Statistics
| Branch: | Revision:

root / block.c @ a74cdab4

History | View | Annotate | Download (81 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 writable.
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, CHANGE_MEDIA);
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, CHANGE_MEDIA);
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
/* make a BlockDriverState anonymous by removing from bdrv_state list.
701
   Also, NULL terminate the device_name to prevent double remove */
702
void bdrv_make_anon(BlockDriverState *bs)
703
{
704
    if (bs->device_name[0] != '\0') {
705
        QTAILQ_REMOVE(&bdrv_states, bs, list);
706
    }
707
    bs->device_name[0] = '\0';
708
}
709

    
710
void bdrv_delete(BlockDriverState *bs)
711
{
712
    assert(!bs->peer);
713

    
714
    /* remove from list, if necessary */
715
    bdrv_make_anon(bs);
716

    
717
    bdrv_close(bs);
718
    if (bs->file != NULL) {
719
        bdrv_delete(bs->file);
720
    }
721

    
722
    assert(bs != bs_snapshots);
723
    qemu_free(bs);
724
}
725

    
726
int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
727
{
728
    if (bs->peer) {
729
        return -EBUSY;
730
    }
731
    bs->peer = qdev;
732
    return 0;
733
}
734

    
735
void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
736
{
737
    assert(bs->peer == qdev);
738
    bs->peer = NULL;
739
}
740

    
741
DeviceState *bdrv_get_attached(BlockDriverState *bs)
742
{
743
    return bs->peer;
744
}
745

    
746
/*
747
 * Run consistency checks on an image
748
 *
749
 * Returns 0 if the check could be completed (it doesn't mean that the image is
750
 * free of errors) or -errno when an internal error occurred. The results of the
751
 * check are stored in res.
752
 */
753
int bdrv_check(BlockDriverState *bs, BdrvCheckResult *res)
754
{
755
    if (bs->drv->bdrv_check == NULL) {
756
        return -ENOTSUP;
757
    }
758

    
759
    memset(res, 0, sizeof(*res));
760
    return bs->drv->bdrv_check(bs, res);
761
}
762

    
763
#define COMMIT_BUF_SECTORS 2048
764

    
765
/* commit COW file into the raw image */
766
int bdrv_commit(BlockDriverState *bs)
767
{
768
    BlockDriver *drv = bs->drv;
769
    BlockDriver *backing_drv;
770
    int64_t sector, total_sectors;
771
    int n, ro, open_flags;
772
    int ret = 0, rw_ret = 0;
773
    uint8_t *buf;
774
    char filename[1024];
775
    BlockDriverState *bs_rw, *bs_ro;
776

    
777
    if (!drv)
778
        return -ENOMEDIUM;
779
    
780
    if (!bs->backing_hd) {
781
        return -ENOTSUP;
782
    }
783

    
784
    if (bs->backing_hd->keep_read_only) {
785
        return -EACCES;
786
    }
787

    
788
    backing_drv = bs->backing_hd->drv;
789
    ro = bs->backing_hd->read_only;
790
    strncpy(filename, bs->backing_hd->filename, sizeof(filename));
791
    open_flags =  bs->backing_hd->open_flags;
792

    
793
    if (ro) {
794
        /* re-open as RW */
795
        bdrv_delete(bs->backing_hd);
796
        bs->backing_hd = NULL;
797
        bs_rw = bdrv_new("");
798
        rw_ret = bdrv_open(bs_rw, filename, open_flags | BDRV_O_RDWR,
799
            backing_drv);
800
        if (rw_ret < 0) {
801
            bdrv_delete(bs_rw);
802
            /* try to re-open read-only */
803
            bs_ro = bdrv_new("");
804
            ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
805
                backing_drv);
806
            if (ret < 0) {
807
                bdrv_delete(bs_ro);
808
                /* drive not functional anymore */
809
                bs->drv = NULL;
810
                return ret;
811
            }
812
            bs->backing_hd = bs_ro;
813
            return rw_ret;
814
        }
815
        bs->backing_hd = bs_rw;
816
    }
817

    
818
    total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
819
    buf = qemu_malloc(COMMIT_BUF_SECTORS * BDRV_SECTOR_SIZE);
820

    
821
    for (sector = 0; sector < total_sectors; sector += n) {
822
        if (drv->bdrv_is_allocated(bs, sector, COMMIT_BUF_SECTORS, &n)) {
823

    
824
            if (bdrv_read(bs, sector, buf, n) != 0) {
825
                ret = -EIO;
826
                goto ro_cleanup;
827
            }
828

    
829
            if (bdrv_write(bs->backing_hd, sector, buf, n) != 0) {
830
                ret = -EIO;
831
                goto ro_cleanup;
832
            }
833
        }
834
    }
835

    
836
    if (drv->bdrv_make_empty) {
837
        ret = drv->bdrv_make_empty(bs);
838
        bdrv_flush(bs);
839
    }
840

    
841
    /*
842
     * Make sure all data we wrote to the backing device is actually
843
     * stable on disk.
844
     */
845
    if (bs->backing_hd)
846
        bdrv_flush(bs->backing_hd);
847

    
848
ro_cleanup:
849
    qemu_free(buf);
850

    
851
    if (ro) {
852
        /* re-open as RO */
853
        bdrv_delete(bs->backing_hd);
854
        bs->backing_hd = NULL;
855
        bs_ro = bdrv_new("");
856
        ret = bdrv_open(bs_ro, filename, open_flags & ~BDRV_O_RDWR,
857
            backing_drv);
858
        if (ret < 0) {
859
            bdrv_delete(bs_ro);
860
            /* drive not functional anymore */
861
            bs->drv = NULL;
862
            return ret;
863
        }
864
        bs->backing_hd = bs_ro;
865
        bs->backing_hd->keep_read_only = 0;
866
    }
867

    
868
    return ret;
869
}
870

    
871
void bdrv_commit_all(void)
872
{
873
    BlockDriverState *bs;
874

    
875
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
876
        bdrv_commit(bs);
877
    }
878
}
879

    
880
/*
881
 * Return values:
882
 * 0        - success
883
 * -EINVAL  - backing format specified, but no file
884
 * -ENOSPC  - can't update the backing file because no space is left in the
885
 *            image file header
886
 * -ENOTSUP - format driver doesn't support changing the backing file
887
 */
888
int bdrv_change_backing_file(BlockDriverState *bs,
889
    const char *backing_file, const char *backing_fmt)
890
{
891
    BlockDriver *drv = bs->drv;
892

    
893
    if (drv->bdrv_change_backing_file != NULL) {
894
        return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
895
    } else {
896
        return -ENOTSUP;
897
    }
898
}
899

    
900
static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
901
                                   size_t size)
902
{
903
    int64_t len;
904

