Statistics
| Branch: | Revision:

root / block.c @ f8a83245

History | View | Annotate | Download (58.7 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
    /* buffer_alignment defaulted to 512, drivers can change this value */
367
    bs->buffer_alignment = 512;
368

    
369
    if (flags & BDRV_O_SNAPSHOT) {
370
        BlockDriverState *bs1;
371
        int64_t total_size;
372
        int is_protocol = 0;
373
        BlockDriver *bdrv_qcow2;
374
        QEMUOptionParameter *options;
375

    
376
        /* if snapshot, we create a temporary backing file and open it
377
           instead of opening 'filename' directly */
378

    
379
        /* if there is a backing file, use it */
380
        bs1 = bdrv_new("");
381
        ret = bdrv_open2(bs1, filename, 0, drv);
382
        if (ret < 0) {
383
            bdrv_delete(bs1);
384
            return ret;
385
        }
386
        total_size = bdrv_getlength(bs1) >> BDRV_SECTOR_BITS;
387

    
388
        if (bs1->drv && bs1->drv->protocol_name)
389
            is_protocol = 1;
390

    
391
        bdrv_delete(bs1);
392

    
393
        get_tmp_filename(tmp_filename, sizeof(tmp_filename));
394

    
395
        /* Real path is meaningless for protocols */
396
        if (is_protocol)
397
            snprintf(backing_filename, sizeof(backing_filename),
398
                     "%s", filename);
399
        else if (!realpath(filename, backing_filename))
400
            return -errno;
401

    
402
        bdrv_qcow2 = bdrv_find_format("qcow2");
403
        options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
404

    
405
        set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size * 512);
406
        set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
407
        if (drv) {
408
            set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
409
                drv->format_name);
410
        }
411

    
412
        ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
413
        if (ret < 0) {
414
            return ret;
415
        }
416

    
417
        filename = tmp_filename;
418
        drv = bdrv_qcow2;
419
        bs->is_temporary = 1;
420
    }
421

    
422
    pstrcpy(bs->filename, sizeof(bs->filename), filename);
423
    if (flags & BDRV_O_FILE) {
424
        drv = find_protocol(filename);
425
    } else if (!drv) {
426
        drv = find_hdev_driver(filename);
427
        if (!drv) {
428
            drv = find_image_format(filename);
429
        }
430
    }
431

    
432
    if (!drv) {
433
        ret = -ENOENT;
434
        goto unlink_and_fail;
435
    }
436
    if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
437
        ret = -ENOTSUP;
438
        goto unlink_and_fail;
439
    }
440

    
441
    bs->drv = drv;
442
    bs->opaque = qemu_mallocz(drv->instance_size);
443

    
444
    /*
445
     * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
446
     * write cache to the guest.  We do need the fdatasync to flush
447
     * out transactions for block allocations, and we maybe have a
448
     * volatile write cache in our backing device to deal with.
449
     */
450
    if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
451
        bs->enable_write_cache = 1;
452

    
453
    bs->read_only = (flags & BDRV_O_RDWR) == 0;
454
    if (!(flags & BDRV_O_FILE)) {
455
        open_flags = (flags & (BDRV_O_RDWR | BDRV_O_CACHE_MASK|BDRV_O_NATIVE_AIO));
456
        if (bs->is_temporary) { /* snapshot should be writeable */
457
            open_flags |= BDRV_O_RDWR;
458
        }
459
    } else {
460
        open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT);
461
    }
462

    
463
    ret = drv->bdrv_open(bs, filename, open_flags);
464
    if (ret < 0) {
465
        goto free_and_fail;
466
    }
467

    
468
    if (drv->bdrv_getlength) {
469
        bs->total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
470
    }
471
#ifndef _WIN32
472
    if (bs->is_temporary) {
473
        unlink(filename);
474
    }
475
#endif
476
    if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
477
        /* if there is a backing file, use it */
478
        BlockDriver *back_drv = NULL;
479
        bs->backing_hd = bdrv_new("");
480
        path_combine(backing_filename, sizeof(backing_filename),
481
                     filename, bs->backing_file);
482
        if (bs->backing_format[0] != '\0')
483
            back_drv = bdrv_find_format(bs->backing_format);
484
        ret = bdrv_open2(bs->backing_hd, backing_filename, open_flags,
485
                         back_drv);
486
        bs->backing_hd->read_only =  (open_flags & BDRV_O_RDWR) == 0;
487
        if (ret < 0) {
488
            bdrv_close(bs);
489
            return ret;
490
        }
491
    }
492

    
493
    if (!bdrv_key_required(bs)) {
494
        /* call the change callback */
495
        bs->media_changed = 1;
496
        if (bs->change_cb)
497
            bs->change_cb(bs->change_opaque);
498
    }
499
    return 0;
500

    
501
free_and_fail:
502
    qemu_free(bs->opaque);
503
    bs->opaque = NULL;
504
    bs->drv = NULL;
505
unlink_and_fail:
506
    if (bs->is_temporary)
507
        unlink(filename);
508
    return ret;
509
}
510

    
511
void bdrv_close(BlockDriverState *bs)
512
{
513
    if (bs->drv) {
514
        if (bs->backing_hd)
515
            bdrv_delete(bs->backing_hd);
516
        bs->drv->bdrv_close(bs);
517
        qemu_free(bs->opaque);
518
#ifdef _WIN32
519
        if (bs->is_temporary) {
520
            unlink(bs->filename);
521
        }
522
#endif
523
        bs->opaque = NULL;
524
        bs->drv = NULL;
525

    
526
        /* call the change callback */
527
        bs->media_changed = 1;
528
        if (bs->change_cb)
529
            bs->change_cb(bs->change_opaque);
530
    }
531
}
532

    
533
void bdrv_delete(BlockDriverState *bs)
534
{
535
    BlockDriverState **pbs;
536

    
537
    pbs = &bdrv_first;
538
    while (*pbs != bs && *pbs != NULL)
539
        pbs = &(*pbs)->next;
540
    if (*pbs == bs)
541
        *pbs = bs->next;
542

    
543
    bdrv_close(bs);
544
    qemu_free(bs);
545
}
546

    
547
/*
548
 * Run consistency checks on an image
549
 *
550
 * Returns the number of errors or -errno when an internal error occurs
551
 */
552
int bdrv_check(BlockDriverState *bs)
553
{
554
    if (bs->drv->bdrv_check == NULL) {
555
        return -ENOTSUP;
556
    }
557

    
558
    return bs->drv->bdrv_check(bs);
559
}
560

    
561
/* commit COW file into the raw image */
562
int bdrv_commit(BlockDriverState *bs)
563
{
564
    BlockDriver *drv = bs->drv;
565
    int64_t i, total_sectors;
566
    int n, j;
567
    int ret = 0;
568
    unsigned char sector[512];
569

    
570
    if (!drv)
571
        return -ENOMEDIUM;
572

    
573
    if (bs->read_only) {
574
        return -EACCES;
575
    }
576

    
577
    if (!bs->backing_hd) {
578
        return -ENOTSUP;
579
    }
580

    
581
    total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
582
    for (i = 0; i < total_sectors;) {
583
        if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
584
            for(j = 0; j < n; j++) {
585
                if (bdrv_read(bs, i, sector, 1) != 0) {
586
                    return -EIO;
587
                }
588

