Statistics
| Branch: | Revision:

root / block.c @ bb8bf76f

History | View | Annotate | Download (77.7 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 "trace.h"
27
#include "monitor.h"
28
#include "block_int.h"
29
#include "module.h"
30
#include "qemu-objects.h"
31

    
32
#ifdef CONFIG_BSD
33
#include <sys/types.h>
34
#include <sys/stat.h>
35
#include <sys/ioctl.h>
36
#include <sys/queue.h>
37
#ifndef __DragonFly__
38
#include <sys/disk.h>
39
#endif
40
#endif
41

    
42
#ifdef _WIN32
43
#include <windows.h>
44
#endif
45

    
46
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
47
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
48
        BlockDriverCompletionFunc *cb, void *opaque);
49
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
50
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
51
        BlockDriverCompletionFunc *cb, void *opaque);
52
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
53
        BlockDriverCompletionFunc *cb, void *opaque);
54
static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
55
        BlockDriverCompletionFunc *cb, void *opaque);
56
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
57
                        uint8_t *buf, int nb_sectors);
58
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
59
                         const uint8_t *buf, int nb_sectors);
60

    
61
static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
62
    QTAILQ_HEAD_INITIALIZER(bdrv_states);
63

    
64
static QLIST_HEAD(, BlockDriver) bdrv_drivers =
65
    QLIST_HEAD_INITIALIZER(bdrv_drivers);
66

    
67
/* The device to use for VM snapshots */
68
static BlockDriverState *bs_snapshots;
69

    
70
/* If non-zero, use only whitelisted block drivers */
71
static int use_bdrv_whitelist;
72

    
73
#ifdef _WIN32
74
static int is_windows_drive_prefix(const char *filename)
75
{
76
    return (((filename[0] >= 'a' && filename[0] <= 'z') ||
77
             (filename[0] >= 'A' && filename[0] <= 'Z')) &&
78
            filename[1] == ':');
79
}
80

    
81
int is_windows_drive(const char *filename)
82
{
83
    if (is_windows_drive_prefix(filename) &&
84
        filename[2] == '\0')
85
        return 1;
86
    if (strstart(filename, "\\\\.\\", NULL) ||
87
        strstart(filename, "//./", NULL))
88
        return 1;
89
    return 0;
90
}
91
#endif
92

    
93
/* check if the path starts with "<protocol>:" */
94
static int path_has_protocol(const char *path)
95
{
96
#ifdef _WIN32
97
    if (is_windows_drive(path) ||
98
        is_windows_drive_prefix(path)) {
99
        return 0;
100
    }
101
#endif
102

    
103
    return strchr(path, ':') != NULL;
104
}
105

    
106
int path_is_absolute(const char *path)
107
{
108
    const char *p;
109
#ifdef _WIN32
110
    /* specific case for names like: "\\.\d:" */
111
    if (*path == '/' || *path == '\\')
112
        return 1;
113
#endif
114
    p = strchr(path, ':');
115
    if (p)
116
        p++;
117
    else
118
        p = path;
119
#ifdef _WIN32
120
    return (*p == '/' || *p == '\\');
121
#else
122
    return (*p == '/');
123
#endif
124
}
125

    
126
/* if filename is absolute, just copy it to dest. Otherwise, build a
127
   path to it by considering it is relative to base_path. URL are
128
   supported. */
129
void path_combine(char *dest, int dest_size,
130
                  const char *base_path,
131
                  const char *filename)
132
{
133
    const char *p, *p1;
134
    int len;
135

    
136
    if (dest_size <= 0)
137
        return;
138
    if (path_is_absolute(filename)) {
139
        pstrcpy(dest, dest_size, filename);
140
    } else {
141
        p = strchr(base_path, ':');
142
        if (p)
143
            p++;
144
        else
145
            p = base_path;
146
        p1 = strrchr(base_path, '/');
147
#ifdef _WIN32
148
        {
149
            const char *p2;
150
            p2 = strrchr(base_path, '\\');
151
            if (!p1 || p2 > p1)
152
                p1 = p2;
153
        }
154
#endif
155
        if (p1)
156
            p1++;
157
        else
158
            p1 = base_path;
159
        if (p1 > p)
160
            p = p1;
161
        len = p - base_path;
162
        if (len > dest_size - 1)
163
            len = dest_size - 1;
164
        memcpy(dest, base_path, len);
165
        dest[len] = '\0';
166
        pstrcat(dest, dest_size, filename);
167
    }
168
}
169

    
170
void bdrv_register(BlockDriver *bdrv)
171
{
172
    if (!bdrv->bdrv_aio_readv) {
173
        /* add AIO emulation layer */
174
        bdrv->bdrv_aio_readv = bdrv_aio_readv_em;
175
        bdrv->bdrv_aio_writev = bdrv_aio_writev_em;
176
    } else if (!bdrv->bdrv_read) {
177
        /* add synchronous IO emulation layer */
178
        bdrv->bdrv_read = bdrv_read_em;
179
        bdrv->bdrv_write = bdrv_write_em;
180
    }
181

    
182
    if (!bdrv->bdrv_aio_flush)
183
        bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
184

    
185
    QLIST_INSERT_HEAD(&bdrv_drivers, bdrv, list);
186
}
187

    
188
/* create a new block device (by default it is empty) */
189
BlockDriverState *bdrv_new(const char *device_name)
190
{
191
    BlockDriverState *bs;
192

    
193
    bs = qemu_mallocz(sizeof(BlockDriverState));
194
    pstrcpy(bs->device_name, sizeof(bs->device_name), device_name);
195
    if (device_name[0] != '\0') {
196
        QTAILQ_INSERT_TAIL(&bdrv_states, bs, list);
197
    }
198
    return bs;
199
}
200

    
201
BlockDriver *bdrv_find_format(const char *format_name)
202
{
203
    BlockDriver *drv1;
204
    QLIST_FOREACH(drv1, &bdrv_drivers, list) {
205
        if (!strcmp(drv1->format_name, format_name)) {
206
            return drv1;
207
        }
208
    }
209
    return NULL;
210
}
211

    
212
static int bdrv_is_whitelisted(BlockDriver *drv)
213
{
214
    static const char *whitelist[] = {
215
        CONFIG_BDRV_WHITELIST
216
    };
217
    const char **p;
218

    
219
    if (!whitelist[0])
220
        return 1;               /* no whitelist, anything goes */
221

    
222
    for (p = whitelist; *p; p++) {
223
        if (!strcmp(drv->format_name, *p)) {
224
            return 1;
225
        }
226
    }
227
    return 0;
228
}
229

    
230
BlockDriver *bdrv_find_whitelisted_format(const char *format_name)
231
{
232
    BlockDriver *drv = bdrv_find_format(format_name);
233
    return drv && bdrv_is_whitelisted(drv) ? drv : NULL;
234
}
235

    
236
int bdrv_create(BlockDriver *drv, const char* filename,
237
    QEMUOptionParameter *options)
238
{
239
    if (!drv->bdrv_create)
240
        return -ENOTSUP;
241

    
242
    return drv->bdrv_create(filename, options);
243
}
244

    
245
int bdrv_create_file(const char* filename, QEMUOptionParameter *options)
246
{
247
    BlockDriver *drv;
248

    
249
    drv = bdrv_find_protocol(filename);
250
    if (drv == NULL) {
251
        return -ENOENT;
252
    }
253

    
254
    return bdrv_create(drv, filename, options);
255
}
256

    
257
#ifdef _WIN32
258
void get_tmp_filename(char *filename, int size)
259
{
260
    char temp_dir[MAX_PATH];
261

    
262
    GetTempPath(MAX_PATH, temp_dir);
263
    GetTempFileName(temp_dir, "qem", 0, filename);
264
}
265
#else
266
void get_tmp_filename(char *filename, int size)
267
{
268
    int fd;
269
    const char *tmpdir;
270
    /* XXX: race condition possible */
271
    tmpdir = getenv("TMPDIR");
272
    if (!tmpdir)
273
        tmpdir = "/tmp";
274
    snprintf(filename, size, "%s/vl.XXXXXX", tmpdir);
275
    fd = mkstemp(filename);
276
    close(fd);
277
}
278
#endif
279

    
280
/*
281
 * Detect host devices. By convention, /dev/cdrom[N] is always
282
 * recognized as a host CDROM.
283
 */
284
static BlockDriver *find_hdev_driver(const char *filename)
285
{
286
    int score_max = 0, score;
287
    BlockDriver *drv = NULL, *d;
288

    
289
    QLIST_FOREACH(d, &bdrv_drivers, list) {
290
        if (d->bdrv_probe_device) {
291
            score = d->bdrv_probe_device(filename);
292
            if (score > score_max) {
293
                score_max = score;
294
                drv = d;
295
            }
296
        }
297
    }
298

    
299
    return drv;
300
}
301

    
302
BlockDriver *bdrv_find_protocol(const char *filename)
303
{
304
    BlockDriver *drv1;
305
    char protocol[128];
306
    int len;
307
    const char *p;
308

    
309
    /* TODO Drivers without bdrv_file_open must be specified explicitly */
310

    
311
    /*
312
     * XXX(hch): we really should not let host device detection
313
     * override an explicit protocol specification, but moving this
314
     * later breaks access to device names with colons in them.
315
     * Thanks to the brain-dead persistent naming schemes on udev-
316
     * based Linux systems those actually are quite common.
317
     */
318
    drv1 = find_hdev_driver(filename);
319
    if (drv1) {
320
        return drv1;
321
    }
322

    
323
    if (!path_has_protocol(filename)) {
324
        return bdrv_find_format("file");
325
    }
326
    p = strchr(filename, ':');
327
    assert(p != NULL);
328
    len = p - filename;
329
    if (len > sizeof(protocol) - 1)
330
        len = sizeof(protocol) - 1;
331
    memcpy(protocol, filename, len);
332
    protocol[len] = '\0';
333
    QLIST_FOREACH(drv1, &bdrv_drivers, list) {
334
        if (drv1->protocol_name &&
335
            !strcmp(drv1->protocol_name, protocol)) {
336
            return drv1;
337
        }
338
    }
339
    return NULL;
340
}
341

    
342
static int find_image_format(const char *filename, BlockDriver **pdrv)
343
{
344
    int ret, score, score_max;
345
    BlockDriver *drv1, *drv;
346
    uint8_t buf[2048];
347
    BlockDriverState *bs;
348

    
349
    ret = bdrv_file_open(&bs, filename, 0);
350
    if (ret < 0) {
351
        *pdrv = NULL;
352
        return ret;
353
    }
354

    
355
    /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
356
    if (bs->sg || !bdrv_is_inserted(bs)) {
357
        bdrv_delete(bs);
358
        drv = bdrv_find_format("raw");
359
        if (!drv) {
360
            ret = -ENOENT;
361
        }
362
        *pdrv = drv;
363
        return ret;
364
    }
365

    
366
    ret = bdrv_pread(bs, 0, buf, sizeof(buf));
367
    bdrv_delete(bs);
368
    if (ret < 0) {
369
        *pdrv = NULL;
370
        return ret;
371
    }
372

    
373
    score_max = 0;
374
    drv = NULL;
375
    QLIST_FOREACH(drv1, &bdrv_drivers, list) {
376
        if (drv1->bdrv_probe) {
377
            score = drv1->bdrv_probe(buf, ret, filename);
378
            if (score > score_max) {
379
                score_max = score;
380
                drv = drv1;
381
            }
382
        }
383
    }
384
    if (!drv) {
385
        ret = -ENOENT;
386
    }
387
    *pdrv = drv;
388
    return ret;
389
}
390