    
905
    if (!bdrv_is_inserted(bs))
906
        return -ENOMEDIUM;
907

    
908
    if (bs->growable)
909
        return 0;
910

    
911
    len = bdrv_getlength(bs);
912

    
913
    if (offset < 0)
914
        return -EIO;
915

    
916
    if ((offset > len) || (len - offset < size))
917
        return -EIO;
918

    
919
    return 0;
920
}
921

    
922
static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
923
                              int nb_sectors)
924
{
925
    return bdrv_check_byte_request(bs, sector_num * BDRV_SECTOR_SIZE,
926
                                   nb_sectors * BDRV_SECTOR_SIZE);
927
}
928

    
929
/* return < 0 if error. See bdrv_write() for the return codes */
930
int bdrv_read(BlockDriverState *bs, int64_t sector_num,
931
              uint8_t *buf, int nb_sectors)
932
{
933
    BlockDriver *drv = bs->drv;
934

    
935
    if (!drv)
936
        return -ENOMEDIUM;
937
    if (bdrv_check_request(bs, sector_num, nb_sectors))
938
        return -EIO;
939

    
940
    return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
941
}
942

    
943
static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
944
                             int nb_sectors, int dirty)
945
{
946
    int64_t start, end;
947
    unsigned long val, idx, bit;
948

    
949
    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
950
    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
951

    
952
    for (; start <= end; start++) {
953
        idx = start / (sizeof(unsigned long) * 8);
954
        bit = start % (sizeof(unsigned long) * 8);
955
        val = bs->dirty_bitmap[idx];
956
        if (dirty) {
957
            if (!(val & (1UL << bit))) {
958
                bs->dirty_count++;
959
                val |= 1UL << bit;
960
            }
961
        } else {
962
            if (val & (1UL << bit)) {
963
                bs->dirty_count--;
964
                val &= ~(1UL << bit);
965
            }
966
        }
967
        bs->dirty_bitmap[idx] = val;
968
    }
969
}
970

    
971
/* Return < 0 if error. Important errors are:
972
  -EIO         generic I/O error (may happen for all errors)
973
  -ENOMEDIUM   No media inserted.
974
  -EINVAL      Invalid sector number or nb_sectors
975
  -EACCES      Trying to write a read-only device
976
*/
977
int bdrv_write(BlockDriverState *bs, int64_t sector_num,
978
               const uint8_t *buf, int nb_sectors)
979
{
980
    BlockDriver *drv = bs->drv;
981
    if (!bs->drv)
982
        return -ENOMEDIUM;
983
    if (bs->read_only)
984
        return -EACCES;
985
    if (bdrv_check_request(bs, sector_num, nb_sectors))
986
        return -EIO;
987

    
988
    if (bs->dirty_bitmap) {
989
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
990
    }
991

    
992
    if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
993
        bs->wr_highest_sector = sector_num + nb_sectors - 1;
994
    }
995

    
996
    return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
997
}
998

    
999
int bdrv_pread(BlockDriverState *bs, int64_t offset,
1000
               void *buf, int count1)
1001
{
1002
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1003
    int len, nb_sectors, count;
1004
    int64_t sector_num;
1005
    int ret;
1006

    
1007
    count = count1;
1008
    /* first read to align to sector start */
1009
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1010
    if (len > count)
1011
        len = count;
1012
    sector_num = offset >> BDRV_SECTOR_BITS;
1013
    if (len > 0) {
1014
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1015
            return ret;
1016
        memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
1017
        count -= len;
1018
        if (count == 0)
1019
            return count1;
1020
        sector_num++;
1021
        buf += len;
1022
    }
1023

    
1024
    /* read the sectors "in place" */
1025
    nb_sectors = count >> BDRV_SECTOR_BITS;
1026
    if (nb_sectors > 0) {
1027
        if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
1028
            return ret;
1029
        sector_num += nb_sectors;
1030
        len = nb_sectors << BDRV_SECTOR_BITS;
1031
        buf += len;
1032
        count -= len;
1033
    }
1034

    
1035
    /* add data from the last sector */
1036
    if (count > 0) {
1037
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1038
            return ret;
1039
        memcpy(buf, tmp_buf, count);
1040
    }
1041
    return count1;
1042
}
1043

    
1044
int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
1045
                const void *buf, int count1)
1046
{
1047
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
1048
    int len, nb_sectors, count;
1049
    int64_t sector_num;
1050
    int ret;
1051

    
1052
    count = count1;
1053
    /* first write to align to sector start */
1054
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
1055
    if (len > count)
1056
        len = count;
1057
    sector_num = offset >> BDRV_SECTOR_BITS;
1058
    if (len > 0) {
1059
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1060
            return ret;
1061
        memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
1062
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1063
            return ret;
1064
        count -= len;
1065
        if (count == 0)
1066
            return count1;
1067
        sector_num++;
1068
        buf += len;
1069
    }
1070

    
1071
    /* write the sectors "in place" */
1072
    nb_sectors = count >> BDRV_SECTOR_BITS;
1073
    if (nb_sectors > 0) {
1074
        if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
1075
            return ret;
1076
        sector_num += nb_sectors;
1077
        len = nb_sectors << BDRV_SECTOR_BITS;
1078
        buf += len;
1079
        count -= len;
1080
    }
1081

    
1082
    /* add data from the last sector */
1083
    if (count > 0) {
1084
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
1085
            return ret;
1086
        memcpy(tmp_buf, buf, count);
1087
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
1088
            return ret;
1089
    }
1090
    return count1;
1091
}
1092

    
1093
/*
1094
 * Writes to the file and ensures that no writes are reordered across this
1095
 * request (acts as a barrier)
1096
 *
1097
 * Returns 0 on success, -errno in error cases.
1098
 */
1099
int bdrv_pwrite_sync(BlockDriverState *bs, int64_t offset,
1100
    const void *buf, int count)
1101
{
1102
    int ret;
1103

    
1104
    ret = bdrv_pwrite(bs, offset, buf, count);
1105
    if (ret < 0) {
1106
        return ret;
1107
    }
1108

    
1109
    /* No flush needed for cache=writethrough, it uses O_DSYNC */
1110
    if ((bs->open_flags & BDRV_O_CACHE_MASK) != 0) {
1111
        bdrv_flush(bs);
1112
    }
1113

    
1114
    return 0;
1115
}
1116

    
1117
/*
1118
 * Writes to the file and ensures that no writes are reordered across this
1119
 * request (acts as a barrier)
1120
 *
1121
 * Returns 0 on success, -errno in error cases.
1122
 */
1123
int bdrv_write_sync(BlockDriverState *bs, int64_t sector_num,
1124
    const uint8_t *buf, int nb_sectors)
1125
{
1126
    return bdrv_pwrite_sync(bs, BDRV_SECTOR_SIZE * sector_num,
1127
        buf, BDRV_SECTOR_SIZE * nb_sectors);
1128
}
1129