    
589
                if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
590
                    return -EIO;
591
                }
592
                i++;
593
            }
594
        } else {
595
            i += n;
596
        }
597
    }
598

    
599
    if (drv->bdrv_make_empty) {
600
        ret = drv->bdrv_make_empty(bs);
601
        bdrv_flush(bs);
602
    }
603

    
604
    /*
605
     * Make sure all data we wrote to the backing device is actually
606
     * stable on disk.
607
     */
608
    if (bs->backing_hd)
609
        bdrv_flush(bs->backing_hd);
610
    return ret;
611
}
612

    
613
/*
614
 * Return values:
615
 * 0        - success
616
 * -EINVAL  - backing format specified, but no file
617
 * -ENOSPC  - can't update the backing file because no space is left in the
618
 *            image file header
619
 * -ENOTSUP - format driver doesn't support changing the backing file
620
 */
621
int bdrv_change_backing_file(BlockDriverState *bs,
622
    const char *backing_file, const char *backing_fmt)
623
{
624
    BlockDriver *drv = bs->drv;
625

    
626
    if (drv->bdrv_change_backing_file != NULL) {
627
        return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
628
    } else {
629
        return -ENOTSUP;
630
    }
631
}
632

    
633
static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
634
                                   size_t size)
635
{
636
    int64_t len;
637

    
638
    if (!bdrv_is_inserted(bs))
639
        return -ENOMEDIUM;
640

    
641
    if (bs->growable)
642
        return 0;
643

    
644
    len = bdrv_getlength(bs);
645

    
646
    if (offset < 0)
647
        return -EIO;
648

    
649
    if ((offset > len) || (len - offset < size))
650
        return -EIO;
651

    
652
    return 0;
653
}
654

    
655
static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
656
                              int nb_sectors)
657
{
658
    return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);
659
}
660

    
661
/* return < 0 if error. See bdrv_write() for the return codes */
662
int bdrv_read(BlockDriverState *bs, int64_t sector_num,
663
              uint8_t *buf, int nb_sectors)
664
{
665
    BlockDriver *drv = bs->drv;
666

    
667
    if (!drv)
668
        return -ENOMEDIUM;
669
    if (bdrv_check_request(bs, sector_num, nb_sectors))
670
        return -EIO;
671

    
672
    return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
673
}
674

    
675
static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
676
                             int nb_sectors, int dirty)
677
{
678
    int64_t start, end;
679
    unsigned long val, idx, bit;
680

    
681
    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
682
    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
683

    
684
    for (; start <= end; start++) {
685
        idx = start / (sizeof(unsigned long) * 8);
686
        bit = start % (sizeof(unsigned long) * 8);
687
        val = bs->dirty_bitmap[idx];
688
        if (dirty) {
689
            val |= 1 << bit;
690
        } else {
691
            val &= ~(1 << bit);
692
        }
693
        bs->dirty_bitmap[idx] = val;
694
    }
695
}
696

    
697
/* Return < 0 if error. Important errors are:
698
  -EIO         generic I/O error (may happen for all errors)
699
  -ENOMEDIUM   No media inserted.
700
  -EINVAL      Invalid sector number or nb_sectors
701
  -EACCES      Trying to write a read-only device
702
*/
703
int bdrv_write(BlockDriverState *bs, int64_t sector_num,
704
               const uint8_t *buf, int nb_sectors)
705
{
706
    BlockDriver *drv = bs->drv;
707
    if (!bs->drv)
708
        return -ENOMEDIUM;
709
    if (bs->read_only)
710
        return -EACCES;
711
    if (bdrv_check_request(bs, sector_num, nb_sectors))
712
        return -EIO;
713

    
714
    if (bs->dirty_bitmap) {
715
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
716
    }
717

    
718
    return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
719
}
720

    
721
int bdrv_pread(BlockDriverState *bs, int64_t offset,
722
               void *buf, int count1)
723
{
724
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
725
    int len, nb_sectors, count;
726
    int64_t sector_num;
727
    int ret;
728

    
729
    count = count1;
730
    /* first read to align to sector start */
731
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
732
    if (len > count)
733
        len = count;
734
    sector_num = offset >> BDRV_SECTOR_BITS;
735
    if (len > 0) {
736
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
737
            return ret;
738
        memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
739
        count -= len;
740
        if (count == 0)
741
            return count1;
742
        sector_num++;
743
        buf += len;
744
    }
745

    
746
    /* read the sectors "in place" */
747
    nb_sectors = count >> BDRV_SECTOR_BITS;
748
    if (nb_sectors > 0) {
749
        if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
750
            return ret;
751
        sector_num += nb_sectors;
752
        len = nb_sectors << BDRV_SECTOR_BITS;
753
        buf += len;
754
        count -= len;
755
    }
756

    
757
    /* add data from the last sector */
758
    if (count > 0) {
759
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
760
            return ret;
761
        memcpy(buf, tmp_buf, count);
762
    }
763
    return count1;
764
}
765

    
766
int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
767
                const void *buf, int count1)
768
{
769
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
770
    int len, nb_sectors, count;
771
    int64_t sector_num;
772
    int ret;
773

    
774
    count = count1;
775
    /* first write to align to sector start */
776
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
777
    if (len > count)
778
        len = count;
779
    sector_num = offset >> BDRV_SECTOR_BITS;
780
    if (len > 0) {
781
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
782
            return ret;
783
        memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
784
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
785
            return ret;
786
        count -= len;
787
        if (count == 0)
788
            return count1;
789
        sector_num++;
790
        buf += len;
791
    }
792

    
793
    /* write the sectors "in place" */
794
    nb_sectors = count >> BDRV_SECTOR_BITS;
795
    if (nb_sectors > 0) {
796
        if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
797
            return ret;
798
        sector_num += nb_sectors;
799
        len = nb_sectors << BDRV_SECTOR_BITS;
800
        buf += len;
801
        count -= len;
802
    }
803

    
804
    /* add data from the last sector */
805
    if (count > 0) {
806
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
807
            return ret;
808
        memcpy(tmp_buf, buf, count);
809
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
810
            return ret;
811
    }
812
    return count1;
813
}
814

    
815
/**
816
 * Truncate file to 'offset' bytes (needed only for file protocols)
817
 */
818
int bdrv_truncate(BlockDriverState *bs, int64_t offset)
819
{
820
    BlockDriver *drv = bs->drv;
821
    if (!drv)
822
        return -ENOMEDIUM;
823
    if (!drv->bdrv_truncate)
824
        return -ENOTSUP;
825
    if (bs->read_only)
826
        return -EACCES;
827
    return drv->bdrv_truncate(bs, offset);
828
}
829

    
830
/**
831
 * Length of a file in bytes. Return < 0 if error or unknown.
832
 */
833
int64_t bdrv_getlength(BlockDriverState *bs)
834
{
835
    BlockDriver *drv = bs->drv;
836
    if (!drv)
837
        return -ENOMEDIUM;
838
    if (!drv->bdrv_getlength) {
839
        /* legacy mode */
840
        return bs->total_sectors * BDRV_SECTOR_SIZE;
841
    }
842
    return drv->bdrv_getlength(bs);
843
}
844

