089
{
1090
    return bs->read_only;
1091
}
1092

    
1093
int bdrv_is_sg(BlockDriverState *bs)
1094
{
1095
    return bs->sg;
1096
}
1097

    
1098
int bdrv_enable_write_cache(BlockDriverState *bs)
1099
{
1100
    return bs->enable_write_cache;
1101
}
1102

    
1103
/* XXX: no longer used */
1104
void bdrv_set_change_cb(BlockDriverState *bs,
1105
                        void (*change_cb)(void *opaque), void *opaque)
1106
{
1107
    bs->change_cb = change_cb;
1108
    bs->change_opaque = opaque;
1109
}
1110

    
1111
int bdrv_is_encrypted(BlockDriverState *bs)
1112
{
1113
    if (bs->backing_hd && bs->backing_hd->encrypted)
1114
        return 1;
1115
    return bs->encrypted;
1116
}
1117

    
1118
int bdrv_key_required(BlockDriverState *bs)
1119
{
1120
    BlockDriverState *backing_hd = bs->backing_hd;
1121

    
1122
    if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1123
        return 1;
1124
    return (bs->encrypted && !bs->valid_key);
1125
}
1126

    
1127
int bdrv_set_key(BlockDriverState *bs, const char *key)
1128
{
1129
    int ret;
1130
    if (bs->backing_hd && bs->backing_hd->encrypted) {
1131
        ret = bdrv_set_key(bs->backing_hd, key);
1132
        if (ret < 0)
1133
            return ret;
1134
        if (!bs->encrypted)
1135
            return 0;
1136
    }
1137
    if (!bs->encrypted) {
1138
        return -EINVAL;
1139
    } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1140
        return -ENOMEDIUM;
1141
    }
1142
    ret = bs->drv->bdrv_set_key(bs, key);
1143
    if (ret < 0) {
1144
        bs->valid_key = 0;
1145
    } else if (!bs->valid_key) {
1146
        bs->valid_key = 1;
1147
        /* call the change callback now, we skipped it on open */
1148
        bs->media_changed = 1;
1149
        if (bs->change_cb)
1150
            bs->change_cb(bs->change_opaque);
1151
    }
1152
    return ret;
1153
}
1154

    
1155
void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1156
{
1157
    if (!bs->drv) {
1158
        buf[0] = '\0';
1159
    } else {
1160
        pstrcpy(buf, buf_size, bs->drv->format_name);
1161
    }
1162
}
1163

    
1164
void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1165
                         void *opaque)
1166
{
1167
    BlockDriver *drv;
1168

    
1169
    for (drv = first_drv; drv != NULL; drv = drv->next) {
1170
        it(opaque, drv->format_name);
1171
    }
1172
}
1173

    
1174
BlockDriverState *bdrv_find(const char *name)
1175
{
1176
    BlockDriverState *bs;
1177

    
1178
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1179
        if (!strcmp(name, bs->device_name))
1180
            return bs;
1181
    }
1182
    return NULL;
1183
}
1184

    
1185
void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1186
{
1187
    BlockDriverState *bs;
1188

    
1189
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1190
        it(opaque, bs);
1191
    }
1192
}
1193

    
1194
const char *bdrv_get_device_name(BlockDriverState *bs)
1195
{
1196
    return bs->device_name;
1197
}
1198

    
1199
void bdrv_flush(BlockDriverState *bs)
1200
{
1201
    if (bs->drv && bs->drv->bdrv_flush)
1202
        bs->drv->bdrv_flush(bs);
1203
}
1204

    
1205
void bdrv_flush_all(void)
1206
{
1207
    BlockDriverState *bs;
1208

    
1209
    for (bs = bdrv_first; bs != NULL; bs = bs->next)
1210
        if (bs->drv && !bdrv_is_read_only(bs) && 
1211
            (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)))
1212
            bdrv_flush(bs);
1213
}
1214

    
1215
/*
1216
 * Returns true iff the specified sector is present in the disk image. Drivers
1217
 * not implementing the functionality are assumed to not support backing files,
1218
 * hence all their sectors are reported as allocated.
1219
 *
1220
 * 'pnum' is set to the number of sectors (including and immediately following
1221
 * the specified sector) that are known to be in the same
1222
 * allocated/unallocated state.
1223
 *
1224
 * 'nb_sectors' is the max value 'pnum' should be set to.
1225
 */
1226
int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1227
        int *pnum)
1228
{
1229
    int64_t n;
1230
    if (!bs->drv->bdrv_is_allocated) {
1231
        if (sector_num >= bs->total_sectors) {
1232
            *pnum = 0;
1233
            return 0;
1234
        }
1235
        n = bs->total_sectors - sector_num;
1236
        *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1237
        return 1;
1238
    }
1239
    return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1240
}
1241