    
1130
/**
1131
 * Truncate file to 'offset' bytes (needed only for file protocols)
1132
 */
1133
int bdrv_truncate(BlockDriverState *bs, int64_t offset)
1134
{
1135
    BlockDriver *drv = bs->drv;
1136
    int ret;
1137
    if (!drv)
1138
        return -ENOMEDIUM;
1139
    if (!drv->bdrv_truncate)
1140
        return -ENOTSUP;
1141
    if (bs->read_only)
1142
        return -EACCES;
1143
    if (bdrv_in_use(bs))
1144
        return -EBUSY;
1145
    ret = drv->bdrv_truncate(bs, offset);
1146
    if (ret == 0) {
1147
        ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
1148
        if (bs->change_cb) {
1149
            bs->change_cb(bs->change_opaque, CHANGE_SIZE);
1150
        }
1151
    }
1152
    return ret;
1153
}
1154

    
1155
/**
1156
 * Length of a file in bytes. Return < 0 if error or unknown.
1157
 */
1158
int64_t bdrv_getlength(BlockDriverState *bs)
1159
{
1160
    BlockDriver *drv = bs->drv;
1161
    if (!drv)
1162
        return -ENOMEDIUM;
1163

    
1164
    if (bs->growable || bs->removable) {
1165
        if (drv->bdrv_getlength) {
1166
            return drv->bdrv_getlength(bs);
1167
        }
1168
    }
1169
    return bs->total_sectors * BDRV_SECTOR_SIZE;
1170
}
1171

    
1172
/* return 0 as number of sectors if no device present or error */
1173
void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
1174
{
1175
    int64_t length;
1176
    length = bdrv_getlength(bs);
1177
    if (length < 0)
1178
        length = 0;
1179
    else
1180
        length = length >> BDRV_SECTOR_BITS;
1181
    *nb_sectors_ptr = length;
1182
}
1183

    
1184
struct partition {
1185
        uint8_t boot_ind;           /* 0x80 - active */
1186
        uint8_t head;               /* starting head */
1187
        uint8_t sector;             /* starting sector */
1188
        uint8_t cyl;                /* starting cylinder */
1189
        uint8_t sys_ind;            /* What partition type */
1190
        uint8_t end_head;           /* end head */
1191
        uint8_t end_sector;         /* end sector */
1192
        uint8_t end_cyl;            /* end cylinder */
1193
        uint32_t start_sect;        /* starting sector counting from 0 */
1194
        uint32_t nr_sects;          /* nr of sectors in partition */
1195
} __attribute__((packed));
1196

    
1197
/* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1198
static int guess_disk_lchs(BlockDriverState *bs,
1199
                           int *pcylinders, int *pheads, int *psectors)
1200
{
1201
    uint8_t buf[BDRV_SECTOR_SIZE];
1202
    int ret, i, heads, sectors, cylinders;
1203
    struct partition *p;
1204
    uint32_t nr_sects;
1205
    uint64_t nb_sectors;
1206

    
1207
    bdrv_get_geometry(bs, &nb_sectors);
1208

    
1209
    ret = bdrv_read(bs, 0, buf, 1);
1210
    if (ret < 0)
1211
        return -1;
1212
    /* test msdos magic */
1213
    if (buf[510] != 0x55 || buf[511] != 0xaa)
1214
        return -1;
1215
    for(i = 0; i < 4; i++) {
1216
        p = ((struct partition *)(buf + 0x1be)) + i;
1217
        nr_sects = le32_to_cpu(p->nr_sects);
1218
        if (nr_sects && p->end_head) {
1219
            /* We make the assumption that the partition terminates on
1220
               a cylinder boundary */
1221
            heads = p->end_head + 1;
1222
            sectors = p->end_sector & 63;
1223
            if (sectors == 0)
1224
                continue;
1225
            cylinders = nb_sectors / (heads * sectors);
1226
            if (cylinders < 1 || cylinders > 16383)
1227
                continue;
1228
            *pheads = heads;
1229
            *psectors = sectors;
1230
            *pcylinders = cylinders;
1231
#if 0
1232
            printf("guessed geometry: LCHS=%d %d %d\n",
1233
                   cylinders, heads, sectors);
1234
#endif
1235
            return 0;
1236
        }
1237
    }
1238
    return -1;
1239
}
1240

    
1241
void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
1242
{
1243
    int translation, lba_detected = 0;
1244
    int cylinders, heads, secs;
1245
    uint64_t nb_sectors;
1246

    
1247
    /* if a geometry hint is available, use it */
1248
    bdrv_get_geometry(bs, &nb_sectors);
1249
    bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
1250
    translation = bdrv_get_translation_hint(bs);
1251
    if (cylinders != 0) {
1252
        *pcyls = cylinders;
1253
        *pheads = heads;
1254
        *psecs = secs;
1255
    } else {
1256
        if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
1257
            if (heads > 16) {
1258
                /* if heads > 16, it means that a BIOS LBA
1259
                   translation was active, so the default
1260
                   hardware geometry is OK */
1261
                lba_detected = 1;
1262
                goto default_geometry;
1263
            } else {
1264
                *pcyls = cylinders;
1265
                *pheads = heads;
1266
                *psecs = secs;
1267
                /* disable any translation to be in sync with
1268
                   the logical geometry */
1269
                if (translation == BIOS_ATA_TRANSLATION_AUTO) {
1270
                    bdrv_set_translation_hint(bs,
1271
                                              BIOS_ATA_TRANSLATION_NONE);
1272
                }
1273
            }
1274
        } else {
1275
        default_geometry:
1276
            /* if no geometry, use a standard physical disk geometry */
1277
            cylinders = nb_sectors / (16 * 63);
1278

    
1279
            if (cylinders > 16383)
1280
                cylinders = 16383;
1281
            else if (cylinders < 2)
1282
                cylinders = 2;
1283
            *pcyls = cylinders;
1284
            *pheads = 16;
1285
            *psecs = 63;
1286
            if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
1287
                if ((*pcyls * *pheads) <= 131072) {
1288
                    bdrv_set_translation_hint(bs,
1289
                                              BIOS_ATA_TRANSLATION_LARGE);
1290
                } else {
1291
                    bdrv_set_translation_hint(bs,
1292
                                              BIOS_ATA_TRANSLATION_LBA);
1293
                }
1294
            }
1295
        }
1296
        bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
1297
    }
1298
}
1299

    
1300
void bdrv_set_geometry_hint(BlockDriverState *bs,
1301
                            int cyls, int heads, int secs)
1302
{
1303
    bs->cyls = cyls;
1304
    bs->heads = heads;
1305
    bs->secs = secs;
1306
}
1307

    
1308
void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1309
{
1310
    bs->translation = translation;
1311
}
1312

    
1313
void bdrv_get_geometry_hint(BlockDriverState *bs,
1314
                            int *pcyls, int *pheads, int *psecs)
1315
{
1316
    *pcyls = bs->cyls;
1317
    *pheads = bs->heads;
1318
    *psecs = bs->secs;
1319
}
1320