    
391
/**
392
 * Set the current 'total_sectors' value
393
 */
394
static int refresh_total_sectors(BlockDriverState *bs, int64_t hint)
395
{
396
    BlockDriver *drv = bs->drv;
397

    
398
    /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
399
    if (bs->sg)
400
        return 0;
401

    
402
    /* query actual device if possible, otherwise just trust the hint */
403
    if (drv->bdrv_getlength) {
404
        int64_t length = drv->bdrv_getlength(bs);
405
        if (length < 0) {
406
            return length;
407
        }
408
        hint = length >> BDRV_SECTOR_BITS;
409
    }
410

    
411
    bs->total_sectors = hint;
412
    return 0;
413
}
414

    
415
/*
416
 * Common part for opening disk images and files
417
 */
418
static int bdrv_open_common(BlockDriverState *bs, const char *filename,
419
    int flags, BlockDriver *drv)
420
{
421
    int ret, open_flags;
422

    
423
    assert(drv != NULL);
424

    
425
    bs->file = NULL;
426
    bs->total_sectors = 0;
427
    bs->encrypted = 0;
428
    bs->valid_key = 0;
429
    bs->open_flags = flags;
430
    /* buffer_alignment defaulted to 512, drivers can change this value */
431
    bs->buffer_alignment = 512;
432

    
433
    pstrcpy(bs->filename, sizeof(bs->filename), filename);
434

    
435
    if (use_bdrv_whitelist && !bdrv_is_whitelisted(drv)) {
436
        return -ENOTSUP;
437
    }
438

    
439
    bs->drv = drv;
440
    bs->opaque = qemu_mallocz(drv->instance_size);
441

    
442
    /*
443
     * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
444
     * write cache to the guest.  We do need the fdatasync to flush
445
     * out transactions for block allocations, and we maybe have a
446
     * volatile write cache in our backing device to deal with.
447
     */
448
    if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
449
        bs->enable_write_cache = 1;
450

    
451
    /*
452
     * Clear flags that are internal to the block layer before opening the
453
     * image.
454
     */
455
    open_flags = flags & ~(BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
456

    
457
    /*
458
     * Snapshots should be writeable.
459
     */
460
    if (bs->is_temporary) {
461
        open_flags |= BDRV_O_RDWR;
462
    }
463

    
464
    /* Open the image, either directly or using a protocol */
465
    if (drv->bdrv_file_open) {
466
        ret = drv->bdrv_file_open(bs, filename, open_flags);
467
    } else {
468
        ret = bdrv_file_open(&bs->file, filename, open_flags);
469
        if (ret >= 0) {
470
            ret = drv->bdrv_open(bs, open_flags);
471
        }
472
    }
473

    
474
    if (ret < 0) {
475
        goto free_and_fail;
476
    }
477

    
478
    bs->keep_read_only = bs->read_only = !(open_flags & BDRV_O_RDWR);
479

    
480
    ret = refresh_total_sectors(bs, bs->total_sectors);
481
    if (ret < 0) {
482
        goto free_and_fail;
483
    }
484

    
485
#ifndef _WIN32
486
    if (bs->is_temporary) {
487
        unlink(filename);
488
    }
489
#endif
490
    return 0;
491

    
492
free_and_fail:
493
    if (bs->file) {
494
        bdrv_delete(bs->file);
495
        bs->file = NULL;
496
    }
497
    qemu_free(bs->opaque);
498
    bs->opaque = NULL;
499
    bs->drv = NULL;
500
    return ret;
501
}
502

    
503
/*
504
 * Opens a file using a protocol (file, host_device, nbd, ...)
505
 */
506
int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags)
507
{
508
    BlockDriverState *bs;
509
    BlockDriver *drv;
510
    int ret;
511

    
512
    drv = bdrv_find_protocol(filename);
513
    if (!drv) {
514
        return -ENOENT;
515
    }
516

    
517
    bs = bdrv_new("");
518
    ret = bdrv_open_common(bs, filename, flags, drv);
519
    if (ret < 0) {
520
        bdrv_delete(bs);
521
        return ret;
522
    }
523
    bs->growable = 1;
524
    *pbs = bs;
525
    return 0;
526
}
527

    
528
/*
529
 * Opens a disk image (raw, qcow2, vmdk, ...)
530
 */
531
int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
532
              BlockDriver *drv)
533
{
534
    int ret;
535

    
536
    if (flags & BDRV_O_SNAPSHOT) {
537
        BlockDriverState *bs1;
538
        int64_t total_size;
539
        int is_protocol = 0;
540
        BlockDriver *bdrv_qcow2;
541
        QEMUOptionParameter *options;
542
        char tmp_filename[PATH_MAX];
543
        char backing_filename[PATH_MAX];
544

    
545
        /* if snapshot, we create a temporary backing file and open it
546
           instead of opening 'filename' directly */
547

    
548
        /* if there is a backing file, use it */
549
        bs1 = bdrv_new("");
550
        ret = bdrv_open(bs1, filename, 0, drv);
551
        if (ret < 0) {
552
            bdrv_delete(bs1);
553
            return ret;
554
        }
555
        total_size = bdrv_getlength(bs1) & BDRV_SECTOR_MASK;
556

    
557
        if (bs1->drv && bs1->drv->protocol_name)
558
            is_protocol = 1;
559

    
560
        bdrv_delete(bs1);
561

    
562
        get_tmp_filename(tmp_filename, sizeof(tmp_filename));
563

    
564
        /* Real path is meaningless for protocols */
565
        if (is_protocol)
566
            snprintf(backing_filename, sizeof(backing_filename),
567
                     "%s", filename);
568
        else if (!realpath(filename, backing_filename))
569
            return -errno;
570

    
571
        bdrv_qcow2 = bdrv_find_format("qcow2");
572
        options = parse_option_parameters("", bdrv_qcow2->create_options, NULL);
573

    
574
        set_option_parameter_int(options, BLOCK_OPT_SIZE, total_size);
575
        set_option_parameter(options, BLOCK_OPT_BACKING_FILE, backing_filename);
576
        if (drv) {
577
            set_option_parameter(options, BLOCK_OPT_BACKING_FMT,
578
                drv->format_name);
579
        }
580

    
581
        ret = bdrv_create(bdrv_qcow2, tmp_filename, options);
582
        free_option_parameters(options);
583
        if (ret < 0) {
584
            return ret;
585
        }
586

    
587
        filename = tmp_filename;
588
        drv = bdrv_qcow2;
589
        bs->is_temporary = 1;
590
    }
591

    
592
    /* Find the right image format driver */
593
    if (!drv) {
594
        ret = find_image_format(filename, &drv);
595
    }
596

    
597
    if (!drv) {
598
        goto unlink_and_fail;
599
    }
600

    
601
    /* Open the image */
602
    ret = bdrv_open_common(bs, filename, flags, drv);
603
    if (ret < 0) {
604
        goto unlink_and_fail;
605
    }
606

    
607
    /* If there is a backing file, use it */
608
    if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
609
        char backing_filename[PATH_MAX];
610
        int back_flags;
611
        BlockDriver *back_drv = NULL;
612

    
613
        bs->backing_hd = bdrv_new("");
614

    
615
        if (path_has_protocol(bs->backing_file)) {
616
            pstrcpy(backing_filename, sizeof(backing_filename),
617
                    bs->backing_file);
618
        } else {
619
            path_combine(backing_filename, sizeof(backing_filename),
620
                         filename, bs->backing_file);
621
        }
622

    
623
        if (bs->backing_format[0] != '\0') {
624
            back_drv = bdrv_find_format(bs->backing_format);
625
        }
626

    
627
        /* backing files always opened read-only */
628
        back_flags =
629
            flags & ~(BDRV_O_RDWR | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
630

    
631
        ret = bdrv_open(bs->backing_hd, backing_filename, back_flags, back_drv);
632
        if (ret < 0) {
633
            bdrv_close(bs);
634
            return ret;
635
        }
636
        if (bs->is_temporary) {
637
            bs->backing_hd->keep_read_only = !(flags & BDRV_O_RDWR);
638
        } else {
639
            /* base image inherits from "parent" */
640
            bs->backing_hd->keep_read_only = bs->keep_read_only;
641
        }
642
    }
643

    
644
    if (!bdrv_key_required(bs)) {
645
        /* call the change callback */
646
        bs->media_changed = 1;
647
        if (bs->change_cb)
648
            bs->change_cb(bs->change_opaque);
649
    }
650

    
651
    return 0;
652

    
653
unlink_and_fail:
654
    if (bs->is_temporary) {
655
        unlink(filename);
656
    }
657
    return ret;
658
}
659

    
660
void bdrv_close(BlockDriverState *bs)
661
{
662
    if (bs->drv) {
663
        if (bs == bs_snapshots) {
664
            bs_snapshots = NULL;
665
        }
666
        if (bs->backing_hd) {
667
            bdrv_delete(bs->backing_hd);
668
            bs->backing_hd = NULL;
669
        }
670
        bs->drv->bdrv_close(bs);
671
        qemu_free(bs->opaque);
672
#ifdef _WIN32
673
        if (bs->is_temporary) {
674
            unlink(bs->filename);
675
        }
676
#endif
677
        bs->opaque = NULL;
678
        bs->drv = NULL;
679

    
680
        if (bs->file != NULL) {
681
            bdrv_close(bs->file);
682
        }
683

    
684
        /* call the change callback */
685
        bs->media_changed = 1;
686
        if (bs->change_cb)
687
            bs->change_cb(bs->change_opaque);
688
    }
689
}
690

    
691
void bdrv_close_all(void)
692
{
693
    BlockDriverState *bs;
694

    
695
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
696
        bdrv_close(bs);
697
    }
698
}
699

    
700
void bdrv_delete(BlockDriverState *bs)
701
{
702
    assert(!bs->peer);
703

    
704
    /* remove from list, if necessary */
705
    if (bs->device_name[0] != '\0') {
706
        QTAILQ_REMOVE(&bdrv_states, bs, list);
707
    }
708

    
709
    bdrv_close(bs);
710
    if (bs->file != NULL) {
711
        bdrv_delete(bs->file);
712
    }
713

    
714
    assert(bs != bs_snapshots);
715
    qemu_free(bs);
716
}
717

    
718
int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
719
{
720
    if (bs->peer) {
721
        return -EBUSY;
722
    }
723
    bs->peer = qdev;
724
    return 0;
725
}
726

    
727
void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
728
{
729
    assert(bs->peer == qdev);
730
    bs->peer = NULL;
731
}
732

    
733
DeviceState *bdrv_get_attached(BlockDriverState *bs)
734
{
735
    return bs->peer;
736
}
737

    
738
/*
739
 * Run consistency checks on an image
740
 *
741
 * Returns 0 if the check could be completed (it doesn't mean that the image is
742
 * free of errors) or -errno when an internal error occured. The results of the
743
 * check are stored in res.
744
 */
745
int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
746
{
747
    if (bs->drv->bdrv_check == NULL) {
748
        return -ENOTSUP;
749
    }
750

    
751
    memset(res, 0, sizeof(*res));
752
    return bs->drv->bdrv_check(bs, res);
753
}
754