    
1242
void bdrv_mon_event(const BlockDriverState *bdrv,
1243
                    BlockMonEventAction action, int is_read)
1244
{
1245
    QObject *data;
1246
    const char *action_str;
1247

    
1248
    switch (action) {
1249
    case BDRV_ACTION_REPORT:
1250
        action_str = "report";
1251
        break;
1252
    case BDRV_ACTION_IGNORE:
1253
        action_str = "ignore";
1254
        break;
1255
    case BDRV_ACTION_STOP:
1256
        action_str = "stop";
1257
        break;
1258
    default:
1259
        abort();
1260
    }
1261

    
1262
    data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1263
                              bdrv->device_name,
1264
                              action_str,
1265
                              is_read ? "read" : "write");
1266
    monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1267

    
1268
    qobject_decref(data);
1269
}
1270

    
1271
static void bdrv_print_dict(QObject *obj, void *opaque)
1272
{
1273
    QDict *bs_dict;
1274
    Monitor *mon = opaque;
1275

    
1276
    bs_dict = qobject_to_qdict(obj);
1277

    
1278
    monitor_printf(mon, "%s: type=%s removable=%d",
1279
                        qdict_get_str(bs_dict, "device"),
1280
                        qdict_get_str(bs_dict, "type"),
1281
                        qdict_get_bool(bs_dict, "removable"));
1282

    
1283
    if (qdict_get_bool(bs_dict, "removable")) {
1284
        monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1285
    }
1286

    
1287
    if (qdict_haskey(bs_dict, "inserted")) {
1288
        QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1289

    
1290
        monitor_printf(mon, " file=");
1291
        monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1292
        if (qdict_haskey(qdict, "backing_file")) {
1293
            monitor_printf(mon, " backing_file=");
1294
            monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1295
        }
1296
        monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1297
                            qdict_get_bool(qdict, "ro"),
1298
                            qdict_get_str(qdict, "drv"),
1299
                            qdict_get_bool(qdict, "encrypted"));
1300
    } else {
1301
        monitor_printf(mon, " [not inserted]");
1302
    }
1303

    
1304
    monitor_printf(mon, "\n");
1305
}
1306

    
1307
void bdrv_info_print(Monitor *mon, const QObject *data)
1308
{
1309
    qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1310
}
1311

    
1312
/**
1313
 * bdrv_info(): Block devices information
1314
 *
1315
 * Each block device information is stored in a QDict and the
1316
 * returned QObject is a QList of all devices.
1317
 *
1318
 * The QDict contains the following:
1319
 *
1320
 * - "device": device name
1321
 * - "type": device type
1322
 * - "removable": true if the device is removable, false otherwise
1323
 * - "locked": true if the device is locked, false otherwise
1324
 * - "inserted": only present if the device is inserted, it is a QDict
1325
 *    containing the following:
1326
 *          - "file": device file name
1327
 *          - "ro": true if read-only, false otherwise
1328
 *          - "drv": driver format name
1329
 *          - "backing_file": backing file name if one is used
1330
 *          - "encrypted": true if encrypted, false otherwise
1331
 *
1332
 * Example:
1333
 *
1334
 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1335
 *     "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1336
 *   { "device": "floppy0", "type": "floppy", "removable": true,
1337
 *     "locked": false } ]
1338
 */
1339
void bdrv_info(Monitor *mon, QObject **ret_data)
1340
{
1341
    QList *bs_list;
1342
    BlockDriverState *bs;
1343

    
1344
    bs_list = qlist_new();
1345

    
1346
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1347
        QObject *bs_obj;
1348
        const char *type = "unknown";
1349

    
1350
        switch(bs->type) {
1351
        case BDRV_TYPE_HD:
1352
            type = "hd";
1353
            break;
1354
        case BDRV_TYPE_CDROM:
1355
            type = "cdrom";
1356
            break;
1357
        case BDRV_TYPE_FLOPPY:
1358
            type = "floppy";
1359
            break;
1360
        }
1361

    
1362
        bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1363
                                    "'removable': %i, 'locked': %i }",
1364
                                    bs->device_name, type, bs->removable,
1365
                                    bs->locked);
1366

    
1367
        if (bs->drv) {
1368
            QObject *obj;
1369
            QDict *bs_dict = qobject_to_qdict(bs_obj);
1370

    
1371
            obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1372
                                     "'encrypted': %i }",
1373
                                     bs->filename, bs->read_only,
1374
                                     bs->drv->format_name,
1375
                                     bdrv_is_encrypted(bs));
1376
            if (bs->backing_file[0] != '\0') {
1377
                QDict *qdict = qobject_to_qdict(obj);
1378
                qdict_put(qdict, "backing_file",
1379
                          qstring_from_str(bs->backing_file));
1380
            }
1381

    
1382
            qdict_put_obj(bs_dict, "inserted", obj);
1383
        }
1384
        qlist_append_obj(bs_list, bs_obj);
1385
    }
1386