    
1321
/* Recognize floppy formats */
1322
typedef struct FDFormat {
1323
    FDriveType drive;
1324
    uint8_t last_sect;
1325
    uint8_t max_track;
1326
    uint8_t max_head;
1327
} FDFormat;
1328

    
1329
static const FDFormat fd_formats[] = {
1330
    /* First entry is default format */
1331
    /* 1.44 MB 3"1/2 floppy disks */
1332
    { FDRIVE_DRV_144, 18, 80, 1, },
1333
    { FDRIVE_DRV_144, 20, 80, 1, },
1334
    { FDRIVE_DRV_144, 21, 80, 1, },
1335
    { FDRIVE_DRV_144, 21, 82, 1, },
1336
    { FDRIVE_DRV_144, 21, 83, 1, },
1337
    { FDRIVE_DRV_144, 22, 80, 1, },
1338
    { FDRIVE_DRV_144, 23, 80, 1, },
1339
    { FDRIVE_DRV_144, 24, 80, 1, },
1340
    /* 2.88 MB 3"1/2 floppy disks */
1341
    { FDRIVE_DRV_288, 36, 80, 1, },
1342
    { FDRIVE_DRV_288, 39, 80, 1, },
1343
    { FDRIVE_DRV_288, 40, 80, 1, },
1344
    { FDRIVE_DRV_288, 44, 80, 1, },
1345
    { FDRIVE_DRV_288, 48, 80, 1, },
1346
    /* 720 kB 3"1/2 floppy disks */
1347
    { FDRIVE_DRV_144,  9, 80, 1, },
1348
    { FDRIVE_DRV_144, 10, 80, 1, },
1349
    { FDRIVE_DRV_144, 10, 82, 1, },
1350
    { FDRIVE_DRV_144, 10, 83, 1, },
1351
    { FDRIVE_DRV_144, 13, 80, 1, },
1352
    { FDRIVE_DRV_144, 14, 80, 1, },
1353
    /* 1.2 MB 5"1/4 floppy disks */
1354
    { FDRIVE_DRV_120, 15, 80, 1, },
1355
    { FDRIVE_DRV_120, 18, 80, 1, },
1356
    { FDRIVE_DRV_120, 18, 82, 1, },
1357
    { FDRIVE_DRV_120, 18, 83, 1, },
1358
    { FDRIVE_DRV_120, 20, 80, 1, },
1359
    /* 720 kB 5"1/4 floppy disks */
1360
    { FDRIVE_DRV_120,  9, 80, 1, },
1361
    { FDRIVE_DRV_120, 11, 80, 1, },
1362
    /* 360 kB 5"1/4 floppy disks */
1363
    { FDRIVE_DRV_120,  9, 40, 1, },
1364
    { FDRIVE_DRV_120,  9, 40, 0, },
1365
    { FDRIVE_DRV_120, 10, 41, 1, },
1366
    { FDRIVE_DRV_120, 10, 42, 1, },
1367
    /* 320 kB 5"1/4 floppy disks */
1368
    { FDRIVE_DRV_120,  8, 40, 1, },
1369
    { FDRIVE_DRV_120,  8, 40, 0, },
1370
    /* 360 kB must match 5"1/4 better than 3"1/2... */
1371
    { FDRIVE_DRV_144,  9, 80, 0, },
1372
    /* end */
1373
    { FDRIVE_DRV_NONE, -1, -1, 0, },
1374
};
1375

    
1376
void bdrv_get_floppy_geometry_hint(BlockDriverState *bs, int *nb_heads,
1377
                                   int *max_track, int *last_sect,
1378
                                   FDriveType drive_in, FDriveType *drive)
1379
{
1380
    const FDFormat *parse;
1381
    uint64_t nb_sectors, size;
1382
    int i, first_match, match;
1383

    
1384
    bdrv_get_geometry_hint(bs, nb_heads, max_track, last_sect);
1385
    if (*nb_heads != 0 && *max_track != 0 && *last_sect != 0) {
1386
        /* User defined disk */
1387
    } else {
1388
        bdrv_get_geometry(bs, &nb_sectors);
1389
        match = -1;
1390
        first_match = -1;
1391
        for (i = 0; ; i++) {
1392
            parse = &fd_formats[i];
1393
            if (parse->drive == FDRIVE_DRV_NONE) {
1394
                break;
1395
            }
1396
            if (drive_in == parse->drive ||
1397
                drive_in == FDRIVE_DRV_NONE) {
1398
                size = (parse->max_head + 1) * parse->max_track *
1399
                    parse->last_sect;
1400
                if (nb_sectors == size) {
1401
                    match = i;
1402
                    break;
1403
                }
1404
                if (first_match == -1) {
1405
                    first_match = i;
1406
                }
1407
            }
1408
        }
1409
        if (match == -1) {
1410
            if (first_match == -1) {
1411
                match = 1;
1412
            } else {
1413
                match = first_match;
1414
            }
1415
            parse = &fd_formats[match];
1416
        }
1417
        *nb_heads = parse->max_head + 1;
1418
        *max_track = parse->max_track;
1419
        *last_sect = parse->last_sect;
1420
        *drive = parse->drive;
1421
    }
1422
}
1423

    
1424
int bdrv_get_translation_hint(BlockDriverState *bs)
1425
{
1426
    return bs->translation;
1427
}
1428

    
1429
void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
1430
                       BlockErrorAction on_write_error)
1431
{
1432
    bs->on_read_error = on_read_error;
1433
    bs->on_write_error = on_write_error;
1434
}
1435

    
1436
BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
1437
{
1438
    return is_read ? bs->on_read_error : bs->on_write_error;
1439
}
1440

    
1441
void bdrv_set_removable(BlockDriverState *bs, int removable)
1442
{
1443
    bs->removable = removable;
1444
    if (removable && bs == bs_snapshots) {
1445
        bs_snapshots = NULL;
1446
    }
1447
}
1448

    
1449
int bdrv_is_removable(BlockDriverState *bs)
1450
{
1451
    return bs->removable;
1452
}
1453

    
1454
int bdrv_is_read_only(BlockDriverState *bs)
1455
{
1456
    return bs->read_only;
1457
}
1458

    
1459
int bdrv_is_sg(BlockDriverState *bs)
1460
{
1461
    return bs->sg;
1462
}
1463

    
1464
int bdrv_enable_write_cache(BlockDriverState *bs)
1465
{
1466
    return bs->enable_write_cache;
1467
}
1468

    
1469
/* XXX: no longer used */
1470
void bdrv_set_change_cb(BlockDriverState *bs,
1471
                        void (*change_cb)(void *opaque, int reason),
1472
                        void *opaque)
1473
{
1474
    bs->change_cb = change_cb;
1475
    bs->change_opaque = opaque;
1476
}
1477

    
1478
int bdrv_is_encrypted(BlockDriverState *bs)
1479
{
1480
    if (bs->backing_hd && bs->backing_hd->encrypted)
1481
        return 1;
1482
    return bs->encrypted;
1483
}
1484