    
845
/* return 0 as number of sectors if no device present or error */
846
void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
847
{
848
    int64_t length;
849
    length = bdrv_getlength(bs);
850
    if (length < 0)
851
        length = 0;
852
    else
853
        length = length >> BDRV_SECTOR_BITS;
854
    *nb_sectors_ptr = length;
855
}
856

    
857
struct partition {
858
        uint8_t boot_ind;           /* 0x80 - active */
859
        uint8_t head;               /* starting head */
860
        uint8_t sector;             /* starting sector */
861
        uint8_t cyl;                /* starting cylinder */
862
        uint8_t sys_ind;            /* What partition type */
863
        uint8_t end_head;           /* end head */
864
        uint8_t end_sector;         /* end sector */
865
        uint8_t end_cyl;            /* end cylinder */
866
        uint32_t start_sect;        /* starting sector counting from 0 */
867
        uint32_t nr_sects;          /* nr of sectors in partition */
868
} __attribute__((packed));
869

    
870
/* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
871
static int guess_disk_lchs(BlockDriverState *bs,
872
                           int *pcylinders, int *pheads, int *psectors)
873
{
874
    uint8_t buf[512];
875
    int ret, i, heads, sectors, cylinders;
876
    struct partition *p;
877
    uint32_t nr_sects;
878
    uint64_t nb_sectors;
879

    
880
    bdrv_get_geometry(bs, &nb_sectors);
881

    
882
    ret = bdrv_read(bs, 0, buf, 1);
883
    if (ret < 0)
884
        return -1;
885
    /* test msdos magic */
886
    if (buf[510] != 0x55 || buf[511] != 0xaa)
887
        return -1;
888
    for(i = 0; i < 4; i++) {
889
        p = ((struct partition *)(buf + 0x1be)) + i;
890
        nr_sects = le32_to_cpu(p->nr_sects);
891
        if (nr_sects && p->end_head) {
892
            /* We make the assumption that the partition terminates on
893
               a cylinder boundary */
894
            heads = p->end_head + 1;
895
            sectors = p->end_sector & 63;
896
            if (sectors == 0)
897
                continue;
898
            cylinders = nb_sectors / (heads * sectors);
899
            if (cylinders < 1 || cylinders > 16383)
900
                continue;
901
            *pheads = heads;
902
            *psectors = sectors;
903
            *pcylinders = cylinders;
904
#if 0
905
            printf("guessed geometry: LCHS=%d %d %d\n",
906
                   cylinders, heads, sectors);
907
#endif
908
            return 0;
909
        }
910
    }
911
    return -1;
912
}
913

    
914
void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
915
{
916
    int translation, lba_detected = 0;
917
    int cylinders, heads, secs;
918
    uint64_t nb_sectors;
919

    
920
    /* if a geometry hint is available, use it */
921
    bdrv_get_geometry(bs, &nb_sectors);
922
    bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
923
    translation = bdrv_get_translation_hint(bs);
924
    if (cylinders != 0) {
925
        *pcyls = cylinders;
926
        *pheads = heads;
927
        *psecs = secs;
928
    } else {
929
        if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
930
            if (heads > 16) {
931
                /* if heads > 16, it means that a BIOS LBA
932
                   translation was active, so the default
933
                   hardware geometry is OK */
934
                lba_detected = 1;
935
                goto default_geometry;
936
            } else {
937
                *pcyls = cylinders;
938
                *pheads = heads;
939
                *psecs = secs;
940
                /* disable any translation to be in sync with
941
                   the logical geometry */
942
                if (translation == BIOS_ATA_TRANSLATION_AUTO) {
943
                    bdrv_set_translation_hint(bs,
944
                                              BIOS_ATA_TRANSLATION_NONE);
945
                }
946
            }
947
        } else {
948
        default_geometry:
949
            /* if no geometry, use a standard physical disk geometry */
950
            cylinders = nb_sectors / (16 * 63);
951

    
952
            if (cylinders > 16383)
953
                cylinders = 16383;
954
            else if (cylinders < 2)
955
                cylinders = 2;
956
            *pcyls = cylinders;
957
            *pheads = 16;
958
            *psecs = 63;
959
            if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
960
                if ((*pcyls * *pheads) <= 131072) {
961
                    bdrv_set_translation_hint(bs,
962
                                              BIOS_ATA_TRANSLATION_LARGE);
963
                } else {
964
                    bdrv_set_translation_hint(bs,
965
                                              BIOS_ATA_TRANSLATION_LBA);
966
                }
967
            }
968
        }
969
        bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
970
    }
971
}
972

    
973
void bdrv_set_geometry_hint(BlockDriverState *bs,
974
                            int cyls, int heads, int secs)
975
{
976
    bs->cyls = cyls;
977
    bs->heads = heads;
978
    bs->secs = secs;
979
}
980

    
981
void bdrv_set_type_hint(BlockDriverState *bs, int type)
982
{
983
    bs->type = type;
984
    bs->removable = ((type == BDRV_TYPE_CDROM ||
985
                      type == BDRV_TYPE_FLOPPY));
986
}
987

    
988
void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
989
{
990
    bs->translation = translation;
991
}
992

    
993
void bdrv_get_geometry_hint(BlockDriverState *bs,
994
                            int *pcyls, int *pheads, int *psecs)
995
{
996
    *pcyls = bs->cyls;
997
    *pheads = bs->heads;
998
    *psecs = bs->secs;
999
}
1000

    
1001
int bdrv_get_type_hint(BlockDriverState *bs)
1002
{
1003
    return bs->type;
1004
}
1005

    
1006
int bdrv_get_translation_hint(BlockDriverState *bs)
1007
{
1008
    return bs->translation;
1009
}
1010

    
1011
int bdrv_is_removable(BlockDriverState *bs)
1012
{
1013
    return bs->removable;
1014
}
1015

    
1016
int bdrv_is_read_only(BlockDriverState *bs)
1017
{
1018
    return bs->read_only;
1019
}
1020

    
1021
int bdrv_is_sg(BlockDriverState *bs)
1022
{
1023
    return bs->sg;
1024
}
1025

    
1026
int bdrv_enable_write_cache(BlockDriverState *bs)
1027
{
1028
    return bs->enable_write_cache;
1029
}
1030

    
1031
/* XXX: no longer used */
1032
void bdrv_set_change_cb(BlockDriverState *bs,
1033
                        void (*change_cb)(void *opaque), void *opaque)
1034
{
1035
    bs->change_cb = change_cb;
1036
    bs->change_opaque = opaque;
1037
}
1038

    
1039
int bdrv_is_encrypted(BlockDriverState *bs)
1040
{
1041
    if (bs->backing_hd && bs->backing_hd->encrypted)
1042
        return 1;
1043
    return bs->encrypted;
1044
}
1045

    
1046
int bdrv_key_required(BlockDriverState *bs)
1047
{
1048
    BlockDriverState *backing_hd = bs->backing_hd;
1049

    
1050
    if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1051
        return 1;
1052
    return (bs->encrypted && !bs->valid_key);
1053
}
1054