    
1387
    *ret_data = QOBJECT(bs_list);
1388
}
1389

    
1390
static void bdrv_stats_iter(QObject *data, void *opaque)
1391
{
1392
    QDict *qdict;
1393
    Monitor *mon = opaque;
1394

    
1395
    qdict = qobject_to_qdict(data);
1396
    monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1397

    
1398
    qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1399
    monitor_printf(mon, " rd_bytes=%" PRId64
1400
                        " wr_bytes=%" PRId64
1401
                        " rd_operations=%" PRId64
1402
                        " wr_operations=%" PRId64
1403
                        "\n",
1404
                        qdict_get_int(qdict, "rd_bytes"),
1405
                        qdict_get_int(qdict, "wr_bytes"),
1406
                        qdict_get_int(qdict, "rd_operations"),
1407
                        qdict_get_int(qdict, "wr_operations"));
1408
}
1409

    
1410
void bdrv_stats_print(Monitor *mon, const QObject *data)
1411
{
1412
    qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1413
}
1414

    
1415
/**
1416
 * bdrv_info_stats(): show block device statistics
1417
 *
1418
 * Each device statistic information is stored in a QDict and
1419
 * the returned QObject is a QList of all devices.
1420
 *
1421
 * The QDict contains the following:
1422
 *
1423
 * - "device": device name
1424
 * - "stats": A QDict with the statistics information, it contains:
1425
 *     - "rd_bytes": bytes read
1426
 *     - "wr_bytes": bytes written
1427
 *     - "rd_operations": read operations
1428
 *     - "wr_operations": write operations
1429
 * 
1430
 * Example:
1431
 *
1432
 * [ { "device": "ide0-hd0",
1433
 *               "stats": { "rd_bytes": 512,
1434
 *                          "wr_bytes": 0,
1435
 *                          "rd_operations": 1,
1436
 *                          "wr_operations": 0 } },
1437
 *   { "device": "ide1-cd0",
1438
 *               "stats": { "rd_bytes": 0,
1439
 *                          "wr_bytes": 0,
1440
 *                          "rd_operations": 0,
1441
 *                          "wr_operations": 0 } } ]
1442
 */
1443
void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1444
{
1445
    QObject *obj;
1446
    QList *devices;
1447
    BlockDriverState *bs;
1448

    
1449
    devices = qlist_new();
1450

    
1451
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1452
        obj = qobject_from_jsonf("{ 'device': %s, 'stats': {"
1453
                                 "'rd_bytes': %" PRId64 ","
1454
                                 "'wr_bytes': %" PRId64 ","
1455
                                 "'rd_operations': %" PRId64 ","
1456
                                 "'wr_operations': %" PRId64
1457
                                 "} }",
1458
                                 bs->device_name,
1459
                                 bs->rd_bytes, bs->wr_bytes,
1460
                                 bs->rd_ops, bs->wr_ops);
1461
        qlist_append_obj(devices, obj);
1462
    }
1463

    
1464
    *ret_data = QOBJECT(devices);
1465
}
1466

    
1467
const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1468
{
1469
    if (bs->backing_hd && bs->backing_hd->encrypted)
1470
        return bs->backing_file;
1471
    else if (bs->encrypted)
1472
        return bs->filename;
1473
    else
1474
        return NULL;
1475
}
1476

    
1477
void bdrv_get_backing_filename(BlockDriverState *bs,
1478
                               char *filename, int filename_size)
1479
{
1480
    if (!bs->backing_file) {
1481
        pstrcpy(filename, filename_size, "");
1482
    } else {
1483
        pstrcpy(filename, filename_size, bs->backing_file);
1484
    }
1485
}
1486

    
1487
int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1488
                          const uint8_t *buf, int nb_sectors)
1489
{
1490
    BlockDriver *drv = bs->drv;
1491
    if (!drv)
1492
        return -ENOMEDIUM;
1493
    if (!drv->bdrv_write_compressed)
1494
        return -ENOTSUP;
1495
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1496
        return -EIO;
1497

    
1498
    if (bs->dirty_bitmap) {
1499
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1500
    }
1501

    
1502
    return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1503
}
1504

    
1505
int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1506
{
1507
    BlockDriver *drv = bs->drv;
1508
    if (!drv)
1509
        return -ENOMEDIUM;
1510
    if (!drv->bdrv_get_info)
1511
        return -ENOTSUP;
1512
    memset(bdi, 0, sizeof(*bdi));
1513
    return drv->bdrv_get_info(bs, bdi);
1514
}
1515

    
1516
int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1517
                      int64_t pos, int size)
1518
{
1519
    BlockDriver *drv = bs->drv;
1520
    if (!drv)
1521
        return -ENOMEDIUM;
1522
    if (!drv->bdrv_save_vmstate)
1523
        return -ENOTSUP;
1524
    return drv->bdrv_save_vmstate(bs, buf, pos, size);
1525
}
1526