    
755
#define COMMIT_BUF_SECTORS 2048
756

    
757
/* commit COW file into the raw image */
758
int bdrv_commit(BlockDriverState *bs)
759
{
760
    BlockDriver *drv = bs->drv;
761
    BlockDriver *backing_drv;
762
    int64_t sector, total_sectors;
763
    int n, ro, open_flags;
764
    int ret = 0, rw_ret = 0;
765
    uint8_t *buf;
766
    char filename[1024];
767
    BlockDriverState *bs_rw, *bs_ro;
768

    
769
    if (!drv)
770
        return -ENOMEDIUM;
771
    
772
    if (!bs->backing_hd) {
773
        return -ENOTSUP;
774
    }
775

    
776
    if (bs->backing_hd->keep_read_only) {
777
        return -EACCES;
778
    }
779

    
780
    backing_drv = bs->backing_hd->drv;
781
    ro = bs->backing_hd->read_only;
782
    strncpy(filename, bs->backing_hd->filename, sizeof(filename));
783
    open_flags =  bs->backing_hd->open_flags;
784

    
785
    if (ro) {
786
        /* re-open as RW */
787
        bdrv_delete(bs->backing_hd);
788
        bs->backing_hd = NULL;
789
        bs_rw = bdrv_new("");
790
        rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
791
            backing_drv);
792
        if (rw_ret < 0) {
793
            bdrv_delete(bs_rw);
794
            /* try to re-open read-only */
795
            bs_ro = bdrv_new("");
796
            ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
797
                backing_drv);
798
            if (ret < 0) {
799
                bdrv_delete(bs_ro);
800
                /* drive not functional anymore */
801
                bs->drv = NULL;
802
                return ret;
803
            }
804
            bs->backing_hd = bs_ro;
805
            return rw_ret;
806
        }
807
        bs->backing_hd = bs_rw;
808
    }
809

    
810
    total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
811
    buf = qemu_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
812

    
813
    for (sector = 0; sector < total_sectors; sector += n) {
814
        if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
815

    
816
            if (bdrv_read(bs, sector, buf, n) != 0) {
817
                ret = -EIO;
818
                goto ro_cleanup;
819
            }
820

    
821
            if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
822
                ret = -EIO;
823
                goto ro_cleanup;
824
            }
825
        }
826
    }
827

    
828
    if (drv->bdrv_make_empty) {
829
        ret = drv->bdrv_make_empty(bs);
830
        bdrv_flush(bs);
831
    }
832

    
833
    /*
834
     * Make sure all data we wrote to the backing device is actually
835
     * stable on disk.
836
     */
837
    if (bs->backing_hd)
838
        bdrv_flush(bs->backing_hd);
839

    
840
ro_cleanup:
841
    qemu_free(buf);
842

    
843
    if (ro) {
844
        /* re-open as RO */
845
        bdrv_delete(bs->backing_hd);
846
        bs->backing_hd = NULL;
847
        bs_ro = bdrv_new("");
848
        ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
849
            backing_drv);
850
        if (ret < 0) {
851
            bdrv_delete(bs_ro);
852
            /* drive not functional anymore */
853
            bs->drv = NULL;
854
            return ret;
855
        }
856
        bs->backing_hd = bs_ro;
857
        bs->backing_hd->keep_read_only = 0;
858
    }
859

    
860
    return ret;
861
}
862

    
863
void bdrv_commit_all(void)
864
{
865
    BlockDriverState *bs;
866

    
867
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
868
        bdrv_commit(bs);
869
    }
870
}
871

    
872
/*
873
 * Return values:
874
 * 0        - success
875
 * -EINVAL  - backing format specified, but no file
876
 * -ENOSPC  - can't update the backing file because no space is left in the
877
 *            image file header
878
 * -ENOTSUP - format driver doesn't support changing the backing file
879
 */
880
int bdrv_change_backing_file(BlockDriverState *bs,
881
    const char *backing_file, const char *backing_fmt)
882
{
883
    BlockDriver *drv = bs->drv;
884

    
885
    if (drv->bdrv_change_backing_file != NULL) {
886
        return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
887
    } else {
888
        return -ENOTSUP;
889
    }
890
}
891

    
892
static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
893
                                   size_t size)
894
{
895
    int64_t len;
896

    
897
    if (!bdrv_is_inserted(bs))
898
        return -ENOMEDIUM;
899

    
900
    if (bs->growable)
901
        return 0;
902

    
903
    len = bdrv_getlength(bs);
904

    
905
    if (offset < 0)
906
        return -EIO;
907

    
908
    if ((offset > len) || (len - offset < size))
909
        return -EIO;
910

    
911
    return 0;
912
}
913

    
914
static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
915
                              int nb_sectors)
916
{
917
    return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
918
                                   nb_sectors * BDRV_SECTOR_SIZE);
919
}
920

    
921
/* return < 0 if error. See bdrv_write() for the return codes */
922
int bdrv_read(BlockDriverState *bs, int64_t sector_num,
923
              uint8_t *buf, int nb_sectors)
924
{
925
    BlockDriver *drv = bs->drv;
926

    
927
    if (!drv)
928
        return -ENOMEDIUM;
929
    if (bdrv_check_request(bs, sector_num, nb_sectors))
930
        return -EIO;
931

    
932
    return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
933
}
934

    
935
static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
936
                             int nb_sectors, int dirty)
937
{
938
    int64_t start, end;
939
    unsigned long val, idx, bit;
940

    
941
    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
942
    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
943

    
944
    for (; start <= end; start++) {
945
        idx = start / (sizeof(unsigned long) * 8);
946
        bit = start % (sizeof(unsigned long) * 8);
947
        val = bs->dirty_bitmap[idx];
948
        if (dirty) {
949
            if (!(val & (1UL << bit))) {
950
                bs->dirty_count++;
951
                val |= 1UL << bit;
952
            }
953
        } else {
954
            if (val & (1UL << bit)) {
955
                bs->dirty_count--;
956
                val &= ~(1UL << bit);
957
            }
958
        }
959
        bs->dirty_bitmap[idx] = val;
960
    }
961
}
962

    
963
/* Return < 0 if error. Important errors are:
964
  -EIO         generic I/O error (may happen for all errors)
965
  -ENOMEDIUM   No media inserted.
966
  -EINVAL      Invalid sector number or nb_sectors
967
  -EACCES      Trying to write a read-only device
968
*/
969
int bdrv_write(BlockDriverState *bs, int64_t sector_num,
970
               const uint8_t *buf, int nb_sectors)
971
{
972
    BlockDriver *drv = bs->drv;
973
    if (!bs->drv)
974
        return -ENOMEDIUM;
975
    if (bs->read_only)
976
        return -EACCES;
977
    if (bdrv_check_request(bs, sector_num, nb_sectors))
978
        return -EIO;
979

    
980
    if (bs->dirty_bitmap) {
981
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
982
    }
983

    
984
    if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
985
        bs->wr_highest_sector = sector_num + nb_sectors - 1;
986
    }
987

    
988
    return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
989
}
990

    
991
int bdrv_pread(BlockDriverState *bs, int64_t offset,
992
               void *buf, int count1)
993
{
994
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
995
    int len, nb_sectors, count;
996
    int64_t sector_num;
997
    int ret;
998

    
999
    count = count1;
1000
    /* first read to align to sector start */
1001
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1002
    if (len > count)
1003
        len = count;
1004
    sector_num = offset >> BDRV_SECTOR_BITS;
1005
    if (len > 0) {
1006
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1007
            return ret;
1008
        memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
1009
        count -= len;
1010
        if (count == 0)
1011
            return count1;
1012
        sector_num++;
1013
        buf += len;
1014
    }
1015

    
1016
    /* read the sectors "in place" */
1017
    nb_sectors = count >> BDRV_SECTOR_BITS;
1018
    if (nb_sectors > 0) {
1019
        if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
1020
            return ret;
1021
        sector_num += nb_sectors;
1022
        len = nb_sectors << BDRV_SECTOR_BITS;
1023
        buf += len;
1024
        count -= len;
1025
    }
1026

    
1027
    /* add data from the last sector */
1028
    if (count > 0) {
1029
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1030
            return ret;
1031
        memcpy(buf, tmp_buf, count);
1032
    }
1033
    return count1;
1034
}
1035

    
1036
int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
1037
                const void *buf, int count1)
1038
{
1039
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1040
    int len, nb_sectors, count;
1041
    int64_t sector_num;
1042
    int ret;
1043

    
1044
    count = count1;
1045
    /* first write to align to sector start */
1046
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1047
    if (len > count)
1048
        len = count;
1049
    sector_num = offset >> BDRV_SECTOR_BITS;
1050
    if (len > 0) {
1051
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1052
            return ret;
1053
        memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
1054
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1055
            return ret;
1056
        count -= len;
1057
        if (count == 0)
1058
            return count1;
1059
        sector_num++;
1060
        buf += len;
1061
    }
1062

    
1063
    /* write the sectors "in place" */
1064
    nb_sectors = count >> BDRV_SECTOR_BITS;
1065
    if (nb_sectors > 0) {
1066
        if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
1067
            return ret;
1068
        sector_num += nb_sectors;
1069
        len = nb_sectors << BDRV_SECTOR_BITS;
1070
        buf += len;
1071
        count -= len;
1072
    }
1073

    
1074
    /* add data from the last sector */
1075
    if (count > 0) {
1076
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1077
            return ret;
1078
        memcpy(tmp_buf, buf, count);
1079
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1080
            return ret;
1081
    }
1082
    return count1;
1083
}
1084

    
1085
/*
1086
 * Writes to the file and ensures that no writes are reordered across this
1087
 * request (acts as a barrier)
1088
 *
1089
 * Returns 0 on success, -errno in error cases.
1090
 */
1091
int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
1092
    const void *buf, int count)
1093
{
1094
    int ret;
1095

    
1096
    ret = bdrv_pwrite(bs, offset, buf, count);
1097
    if (ret < 0) {
1098
        return ret;
1099
    }
1100

    
1101
    /* No flush needed for cache=writethrough, it uses O_DSYNC */
1102
    if ((bs->open_flags & BDRV_O_CACHE_MASK) != 0) {
1103
        bdrv_flush(bs);
1104
    }
1105

    
1106
    return 0;
1107
}
1108

    
1109
/*
1110
 * Writes to the file and ensures that no writes are reordered across this
1111
 * request (acts as a barrier)
1112
 *
1113
 * Returns 0 on success, -errno in error cases.
1114
 */
1115
int bdrv_write_sync(BlockDriverState *bs, int64_t sector_num,
1116
    const uint8_t *buf, int nb_sectors)
1117
{
1118
    return bdrv_pwrite_sync(bs, BDRV_SECTOR_SIZE * sector_num,
1119
        buf, BDRV_SECTOR_SIZE * nb_sectors);
1120
}
1121

    
1122
/**
1123
 * Truncate file to 'offset' bytes (needed only for file protocols)
1124
 */
1125
int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1126
{
1127
    BlockDriver *drv = bs->drv;
1128
    int ret;
1129
    if (!drv)
1130
        return -ENOMEDIUM;
1131
    if (!drv->bdrv_truncate)
1132
        return -ENOTSUP;
1133
    if (bs->read_only)
1134
        return -EACCES;
1135
    ret = drv->bdrv_truncate(bs, offset);
1136
    if (ret == 0) {
1137
        ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1138
    }
1139
    return ret;
1140
}
1141

    
1142
/**
1143
 * Length of a file in bytes. Return < 0 if error or unknown.
1144
 */
1145
int64_t bdrv_getlength(BlockDriverState *bs)
1146
{
1147
    BlockDriver *drv = bs->drv;
1148
    if (!drv)
1149
        return -ENOMEDIUM;
1150