    
1055
int bdrv_set_key(BlockDriverState *bs, const char *key)
1056
{
1057
    int ret;
1058
    if (bs->backing_hd && bs->backing_hd->encrypted) {
1059
        ret = bdrv_set_key(bs->backing_hd, key);
1060
        if (ret < 0)
1061
            return ret;
1062
        if (!bs->encrypted)
1063
            return 0;
1064
    }
1065
    if (!bs->encrypted || !bs->drv || !bs->drv->bdrv_set_key)
1066
        return -1;
1067
    ret = bs->drv->bdrv_set_key(bs, key);
1068
    if (ret < 0) {
1069
        bs->valid_key = 0;
1070
    } else if (!bs->valid_key) {
1071
        bs->valid_key = 1;
1072
        /* call the change callback now, we skipped it on open */
1073
        bs->media_changed = 1;
1074
        if (bs->change_cb)
1075
            bs->change_cb(bs->change_opaque);
1076
    }
1077
    return ret;
1078
}
1079

    
1080
void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1081
{
1082
    if (!bs->drv) {
1083
        buf[0] = '\0';
1084
    } else {
1085
        pstrcpy(buf, buf_size, bs->drv->format_name);
1086
    }
1087
}
1088

    
1089
void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1090
                         void *opaque)
1091
{
1092
    BlockDriver *drv;
1093

    
1094
    for (drv = first_drv; drv != NULL; drv = drv->next) {
1095
        it(opaque, drv->format_name);
1096
    }
1097
}
1098

    
1099
BlockDriverState *bdrv_find(const char *name)
1100
{
1101
    BlockDriverState *bs;
1102

    
1103
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1104
        if (!strcmp(name, bs->device_name))
1105
            return bs;
1106
    }
1107
    return NULL;
1108
}
1109

    
1110
void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1111
{
1112
    BlockDriverState *bs;
1113

    
1114
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1115
        it(opaque, bs);
1116
    }
1117
}
1118

    
1119
const char *bdrv_get_device_name(BlockDriverState *bs)
1120
{
1121
    return bs->device_name;
1122
}
1123

    
1124
void bdrv_flush(BlockDriverState *bs)
1125
{
1126
    if (bs->drv && bs->drv->bdrv_flush)
1127
        bs->drv->bdrv_flush(bs);
1128
}
1129

    
1130
void bdrv_flush_all(void)
1131
{
1132
    BlockDriverState *bs;
1133

    
1134
    for (bs = bdrv_first; bs != NULL; bs = bs->next)
1135
        if (bs->drv && !bdrv_is_read_only(bs) && 
1136
            (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)))
1137
            bdrv_flush(bs);
1138
}
1139

    
1140
/*
1141
 * Returns true iff the specified sector is present in the disk image. Drivers
1142
 * not implementing the functionality are assumed to not support backing files,
1143
 * hence all their sectors are reported as allocated.
1144
 *
1145
 * 'pnum' is set to the number of sectors (including and immediately following
1146
 * the specified sector) that are known to be in the same
1147
 * allocated/unallocated state.
1148
 *
1149
 * 'nb_sectors' is the max value 'pnum' should be set to.
1150
 */
1151
int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1152
        int *pnum)
1153
{
1154
    int64_t n;
1155
    if (!bs->drv->bdrv_is_allocated) {
1156
        if (sector_num >= bs->total_sectors) {
1157
            *pnum = 0;
1158
            return 0;
1159
        }
1160
        n = bs->total_sectors - sector_num;
1161
        *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1162
        return 1;
1163
    }
1164
    return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1165
}
1166

    
1167
static void bdrv_print_dict(QObject *obj, void *opaque)
1168
{
1169
    QDict *bs_dict;
1170
    Monitor *mon = opaque;
1171

    
1172
    bs_dict = qobject_to_qdict(obj);
1173

    
1174
    monitor_printf(mon, "%s: type=%s removable=%d",
1175
                        qdict_get_str(bs_dict, "device"),
1176
                        qdict_get_str(bs_dict, "type"),
1177
                        qdict_get_bool(bs_dict, "removable"));
1178

    
1179
    if (qdict_get_bool(bs_dict, "removable")) {
1180
        monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1181
    }
1182

    
1183
    if (qdict_haskey(bs_dict, "inserted")) {
1184
        QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1185

    
1186
        monitor_printf(mon, " file=");
1187
        monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1188
        if (qdict_haskey(qdict, "backing_file")) {
1189
            monitor_printf(mon, " backing_file=");
1190
            monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1191
        }
1192
        monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1193
                            qdict_get_bool(qdict, "ro"),
1194
                            qdict_get_str(qdict, "drv"),
1195
                            qdict_get_bool(qdict, "encrypted"));
1196
    } else {
1197
        monitor_printf(mon, " [not inserted]");
1198
    }
1199

    
1200
    monitor_printf(mon, "\n");
1201
}
1202

    
1203
void bdrv_info_print(Monitor *mon, const QObject *data)
1204
{
1205
    qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1206
}
1207

    
1208
/**
1209
 * bdrv_info(): Block devices information
1210
 *
1211
 * Each block device information is stored in a QDict and the
1212
 * returned QObject is a QList of all devices.
1213
 *
1214
 * The QDict contains the following:
1215
 *
1216
 * - "device": device name
1217
 * - "type": device type
1218
 * - "removable": true if the device is removable, false otherwise
1219
 * - "locked": true if the device is locked, false otherwise
1220
 * - "inserted": only present if the device is inserted, it is a QDict
1221
 *    containing the following:
1222
 *          - "file": device file name
1223
 *          - "ro": true if read-only, false otherwise
1224
 *          - "drv": driver format name
1225
 *          - "backing_file": backing file name if one is used
1226
 *          - "encrypted": true if encrypted, false otherwise
1227
 *
1228
 * Example:
1229
 *
1230
 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1231
 *     "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1232
 *   { "device": "floppy0", "type": "floppy", "removable": true,
1233
 *     "locked": false } ]
1234
 */
1235
void bdrv_info(Monitor *mon, QObject **ret_data)
1236
{
1237
    QList *bs_list;
1238
    BlockDriverState *bs;
1239

    
1240
    bs_list = qlist_new();
1241

    
1242
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1243
        QObject *bs_obj;
1244
        const char *type = "unknown";
1245

    
1246
        switch(bs->type) {
1247
        case BDRV_TYPE_HD:
1248
            type = "hd";
1249
            break;
1250
        case BDRV_TYPE_CDROM:
1251
            type = "cdrom";
1252
            break;
1253
        case BDRV_TYPE_FLOPPY:
1254
            type = "floppy";
1255
            break;
1256
        }
1257

    
1258
        bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1259
                                    "'removable': %i, 'locked': %i }",
1260
                                    bs->device_name, type, bs->removable,
1261
                                    bs->locked);
1262
        assert(bs_obj != NULL);
1263

    
1264
        if (bs->drv) {
1265
            QObject *obj;
1266
            QDict *bs_dict = qobject_to_qdict(bs_obj);
1267

    
1268
            obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1269
                                     "'encrypted': %i }",
1270
                                     bs->filename, bs->read_only,
1271
                                     bs->drv->format_name,
1272
                                     bdrv_is_encrypted(bs));
1273
            assert(obj != NULL);
1274
            if (bs->backing_file[0] != '\0') {
1275
                QDict *qdict = qobject_to_qdict(obj);
1276
                qdict_put(qdict, "backing_file",
1277
                          qstring_from_str(bs->backing_file));
1278
            }
1279