    
1485
int bdrv_key_required(BlockDriverState *bs)
1486
{
1487
    BlockDriverState *backing_hd = bs->backing_hd;
1488

    
1489
    if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1490
        return 1;
1491
    return (bs->encrypted && !bs->valid_key);
1492
}
1493

    
1494
int bdrv_set_key(BlockDriverState *bs, const char *key)
1495
{
1496
    int ret;
1497
    if (bs->backing_hd && bs->backing_hd->encrypted) {
1498
        ret = bdrv_set_key(bs->backing_hd, key);
1499
        if (ret < 0)
1500
            return ret;
1501
        if (!bs->encrypted)
1502
            return 0;
1503
    }
1504
    if (!bs->encrypted) {
1505
        return -EINVAL;
1506
    } else if (!bs->drv || !bs->drv->bdrv_set_key) {
1507
        return -ENOMEDIUM;
1508
    }
1509
    ret = bs->drv->bdrv_set_key(bs, key);
1510
    if (ret < 0) {
1511
        bs->valid_key = 0;
1512
    } else if (!bs->valid_key) {
1513
        bs->valid_key = 1;
1514
        /* call the change callback now, we skipped it on open */
1515
        bs->media_changed = 1;
1516
        if (bs->change_cb)
1517
            bs->change_cb(bs->change_opaque, CHANGE_MEDIA);
1518
    }
1519
    return ret;
1520
}
1521

    
1522
void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1523
{
1524
    if (!bs->drv) {
1525
        buf[0] = '\0';
1526
    } else {
1527
        pstrcpy(buf, buf_size, bs->drv->format_name);
1528
    }
1529
}
1530

    
1531
void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1532
                         void *opaque)
1533
{
1534
    BlockDriver *drv;
1535

    
1536
    QLIST_FOREACH(drv, &bdrv_drivers, list) {
1537
        it(opaque, drv->format_name);
1538
    }
1539
}
1540

    
1541
BlockDriverState *bdrv_find(const char *name)
1542
{
1543
    BlockDriverState *bs;
1544

    
1545
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1546
        if (!strcmp(name, bs->device_name)) {
1547
            return bs;
1548
        }
1549
    }
1550
    return NULL;
1551
}
1552

    
1553
BlockDriverState *bdrv_next(BlockDriverState *bs)
1554
{
1555
    if (!bs) {
1556
        return QTAILQ_FIRST(&bdrv_states);
1557
    }
1558
    return QTAILQ_NEXT(bs, list);
1559
}
1560

    
1561
void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1562
{
1563
    BlockDriverState *bs;
1564

    
1565
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1566
        it(opaque, bs);
1567
    }
1568
}
1569

    
1570
const char *bdrv_get_device_name(BlockDriverState *bs)
1571
{
1572
    return bs->device_name;
1573
}
1574

    
1575
int bdrv_flush(BlockDriverState *bs)
1576
{
1577
    if (bs->open_flags & BDRV_O_NO_FLUSH) {
1578
        return 0;
1579
    }
1580

    
1581
    if (bs->drv && bs->drv->bdrv_flush) {
1582
        return bs->drv->bdrv_flush(bs);
1583
    }
1584

    
1585
    /*
1586
     * Some block drivers always operate in either writethrough or unsafe mode
1587
     * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1588
     * the server works (because the behaviour is hardcoded or depends on
1589
     * server-side configuration), so we can't ensure that everything is safe
1590
     * on disk. Returning an error doesn't work because that would break guests
1591
     * even if the server operates in writethrough mode.
1592
     *
1593
     * Let's hope the user knows what he's doing.
1594
     */
1595
    return 0;
1596
}
1597

    
1598
void bdrv_flush_all(void)
1599
{
1600
    BlockDriverState *bs;
1601

    
1602
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1603
        if (bs->drv && !bdrv_is_read_only(bs) &&
1604
            (!bdrv_is_removable(bs) || bdrv_is_inserted(bs))) {
1605
            bdrv_flush(bs);
1606
        }
1607
    }
1608
}
1609

    
1610
int bdrv_has_zero_init(BlockDriverState *bs)
1611
{
1612
    assert(bs->drv);
1613

    
1614
    if (bs->drv->bdrv_has_zero_init) {
1615
        return bs->drv->bdrv_has_zero_init(bs);
1616
    }
1617

    
1618
    return 1;
1619
}
1620

    
1621
int bdrv_discard(BlockDriverState *bs, int64_t sector_num, int nb_sectors)
1622
{
1623
    if (!bs->drv) {
1624
        return -ENOMEDIUM;
1625
    }
1626
    if (!bs->drv->bdrv_discard) {
1627
        return 0;
1628
    }
1629
    return bs->drv->bdrv_discard(bs, sector_num, nb_sectors);
1630
}
1631

    
1632
/*
1633
 * Returns true iff the specified sector is present in the disk image. Drivers
1634
 * not implementing the functionality are assumed to not support backing files,
1635
 * hence all their sectors are reported as allocated.
1636
 *
1637
 * 'pnum' is set to the number of sectors (including and immediately following
1638
 * the specified sector) that are known to be in the same
1639
 * allocated/unallocated state.
1640
 *
1641
 * 'nb_sectors' is the max value 'pnum' should be set to.
1642
 */
1643
int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1644
        int *pnum)
1645
{
1646
    int64_t n;
1647
    if (!bs->drv->bdrv_is_allocated) {
1648
        if (sector_num >= bs->total_sectors) {
1649
            *pnum = 0;
1650
            return 0;
1651
        }
1652
        n = bs->total_sectors - sector_num;
1653
        *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1654
        return 1;
1655
    }
1656
    return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1657
}
1658

    
1659
void bdrv_mon_event(const BlockDriverState *bdrv,
1660
                    BlockMonEventAction action, int is_read)
1661
{
1662
    QObject *data;
1663
    const char *action_str;
1664

    
1665
    switch (action) {
1666
    case BDRV_ACTION_REPORT:
1667
        action_str = "report";
1668
        break;
1669
    case BDRV_ACTION_IGNORE:
1670
        action_str = "ignore";
1671
        break;
1672
    case BDRV_ACTION_STOP:
1673
        action_str = "stop";
1674
        break;
1675
    default:
1676
        abort();
1677
    }
1678

    
1679
    data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1680
                              bdrv->device_name,
1681
                              action_str,
1682
                              is_read ? "read" : "write");
1683
    monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1684

    
1685
    qobject_decref(data);
1686
}
1687

    
1688
static void bdrv_print_dict(QObject *obj, void *opaque)
1689
{
1690
    QDict *bs_dict;
1691
    Monitor *mon = opaque;
1692

    
1693
    bs_dict = qobject_to_qdict(obj);
1694

    
1695
    monitor_printf(mon, "%s: removable=%d",
1696
                        qdict_get_str(bs_dict, "device"),
1697
                        qdict_get_bool(bs_dict, "removable"));
1698