    
1151
    /* Fixed size devices use the total_sectors value for speed instead of
1152
       issuing a length query (like lseek) on each call.  Also, legacy block
1153
       drivers don't provide a bdrv_getlength function and must use
1154
       total_sectors. */
1155
    if (!bs->growable || !drv->bdrv_getlength) {
1156
        return bs->total_sectors * BDRV_SECTOR_SIZE;
1157
    }
1158
    return drv->bdrv_getlength(bs);
1159
}
1160

    
1161
/* return 0 as number of sectors if no device present or error */
1162
void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1163
{
1164
    int64_t length;
1165
    length = bdrv_getlength(bs);
1166
    if (length < 0)
1167
        length = 0;
1168
    else
1169
        length = length >> BDRV_SECTOR_BITS;
1170
    *nb_sectors_ptr = length;
1171
}
1172

    
1173
struct partition {
1174
        uint8_t boot_ind;           /* 0x80 - active */
1175
        uint8_t head;               /* starting head */
1176
        uint8_t sector;             /* starting sector */
1177
        uint8_t cyl;                /* starting cylinder */
1178
        uint8_t sys_ind;            /* What partition type */
1179
        uint8_t end_head;           /* end head */
1180
        uint8_t end_sector;         /* end sector */
1181
        uint8_t end_cyl;            /* end cylinder */
1182
        uint32_t start_sect;        /* starting sector counting from 0 */
1183
        uint32_t nr_sects;          /* nr of sectors in partition */
1184
} __attribute__((packed));
1185

    
1186
/* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1187
static int guess_disk_lchs(BlockDriverState *bs,
1188
                           int *pcylinders, int *pheads, int *psectors)
1189
{
1190
    uint8_t buf[BDRV_SECTOR_SIZE];
1191
    int ret, i, heads, sectors, cylinders;
1192
    struct partition *p;
1193
    uint32_t nr_sects;
1194
    uint64_t nb_sectors;
1195

    
1196
    bdrv_get_geometry(bs, &nb_sectors);
1197

    
1198
    ret = bdrv_read(bs, 0, buf, 1);
1199
    if (ret < 0)
1200
        return -1;
1201
    /* test msdos magic */
1202
    if (buf[510] != 0x55 || buf[511] != 0xaa)
1203
        return -1;
1204
    for(i = 0; i < 4; i++) {
1205
        p = ((struct partition *)(buf + 0x1be)) + i;
1206
        nr_sects = le32_to_cpu(p->nr_sects);
1207
        if (nr_sects && p->end_head) {
1208
            /* We make the assumption that the partition terminates on
1209
               a cylinder boundary */
1210
            heads = p->end_head + 1;
1211
            sectors = p->end_sector & 63;
1212
            if (sectors == 0)
1213
                continue;
1214
            cylinders = nb_sectors / (heads * sectors);
1215
            if (cylinders < 1 || cylinders > 16383)
1216
                continue;
1217
            *pheads = heads;
1218
            *psectors = sectors;
1219
            *pcylinders = cylinders;
1220
#if 0
1221
            printf("guessed geometry: LCHS=%d %d %d\n",
1222
                   cylinders, heads, sectors);
1223
#endif
1224
            return 0;
1225
        }
1226
    }
1227
    return -1;
1228
}
1229

    
1230
void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1231
{
1232
    int translation, lba_detected = 0;
1233
    int cylinders, heads, secs;
1234
    uint64_t nb_sectors;
1235

    
1236
    /* if a geometry hint is available, use it */
1237
    bdrv_get_geometry(bs, &nb_sectors);
1238
    bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1239
    translation = bdrv_get_translation_hint(bs);
1240
    if (cylinders != 0) {
1241
        *pcyls = cylinders;
1242
        *pheads = heads;
1243
        *psecs = secs;
1244
    } else {
1245
        if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1246
            if (heads > 16) {
1247
                /* if heads > 16, it means that a BIOS LBA
1248
                   translation was active, so the default
1249
                   hardware geometry is OK */
1250
                lba_detected = 1;
1251
                goto default_geometry;
1252
            } else {
1253
                *pcyls = cylinders;
1254
                *pheads = heads;
1255
                *psecs = secs;
1256
                /* disable any translation to be in sync with
1257
                   the logical geometry */
1258
                if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1259
                    bdrv_set_translation_hint(bs,
1260
                                              BIOS_ATA_TRANSLATION_NONE);
1261
                }
1262
            }
1263
        } else {
1264
        default_geometry:
1265
            /* if no geometry, use a standard physical disk geometry */
1266
            cylinders = nb_sectors / (16 * 63);
1267

    
1268
            if (cylinders > 16383)
1269
                cylinders = 16383;
1270
            else if (cylinders < 2)
1271
                cylinders = 2;
1272
            *pcyls = cylinders;
1273
            *pheads = 16;
1274
            *psecs = 63;
1275
            if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1276
                if ((*pcyls * *pheads) <= 131072) {
1277
                    bdrv_set_translation_hint(bs,
1278
                                              BIOS_ATA_TRANSLATION_LARGE);
1279
                } else {
1280
                    bdrv_set_translation_hint(bs,
1281
                                              BIOS_ATA_TRANSLATION_LBA);
1282
                }
1283
            }
1284
        }
1285
        bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1286
    }
1287
}
1288

    
1289
void bdrv_set_geometry_hint(BlockDriverState *bs,
1290
                            int cyls, int heads, int secs)
1291
{
1292
    bs->cyls = cyls;
1293
    bs->heads = heads;
1294
    bs->secs = secs;
1295
}
1296

    
1297
void bdrv_set_type_hint(BlockDriverState *bs, int type)
1298
{
1299
    bs->type = type;
1300
    bs->removable = ((type == BDRV_TYPE_CDROM ||
1301
                      type == BDRV_TYPE_FLOPPY));
1302
}
1303

    
1304
void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1305
{
1306
    bs->translation = translation;
1307
}
1308

    
1309
void bdrv_get_geometry_hint(BlockDriverState *bs,
1310
                            int *pcyls, int *pheads, int *psecs)
1311
{
1312
    *pcyls = bs->cyls;
1313
    *pheads = bs->heads;
1314
    *psecs = bs->secs;
1315
}
1316

    
1317
int bdrv_get_type_hint(BlockDriverState *bs)
1318
{
1319
    return bs->type;
1320
}
1321

    
1322
int bdrv_get_translation_hint(BlockDriverState *bs)
1323
{
1324
    return bs->translation;
1325
}
1326

    
1327
void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
1328
                       BlockErrorAction on_write_error)
1329
{
1330
    bs->on_read_error = on_read_error;
1331
    bs->on_write_error = on_write_error;
1332
}
1333

    
1334
BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
1335
{
1336
    return is_read ? bs->on_read_error : bs->on_write_error;
1337
}
1338

    
1339
void bdrv_set_removable(BlockDriverState *bs, int removable)
1340
{
1341
    bs->removable = removable;
1342
    if (removable && bs == bs_snapshots) {
1343
        bs_snapshots = NULL;
1344
    }
1345
}
1346

    
1347
int bdrv_is_removable(BlockDriverState *bs)
1348
{
1349
    return bs->removable;
1350
}
1351

    
1352
int bdrv_is_read_only(BlockDriverState *bs)
1353
{
1354
    return bs->read_only;
1355
}
1356

    
1357
int bdrv_is_sg(BlockDriverState *bs)
1358
{
1359
    return bs->sg;
1360
}
1361

    
1362
int bdrv_enable_write_cache(BlockDriverState *bs)
1363
{
1364
    return bs->enable_write_cache;
1365
}
1366

    
1367
/* XXX: no longer used */
1368
void bdrv_set_change_cb(BlockDriverState *bs,
1369
                        void (*change_cb)(void *opaque), void *opaque)
1370
{
1371
    bs->change_cb = change_cb;
1372
    bs->change_opaque = opaque;
1373
}
1374

    
1375
int bdrv_is_encrypted(BlockDriverState *bs)
1376
{
1377
    if (bs->backing_hd && bs->backing_hd->encrypted)
1378
        return 1;
1379
    return bs->encrypted;
1380
}
1381

    
1382
int bdrv_key_required(BlockDriverState *bs)
1383
{
1384
    BlockDriverState *backing_hd = bs->backing_hd;
1385

    
1386
    if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1387
        return 1;
1388
    return (bs->encrypted && !bs->valid_key);
1389
}
1390

    
1391
int bdrv_set_key(BlockDriverState *bs, const char *key)
1392
{
1393
    int ret;
1394
    if (bs->backing_hd && bs->backing_hd->encrypted) {
1395
        ret = bdrv_set_key(bs->backing_hd, key);
1396
        if (ret < 0)
1397
            return ret;
1398
        if (!bs->encrypted)
1399
            return 0;
1400
    }
1401
    if (!bs->encrypted) {
1402
        return -EINVAL;
1403
    } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1404
        return -ENOMEDIUM;
1405
    }
1406
    ret = bs->drv->bdrv_set_key(bs, key);
1407
    if (ret < 0) {
1408
        bs->valid_key = 0;
1409
    } else if (!bs->valid_key) {
1410
        bs->valid_key = 1;
1411
        /* call the change callback now, we skipped it on open */
1412
        bs->media_changed = 1;
1413
        if (bs->change_cb)
1414
            bs->change_cb(bs->change_opaque);
1415
    }
1416
    return ret;
1417
}
1418

    
1419
void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1420
{
1421
    if (!bs->drv) {
1422
        buf[0] = '\0';
1423
    } else {
1424
        pstrcpy(buf, buf_size, bs->drv->format_name);
1425
    }
1426
}
1427

    
1428
void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1429
                         void *opaque)
1430
{
1431
    BlockDriver *drv;
1432

    
1433
    QLIST_FOREACH(drv, &bdrv_drivers, list) {
1434
        it(opaque, drv->format_name);
1435
    }
1436
}
1437

    
1438
BlockDriverState *bdrv_find(const char *name)
1439
{
1440
    BlockDriverState *bs;
1441

    
1442
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1443
        if (!strcmp(name, bs->device_name)) {
1444
            return bs;
1445
        }
1446
    }
1447
    return NULL;
1448
}
1449

    
1450
BlockDriverState *bdrv_next(BlockDriverState *bs)
1451
{
1452
    if (!bs) {
1453
        return QTAILQ_FIRST(&bdrv_states);
1454
    }
1455
    return QTAILQ_NEXT(bs, list);
1456
}
1457

    
1458
void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1459
{
1460
    BlockDriverState *bs;
1461

    
1462
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1463
        it(opaque, bs);
1464
    }
1465
}
1466

    
1467
const char *bdrv_get_device_name(BlockDriverState *bs)
1468
{
1469
    return bs->device_name;
1470
}
1471

    
1472
int bdrv_flush(BlockDriverState *bs)
1473
{
1474
    if (bs->open_flags & BDRV_O_NO_FLUSH) {
1475
        return 0;
1476
    }
1477

    
1478
    if (bs->drv && bs->drv->bdrv_flush) {
1479
        return bs->drv->bdrv_flush(bs);
1480
    }
1481

    
1482
    /*
1483
     * Some block drivers always operate in either writethrough or unsafe mode
1484
     * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1485
     * the server works (because the behaviour is hardcoded or depends on
1486
     * server-side configuration), so we can't ensure that everything is safe
1487
     * on disk. Returning an error doesn't work because that would break guests
1488
     * even if the server operates in writethrough mode.
1489
     *
1490
     * Let's hope the user knows what he's doing.
1491
     */
1492
    return 0;
1493
}
1494