    
1280
            qdict_put_obj(bs_dict, "inserted", obj);
1281
        }
1282
        qlist_append_obj(bs_list, bs_obj);
1283
    }
1284

    
1285
    *ret_data = QOBJECT(bs_list);
1286
}
1287

    
1288
static void bdrv_stats_iter(QObject *data, void *opaque)
1289
{
1290
    QDict *qdict;
1291
    Monitor *mon = opaque;
1292

    
1293
    qdict = qobject_to_qdict(data);
1294
    monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1295

    
1296
    qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1297
    monitor_printf(mon, " rd_bytes=%" PRId64
1298
                        " wr_bytes=%" PRId64
1299
                        " rd_operations=%" PRId64
1300
                        " wr_operations=%" PRId64
1301
                        "\n",
1302
                        qdict_get_int(qdict, "rd_bytes"),
1303
                        qdict_get_int(qdict, "wr_bytes"),
1304
                        qdict_get_int(qdict, "rd_operations"),
1305
                        qdict_get_int(qdict, "wr_operations"));
1306
}
1307

    
1308
void bdrv_stats_print(Monitor *mon, const QObject *data)
1309
{
1310
    qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1311
}
1312

    
1313
/**
1314
 * bdrv_info_stats(): show block device statistics
1315
 *
1316
 * Each device statistic information is stored in a QDict and
1317
 * the returned QObject is a QList of all devices.
1318
 *
1319
 * The QDict contains the following:
1320
 *
1321
 * - "device": device name
1322
 * - "stats": A QDict with the statistics information, it contains:
1323
 *     - "rd_bytes": bytes read
1324
 *     - "wr_bytes": bytes written
1325
 *     - "rd_operations": read operations
1326
 *     - "wr_operations": write operations
1327
 * 
1328
 * Example:
1329
 *
1330
 * [ { "device": "ide0-hd0",
1331
 *               "stats": { "rd_bytes": 512,
1332
 *                          "wr_bytes": 0,
1333
 *                          "rd_operations": 1,
1334
 *                          "wr_operations": 0 } },
1335
 *   { "device": "ide1-cd0",
1336
 *               "stats": { "rd_bytes": 0,
1337
 *                          "wr_bytes": 0,
1338
 *                          "rd_operations": 0,
1339
 *                          "wr_operations": 0 } } ]
1340
 */
1341
void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1342
{
1343
    QObject *obj;
1344
    QList *devices;
1345
    BlockDriverState *bs;
1346

    
1347
    devices = qlist_new();
1348

    
1349
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1350
        obj = qobject_from_jsonf("{ 'device': %s, 'stats': {"
1351
                                 "'rd_bytes': %" PRId64 ","
1352
                                 "'wr_bytes': %" PRId64 ","
1353
                                 "'rd_operations': %" PRId64 ","
1354
                                 "'wr_operations': %" PRId64
1355
                                 "} }",
1356
                                 bs->device_name,
1357
                                 bs->rd_bytes, bs->wr_bytes,
1358
                                 bs->rd_ops, bs->wr_ops);
1359
        assert(obj != NULL);
1360
        qlist_append_obj(devices, obj);
1361
    }
1362

    
1363
    *ret_data = QOBJECT(devices);
1364
}
1365

    
1366
const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1367
{
1368
    if (bs->backing_hd && bs->backing_hd->encrypted)
1369
        return bs->backing_file;
1370
    else if (bs->encrypted)
1371
        return bs->filename;
1372
    else
1373
        return NULL;
1374
}
1375

    
1376
void bdrv_get_backing_filename(BlockDriverState *bs,
1377
                               char *filename, int filename_size)
1378
{
1379
    if (!bs->backing_file) {
1380
        pstrcpy(filename, filename_size, "");
1381
    } else {
1382
        pstrcpy(filename, filename_size, bs->backing_file);
1383
    }
1384
}
1385

    
1386
int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1387
                          const uint8_t *buf, int nb_sectors)
1388
{
1389
    BlockDriver *drv = bs->drv;
1390
    if (!drv)
1391
        return -ENOMEDIUM;
1392
    if (!drv->bdrv_write_compressed)
1393
        return -ENOTSUP;
1394
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1395
        return -EIO;
1396

    
1397
    if (bs->dirty_bitmap) {
1398
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1399
    }
1400

    
1401
    return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1402
}
1403

    
1404
int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1405
{
1406
    BlockDriver *drv = bs->drv;
1407
    if (!drv)
1408
        return -ENOMEDIUM;
1409
    if (!drv->bdrv_get_info)
1410
        return -ENOTSUP;
1411
    memset(bdi, 0, sizeof(*bdi));
1412
    return drv->bdrv_get_info(bs, bdi);
1413
}
1414

    
1415
int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1416
                      int64_t pos, int size)
1417
{
1418
    BlockDriver *drv = bs->drv;
1419
    if (!drv)
1420
        return -ENOMEDIUM;
1421
    if (!drv->bdrv_save_vmstate)
1422
        return -ENOTSUP;
1423
    return drv->bdrv_save_vmstate(bs, buf, pos, size);
1424
}
1425

    
1426
int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1427
                      int64_t pos, int size)
1428
{
1429
    BlockDriver *drv = bs->drv;
1430
    if (!drv)
1431
        return -ENOMEDIUM;
1432
    if (!drv->bdrv_load_vmstate)
1433
        return -ENOTSUP;
1434
    return drv->bdrv_load_vmstate(bs, buf, pos, size);
1435
}
1436

    
1437
/**************************************************************/
1438
/* handling of snapshots */
1439

    
1440
int bdrv_snapshot_create(BlockDriverState *bs,
1441
                         QEMUSnapshotInfo *sn_info)
1442
{
1443
    BlockDriver *drv = bs->drv;
1444
    if (!drv)
1445
        return -ENOMEDIUM;
1446
    if (!drv->bdrv_snapshot_create)
1447
        return -ENOTSUP;
1448
    return drv->bdrv_snapshot_create(bs, sn_info);
1449
}
1450

    
1451
int bdrv_snapshot_goto(BlockDriverState *bs,
1452
                       const char *snapshot_id)
1453
{
1454
    BlockDriver *drv = bs->drv;
1455
    if (!drv)
1456
        return -ENOMEDIUM;
1457
    if (!drv->bdrv_snapshot_goto)
1458
        return -ENOTSUP;
1459
    return drv->bdrv_snapshot_goto(bs, snapshot_id);
1460
}
1461

    
1462
int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1463
{
1464
    BlockDriver *drv = bs->drv;
1465
    if (!drv)
1466
        return -ENOMEDIUM;
1467
    if (!drv->bdrv_snapshot_delete)
1468
        return -ENOTSUP;
1469
    return drv->bdrv_snapshot_delete(bs, snapshot_id);
1470
}
1471

    
1472
int bdrv_snapshot_list(BlockDriverState *bs,
1473
                       QEMUSnapshotInfo **psn_info)
1474
{
1475
    BlockDriver *drv = bs->drv;
1476
    if (!drv)
1477
        return -ENOMEDIUM;
1478
    if (!drv->bdrv_snapshot_list)
1479
        return -ENOTSUP;
1480
    return drv->bdrv_snapshot_list(bs, psn_info);
1481
}
1482