    
1699
    if (qdict_get_bool(bs_dict, "removable")) {
1700
        monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1701
    }
1702

    
1703
    if (qdict_haskey(bs_dict, "inserted")) {
1704
        QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1705

    
1706
        monitor_printf(mon, " file=");
1707
        monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1708
        if (qdict_haskey(qdict, "backing_file")) {
1709
            monitor_printf(mon, " backing_file=");
1710
            monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1711
        }
1712
        monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1713
                            qdict_get_bool(qdict, "ro"),
1714
                            qdict_get_str(qdict, "drv"),
1715
                            qdict_get_bool(qdict, "encrypted"));
1716
    } else {
1717
        monitor_printf(mon, " [not inserted]");
1718
    }
1719

    
1720
    monitor_printf(mon, "\n");
1721
}
1722

    
1723
void bdrv_info_print(Monitor *mon, const QObject *data)
1724
{
1725
    qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1726
}
1727

    
1728
void bdrv_info(Monitor *mon, QObject **ret_data)
1729
{
1730
    QList *bs_list;
1731
    BlockDriverState *bs;
1732

    
1733
    bs_list = qlist_new();
1734

    
1735
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1736
        QObject *bs_obj;
1737

    
1738
        bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
1739
                                    "'removable': %i, 'locked': %i }",
1740
                                    bs->device_name, bs->removable,
1741
                                    bs->locked);
1742

    
1743
        if (bs->drv) {
1744
            QObject *obj;
1745
            QDict *bs_dict = qobject_to_qdict(bs_obj);
1746

    
1747
            obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1748
                                     "'encrypted': %i }",
1749
                                     bs->filename, bs->read_only,
1750
                                     bs->drv->format_name,
1751
                                     bdrv_is_encrypted(bs));
1752
            if (bs->backing_file[0] != '\0') {
1753
                QDict *qdict = qobject_to_qdict(obj);
1754
                qdict_put(qdict, "backing_file",
1755
                          qstring_from_str(bs->backing_file));
1756
            }
1757

    
1758
            qdict_put_obj(bs_dict, "inserted", obj);
1759
        }
1760
        qlist_append_obj(bs_list, bs_obj);
1761
    }
1762

    
1763
    *ret_data = QOBJECT(bs_list);
1764
}
1765

    
1766
static void bdrv_stats_iter(QObject *data, void *opaque)
1767
{
1768
    QDict *qdict;
1769
    Monitor *mon = opaque;
1770

    
1771
    qdict = qobject_to_qdict(data);
1772
    monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1773

    
1774
    qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1775
    monitor_printf(mon, " rd_bytes=%" PRId64
1776
                        " wr_bytes=%" PRId64
1777
                        " rd_operations=%" PRId64
1778
                        " wr_operations=%" PRId64
1779
                        "\n",
1780
                        qdict_get_int(qdict, "rd_bytes"),
1781
                        qdict_get_int(qdict, "wr_bytes"),
1782
                        qdict_get_int(qdict, "rd_operations"),
1783
                        qdict_get_int(qdict, "wr_operations"));
1784
}
1785

    
1786
void bdrv_stats_print(Monitor *mon, const QObject *data)
1787
{
1788
    qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1789
}
1790

    
1791
static QObject* bdrv_info_stats_bs(BlockDriverState *bs)
1792
{
1793
    QObject *res;
1794
    QDict *dict;
1795

    
1796
    res = qobject_from_jsonf("{ 'stats': {"
1797
                             "'rd_bytes': %" PRId64 ","
1798
                             "'wr_bytes': %" PRId64 ","
1799
                             "'rd_operations': %" PRId64 ","
1800
                             "'wr_operations': %" PRId64 ","
1801
                             "'wr_highest_offset': %" PRId64
1802
                             "} }",
1803
                             bs->rd_bytes, bs->wr_bytes,
1804
                             bs->rd_ops, bs->wr_ops,
1805
                             bs->wr_highest_sector *
1806
                             (uint64_t)BDRV_SECTOR_SIZE);
1807
    dict  = qobject_to_qdict(res);
1808

    
1809
    if (*bs->device_name) {
1810
        qdict_put(dict, "device", qstring_from_str(bs->device_name));
1811
    }
1812

    
1813
    if (bs->file) {
1814
        QObject *parent = bdrv_info_stats_bs(bs->file);
1815
        qdict_put_obj(dict, "parent", parent);
1816
    }
1817

    
1818
    return res;
1819
}
1820

    
1821
void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1822
{
1823
    QObject *obj;
1824
    QList *devices;
1825
    BlockDriverState *bs;
1826

    
1827
    devices = qlist_new();
1828

    
1829
    QTAILQ_FOREACH(bs, &bdrv_states, list) {
1830
        obj = bdrv_info_stats_bs(bs);
1831
        qlist_append_obj(devices, obj);
1832
    }
1833

    
1834
    *ret_data = QOBJECT(devices);
1835
}
1836

    
1837
const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1838
{
1839
    if (bs->backing_hd && bs->backing_hd->encrypted)
1840
        return bs->backing_file;
1841
    else if (bs->encrypted)
1842
        return bs->filename;
1843
    else
1844
        return NULL;
1845
}
1846

    
1847
void bdrv_get_backing_filename(BlockDriverState *bs,
1848
                               char *filename, int filename_size)
1849
{
1850
    if (!bs->backing_file) {
1851
        pstrcpy(filename, filename_size, "");
1852
    } else {
1853
        pstrcpy(filename, filename_size, bs->backing_file);
1854
    }
1855
}
1856

    
1857
int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1858
                          const uint8_t *buf, int nb_sectors)
1859
{
1860
    BlockDriver *drv = bs->drv;
1861
    if (!drv)
1862
        return -ENOMEDIUM;
1863
    if (!drv->bdrv_write_compressed)
1864
        return -ENOTSUP;
1865
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1866
        return -EIO;
1867

    
1868
    if (bs->dirty_bitmap) {
1869
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1870
    }
1871

    
1872
    return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1873
}
1874

    
1875
int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1876
{
1877
    BlockDriver *drv = bs->drv;
1878
    if (!drv)
1879
        return -ENOMEDIUM;
1880
    if (!drv->bdrv_get_info)
1881
        return -ENOTSUP;
1882
    memset(bdi, 0, sizeof(*bdi));
1883
    return drv->bdrv_get_info(bs, bdi);
1884
}
1885

    
1886
int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1887
                      int64_t pos, int size)
1888
{
1889
    BlockDriver *drv = bs->drv;
1890
    if (!drv)
1891
        return -ENOMEDIUM;
1892
    if (drv->bdrv_save_vmstate)
1893
        return drv->bdrv_save_vmstate(bs, buf, pos, size);
1894
    if (bs->file)
1895
        return bdrv_save_vmstate(bs->file, buf, pos, size);
1896
    return -ENOTSUP;
1897
}
1898