    
1495
void bdrv_flush_all(void)
1496
{
1497
    BlockDriverState *bs;
1498

    
1499
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1500
        if (bs->drv && !bdrv_is_read_only(bs) &&
1501
            (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1502
            bdrv_flush(bs);
1503
        }
1504
    }
1505
}
1506

    
1507
int bdrv_has_zero_init(BlockDriverState *bs)
1508
{
1509
    assert(bs->drv);
1510

    
1511
    if (bs->drv->bdrv_has_zero_init) {
1512
        return bs->drv->bdrv_has_zero_init(bs);
1513
    }
1514

    
1515
    return 1;
1516
}
1517

    
1518
int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
1519
{
1520
    if (!bs->drv) {
1521
        return -ENOMEDIUM;
1522
    }
1523
    if (!bs->drv->bdrv_discard) {
1524
        return 0;
1525
    }
1526
    return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
1527
}
1528

    
1529
/*
1530
 * Returns true iff the specified sector is present in the disk image. Drivers
1531
 * not implementing the functionality are assumed to not support backing files,
1532
 * hence all their sectors are reported as allocated.
1533
 *
1534
 * 'pnum' is set to the number of sectors (including and immediately following
1535
 * the specified sector) that are known to be in the same
1536
 * allocated/unallocated state.
1537
 *
1538
 * 'nb_sectors' is the max value 'pnum' should be set to.
1539
 */
1540
int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1541
        int *pnum)
1542
{
1543
    int64_t n;
1544
    if (!bs->drv->bdrv_is_allocated) {
1545
        if (sector_num >= bs->total_sectors) {
1546
            *pnum = 0;
1547
            return 0;
1548
        }
1549
        n = bs->total_sectors - sector_num;
1550
        *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1551
        return 1;
1552
    }
1553
    return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1554
}
1555

    
1556
void bdrv_mon_event(const BlockDriverState *bdrv,
1557
                    BlockMonEventAction action, int is_read)
1558
{
1559
    QObject *data;
1560
    const char *action_str;
1561

    
1562
    switch (action) {
1563
    case BDRV_ACTION_REPORT:
1564
        action_str = "report";
1565
        break;
1566
    case BDRV_ACTION_IGNORE:
1567
        action_str = "ignore";
1568
        break;
1569
    case BDRV_ACTION_STOP:
1570
        action_str = "stop";
1571
        break;
1572
    default:
1573
        abort();
1574
    }
1575

    
1576
    data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1577
                              bdrv->device_name,
1578
                              action_str,
1579
                              is_read ? "read" : "write");
1580
    monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1581

    
1582
    qobject_decref(data);
1583
}
1584

    
1585
static void bdrv_print_dict(QObject *obj, void *opaque)
1586
{
1587
    QDict *bs_dict;
1588
    Monitor *mon = opaque;
1589

    
1590
    bs_dict = qobject_to_qdict(obj);
1591

    
1592
    monitor_printf(mon, "%s: type=%s removable=%d",
1593
                        qdict_get_str(bs_dict, "device"),
1594
                        qdict_get_str(bs_dict, "type"),
1595
                        qdict_get_bool(bs_dict, "removable"));
1596

    
1597
    if (qdict_get_bool(bs_dict, "removable")) {
1598
        monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1599
    }
1600

    
1601
    if (qdict_haskey(bs_dict, "inserted")) {
1602
        QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1603

    
1604
        monitor_printf(mon, " file=");
1605
        monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1606
        if (qdict_haskey(qdict, "backing_file")) {
1607
            monitor_printf(mon, " backing_file=");
1608
            monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1609
        }
1610
        monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1611
                            qdict_get_bool(qdict, "ro"),
1612
                            qdict_get_str(qdict, "drv"),
1613
                            qdict_get_bool(qdict, "encrypted"));
1614
    } else {
1615
        monitor_printf(mon, " [not inserted]");
1616
    }
1617

    
1618
    monitor_printf(mon, "\n");
1619
}
1620

    
1621
void bdrv_info_print(Monitor *mon, const QObject *data)
1622
{
1623
    qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1624
}
1625

    
1626
void bdrv_info(Monitor *mon, QObject **ret_data)
1627
{
1628
    QList *bs_list;
1629
    BlockDriverState *bs;
1630

    
1631
    bs_list = qlist_new();
1632

    
1633
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1634
        QObject *bs_obj;
1635
        const char *type = "unknown";
1636

    
1637
        switch(bs->type) {
1638
        case BDRV_TYPE_HD:
1639
            type = "hd";
1640
            break;
1641
        case BDRV_TYPE_CDROM:
1642
            type = "cdrom";
1643
            break;
1644
        case BDRV_TYPE_FLOPPY:
1645
            type = "floppy";
1646
            break;
1647
        }
1648

    
1649
        bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1650
                                    "'removable': %i, 'locked': %i }",
1651
                                    bs->device_name, type, bs->removable,
1652
                                    bs->locked);
1653

    
1654
        if (bs->drv) {
1655
            QObject *obj;
1656
            QDict *bs_dict = qobject_to_qdict(bs_obj);
1657

    
1658
            obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1659
                                     "'encrypted': %i }",
1660
                                     bs->filename, bs->read_only,
1661
                                     bs->drv->format_name,
1662
                                     bdrv_is_encrypted(bs));
1663
            if (bs->backing_file[0] != '\0') {
1664
                QDict *qdict = qobject_to_qdict(obj);
1665
                qdict_put(qdict, "backing_file",
1666
                          qstring_from_str(bs->backing_file));
1667
            }
1668

    
1669
            qdict_put_obj(bs_dict, "inserted", obj);
1670
        }
1671
        qlist_append_obj(bs_list, bs_obj);
1672
    }
1673

    
1674
    *ret_data = QOBJECT(bs_list);
1675
}
1676

    
1677
static void bdrv_stats_iter(QObject *data, void *opaque)
1678
{
1679
    QDict *qdict;
1680
    Monitor *mon = opaque;
1681

    
1682
    qdict = qobject_to_qdict(data);
1683
    monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1684

    
1685
    qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1686
    monitor_printf(mon, " rd_bytes=%" PRId64
1687
                        " wr_bytes=%" PRId64
1688
                        " rd_operations=%" PRId64
1689
                        " wr_operations=%" PRId64
1690
                        "\n",
1691
                        qdict_get_int(qdict, "rd_bytes"),
1692
                        qdict_get_int(qdict, "wr_bytes"),
1693
                        qdict_get_int(qdict, "rd_operations"),
1694
                        qdict_get_int(qdict, "wr_operations"));
1695
}
1696

    
1697
void bdrv_stats_print(Monitor *mon, const QObject *data)
1698
{
1699
    qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1700
}
1701

    
1702
static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1703
{
1704
    QObject *res;
1705
    QDict *dict;
1706

    
1707
    res = qobject_from_jsonf("{ 'stats': {"
1708
                             "'rd_bytes': %" PRId64 ","
1709
                             "'wr_bytes': %" PRId64 ","
1710
                             "'rd_operations': %" PRId64 ","
1711
                             "'wr_operations': %" PRId64 ","
1712
                             "'wr_highest_offset': %" PRId64
1713
                             "} }",
1714
                             bs->rd_bytes, bs->wr_bytes,
1715
                             bs->rd_ops, bs->wr_ops,
1716
                             bs->wr_highest_sector *
1717
                             (uint64_t)BDRV_SECTOR_SIZE);
1718
    dict  = qobject_to_qdict(res);
1719

    
1720
    if (*bs->device_name) {
1721
        qdict_put(dict, "device", qstring_from_str(bs->device_name));
1722
    }
1723

    
1724
    if (bs->file) {
1725
        QObject *parent = bdrv_info_stats_bs(bs->file);
1726
        qdict_put_obj(dict, "parent", parent);
1727
    }
1728

    
1729
    return res;
1730
}
1731

    
1732
void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1733
{
1734
    QObject *obj;
1735
    QList *devices;
1736
    BlockDriverState *bs;
1737

    
1738
    devices = qlist_new();
1739

    
1740
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1741
        obj = bdrv_info_stats_bs(bs);
1742
        qlist_append_obj(devices, obj);
1743
    }
1744

    
1745
    *ret_data = QOBJECT(devices);
1746
}
1747

    
1748
const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1749
{
1750
    if (bs->backing_hd && bs->backing_hd->encrypted)
1751
        return bs->backing_file;
1752
    else if (bs->encrypted)
1753
        return bs->filename;
1754
    else
1755
        return NULL;
1756
}
1757

    
1758
void bdrv_get_backing_filename(BlockDriverState *bs,
1759
                               char *filename, int filename_size)
1760
{
1761
    if (!bs->backing_file) {
1762
        pstrcpy(filename, filename_size, "");
1763
    } else {
1764
        pstrcpy(filename, filename_size, bs->backing_file);
1765
    }
1766
}
1767

    
1768
int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1769
                          const uint8_t *buf, int nb_sectors)
1770
{
1771
    BlockDriver *drv = bs->drv;
1772
    if (!drv)
1773
        return -ENOMEDIUM;
1774
    if (!drv->bdrv_write_compressed)
1775
        return -ENOTSUP;
1776
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1777
        return -EIO;
1778

    
1779
    if (bs->dirty_bitmap) {
1780
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1781
    }
1782

    
1783
    return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1784
}
1785

    
1786
int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1787
{
1788
    BlockDriver *drv = bs->drv;
1789
    if (!drv)
1790
        return -ENOMEDIUM;
1791
    if (!drv->bdrv_get_info)
1792
        return -ENOTSUP;
1793
    memset(bdi, 0, sizeof(*bdi));
1794
    return drv->bdrv_get_info(bs, bdi);
1795
}
1796

    
1797
int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1798
                      int64_t pos, int size)
1799
{
1800
    BlockDriver *drv = bs->drv;
1801
    if (!drv)
1802
        return -ENOMEDIUM;
1803
    if (drv->bdrv_save_vmstate)
1804
        return drv->bdrv_save_vmstate(bs, buf, pos, size);
1805
    if (bs->file)
1806
        return bdrv_save_vmstate(bs->file, buf, pos, size);
1807
    return -ENOTSUP;
1808
}
1809

    
1810
int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1811
                      int64_t pos, int size)
1812
{
1813
    BlockDriver *drv = bs->drv;
1814
    if (!drv)
1815
        return -ENOMEDIUM;
1816
    if (drv->bdrv_load_vmstate)
1817
        return drv->bdrv_load_vmstate(bs, buf, pos, size);
1818
    if (bs->file)
1819
        return bdrv_load_vmstate(bs->file, buf, pos, size);
1820
    return -ENOTSUP;
1821
}
1822

    
1823
void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1824
{
1825
    BlockDriver *drv = bs->drv;
1826

    
1827
    if (!drv || !drv->bdrv_debug_event) {
1828
        return;
1829
    }
1830

    
1831
    return drv->bdrv_debug_event(bs, event);
1832

    
1833
}
1834

    
1835
/**************************************************************/
1836
/* handling of snapshots */
1837

    
1838
int bdrv_can_snapshot(BlockDriverState *bs)
1839
{
1840
    BlockDriver *drv = bs->drv;
1841
    if (!drv || bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1842
        return 0;
1843
    }
1844

    
1845
    if (!drv->bdrv_snapshot_create) {
1846
        if (bs->file != NULL) {
1847
            return bdrv_can_snapshot(bs->file);
1848
        }
1849
        return 0;
1850
    }
1851