    
1527
int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1528
                      int64_t pos, int size)
1529
{
1530
    BlockDriver *drv = bs->drv;
1531
    if (!drv)
1532
        return -ENOMEDIUM;
1533
    if (!drv->bdrv_load_vmstate)
1534
        return -ENOTSUP;
1535
    return drv->bdrv_load_vmstate(bs, buf, pos, size);
1536
}
1537

    
1538
/**************************************************************/
1539
/* handling of snapshots */
1540

    
1541
int bdrv_snapshot_create(BlockDriverState *bs,
1542
                         QEMUSnapshotInfo *sn_info)
1543
{
1544
    BlockDriver *drv = bs->drv;
1545
    if (!drv)
1546
        return -ENOMEDIUM;
1547
    if (!drv->bdrv_snapshot_create)
1548
        return -ENOTSUP;
1549
    return drv->bdrv_snapshot_create(bs, sn_info);
1550
}
1551

    
1552
int bdrv_snapshot_goto(BlockDriverState *bs,
1553
                       const char *snapshot_id)
1554
{
1555
    BlockDriver *drv = bs->drv;
1556
    if (!drv)
1557
        return -ENOMEDIUM;
1558
    if (!drv->bdrv_snapshot_goto)
1559
        return -ENOTSUP;
1560
    return drv->bdrv_snapshot_goto(bs, snapshot_id);
1561
}
1562

    
1563
int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1564
{
1565
    BlockDriver *drv = bs->drv;
1566
    if (!drv)
1567
        return -ENOMEDIUM;
1568
    if (!drv->bdrv_snapshot_delete)
1569
        return -ENOTSUP;
1570
    return drv->bdrv_snapshot_delete(bs, snapshot_id);
1571
}
1572

    
1573
int bdrv_snapshot_list(BlockDriverState *bs,
1574
                       QEMUSnapshotInfo **psn_info)
1575
{
1576
    BlockDriver *drv = bs->drv;
1577
    if (!drv)
1578
        return -ENOMEDIUM;
1579
    if (!drv->bdrv_snapshot_list)
1580
        return -ENOTSUP;
1581
    return drv->bdrv_snapshot_list(bs, psn_info);
1582
}
1583

    
1584
#define NB_SUFFIXES 4
1585

    
1586
char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1587
{
1588
    static const char suffixes[NB_SUFFIXES] = "KMGT";
1589
    int64_t base;
1590
    int i;
1591

    
1592
    if (size <= 999) {
1593
        snprintf(buf, buf_size, "%" PRId64, size);
1594
    } else {
1595
        base = 1024;
1596
        for(i = 0; i < NB_SUFFIXES; i++) {
1597
            if (size < (10 * base)) {
1598
                snprintf(buf, buf_size, "%0.1f%c",
1599
                         (double)size / base,
1600
                         suffixes[i]);
1601
                break;
1602
            } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1603
                snprintf(buf, buf_size, "%" PRId64 "%c",
1604
                         ((size + (base >> 1)) / base),
1605
                         suffixes[i]);
1606
                break;
1607
            }
1608
            base = base * 1024;
1609
        }
1610
    }
1611
    return buf;
1612
}
1613

    
1614
char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1615
{
1616
    char buf1[128], date_buf[128], clock_buf[128];
1617
#ifdef _WIN32
1618
    struct tm *ptm;
1619
#else
1620
    struct tm tm;
1621
#endif
1622
    time_t ti;
1623
    int64_t secs;
1624

    
1625
    if (!sn) {
1626
        snprintf(buf, buf_size,
1627
                 "%-10s%-20s%7s%20s%15s",
1628
                 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1629
    } else {
1630
        ti = sn->date_sec;
1631
#ifdef _WIN32
1632
        ptm = localtime(&ti);
1633
        strftime(date_buf, sizeof(date_buf),
1634
                 "%Y-%m-%d %H:%M:%S", ptm);
1635
#else
1636
        localtime_r(&ti, &tm);
1637
        strftime(date_buf, sizeof(date_buf),
1638
                 "%Y-%m-%d %H:%M:%S", &tm);
1639
#endif
1640
        secs = sn->vm_clock_nsec / 1000000000;
1641
        snprintf(clock_buf, sizeof(clock_buf),
1642
                 "%02d:%02d:%02d.%03d",
1643
                 (int)(secs / 3600),
1644
                 (int)((secs / 60) % 60),
1645
                 (int)(secs % 60),
1646
                 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1647
        snprintf(buf, buf_size,
1648
                 "%-10s%-20s%7s%20s%15s",
1649
                 sn->id_str, sn->name,
1650
                 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1651
                 date_buf,
1652
                 clock_buf);
1653
    }
1654
    return buf;
1655
}
1656