    
1899
int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1900
                      int64_t pos, int size)
1901
{
1902
    BlockDriver *drv = bs->drv;
1903
    if (!drv)
1904
        return -ENOMEDIUM;
1905
    if (drv->bdrv_load_vmstate)
1906
        return drv->bdrv_load_vmstate(bs, buf, pos, size);
1907
    if (bs->file)
1908
        return bdrv_load_vmstate(bs->file, buf, pos, size);
1909
    return -ENOTSUP;
1910
}
1911

    
1912
void bdrv_debug_event(BlockDriverState *bs, BlkDebugEvent event)
1913
{
1914
    BlockDriver *drv = bs->drv;
1915

    
1916
    if (!drv || !drv->bdrv_debug_event) {
1917
        return;
1918
    }
1919

    
1920
    return drv->bdrv_debug_event(bs, event);
1921

    
1922
}
1923

    
1924
/**************************************************************/
1925
/* handling of snapshots */
1926

    
1927
int bdrv_can_snapshot(BlockDriverState *bs)
1928
{
1929
    BlockDriver *drv = bs->drv;
1930
    if (!drv || bdrv_is_removable(bs) || bdrv_is_read_only(bs)) {
1931
        return 0;
1932
    }
1933

    
1934
    if (!drv->bdrv_snapshot_create) {
1935
        if (bs->file != NULL) {
1936
            return bdrv_can_snapshot(bs->file);
1937
        }
1938
        return 0;
1939
    }
1940

    
1941
    return 1;
1942
}
1943

    
1944
int bdrv_is_snapshot(BlockDriverState *bs)
1945
{
1946
    return !!(bs->open_flags & BDRV_O_SNAPSHOT);
1947
}
1948

    
1949
BlockDriverState *bdrv_snapshots(void)
1950
{
1951
    BlockDriverState *bs;
1952

    
1953
    if (bs_snapshots) {
1954
        return bs_snapshots;
1955
    }
1956

    
1957
    bs = NULL;
1958
    while ((bs = bdrv_next(bs))) {
1959
        if (bdrv_can_snapshot(bs)) {
1960
            bs_snapshots = bs;
1961
            return bs;
1962
        }
1963
    }
1964
    return NULL;
1965
}
1966

    
1967
int bdrv_snapshot_create(BlockDriverState *bs,
1968
                         QEMUSnapshotInfo *sn_info)
1969
{
1970
    BlockDriver *drv = bs->drv;
1971
    if (!drv)
1972
        return -ENOMEDIUM;
1973
    if (drv->bdrv_snapshot_create)
1974
        return drv->bdrv_snapshot_create(bs, sn_info);
1975
    if (bs->file)
1976
        return bdrv_snapshot_create(bs->file, sn_info);
1977
    return -ENOTSUP;
1978
}
1979

    
1980
int bdrv_snapshot_goto(BlockDriverState *bs,
1981
                       const char *snapshot_id)
1982
{
1983
    BlockDriver *drv = bs->drv;
1984
    int ret, open_ret;
1985

    
1986
    if (!drv)
1987
        return -ENOMEDIUM;
1988
    if (drv->bdrv_snapshot_goto)
1989
        return drv->bdrv_snapshot_goto(bs, snapshot_id);
1990

    
1991
    if (bs->file) {
1992
        drv->bdrv_close(bs);
1993
        ret = bdrv_snapshot_goto(bs->file, snapshot_id);
1994
        open_ret = drv->bdrv_open(bs, bs->open_flags);
1995
        if (open_ret < 0) {
1996
            bdrv_delete(bs->file);
1997
            bs->drv = NULL;
1998
            return open_ret;
1999
        }
2000
        return ret;
2001
    }
2002

    
2003
    return -ENOTSUP;
2004
}
2005

    
2006
int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
2007
{
2008
    BlockDriver *drv = bs->drv;
2009
    if (!drv)
2010
        return -ENOMEDIUM;
2011
    if (drv->bdrv_snapshot_delete)
2012
        return drv->bdrv_snapshot_delete(bs, snapshot_id);
2013
    if (bs->file)
2014
        return bdrv_snapshot_delete(bs->file, snapshot_id);
2015
    return -ENOTSUP;
2016
}
2017

    
2018
int bdrv_snapshot_list(BlockDriverState *bs,
2019
                       QEMUSnapshotInfo **psn_info)
2020
{
2021
    BlockDriver *drv = bs->drv;
2022
    if (!drv)
2023
        return -ENOMEDIUM;
2024
    if (drv->bdrv_snapshot_list)
2025
        return drv->bdrv_snapshot_list(bs, psn_info);
2026
    if (bs->file)
2027
        return bdrv_snapshot_list(bs->file, psn_info);
2028
    return -ENOTSUP;
2029
}
2030

    
2031
int bdrv_snapshot_load_tmp(BlockDriverState *bs,
2032
        const char *snapshot_name)
2033
{
2034
    BlockDriver *drv = bs->drv;
2035
    if (!drv) {
2036
        return -ENOMEDIUM;
2037
    }
2038
    if (!bs->read_only) {
2039
        return -EINVAL;
2040
    }
2041
    if (drv->bdrv_snapshot_load_tmp) {
2042
        return drv->bdrv_snapshot_load_tmp(bs, snapshot_name);
2043
    }
2044
    return -ENOTSUP;
2045
}
2046

    
2047
#define NB_SUFFIXES 4
2048

    
2049
char *get_human_readable_size(char *buf, int buf_size, int64_t size)
2050
{
2051
    static const char suffixes[NB_SUFFIXES] = "KMGT";
2052
    int64_t base;
2053
    int i;
2054

    
2055
    if (size <= 999) {
2056
        snprintf(buf, buf_size, "%" PRId64, size);
2057
    } else {
2058
        base = 1024;
2059
        for(i = 0; i < NB_SUFFIXES; i++) {
2060
            if (size < (10 * base)) {
2061
                snprintf(buf, buf_size, "%0.1f%c",
2062
                         (double)size / base,
2063
                         suffixes[i]);
2064
                break;
2065
            } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
2066
                snprintf(buf, buf_size, "%" PRId64 "%c",
2067
                         ((size + (base >> 1)) / base),
2068
                         suffixes[i]);
2069
                break;
2070
            }
2071
            base = base * 1024;
2072
        }
2073
    }
2074
    return buf;
2075
}
2076

    
2077
char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
2078
{
2079
    char buf1[128], date_buf[128], clock_buf[128];
2080
#ifdef _WIN32
2081
    struct tm *ptm;
2082
#else
2083
    struct tm tm;
2084
#endif
2085
    time_t ti;
2086
    int64_t secs;
2087