    
1483
#define NB_SUFFIXES 4
1484

    
1485
char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1486
{
1487
    static const char suffixes[NB_SUFFIXES] = "KMGT";
1488
    int64_t base;
1489
    int i;
1490

    
1491
    if (size <= 999) {
1492
        snprintf(buf, buf_size, "%" PRId64, size);
1493
    } else {
1494
        base = 1024;
1495
        for(i = 0; i < NB_SUFFIXES; i++) {
1496
            if (size < (10 * base)) {
1497
                snprintf(buf, buf_size, "%0.1f%c",
1498
                         (double)size / base,
1499
                         suffixes[i]);
1500
                break;
1501
            } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1502
                snprintf(buf, buf_size, "%" PRId64 "%c",
1503
                         ((size + (base >> 1)) / base),
1504
                         suffixes[i]);
1505
                break;
1506
            }
1507
            base = base * 1024;
1508
        }
1509
    }
1510
    return buf;
1511
}
1512

    
1513
char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1514
{
1515
    char buf1[128], date_buf[128], clock_buf[128];
1516
#ifdef _WIN32
1517
    struct tm *ptm;
1518
#else
1519
    struct tm tm;
1520
#endif
1521
    time_t ti;
1522
    int64_t secs;
1523

    
1524
    if (!sn) {
1525
        snprintf(buf, buf_size,
1526
                 "%-10s%-20s%7s%20s%15s",
1527
                 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1528
    } else {
1529
        ti = sn->date_sec;
1530
#ifdef _WIN32
1531
        ptm = localtime(&ti);
1532
        strftime(date_buf, sizeof(date_buf),
1533
                 "%Y-%m-%d %H:%M:%S", ptm);
1534
#else
1535
        localtime_r(&ti, &tm);
1536
        strftime(date_buf, sizeof(date_buf),
1537
                 "%Y-%m-%d %H:%M:%S", &tm);
1538
#endif
1539
        secs = sn->vm_clock_nsec / 1000000000;
1540
        snprintf(clock_buf, sizeof(clock_buf),
1541
                 "%02d:%02d:%02d.%03d",
1542
                 (int)(secs / 3600),
1543
                 (int)((secs / 60) % 60),
1544
                 (int)(secs % 60),
1545
                 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1546
        snprintf(buf, buf_size,
1547
                 "%-10s%-20s%7s%20s%15s",
1548
                 sn->id_str, sn->name,
1549
                 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1550
                 date_buf,
1551
                 clock_buf);
1552
    }
1553
    return buf;
1554
}
1555

    
1556

    
1557
/**************************************************************/
1558
/* async I/Os */
1559

    
1560
BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1561
                                 QEMUIOVector *qiov, int nb_sectors,
1562
                                 BlockDriverCompletionFunc *cb, void *opaque)
1563
{
1564
    BlockDriver *drv = bs->drv;
1565
    BlockDriverAIOCB *ret;
1566

    
1567
    if (!drv)
1568
        return NULL;
1569
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1570
        return NULL;
1571

    
1572
    ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1573
                              cb, opaque);
1574

    
1575
    if (ret) {
1576
        /* Update stats even though technically transfer has not happened. */
1577
        bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1578
        bs->rd_ops ++;
1579
    }
1580

    
1581
    return ret;
1582
}
1583

    
1584
BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1585
                                  QEMUIOVector *qiov, int nb_sectors,
1586
                                  BlockDriverCompletionFunc *cb, void *opaque)
1587
{
1588
    BlockDriver *drv = bs->drv;
1589
    BlockDriverAIOCB *ret;
1590

    
1591
    if (!drv)
1592
        return NULL;
1593
    if (bs->read_only)
1594
        return NULL;
1595
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1596
        return NULL;
1597

    
1598
    if (bs->dirty_bitmap) {
1599
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1600
    }
1601

    
1602
    ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1603
                               cb, opaque);
1604

    
1605
    if (ret) {
1606
        /* Update stats even though technically transfer has not happened. */
1607
        bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1608
        bs->wr_ops ++;
1609
    }
1610

    
1611
    return ret;
1612
}
1613

    
1614

    
1615
typedef struct MultiwriteCB {
1616
    int error;
1617
    int num_requests;
1618
    int num_callbacks;
1619
    struct {
1620
        BlockDriverCompletionFunc *cb;
1621
        void *opaque;
1622
        QEMUIOVector *free_qiov;
1623
        void *free_buf;
1624
    } callbacks[];
1625
} MultiwriteCB;
1626

    
1627
static void multiwrite_user_cb(MultiwriteCB *mcb)
1628
{
1629
    int i;
1630

    
1631
    for (i = 0; i < mcb->num_callbacks; i++) {
1632
        mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1633
        qemu_free(mcb->callbacks[i].free_qiov);
1634
        qemu_vfree(mcb->callbacks[i].free_buf);
1635
    }
1636
}
1637

    
1638
static void multiwrite_cb(void *opaque, int ret)
1639
{
1640
    MultiwriteCB *mcb = opaque;
1641

    
1642
    if (ret < 0) {
1643
        mcb->error = ret;
1644
        multiwrite_user_cb(mcb);
1645
    }
1646

    
1647
    mcb->num_requests--;
1648
    if (mcb->num_requests == 0) {
1649
        if (mcb->error == 0) {
1650
            multiwrite_user_cb(mcb);
1651
        }
1652
        qemu_free(mcb);
1653
    }
1654
}
1655

    
1656
static int multiwrite_req_compare(const void *a, const void *b)
1657
{
1658
    return (((BlockRequest*) a)->sector - ((BlockRequest*) b)->sector);
1659
}
1660

    
1661
/*
1662
 * Takes a bunch of requests and tries to merge them. Returns the number of
1663
 * requests that remain after merging.
1664
 */
1665
static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1666
    int num_reqs, MultiwriteCB *mcb)
1667
{
1668
    int i, outidx;
1669

    
1670
    // Sort requests by start sector
1671
    qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1672

    
1673
    // Check if adjacent requests touch the same clusters. If so, combine them,
1674
    // filling up gaps with zero sectors.
1675
    outidx = 0;
1676
    for (i = 1; i < num_reqs; i++) {
1677
        int merge = 0;
1678
        int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1679

    
1680
        // This handles the cases that are valid for all block drivers, namely
1681
        // exactly sequential writes and overlapping writes.
1682
        if (reqs[i].sector <= oldreq_last) {
1683
            merge = 1;
1684
        }
1685

    
1686
        // The block driver may decide that it makes sense to combine requests
1687
        // even if there is a gap of some sectors between them. In this case,
1688
        // the gap is filled with zeros (therefore only applicable for yet
1689
        // unused space in format like qcow2).
1690
        if (!merge && bs->drv->bdrv_merge_requests) {
1691
            merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1692
        }
1693

    
1694
        if (merge) {
1695
            size_t size;
1696
            QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1697
            qemu_iovec_init(qiov,
1698
                reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1699