    
1657

    
1658
/**************************************************************/
1659
/* async I/Os */
1660

    
1661
BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1662
                                 QEMUIOVector *qiov, int nb_sectors,
1663
                                 BlockDriverCompletionFunc *cb, void *opaque)
1664
{
1665
    BlockDriver *drv = bs->drv;
1666
    BlockDriverAIOCB *ret;
1667

    
1668
    if (!drv)
1669
        return NULL;
1670
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1671
        return NULL;
1672

    
1673
    ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1674
                              cb, opaque);
1675

    
1676
    if (ret) {
1677
        /* Update stats even though technically transfer has not happened. */
1678
        bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1679
        bs->rd_ops ++;
1680
    }
1681

    
1682
    return ret;
1683
}
1684

    
1685
BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1686
                                  QEMUIOVector *qiov, int nb_sectors,
1687
                                  BlockDriverCompletionFunc *cb, void *opaque)
1688
{
1689
    BlockDriver *drv = bs->drv;
1690
    BlockDriverAIOCB *ret;
1691

    
1692
    if (!drv)
1693
        return NULL;
1694
    if (bs->read_only)
1695
        return NULL;
1696
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1697
        return NULL;
1698

    
1699
    if (bs->dirty_bitmap) {
1700
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1701
    }
1702

    
1703
    ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1704
                               cb, opaque);
1705

    
1706
    if (ret) {
1707
        /* Update stats even though technically transfer has not happened. */
1708
        bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1709
        bs->wr_ops ++;
1710
    }
1711

    
1712
    return ret;
1713
}
1714

    
1715

    
1716
typedef struct MultiwriteCB {
1717
    int error;
1718
    int num_requests;
1719
    int num_callbacks;
1720
    struct {
1721
        BlockDriverCompletionFunc *cb;
1722
        void *opaque;
1723
        QEMUIOVector *free_qiov;
1724
        void *free_buf;
1725
    } callbacks[];
1726
} MultiwriteCB;
1727

    
1728
static void multiwrite_user_cb(MultiwriteCB *mcb)
1729
{
1730
    int i;
1731

    
1732
    for (i = 0; i < mcb->num_callbacks; i++) {
1733
        mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1734
        qemu_free(mcb->callbacks[i].free_qiov);
1735
        qemu_vfree(mcb->callbacks[i].free_buf);
1736
    }
1737
}
1738

    
1739
static void multiwrite_cb(void *opaque, int ret)
1740
{
1741
    MultiwriteCB *mcb = opaque;
1742

    
1743
    if (ret < 0) {
1744
        mcb->error = ret;
1745
        multiwrite_user_cb(mcb);
1746
    }
1747

    
1748
    mcb->num_requests--;
1749
    if (mcb->num_requests == 0) {
1750
        if (mcb->error == 0) {
1751
            multiwrite_user_cb(mcb);
1752
        }
1753
        qemu_free(mcb);
1754
    }
1755
}
1756

    
1757
static int multiwrite_req_compare(const void *a, const void *b)
1758
{
1759
    return (((BlockRequest*) a)->sector - ((BlockRequest*) b)->sector);
1760
}
1761

    
1762
/*
1763
 * Takes a bunch of requests and tries to merge them. Returns the number of
1764
 * requests that remain after merging.
1765
 */
1766
static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1767
    int num_reqs, MultiwriteCB *mcb)
1768
{
1769
    int i, outidx;
1770

    
1771
    // Sort requests by start sector
1772
    qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1773

    
1774
    // Check if adjacent requests touch the same clusters. If so, combine them,
1775
    // filling up gaps with zero sectors.
1776
    outidx = 0;
1777
    for (i = 1; i < num_reqs; i++) {
1778
        int merge = 0;
1779
        int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1780

    
1781
        // This handles the cases that are valid for all block drivers, namely
1782
        // exactly sequential writes and overlapping writes.
1783
        if (reqs[i].sector <= oldreq_last) {
1784
            merge = 1;
1785
        }
1786

    
1787
        // The block driver may decide that it makes sense to combine requests
1788
        // even if there is a gap of some sectors between them. In this case,
1789
        // the gap is filled with zeros (therefore only applicable for yet
1790
        // unused space in format like qcow2).
1791
        if (!merge && bs->drv->bdrv_merge_requests) {
1792
            merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1793
        }
1794

    
1795
        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
1796
            merge = 0;
1797
        }
1798

    
1799
        if (merge) {
1800
            size_t size;
1801
            QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1802
            qemu_iovec_init(qiov,
1803
                reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1804

    
1805
            // Add the first request to the merged one. If the requests are
1806
            // overlapping, drop the last sectors of the first request.
1807
            size = (reqs[i].sector - reqs[outidx].sector) << 9;
1808
            qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1809