    
1852
    return 1;
1853
}
1854

    
1855
int bdrv_is_snapshot(BlockDriverState *bs)
1856
{
1857
    return !!(bs->open_flags & BDRV_O_SNAPSHOT);
1858
}
1859

    
1860
BlockDriverState *bdrv_snapshots(void)
1861
{
1862
    BlockDriverState *bs;
1863

    
1864
    if (bs_snapshots) {
1865
        return bs_snapshots;
1866
    }
1867

    
1868
    bs = NULL;
1869
    while ((bs = bdrv_next(bs))) {
1870
        if (bdrv_can_snapshot(bs)) {
1871
            bs_snapshots = bs;
1872
            return bs;
1873
        }
1874
    }
1875
    return NULL;
1876
}
1877

    
1878
int bdrv_snapshot_create(BlockDriverState *bs,
1879
                         QEMUSnapshotInfo *sn_info)
1880
{
1881
    BlockDriver *drv = bs->drv;
1882
    if (!drv)
1883
        return -ENOMEDIUM;
1884
    if (drv->bdrv_snapshot_create)
1885
        return drv->bdrv_snapshot_create(bs, sn_info);
1886
    if (bs->file)
1887
        return bdrv_snapshot_create(bs->file, sn_info);
1888
    return -ENOTSUP;
1889
}
1890

    
1891
int bdrv_snapshot_goto(BlockDriverState *bs,
1892
                       const char *snapshot_id)
1893
{
1894
    BlockDriver *drv = bs->drv;
1895
    int ret, open_ret;
1896

    
1897
    if (!drv)
1898
        return -ENOMEDIUM;
1899
    if (drv->bdrv_snapshot_goto)
1900
        return drv->bdrv_snapshot_goto(bs, snapshot_id);
1901

    
1902
    if (bs->file) {
1903
        drv->bdrv_close(bs);
1904
        ret = bdrv_snapshot_goto(bs->file, snapshot_id);
1905
        open_ret = drv->bdrv_open(bs, bs->open_flags);
1906
        if (open_ret < 0) {
1907
            bdrv_delete(bs->file);
1908
            bs->drv = NULL;
1909
            return open_ret;
1910
        }
1911
        return ret;
1912
    }
1913

    
1914
    return -ENOTSUP;
1915
}
1916

    
1917
int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1918
{
1919
    BlockDriver *drv = bs->drv;
1920
    if (!drv)
1921
        return -ENOMEDIUM;
1922
    if (drv->bdrv_snapshot_delete)
1923
        return drv->bdrv_snapshot_delete(bs, snapshot_id);
1924
    if (bs->file)
1925
        return bdrv_snapshot_delete(bs->file, snapshot_id);
1926
    return -ENOTSUP;
1927
}
1928

    
1929
int bdrv_snapshot_list(BlockDriverState *bs,
1930
                       QEMUSnapshotInfo **psn_info)
1931
{
1932
    BlockDriver *drv = bs->drv;
1933
    if (!drv)
1934
        return -ENOMEDIUM;
1935
    if (drv->bdrv_snapshot_list)
1936
        return drv->bdrv_snapshot_list(bs, psn_info);
1937
    if (bs->file)
1938
        return bdrv_snapshot_list(bs->file, psn_info);
1939
    return -ENOTSUP;
1940
}
1941

    
1942
int bdrv_snapshot_load_tmp(BlockDriverState *bs,
1943
        const char *snapshot_name)
1944
{
1945
    BlockDriver *drv = bs->drv;
1946
    if (!drv) {
1947
        return -ENOMEDIUM;
1948
    }
1949
    if (!bs->read_only) {
1950
        return -EINVAL;
1951
    }
1952
    if (drv->bdrv_snapshot_load_tmp) {
1953
        return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
1954
    }
1955
    return -ENOTSUP;
1956
}
1957

    
1958
#define NB_SUFFIXES 4
1959

    
1960
char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1961
{
1962
    static const char suffixes[NB_SUFFIXES] = "KMGT";
1963
    int64_t base;
1964
    int i;
1965

    
1966
    if (size <= 999) {
1967
        snprintf(buf, buf_size, "%" PRId64, size);
1968
    } else {
1969
        base = 1024;
1970
        for(i = 0; i < NB_SUFFIXES; i++) {
1971
            if (size < (10 * base)) {
1972
                snprintf(buf, buf_size, "%0.1f%c",
1973
                         (double)size / base,
1974
                         suffixes[i]);
1975
                break;
1976
            } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1977
                snprintf(buf, buf_size, "%" PRId64 "%c",
1978
                         ((size + (base >> 1)) / base),
1979
                         suffixes[i]);
1980
                break;
1981
            }
1982
            base = base * 1024;
1983
        }
1984
    }
1985
    return buf;
1986
}
1987

    
1988
char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1989
{
1990
    char buf1[128], date_buf[128], clock_buf[128];
1991
#ifdef _WIN32
1992
    struct tm *ptm;
1993
#else
1994
    struct tm tm;
1995
#endif
1996
    time_t ti;
1997
    int64_t secs;
1998

    
1999
    if (!sn) {
2000
        snprintf(buf, buf_size,
2001
                 "%-10s%-20s%7s%20s%15s",
2002
                 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2003
    } else {
2004
        ti = sn->date_sec;
2005
#ifdef _WIN32
2006
        ptm = localtime(&ti);
2007
        strftime(date_buf, sizeof(date_buf),
2008
                 "%Y-%m-%d %H:%M:%S", ptm);
2009
#else
2010
        localtime_r(&ti, &tm);
2011
        strftime(date_buf, sizeof(date_buf),
2012
                 "%Y-%m-%d %H:%M:%S", &tm);
2013
#endif
2014
        secs = sn->vm_clock_nsec / 1000000000;
2015
        snprintf(clock_buf, sizeof(clock_buf),
2016
                 "%02d:%02d:%02d.%03d",
2017
                 (int)(secs / 3600),
2018
                 (int)((secs / 60) % 60),
2019
                 (int)(secs % 60),
2020
                 (int)((sn->vm_clock_nsec / 1000000) % 1000));
2021
        snprintf(buf, buf_size,
2022
                 "%-10s%-20s%7s%20s%15s",
2023
                 sn->id_str, sn->name,
2024
                 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
2025
                 date_buf,
2026
                 clock_buf);
2027
    }
2028
    return buf;
2029
}
2030

    
2031

    
2032
/**************************************************************/
2033
/* async I/Os */
2034

    
2035
BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
2036
                                 QEMUIOVector *qiov, int nb_sectors,
2037
                                 BlockDriverCompletionFunc *cb, void *opaque)
2038
{
2039
    BlockDriver *drv = bs->drv;
2040
    BlockDriverAIOCB *ret;
2041

    
2042
    trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
2043

    
2044
    if (!drv)
2045
        return NULL;
2046
    if (bdrv_check_request(bs, sector_num, nb_sectors))
2047
        return NULL;
2048

    
2049
    ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
2050
                              cb, opaque);
2051

    
2052
    if (ret) {
2053
        /* Update stats even though technically transfer has not happened. */
2054
        bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2055
        bs->rd_ops ++;
2056
    }
2057

    
2058
    return ret;
2059
}
2060

    
2061
typedef struct BlockCompleteData {
2062
    BlockDriverCompletionFunc *cb;
2063
    void *opaque;
2064
    BlockDriverState *bs;
2065
    int64_t sector_num;
2066
    int nb_sectors;
2067
} BlockCompleteData;
2068

    
2069
static void block_complete_cb(void *opaque, int ret)
2070
{
2071
    BlockCompleteData *b = opaque;
2072

    
2073
    if (b->bs->dirty_bitmap) {
2074
        set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
2075
    }
2076
    b->cb(b->opaque, ret);
2077
    qemu_free(b);
2078
}
2079

    
2080
static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
2081
                                             int64_t sector_num,
2082
                                             int nb_sectors,
2083
                                             BlockDriverCompletionFunc *cb,
2084
                                             void *opaque)
2085
{
2086
    BlockCompleteData *blkdata = qemu_mallocz(sizeof(BlockCompleteData));
2087

    
2088
    blkdata->bs = bs;
2089
    blkdata->cb = cb;
2090
    blkdata->opaque = opaque;
2091
    blkdata->sector_num = sector_num;
2092
    blkdata->nb_sectors = nb_sectors;
2093

    
2094
    return blkdata;
2095
}
2096

    
2097
BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
2098
                                  QEMUIOVector *qiov, int nb_sectors,
2099
                                  BlockDriverCompletionFunc *cb, void *opaque)
2100
{
2101
    BlockDriver *drv = bs->drv;
2102
    BlockDriverAIOCB *ret;
2103
    BlockCompleteData *blk_cb_data;
2104

    
2105
    trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
2106

    
2107
    if (!drv)
2108
        return NULL;
2109
    if (bs->read_only)
2110
        return NULL;
2111
    if (bdrv_check_request(bs, sector_num, nb_sectors))
2112
        return NULL;
2113

    
2114
    if (bs->dirty_bitmap) {
2115
        blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
2116
                                         opaque);
2117
        cb = &block_complete_cb;
2118
        opaque = blk_cb_data;
2119
    }
2120

    
2121
    ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
2122
                               cb, opaque);
2123

    
2124
    if (ret) {
2125
        /* Update stats even though technically transfer has not happened. */
2126
        bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2127
        bs->wr_ops ++;
2128
        if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
2129
            bs->wr_highest_sector = sector_num + nb_sectors - 1;
2130
        }
2131
    }
2132

    
2133
    return ret;
2134
}
2135

    
2136

    
2137
typedef struct MultiwriteCB {
2138
    int error;
2139
    int num_requests;
2140
    int num_callbacks;
2141
    struct {
2142
        BlockDriverCompletionFunc *cb;
2143
        void *opaque;
2144
        QEMUIOVector *free_qiov;
2145
        void *free_buf;
2146
    } callbacks[];
2147
} MultiwriteCB;
2148

    
2149
static void multiwrite_user_cb(MultiwriteCB *mcb)
2150
{
2151
    int i;
2152

    
2153
    for (i = 0; i < mcb->num_callbacks; i++) {
2154
        mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
2155
        if (mcb->callbacks[i].free_qiov) {
2156
            qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
2157
        }
2158
        qemu_free(mcb->callbacks[i].free_qiov);
2159
        qemu_vfree(mcb->callbacks[i].free_buf);
2160
    }
2161
}
2162

    
2163
static void multiwrite_cb(void *opaque, int ret)
2164
{
2165
    MultiwriteCB *mcb = opaque;
2166

    
2167
    trace_multiwrite_cb(mcb, ret);
2168

    
2169
    if (ret < 0 && !mcb->error) {
2170
        mcb->error = ret;
2171
    }
2172

    
2173
    mcb->num_requests--;
2174
    if (mcb->num_requests == 0) {
2175
        multiwrite_user_cb(mcb);
2176
        qemu_free(mcb);
2177
    }
2178
}
2179

    
2180
static int multiwrite_req_compare(const void *a, const void *b)
2181
{
2182
    const BlockRequest *req1 = a, *req2 = b;
2183

    
2184
    /*
2185
     * Note that we can't simply subtract req2->sector from req1->sector
2186
     * here as that could overflow the return value.
2187
     */
2188
    if (req1->sector > req2->sector) {
2189
        return 1;
2190
    } else if (req1->sector < req2->sector) {
2191
        return -1;
2192
    } else {
2193
        return 0;
2194
    }
2195
}
2196