    
1700
            // Add the first request to the merged one. If the requests are
1701
            // overlapping, drop the last sectors of the first request.
1702
            size = (reqs[i].sector - reqs[outidx].sector) << 9;
1703
            qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1704

    
1705
            // We might need to add some zeros between the two requests
1706
            if (reqs[i].sector > oldreq_last) {
1707
                size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1708
                uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1709
                memset(buf, 0, zero_bytes);
1710
                qemu_iovec_add(qiov, buf, zero_bytes);
1711
                mcb->callbacks[i].free_buf = buf;
1712
            }
1713

    
1714
            // Add the second request
1715
            qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
1716

    
1717
            reqs[outidx].nb_sectors += reqs[i].nb_sectors;
1718
            reqs[outidx].qiov = qiov;
1719

    
1720
            mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
1721
        } else {
1722
            outidx++;
1723
            reqs[outidx].sector     = reqs[i].sector;
1724
            reqs[outidx].nb_sectors = reqs[i].nb_sectors;
1725
            reqs[outidx].qiov       = reqs[i].qiov;
1726
        }
1727
    }
1728

    
1729
    return outidx + 1;
1730
}
1731

    
1732
/*
1733
 * Submit multiple AIO write requests at once.
1734
 *
1735
 * On success, the function returns 0 and all requests in the reqs array have
1736
 * been submitted. In error case this function returns -1, and any of the
1737
 * requests may or may not be submitted yet. In particular, this means that the
1738
 * callback will be called for some of the requests, for others it won't. The
1739
 * caller must check the error field of the BlockRequest to wait for the right
1740
 * callbacks (if error != 0, no callback will be called).
1741
 *
1742
 * The implementation may modify the contents of the reqs array, e.g. to merge
1743
 * requests. However, the fields opaque and error are left unmodified as they
1744
 * are used to signal failure for a single request to the caller.
1745
 */
1746
int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
1747
{
1748
    BlockDriverAIOCB *acb;
1749
    MultiwriteCB *mcb;
1750
    int i;
1751

    
1752
    if (num_reqs == 0) {
1753
        return 0;
1754
    }
1755

    
1756
    // Create MultiwriteCB structure
1757
    mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
1758
    mcb->num_requests = 0;
1759
    mcb->num_callbacks = num_reqs;
1760

    
1761
    for (i = 0; i < num_reqs; i++) {
1762
        mcb->callbacks[i].cb = reqs[i].cb;
1763
        mcb->callbacks[i].opaque = reqs[i].opaque;
1764
    }
1765

    
1766
    // Check for mergable requests
1767
    num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
1768

    
1769
    // Run the aio requests
1770
    for (i = 0; i < num_reqs; i++) {
1771
        acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
1772
            reqs[i].nb_sectors, multiwrite_cb, mcb);
1773

    
1774
        if (acb == NULL) {
1775
            // We can only fail the whole thing if no request has been
1776
            // submitted yet. Otherwise we'll wait for the submitted AIOs to
1777
            // complete and report the error in the callback.
1778
            if (mcb->num_requests == 0) {
1779
                reqs[i].error = EIO;
1780
                goto fail;
1781
            } else {
1782
                mcb->error = EIO;
1783
                break;
1784
            }
1785
        } else {
1786
            mcb->num_requests++;
1787
        }
1788
    }
1789

    
1790
    return 0;
1791

    
1792
fail:
1793
    free(mcb);
1794
    return -1;
1795
}
1796

    
1797
BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
1798
        BlockDriverCompletionFunc *cb, void *opaque)
1799
{
1800
    BlockDriver *drv = bs->drv;
1801

    
1802
    if (!drv)
1803
        return NULL;
1804
    return drv->bdrv_aio_flush(bs, cb, opaque);
1805
}
1806

    
1807
void bdrv_aio_cancel(BlockDriverAIOCB *acb)
1808
{
1809
    acb->pool->cancel(acb);
1810
}
1811

    
1812

    
1813
/**************************************************************/
1814
/* async block device emulation */
1815

    
1816
typedef struct BlockDriverAIOCBSync {
1817
    BlockDriverAIOCB common;
1818
    QEMUBH *bh;
1819
    int ret;
1820
    /* vector translation state */
1821
    QEMUIOVector *qiov;
1822
    uint8_t *bounce;
1823
    int is_write;
1824
} BlockDriverAIOCBSync;
1825

    
1826
static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
1827
{
1828
    BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
1829
    qemu_bh_delete(acb->bh);
1830
    acb->bh = NULL;
1831
    qemu_aio_release(acb);
1832
}
1833

    
1834
static AIOPool bdrv_em_aio_pool = {
1835
    .aiocb_size         = sizeof(BlockDriverAIOCBSync),
1836
    .cancel             = bdrv_aio_cancel_em,
1837
};
1838

    
1839
static void bdrv_aio_bh_cb(void *opaque)
1840
{
1841
    BlockDriverAIOCBSync *acb = opaque;
1842

    
1843
    if (!acb->is_write)
1844
        qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
1845
    qemu_vfree(acb->bounce);
1846
    acb->common.cb(acb->common.opaque, acb->ret);
1847
    qemu_bh_delete(acb->bh);
1848
    acb->bh = NULL;
1849
    qemu_aio_release(acb);
1850
}
1851

    
1852
static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
1853
                                            int64_t sector_num,
1854
                                            QEMUIOVector *qiov,
1855
                                            int nb_sectors,
1856
                                            BlockDriverCompletionFunc *cb,
1857
                                            void *opaque,
1858
                                            int is_write)
1859

    
1860
{
1861
    BlockDriverAIOCBSync *acb;
1862

    
1863
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1864
    acb->is_write = is_write;
1865
    acb->qiov = qiov;
1866
    acb->bounce = qemu_blockalign(bs, qiov->size);
1867

    
1868
    if (!acb->bh)
1869
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1870

    
1871
    if (is_write) {
1872
        qemu_iovec_to_buffer(acb->qiov, acb->bounce);
1873
        acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
1874
    } else {
1875
        acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
1876
    }
1877

    
1878
    qemu_bh_schedule(acb->bh);
1879

    
1880
    return &acb->common;
1881
}
1882

    
1883
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
1884
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1885
        BlockDriverCompletionFunc *cb, void *opaque)
1886
{
1887
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
1888
}
1889

    
1890
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
1891
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1892
        BlockDriverCompletionFunc *cb, void *opaque)
1893
{
1894
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
1895
}
1896

    
1897
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
1898
        BlockDriverCompletionFunc *cb, void *opaque)
1899
{
1900
    BlockDriverAIOCBSync *acb;
1901

    
1902
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1903
    acb->is_write = 1; /* don't bounce in the completion hadler */
1904
    acb->qiov = NULL;
1905
    acb->bounce = NULL;
1906
    acb->ret = 0;
1907

    
1908
    if (!acb->bh)
1909
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1910

    
1911
    bdrv_flush(bs);
1912
    qemu_bh_schedule(acb->bh);
1913
    return &acb->common;
1914
}
1915

    
1916
/**************************************************************/
1917
/* sync block device emulation */
1918

    
1919
static void bdrv_rw_em_cb(void *opaque, int ret)
1920
{
1921
    *(int *)opaque = ret;
1922
}
1923