    
1810
            // We might need to add some zeros between the two requests
1811
            if (reqs[i].sector > oldreq_last) {
1812
                size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1813
                uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1814
                memset(buf, 0, zero_bytes);
1815
                qemu_iovec_add(qiov, buf, zero_bytes);
1816
                mcb->callbacks[i].free_buf = buf;
1817
            }
1818

    
1819
            // Add the second request
1820
            qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
1821

    
1822
            reqs[outidx].nb_sectors += reqs[i].nb_sectors;
1823
            reqs[outidx].qiov = qiov;
1824

    
1825
            mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
1826
        } else {
1827
            outidx++;
1828
            reqs[outidx].sector     = reqs[i].sector;
1829
            reqs[outidx].nb_sectors = reqs[i].nb_sectors;
1830
            reqs[outidx].qiov       = reqs[i].qiov;
1831
        }
1832
    }
1833

    
1834
    return outidx + 1;
1835
}
1836

    
1837
/*
1838
 * Submit multiple AIO write requests at once.
1839
 *
1840
 * On success, the function returns 0 and all requests in the reqs array have
1841
 * been submitted. In error case this function returns -1, and any of the
1842
 * requests may or may not be submitted yet. In particular, this means that the
1843
 * callback will be called for some of the requests, for others it won't. The
1844
 * caller must check the error field of the BlockRequest to wait for the right
1845
 * callbacks (if error != 0, no callback will be called).
1846
 *
1847
 * The implementation may modify the contents of the reqs array, e.g. to merge
1848
 * requests. However, the fields opaque and error are left unmodified as they
1849
 * are used to signal failure for a single request to the caller.
1850
 */
1851
int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
1852
{
1853
    BlockDriverAIOCB *acb;
1854
    MultiwriteCB *mcb;
1855
    int i;
1856

    
1857
    if (num_reqs == 0) {
1858
        return 0;
1859
    }
1860

    
1861
    // Create MultiwriteCB structure
1862
    mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
1863
    mcb->num_requests = 0;
1864
    mcb->num_callbacks = num_reqs;
1865

    
1866
    for (i = 0; i < num_reqs; i++) {
1867
        mcb->callbacks[i].cb = reqs[i].cb;
1868
        mcb->callbacks[i].opaque = reqs[i].opaque;
1869
    }
1870

    
1871
    // Check for mergable requests
1872
    num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
1873

    
1874
    // Run the aio requests
1875
    for (i = 0; i < num_reqs; i++) {
1876
        acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
1877
            reqs[i].nb_sectors, multiwrite_cb, mcb);
1878

    
1879
        if (acb == NULL) {
1880
            // We can only fail the whole thing if no request has been
1881
            // submitted yet. Otherwise we'll wait for the submitted AIOs to
1882
            // complete and report the error in the callback.
1883
            if (mcb->num_requests == 0) {
1884
                reqs[i].error = EIO;
1885
                goto fail;
1886
            } else {
1887
                mcb->error = EIO;
1888
                break;
1889
            }
1890
        } else {
1891
            mcb->num_requests++;
1892
        }
1893
    }
1894

    
1895
    return 0;
1896

    
1897
fail:
1898
    free(mcb);
1899
    return -1;
1900
}
1901

    
1902
BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
1903
        BlockDriverCompletionFunc *cb, void *opaque)
1904
{
1905
    BlockDriver *drv = bs->drv;
1906

    
1907
    if (!drv)
1908
        return NULL;
1909
    return drv->bdrv_aio_flush(bs, cb, opaque);
1910
}
1911

    
1912
void bdrv_aio_cancel(BlockDriverAIOCB *acb)
1913
{
1914
    acb->pool->cancel(acb);
1915
}
1916

    
1917

    
1918
/**************************************************************/
1919
/* async block device emulation */
1920

    
1921
typedef struct BlockDriverAIOCBSync {
1922
    BlockDriverAIOCB common;
1923
    QEMUBH *bh;
1924
    int ret;
1925
    /* vector translation state */
1926
    QEMUIOVector *qiov;
1927
    uint8_t *bounce;
1928
    int is_write;
1929
} BlockDriverAIOCBSync;
1930

    
1931
static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
1932
{
1933
    BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
1934
    qemu_bh_delete(acb->bh);
1935
    acb->bh = NULL;
1936
    qemu_aio_release(acb);
1937
}
1938

    
1939
static AIOPool bdrv_em_aio_pool = {
1940
    .aiocb_size         = sizeof(BlockDriverAIOCBSync),
1941
    .cancel             = bdrv_aio_cancel_em,
1942
};
1943

    
1944
static void bdrv_aio_bh_cb(void *opaque)
1945
{
1946
    BlockDriverAIOCBSync *acb = opaque;
1947