    
2197
/*
2198
 * Takes a bunch of requests and tries to merge them. Returns the number of
2199
 * requests that remain after merging.
2200
 */
2201
static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
2202
    int num_reqs, MultiwriteCB *mcb)
2203
{
2204
    int i, outidx;
2205

    
2206
    // Sort requests by start sector
2207
    qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
2208

    
2209
    // Check if adjacent requests touch the same clusters. If so, combine them,
2210
    // filling up gaps with zero sectors.
2211
    outidx = 0;
2212
    for (i = 1; i < num_reqs; i++) {
2213
        int merge = 0;
2214
        int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
2215

    
2216
        // This handles the cases that are valid for all block drivers, namely
2217
        // exactly sequential writes and overlapping writes.
2218
        if (reqs[i].sector <= oldreq_last) {
2219
            merge = 1;
2220
        }
2221

    
2222
        // The block driver may decide that it makes sense to combine requests
2223
        // even if there is a gap of some sectors between them. In this case,
2224
        // the gap is filled with zeros (therefore only applicable for yet
2225
        // unused space in format like qcow2).
2226
        if (!merge && bs->drv->bdrv_merge_requests) {
2227
            merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
2228
        }
2229

    
2230
        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
2231
            merge = 0;
2232
        }
2233

    
2234
        if (merge) {
2235
            size_t size;
2236
            QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
2237
            qemu_iovec_init(qiov,
2238
                reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2239

    
2240
            // Add the first request to the merged one. If the requests are
2241
            // overlapping, drop the last sectors of the first request.
2242
            size = (reqs[i].sector - reqs[outidx].sector) << 9;
2243
            qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2244

    
2245
            // We might need to add some zeros between the two requests
2246
            if (reqs[i].sector > oldreq_last) {
2247
                size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2248
                uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2249
                memset(buf, 0, zero_bytes);
2250
                qemu_iovec_add(qiov, buf, zero_bytes);
2251
                mcb->callbacks[i].free_buf = buf;
2252
            }
2253

    
2254
            // Add the second request
2255
            qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2256

    
2257
            reqs[outidx].nb_sectors = qiov->size >> 9;
2258
            reqs[outidx].qiov = qiov;
2259

    
2260
            mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2261
        } else {
2262
            outidx++;
2263
            reqs[outidx].sector     = reqs[i].sector;
2264
            reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2265
            reqs[outidx].qiov       = reqs[i].qiov;
2266
        }
2267
    }
2268

    
2269
    return outidx + 1;
2270
}
2271

    
2272
/*
2273
 * Submit multiple AIO write requests at once.
2274
 *
2275
 * On success, the function returns 0 and all requests in the reqs array have
2276
 * been submitted. In error case this function returns -1, and any of the
2277
 * requests may or may not be submitted yet. In particular, this means that the
2278
 * callback will be called for some of the requests, for others it won't. The
2279
 * caller must check the error field of the BlockRequest to wait for the right
2280
 * callbacks (if error != 0, no callback will be called).
2281
 *
2282
 * The implementation may modify the contents of the reqs array, e.g. to merge
2283
 * requests. However, the fields opaque and error are left unmodified as they
2284
 * are used to signal failure for a single request to the caller.
2285
 */
2286
int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2287
{
2288
    BlockDriverAIOCB *acb;
2289
    MultiwriteCB *mcb;
2290
    int i;
2291

    
2292
    if (num_reqs == 0) {
2293
        return 0;
2294
    }
2295

    
2296
    // Create MultiwriteCB structure
2297
    mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2298
    mcb->num_requests = 0;
2299
    mcb->num_callbacks = num_reqs;
2300

    
2301
    for (i = 0; i < num_reqs; i++) {
2302
        mcb->callbacks[i].cb = reqs[i].cb;
2303
        mcb->callbacks[i].opaque = reqs[i].opaque;
2304
    }
2305

    
2306
    // Check for mergable requests
2307
    num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2308

    
2309
    trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
2310

    
2311
    /*
2312
     * Run the aio requests. As soon as one request can't be submitted
2313
     * successfully, fail all requests that are not yet submitted (we must
2314
     * return failure for all requests anyway)
2315
     *
2316
     * num_requests cannot be set to the right value immediately: If
2317
     * bdrv_aio_writev fails for some request, num_requests would be too high
2318
     * and therefore multiwrite_cb() would never recognize the multiwrite
2319
     * request as completed. We also cannot use the loop variable i to set it
2320
     * when the first request fails because the callback may already have been
2321
     * called for previously submitted requests. Thus, num_requests must be
2322
     * incremented for each request that is submitted.
2323
     *
2324
     * The problem that callbacks may be called early also means that we need
2325
     * to take care that num_requests doesn't become 0 before all requests are
2326
     * submitted - multiwrite_cb() would consider the multiwrite request
2327
     * completed. A dummy request that is "completed" by a manual call to
2328
     * multiwrite_cb() takes care of this.
2329
     */
2330
    mcb->num_requests = 1;
2331

    
2332
    // Run the aio requests
2333
    for (i = 0; i < num_reqs; i++) {
2334
        mcb->num_requests++;
2335
        acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2336
            reqs[i].nb_sectors, multiwrite_cb, mcb);
2337

    
2338
        if (acb == NULL) {
2339
            // We can only fail the whole thing if no request has been
2340
            // submitted yet. Otherwise we'll wait for the submitted AIOs to
2341
            // complete and report the error in the callback.
2342
            if (i == 0) {
2343
                trace_bdrv_aio_multiwrite_earlyfail(mcb);
2344
                goto fail;
2345
            } else {
2346
                trace_bdrv_aio_multiwrite_latefail(mcb, i);
2347
                multiwrite_cb(mcb, -EIO);
2348
                break;
2349
            }
2350
        }
2351
    }
2352

    
2353
    /* Complete the dummy request */
2354
    multiwrite_cb(mcb, 0);
2355

    
2356
    return 0;
2357

    
2358
fail:
2359
    for (i = 0; i < mcb->num_callbacks; i++) {
2360
        reqs[i].error = -EIO;
2361
    }
2362
    qemu_free(mcb);
2363
    return -1;
2364
}
2365

    
2366
BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2367
        BlockDriverCompletionFunc *cb, void *opaque)
2368
{
2369
    BlockDriver *drv = bs->drv;
2370

    
2371
    if (bs->open_flags & BDRV_O_NO_FLUSH) {
2372
        return bdrv_aio_noop_em(bs, cb, opaque);
2373
    }
2374

    
2375
    if (!drv)
2376
        return NULL;
2377
    return drv->bdrv_aio_flush(bs, cb, opaque);
2378
}
2379

    
2380
void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2381
{
2382
    acb->pool->cancel(acb);
2383
}
2384

    
2385

    
2386
/**************************************************************/
2387
/* async block device emulation */
2388

    
2389
typedef struct BlockDriverAIOCBSync {
2390
    BlockDriverAIOCB common;
2391
    QEMUBH *bh;
2392
    int ret;
2393
    /* vector translation state */
2394
    QEMUIOVector *qiov;
2395
    uint8_t *bounce;
2396
    int is_write;
2397
} BlockDriverAIOCBSync;
2398

    
2399
static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2400
{
2401
    BlockDriverAIOCBSync *acb =
2402
        container_of(blockacb, BlockDriverAIOCBSync, common);
2403
    qemu_bh_delete(acb->bh);
2404
    acb->bh = NULL;
2405
    qemu_aio_release(acb);
2406
}
2407

    
2408
static AIOPool bdrv_em_aio_pool = {
2409
    .aiocb_size         = sizeof(BlockDriverAIOCBSync),
2410
    .cancel             = bdrv_aio_cancel_em,
2411
};
2412

    
2413
static void bdrv_aio_bh_cb(void *opaque)
2414
{
2415
    BlockDriverAIOCBSync *acb = opaque;
2416

    
2417
    if (!acb->is_write)
2418
        qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2419
    qemu_vfree(acb->bounce);
2420
    acb->common.cb(acb->common.opaque, acb->ret);
2421
    qemu_bh_delete(acb->bh);
2422
    acb->bh = NULL;
2423
    qemu_aio_release(acb);
2424
}
2425

    
2426
static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2427
                                            int64_t sector_num,
2428
                                            QEMUIOVector *qiov,
2429
                                            int nb_sectors,
2430
                                            BlockDriverCompletionFunc *cb,
2431
                                            void *opaque,
2432
                                            int is_write)
2433

    
2434
{
2435
    BlockDriverAIOCBSync *acb;
2436

    
2437
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2438
    acb->is_write = is_write;
2439
    acb->qiov = qiov;
2440
    acb->bounce = qemu_blockalign(bs, qiov->size);
2441

    
2442
    if (!acb->bh)
2443
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2444

    
2445
    if (is_write) {
2446
        qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2447
        acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2448
    } else {
2449
        acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2450
    }
2451

    
2452
    qemu_bh_schedule(acb->bh);
2453

    
2454
    return &acb->common;
2455
}
2456

    
2457
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2458
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2459
        BlockDriverCompletionFunc *cb, void *opaque)
2460
{
2461
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2462
}
2463

    
2464
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2465
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2466
        BlockDriverCompletionFunc *cb, void *opaque)
2467
{
2468
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2469
}
2470

    
2471
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2472
        BlockDriverCompletionFunc *cb, void *opaque)
2473
{
2474
    BlockDriverAIOCBSync *acb;
2475

    
2476
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2477
    acb->is_write = 1; /* don't bounce in the completion hadler */
2478
    acb->qiov = NULL;
2479
    acb->bounce = NULL;
2480
    acb->ret = 0;
2481

    
2482
    if (!acb->bh)
2483
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2484

    
2485
    bdrv_flush(bs);
2486
    qemu_bh_schedule(acb->bh);
2487
    return &acb->common;
2488
}
2489

    
2490
static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2491
        BlockDriverCompletionFunc *cb, void *opaque)
2492
{
2493
    BlockDriverAIOCBSync *acb;
2494

    
2495
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2496
    acb->is_write = 1; /* don't bounce in the completion handler */
2497
    acb->qiov = NULL;
2498
    acb->bounce = NULL;
2499
    acb->ret = 0;
2500

    
2501
    if (!acb->bh) {
2502
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2503
    }
2504

    
2505
    qemu_bh_schedule(acb->bh);
2506
    return &acb->common;
2507
}
2508

    
2509
/**************************************************************/
2510
/* sync block device emulation */
2511

    
2512
static void bdrv_rw_em_cb(void *opaque, int ret)
2513
{
2514
    *(int *)opaque = ret;
2515
}
2516

    
2517
#define NOT_DONE 0x7fffffff
2518

    
2519
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2520
                        uint8_t *buf, int nb_sectors)
2521
{
2522
    int async_ret;
2523
    BlockDriverAIOCB *acb;
2524
    struct iovec iov;
2525
    QEMUIOVector qiov;
2526

    
2527
    async_context_push();
2528

    
2529
    async_ret = NOT_DONE;
2530
    iov.iov_base = (void *)buf;
2531
    iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2532
    qemu_iovec_init_external(&qiov, &iov, 1);
2533
    acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2534
        bdrv_rw_em_cb, &async_ret);
2535
    if (acb == NULL) {
2536
        async_ret = -1;
2537
        goto fail;
2538
    }
2539

    
2540
    while (async_ret == NOT_DONE) {
2541
        qemu_aio_wait();
2542
    }
2543