    
1924
#define NOT_DONE 0x7fffffff
1925

    
1926
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
1927
                        uint8_t *buf, int nb_sectors)
1928
{
1929
    int async_ret;
1930
    BlockDriverAIOCB *acb;
1931
    struct iovec iov;
1932
    QEMUIOVector qiov;
1933

    
1934
    async_context_push();
1935

    
1936
    async_ret = NOT_DONE;
1937
    iov.iov_base = (void *)buf;
1938
    iov.iov_len = nb_sectors * 512;
1939
    qemu_iovec_init_external(&qiov, &iov, 1);
1940
    acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
1941
        bdrv_rw_em_cb, &async_ret);
1942
    if (acb == NULL) {
1943
        async_ret = -1;
1944
        goto fail;
1945
    }
1946

    
1947
    while (async_ret == NOT_DONE) {
1948
        qemu_aio_wait();
1949
    }
1950

    
1951

    
1952
fail:
1953
    async_context_pop();
1954
    return async_ret;
1955
}
1956

    
1957
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
1958
                         const uint8_t *buf, int nb_sectors)
1959
{
1960
    int async_ret;
1961
    BlockDriverAIOCB *acb;
1962
    struct iovec iov;
1963
    QEMUIOVector qiov;
1964

    
1965
    async_context_push();
1966

    
1967
    async_ret = NOT_DONE;
1968
    iov.iov_base = (void *)buf;
1969
    iov.iov_len = nb_sectors * 512;
1970
    qemu_iovec_init_external(&qiov, &iov, 1);
1971
    acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
1972
        bdrv_rw_em_cb, &async_ret);
1973
    if (acb == NULL) {
1974
        async_ret = -1;
1975
        goto fail;
1976
    }
1977
    while (async_ret == NOT_DONE) {
1978
        qemu_aio_wait();
1979
    }
1980

    
1981
fail:
1982
    async_context_pop();
1983
    return async_ret;
1984
}
1985

    
1986
void bdrv_init(void)
1987
{
1988
    module_call_init(MODULE_INIT_BLOCK);
1989
}
1990

    
1991
void bdrv_init_with_whitelist(void)
1992
{
1993
    use_bdrv_whitelist = 1;
1994
    bdrv_init();
1995
}
1996

    
1997
void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
1998
                   BlockDriverCompletionFunc *cb, void *opaque)
1999
{
2000
    BlockDriverAIOCB *acb;
2001

    
2002
    if (pool->free_aiocb) {
2003
        acb = pool->free_aiocb;
2004
        pool->free_aiocb = acb->next;
2005
    } else {
2006
        acb = qemu_mallocz(pool->aiocb_size);
2007
        acb->pool = pool;
2008
    }
2009
    acb->bs = bs;
2010
    acb->cb = cb;
2011
    acb->opaque = opaque;
2012
    return acb;
2013
}
2014

    
2015
void qemu_aio_release(void *p)
2016
{
2017
    BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2018
    AIOPool *pool = acb->pool;
2019
    acb->next = pool->free_aiocb;
2020
    pool->free_aiocb = acb;
2021
}
2022

    
2023
/**************************************************************/
2024
/* removable device support */
2025

    
2026
/**
2027
 * Return TRUE if the media is present
2028
 */
2029
int bdrv_is_inserted(BlockDriverState *bs)
2030
{
2031
    BlockDriver *drv = bs->drv;
2032
    int ret;
2033
    if (!drv)
2034
        return 0;
2035
    if (!drv->bdrv_is_inserted)
2036
        return 1;
2037
    ret = drv->bdrv_is_inserted(bs);
2038
    return ret;
2039
}
2040

    
2041
/**
2042
 * Return TRUE if the media changed since the last call to this
2043
 * function. It is currently only used for floppy disks
2044
 */
2045
int bdrv_media_changed(BlockDriverState *bs)
2046
{
2047
    BlockDriver *drv = bs->drv;
2048
    int ret;
2049

    
2050
    if (!drv || !drv->bdrv_media_changed)
2051
        ret = -ENOTSUP;
2052
    else
2053
        ret = drv->bdrv_media_changed(bs);
2054
    if (ret == -ENOTSUP)
2055
        ret = bs->media_changed;
2056
    bs->media_changed = 0;
2057
    return ret;
2058
}
2059

    
2060
/**
2061
 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2062
 */
2063
int bdrv_eject(BlockDriverState *bs, int eject_flag)
2064
{
2065
    BlockDriver *drv = bs->drv;
2066
    int ret;
2067

    
2068
    if (bs->locked) {
2069
        return -EBUSY;
2070
    }
2071

    
2072
    if (!drv || !drv->bdrv_eject) {
2073
        ret = -ENOTSUP;
2074
    } else {
2075
        ret = drv->bdrv_eject(bs, eject_flag);
2076
    }
2077
    if (ret == -ENOTSUP) {
2078
        if (eject_flag)
2079
            bdrv_close(bs);
2080
        ret = 0;
2081
    }
2082

    
2083
    return ret;
2084
}
2085

    
2086
int bdrv_is_locked(BlockDriverState *bs)
2087
{
2088
    return bs->locked;
2089
}
2090

    
2091
/**
2092
 * Lock or unlock the media (if it is locked, the user won't be able
2093
 * to eject it manually).
2094
 */
2095
void bdrv_set_locked(BlockDriverState *bs, int locked)
2096
{
2097
    BlockDriver *drv = bs->drv;
2098

    
2099
    bs->locked = locked;
2100
    if (drv && drv->bdrv_set_locked) {
2101
        drv->bdrv_set_locked(bs, locked);
2102
    }
2103
}
2104

    
2105
/* needed for generic scsi interface */
2106

    
2107
int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2108
{
2109
    BlockDriver *drv = bs->drv;
2110

    
2111
    if (drv && drv->bdrv_ioctl)
2112
        return drv->bdrv_ioctl(bs, req, buf);
2113
    return -ENOTSUP;
2114
}
2115

    
2116
BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2117
        unsigned long int req, void *buf,
2118
        BlockDriverCompletionFunc *cb, void *opaque)
2119
{
2120
    BlockDriver *drv = bs->drv;
2121

    
2122
    if (drv && drv->bdrv_aio_ioctl)
2123
        return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2124
    return NULL;
2125
}
2126

    
2127

    
2128

    
2129
void *qemu_blockalign(BlockDriverState *bs, size_t size)
2130
{
2131
    return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2132
}
2133

    
2134
void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2135
{
2136
    int64_t bitmap_size;
2137

    
2138
    if (enable) {
2139
        if (!bs->dirty_bitmap) {
2140
            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2141
                    BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2142
            bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2143

    
2144
            bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2145
        }
2146
    } else {
2147
        if (bs->dirty_bitmap) {
2148
            qemu_free(bs->dirty_bitmap);
2149
            bs->dirty_bitmap = NULL;
2150
        }
2151
    }
2152
}
2153

    
2154
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2155
{
2156
    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2157

    
2158
    if (bs->dirty_bitmap &&
2159
        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2160
        return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2161
            (1 << (chunk % (sizeof(unsigned long) * 8)));
2162
    } else {
2163
        return 0;
2164
    }
2165
}
2166

    
2167
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2168
                      int nr_sectors)
2169
{
2170
    set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2171
}