    
1948
    if (!acb->is_write)
1949
        qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
1950
    qemu_vfree(acb->bounce);
1951
    acb->common.cb(acb->common.opaque, acb->ret);
1952
    qemu_bh_delete(acb->bh);
1953
    acb->bh = NULL;
1954
    qemu_aio_release(acb);
1955
}
1956

    
1957
static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
1958
                                            int64_t sector_num,
1959
                                            QEMUIOVector *qiov,
1960
                                            int nb_sectors,
1961
                                            BlockDriverCompletionFunc *cb,
1962
                                            void *opaque,
1963
                                            int is_write)
1964

    
1965
{
1966
    BlockDriverAIOCBSync *acb;
1967

    
1968
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1969
    acb->is_write = is_write;
1970
    acb->qiov = qiov;
1971
    acb->bounce = qemu_blockalign(bs, qiov->size);
1972

    
1973
    if (!acb->bh)
1974
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1975

    
1976
    if (is_write) {
1977
        qemu_iovec_to_buffer(acb->qiov, acb->bounce);
1978
        acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
1979
    } else {
1980
        acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
1981
    }
1982

    
1983
    qemu_bh_schedule(acb->bh);
1984

    
1985
    return &acb->common;
1986
}
1987

    
1988
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
1989
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1990
        BlockDriverCompletionFunc *cb, void *opaque)
1991
{
1992
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
1993
}
1994

    
1995
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
1996
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1997
        BlockDriverCompletionFunc *cb, void *opaque)
1998
{
1999
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2000
}
2001

    
2002
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2003
        BlockDriverCompletionFunc *cb, void *opaque)
2004
{
2005
    BlockDriverAIOCBSync *acb;
2006

    
2007
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2008
    acb->is_write = 1; /* don't bounce in the completion hadler */
2009
    acb->qiov = NULL;
2010
    acb->bounce = NULL;
2011
    acb->ret = 0;
2012

    
2013
    if (!acb->bh)
2014
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2015

    
2016
    bdrv_flush(bs);
2017
    qemu_bh_schedule(acb->bh);
2018
    return &acb->common;
2019
}
2020

    
2021
/**************************************************************/
2022
/* sync block device emulation */
2023

    
2024
static void bdrv_rw_em_cb(void *opaque, int ret)
2025
{
2026
    *(int *)opaque = ret;
2027
}
2028

    
2029
#define NOT_DONE 0x7fffffff
2030

    
2031
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2032
                        uint8_t *buf, int nb_sectors)
2033
{
2034
    int async_ret;
2035
    BlockDriverAIOCB *acb;
2036
    struct iovec iov;
2037
    QEMUIOVector qiov;
2038

    
2039
    async_context_push();
2040

    
2041
    async_ret = NOT_DONE;
2042
    iov.iov_base = (void *)buf;
2043
    iov.iov_len = nb_sectors * 512;
2044
    qemu_iovec_init_external(&qiov, &iov, 1);
2045
    acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2046
        bdrv_rw_em_cb, &async_ret);
2047
    if (acb == NULL) {
2048
        async_ret = -1;
2049
        goto fail;
2050
    }
2051

    
2052
    while (async_ret == NOT_DONE) {
2053
        qemu_aio_wait();
2054
    }
2055

    
2056

    
2057
fail:
2058
    async_context_pop();
2059
    return async_ret;
2060
}
2061

    
2062
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2063
                         const uint8_t *buf, int nb_sectors)
2064
{
2065
    int async_ret;
2066
    BlockDriverAIOCB *acb;
2067
    struct iovec iov;
2068
    QEMUIOVector qiov;
2069

    
2070
    async_context_push();
2071

    
2072
    async_ret = NOT_DONE;
2073
    iov.iov_base = (void *)buf;
2074
    iov.iov_len = nb_sectors * 512;
2075
    qemu_iovec_init_external(&qiov, &iov, 1);
2076
    acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2077
        bdrv_rw_em_cb, &async_ret);
2078
    if (acb == NULL) {
2079
        async_ret = -1;
2080
        goto fail;
2081
    }
2082
    while (async_ret == NOT_DONE) {
2083
        qemu_aio_wait();
2084
    }
2085

    
2086
fail:
2087
    async_context_pop();
2088
    return async_ret;
2089
}
2090

    
2091
void bdrv_init(void)
2092
{
2093
    module_call_init(MODULE_INIT_BLOCK);
2094
}
2095

    
2096
void bdrv_init_with_whitelist(void)
2097
{
2098
    use_bdrv_whitelist = 1;
2099
    bdrv_init();
2100
}
2101

    
2102
void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2103
                   BlockDriverCompletionFunc *cb, void *opaque)
2104
{
2105
    BlockDriverAIOCB *acb;
2106

    
2107
    if (pool->free_aiocb) {
2108
        acb = pool->free_aiocb;
2109
        pool->free_aiocb = acb->next;
2110
    } else {
2111
        acb = qemu_mallocz(pool->aiocb_size);
2112
        acb->pool = pool;
2113
    }
2114
    acb->bs = bs;
2115
    acb->cb = cb;
2116
    acb->opaque = opaque;
2117
    return acb;
2118
}
2119