    
2544

    
2545
fail:
2546
    async_context_pop();
2547
    return async_ret;
2548
}
2549

    
2550
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2551
                         const uint8_t *buf, int nb_sectors)
2552
{
2553
    int async_ret;
2554
    BlockDriverAIOCB *acb;
2555
    struct iovec iov;
2556
    QEMUIOVector qiov;
2557

    
2558
    async_context_push();
2559

    
2560
    async_ret = NOT_DONE;
2561
    iov.iov_base = (void *)buf;
2562
    iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2563
    qemu_iovec_init_external(&qiov, &iov, 1);
2564
    acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2565
        bdrv_rw_em_cb, &async_ret);
2566
    if (acb == NULL) {
2567
        async_ret = -1;
2568
        goto fail;
2569
    }
2570
    while (async_ret == NOT_DONE) {
2571
        qemu_aio_wait();
2572
    }
2573

    
2574
fail:
2575
    async_context_pop();
2576
    return async_ret;
2577
}
2578

    
2579
void bdrv_init(void)
2580
{
2581
    module_call_init(MODULE_INIT_BLOCK);
2582
}
2583

    
2584
void bdrv_init_with_whitelist(void)
2585
{
2586
    use_bdrv_whitelist = 1;
2587
    bdrv_init();
2588
}
2589

    
2590
void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2591
                   BlockDriverCompletionFunc *cb, void *opaque)
2592
{
2593
    BlockDriverAIOCB *acb;
2594

    
2595
    if (pool->free_aiocb) {
2596
        acb = pool->free_aiocb;
2597
        pool->free_aiocb = acb->next;
2598
    } else {
2599
        acb = qemu_mallocz(pool->aiocb_size);
2600
        acb->pool = pool;
2601
    }
2602
    acb->bs = bs;
2603
    acb->cb = cb;
2604
    acb->opaque = opaque;
2605
    return acb;
2606
}
2607

    
2608
void qemu_aio_release(void *p)
2609
{
2610
    BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2611
    AIOPool *pool = acb->pool;
2612
    acb->next = pool->free_aiocb;
2613
    pool->free_aiocb = acb;
2614
}
2615

    
2616
/**************************************************************/
2617
/* removable device support */
2618

    
2619
/**
2620
 * Return TRUE if the media is present
2621
 */
2622
int bdrv_is_inserted(BlockDriverState *bs)
2623
{
2624
    BlockDriver *drv = bs->drv;
2625
    int ret;
2626
    if (!drv)
2627
        return 0;
2628
    if (!drv->bdrv_is_inserted)
2629
        return !bs->tray_open;
2630
    ret = drv->bdrv_is_inserted(bs);
2631
    return ret;
2632
}
2633

    
2634
/**
2635
 * Return TRUE if the media changed since the last call to this
2636
 * function. It is currently only used for floppy disks
2637
 */
2638
int bdrv_media_changed(BlockDriverState *bs)
2639
{
2640
    BlockDriver *drv = bs->drv;
2641
    int ret;
2642

    
2643
    if (!drv || !drv->bdrv_media_changed)
2644
        ret = -ENOTSUP;
2645
    else
2646
        ret = drv->bdrv_media_changed(bs);
2647
    if (ret == -ENOTSUP)
2648
        ret = bs->media_changed;
2649
    bs->media_changed = 0;
2650
    return ret;
2651
}
2652

    
2653
/**
2654
 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2655
 */
2656
int bdrv_eject(BlockDriverState *bs, int eject_flag)
2657
{
2658
    BlockDriver *drv = bs->drv;
2659
    int ret;
2660

    
2661
    if (bs->locked) {
2662
        return -EBUSY;
2663
    }
2664

    
2665
    if (!drv || !drv->bdrv_eject) {
2666
        ret = -ENOTSUP;
2667
    } else {
2668
        ret = drv->bdrv_eject(bs, eject_flag);
2669
    }
2670
    if (ret == -ENOTSUP) {
2671
        ret = 0;
2672
    }
2673
    if (ret >= 0) {
2674
        bs->tray_open = eject_flag;
2675
    }
2676

    
2677
    return ret;
2678
}
2679

    
2680
int bdrv_is_locked(BlockDriverState *bs)
2681
{
2682
    return bs->locked;
2683
}
2684

    
2685
/**
2686
 * Lock or unlock the media (if it is locked, the user won't be able
2687
 * to eject it manually).
2688
 */
2689
void bdrv_set_locked(BlockDriverState *bs, int locked)
2690
{
2691
    BlockDriver *drv = bs->drv;
2692

    
2693
    bs->locked = locked;
2694
    if (drv && drv->bdrv_set_locked) {
2695
        drv->bdrv_set_locked(bs, locked);
2696
    }
2697
}
2698

    
2699
/* needed for generic scsi interface */
2700

    
2701
int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2702
{
2703
    BlockDriver *drv = bs->drv;
2704

    
2705
    if (drv && drv->bdrv_ioctl)
2706
        return drv->bdrv_ioctl(bs, req, buf);
2707
    return -ENOTSUP;
2708
}
2709

    
2710
BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2711
        unsigned long int req, void *buf,
2712
        BlockDriverCompletionFunc *cb, void *opaque)
2713
{
2714
    BlockDriver *drv = bs->drv;
2715

    
2716
    if (drv && drv->bdrv_aio_ioctl)
2717
        return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2718
    return NULL;
2719
}
2720

    
2721

    
2722

    
2723
void *qemu_blockalign(BlockDriverState *bs, size_t size)
2724
{
2725
    return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2726
}
2727

    
2728
void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2729
{
2730
    int64_t bitmap_size;
2731

    
2732
    bs->dirty_count = 0;
2733
    if (enable) {
2734
        if (!bs->dirty_bitmap) {
2735
            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2736
                    BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2737
            bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2738

    
2739
            bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2740
        }
2741
    } else {
2742
        if (bs->dirty_bitmap) {
2743
            qemu_free(bs->dirty_bitmap);
2744
            bs->dirty_bitmap = NULL;
2745
        }
2746
    }
2747
}
2748

    
2749
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2750
{
2751
    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2752

    
2753
    if (bs->dirty_bitmap &&
2754
        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2755
        return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2756
            (1UL << (chunk % (sizeof(unsigned long) * 8))));
2757
    } else {
2758
        return 0;
2759
    }
2760
}
2761

    
2762
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2763
                      int nr_sectors)
2764
{
2765
    set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2766
}
2767

    
2768
int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2769
{
2770
    return bs->dirty_count;
2771
}
2772

    
2773
int bdrv_img_create(const char *filename, const char *fmt,
2774
                    const char *base_filename, const char *base_fmt,
2775
                    char *options, uint64_t img_size, int flags)
2776
{
2777
    QEMUOptionParameter *param = NULL, *create_options = NULL;
2778
    QEMUOptionParameter *backing_fmt, *backing_file;
2779
    BlockDriverState *bs = NULL;
2780
    BlockDriver *drv, *proto_drv;
2781
    int ret = 0;
2782

    
2783
    /* Find driver and parse its options */
2784
    drv = bdrv_find_format(fmt);
2785
    if (!drv) {
2786
        error_report("Unknown file format '%s'", fmt);
2787
        ret = -EINVAL;
2788
        goto out;
2789
    }
2790

    
2791
    proto_drv = bdrv_find_protocol(filename);
2792
    if (!proto_drv) {
2793
        error_report("Unknown protocol '%s'", filename);
2794
        ret = -EINVAL;
2795
        goto out;
2796
    }
2797

    
2798
    create_options = append_option_parameters(create_options,
2799
                                              drv->create_options);
2800
    create_options = append_option_parameters(create_options,
2801
                                              proto_drv->create_options);
2802

    
2803
    /* Create parameter list with default values */
2804
    param = parse_option_parameters("", create_options, param);
2805

    
2806
    set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
2807

    
2808
    /* Parse -o options */
2809
    if (options) {
2810
        param = parse_option_parameters(options, create_options, param);
2811
        if (param == NULL) {
2812
            error_report("Invalid options for file format '%s'.", fmt);
2813
            ret = -EINVAL;
2814
            goto out;
2815
        }
2816
    }
2817

    
2818
    if (base_filename) {
2819
        if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
2820
                                 base_filename)) {
2821
            error_report("Backing file not supported for file format '%s'",
2822
                         fmt);
2823
            ret = -EINVAL;
2824
            goto out;
2825
        }
2826
    }
2827

    
2828
    if (base_fmt) {
2829
        if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
2830
            error_report("Backing file format not supported for file "
2831
                         "format '%s'", fmt);
2832
            ret = -EINVAL;
2833
            goto out;
2834
        }
2835
    }
2836

    
2837
    backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
2838
    if (backing_file && backing_file->value.s) {
2839
        if (!strcmp(filename, backing_file->value.s)) {
2840
            error_report("Error: Trying to create an image with the "
2841
                         "same filename as the backing file");
2842
            ret = -EINVAL;
2843
            goto out;
2844
        }
2845
    }
2846

    
2847
    backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
2848
    if (backing_fmt && backing_fmt->value.s) {
2849
        if (!bdrv_find_format(backing_fmt->value.s)) {
2850
            error_report("Unknown backing file format '%s'",
2851
                         backing_fmt->value.s);
2852
            ret = -EINVAL;
2853
            goto out;
2854
        }
2855
    }
2856

    
2857
    // The size for the image must always be specified, with one exception:
2858
    // If we are using a backing file, we can obtain the size from there
2859
    if (get_option_parameter(param, BLOCK_OPT_SIZE)->value.n == -1) {
2860
        if (backing_file && backing_file->value.s) {
2861
            uint64_t size;
2862
            const char *fmt = NULL;
2863
            char buf[32];
2864

    
2865
            if (backing_fmt && backing_fmt->value.s) {
2866
                fmt = backing_fmt->value.s;
2867
            }
2868

    
2869
            bs = bdrv_new("");
2870

    
2871
            ret = bdrv_open(bs, backing_file->value.s, flags, drv);
2872
            if (ret < 0) {
2873
                error_report("Could not open '%s'", filename);
2874
                goto out;
2875
            }
2876
            bdrv_get_geometry(bs, &size);
2877
            size *= 512;
2878

    
2879
            snprintf(buf, sizeof(buf), "%" PRId64, size);
2880
            set_option_parameter(param, BLOCK_OPT_SIZE, buf);
2881
        } else {
2882
            error_report("Image creation needs a size parameter");
2883
            ret = -EINVAL;
2884
            goto out;
2885
        }
2886
    }
2887

    
2888
    printf("Formatting '%s', fmt=%s ", filename, fmt);
2889
    print_option_parameters(param);
2890
    puts("");
2891

    
2892
    ret = bdrv_create(drv, filename, param);
2893

    
2894
    if (ret < 0) {
2895
        if (ret == -ENOTSUP) {
2896
            error_report("Formatting or formatting option not supported for "
2897
                         "file format '%s'", fmt);
2898
        } else if (ret == -EFBIG) {
2899
            error_report("The image size is too large for file format '%s'",
2900
                         fmt);
2901
        } else {
2902
            error_report("%s: error while creating %s: %s", filename, fmt,
2903
                         strerror(-ret));
2904
        }
2905
    }
2906

    
2907
out:
2908
    free_option_parameters(create_options);
2909
    free_option_parameters(param);
2910

    
2911
    if (bs) {
2912
        bdrv_delete(bs);
2913
    }
2914

    
2915
    return ret;
2916
}