    
2120
void qemu_aio_release(void *p)
2121
{
2122
    BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2123
    AIOPool *pool = acb->pool;
2124
    acb->next = pool->free_aiocb;
2125
    pool->free_aiocb = acb;
2126
}
2127

    
2128
/**************************************************************/
2129
/* removable device support */
2130

    
2131
/**
2132
 * Return TRUE if the media is present
2133
 */
2134
int bdrv_is_inserted(BlockDriverState *bs)
2135
{
2136
    BlockDriver *drv = bs->drv;
2137
    int ret;
2138
    if (!drv)
2139
        return 0;
2140
    if (!drv->bdrv_is_inserted)
2141
        return 1;
2142
    ret = drv->bdrv_is_inserted(bs);
2143
    return ret;
2144
}
2145

    
2146
/**
2147
 * Return TRUE if the media changed since the last call to this
2148
 * function. It is currently only used for floppy disks
2149
 */
2150
int bdrv_media_changed(BlockDriverState *bs)
2151
{
2152
    BlockDriver *drv = bs->drv;
2153
    int ret;
2154

    
2155
    if (!drv || !drv->bdrv_media_changed)
2156
        ret = -ENOTSUP;
2157
    else
2158
        ret = drv->bdrv_media_changed(bs);
2159
    if (ret == -ENOTSUP)
2160
        ret = bs->media_changed;
2161
    bs->media_changed = 0;
2162
    return ret;
2163
}
2164

    
2165
/**
2166
 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2167
 */
2168
int bdrv_eject(BlockDriverState *bs, int eject_flag)
2169
{
2170
    BlockDriver *drv = bs->drv;
2171
    int ret;
2172

    
2173
    if (bs->locked) {
2174
        return -EBUSY;
2175
    }
2176

    
2177
    if (!drv || !drv->bdrv_eject) {
2178
        ret = -ENOTSUP;
2179
    } else {
2180
        ret = drv->bdrv_eject(bs, eject_flag);
2181
    }
2182
    if (ret == -ENOTSUP) {
2183
        if (eject_flag)
2184
            bdrv_close(bs);
2185
        ret = 0;
2186
    }
2187

    
2188
    return ret;
2189
}
2190

    
2191
int bdrv_is_locked(BlockDriverState *bs)
2192
{
2193
    return bs->locked;
2194
}
2195

    
2196
/**
2197
 * Lock or unlock the media (if it is locked, the user won't be able
2198
 * to eject it manually).
2199
 */
2200
void bdrv_set_locked(BlockDriverState *bs, int locked)
2201
{
2202
    BlockDriver *drv = bs->drv;
2203

    
2204
    bs->locked = locked;
2205
    if (drv && drv->bdrv_set_locked) {
2206
        drv->bdrv_set_locked(bs, locked);
2207
    }
2208
}
2209

    
2210
/* needed for generic scsi interface */
2211

    
2212
int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2213
{
2214
    BlockDriver *drv = bs->drv;
2215

    
2216
    if (drv && drv->bdrv_ioctl)
2217
        return drv->bdrv_ioctl(bs, req, buf);
2218
    return -ENOTSUP;
2219
}
2220

    
2221
BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2222
        unsigned long int req, void *buf,
2223
        BlockDriverCompletionFunc *cb, void *opaque)
2224
{
2225
    BlockDriver *drv = bs->drv;
2226

    
2227
    if (drv && drv->bdrv_aio_ioctl)
2228
        return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2229
    return NULL;
2230
}
2231

    
2232

    
2233

    
2234
void *qemu_blockalign(BlockDriverState *bs, size_t size)
2235
{
2236
    return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2237
}
2238

    
2239
void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2240
{
2241
    int64_t bitmap_size;
2242

    
2243
    bs->dirty_count = 0;
2244
    if (enable) {
2245
        if (!bs->dirty_bitmap) {
2246
            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2247
                    BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2248
            bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2249

    
2250
            bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2251
        }
2252
    } else {
2253
        if (bs->dirty_bitmap) {
2254
            qemu_free(bs->dirty_bitmap);
2255
            bs->dirty_bitmap = NULL;
2256
        }
2257
    }
2258
}
2259

    
2260
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2261
{
2262
    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2263

    
2264
    if (bs->dirty_bitmap &&
2265
        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2266
        return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2267
            (1 << (chunk % (sizeof(unsigned long) * 8)));
2268
    } else {
2269
        return 0;
2270
    }
2271
}
2272

    
2273
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2274
                      int nr_sectors)
2275
{
2276
    set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2277
}
2278

    
2279
int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2280
{
2281
    return bs->dirty_count;
2282
}