    
2088
    if (!sn) {
2089
        snprintf(buf, buf_size,
2090
                 "%-10s%-20s%7s%20s%15s",
2091
                 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2092
    } else {
2093
        ti = sn->date_sec;
2094
#ifdef _WIN32
2095
        ptm = localtime(&ti);
2096
        strftime(date_buf, sizeof(date_buf),
2097
                 "%Y-%m-%d %H:%M:%S", ptm);
2098
#else
2099
        localtime_r(&ti, &tm);
2100
        strftime(date_buf, sizeof(date_buf),
2101
                 "%Y-%m-%d %H:%M:%S", &tm);
2102
#endif
2103
        secs = sn->vm_clock_nsec / 1000000000;
2104
        snprintf(clock_buf, sizeof(clock_buf),
2105
                 "%02d:%02d:%02d.%03d",
2106
                 (int)(secs / 3600),
2107
                 (int)((secs / 60) % 60),
2108
                 (int)(secs % 60),
2109
                 (int)((sn->vm_clock_nsec / 1000000) % 1000));
2110
        snprintf(buf, buf_size,
2111
                 "%-10s%-20s%7s%20s%15s",
2112
                 sn->id_str, sn->name,
2113
                 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
2114
                 date_buf,
2115
                 clock_buf);
2116
    }
2117
    return buf;
2118
}
2119

    
2120

    
2121
/**************************************************************/
2122
/* async I/Os */
2123

    
2124
BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
2125
                                 QEMUIOVector *qiov, int nb_sectors,
2126
                                 BlockDriverCompletionFunc *cb, void *opaque)
2127
{
2128
    BlockDriver *drv = bs->drv;
2129
    BlockDriverAIOCB *ret;
2130

    
2131
    trace_bdrv_aio_readv(bs, sector_num, nb_sectors, opaque);
2132

    
2133
    if (!drv)
2134
        return NULL;
2135
    if (bdrv_check_request(bs, sector_num, nb_sectors))
2136
        return NULL;
2137

    
2138
    ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
2139
                              cb, opaque);
2140

    
2141
    if (ret) {
2142
        /* Update stats even though technically transfer has not happened. */
2143
        bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2144
        bs->rd_ops ++;
2145
    }
2146

    
2147
    return ret;
2148
}
2149

    
2150
typedef struct BlockCompleteData {
2151
    BlockDriverCompletionFunc *cb;
2152
    void *opaque;
2153
    BlockDriverState *bs;
2154
    int64_t sector_num;
2155
    int nb_sectors;
2156
} BlockCompleteData;
2157

    
2158
static void block_complete_cb(void *opaque, int ret)
2159
{
2160
    BlockCompleteData *b = opaque;
2161

    
2162
    if (b->bs->dirty_bitmap) {
2163
        set_dirty_bitmap(b->bs, b->sector_num, b->nb_sectors, 1);
2164
    }
2165
    b->cb(b->opaque, ret);
2166
    qemu_free(b);
2167
}
2168

    
2169
static BlockCompleteData *blk_dirty_cb_alloc(BlockDriverState *bs,
2170
                                             int64_t sector_num,
2171
                                             int nb_sectors,
2172
                                             BlockDriverCompletionFunc *cb,
2173
                                             void *opaque)
2174
{
2175
    BlockCompleteData *blkdata = qemu_mallocz(sizeof(BlockCompleteData));
2176

    
2177
    blkdata->bs = bs;
2178
    blkdata->cb = cb;
2179
    blkdata->opaque = opaque;
2180
    blkdata->sector_num = sector_num;
2181
    blkdata->nb_sectors = nb_sectors;
2182

    
2183
    return blkdata;
2184
}
2185

    
2186
BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
2187
                                  QEMUIOVector *qiov, int nb_sectors,
2188
                                  BlockDriverCompletionFunc *cb, void *opaque)
2189
{
2190
    BlockDriver *drv = bs->drv;
2191
    BlockDriverAIOCB *ret;
2192
    BlockCompleteData *blk_cb_data;
2193

    
2194
    trace_bdrv_aio_writev(bs, sector_num, nb_sectors, opaque);
2195

    
2196
    if (!drv)
2197
        return NULL;
2198
    if (bs->read_only)
2199
        return NULL;
2200
    if (bdrv_check_request(bs, sector_num, nb_sectors))
2201
        return NULL;
2202

    
2203
    if (bs->dirty_bitmap) {
2204
        blk_cb_data = blk_dirty_cb_alloc(bs, sector_num, nb_sectors, cb,
2205
                                         opaque);
2206
        cb = &block_complete_cb;
2207
        opaque = blk_cb_data;
2208
    }
2209

    
2210
    ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
2211
                               cb, opaque);
2212

    
2213
    if (ret) {
2214
        /* Update stats even though technically transfer has not happened. */
2215
        bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
2216
        bs->wr_ops ++;
2217
        if (bs->wr_highest_sector < sector_num + nb_sectors - 1) {
2218
            bs->wr_highest_sector = sector_num + nb_sectors - 1;
2219
        }
2220
    }
2221

    
2222
    return ret;
2223
}
2224

    
2225

    
2226
typedef struct MultiwriteCB {
2227
    int error;
2228
    int num_requests;
2229
    int num_callbacks;
2230
    struct {
2231
        BlockDriverCompletionFunc *cb;
2232
        void *opaque;
2233
        QEMUIOVector *free_qiov;
2234
        void *free_buf;
2235
    } callbacks[];
2236
} MultiwriteCB;
2237

    
2238
static void multiwrite_user_cb(MultiwriteCB *mcb)
2239
{
2240
    int i;
2241

    
2242
    for (i = 0; i < mcb->num_callbacks; i++) {
2243
        mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
2244
        if (mcb->callbacks[i].free_qiov) {
2245
            qemu_iovec_destroy(mcb->callbacks[i].free_qiov);
2246
        }
2247
        qemu_free(mcb->callbacks[i].free_qiov);
2248
        qemu_vfree(mcb->callbacks[i].free_buf);
2249
    }
2250
}
2251

    
2252
static void multiwrite_cb(void *opaque, int ret)
2253
{
2254
    MultiwriteCB *mcb = opaque;
2255

    
2256
    trace_multiwrite_cb(mcb, ret);
2257

    
2258
    if (ret < 0 && !mcb->error) {
2259
        mcb->error = ret;
2260
    }
2261

    
2262
    mcb->num_requests--;
2263
    if (mcb->num_requests == 0) {
2264
        multiwrite_user_cb(mcb);
2265
        qemu_free(mcb);
2266
    }
2267
}
2268

    
2269
static int multiwrite_req_compare(const void *a, const void *b)
2270
{
2271
    const BlockRequest *req1 = a, *req2 = b;
2272

    
2273
    /*
2274
     * Note that we can't simply subtract req2->sector from req1->sector
2275
     * here as that could overflow the return value.
2276
     */
2277
    if (req1->sector > req2->sector) {
2278
        return 1;
2279
    } else if (req1->sector < req2->sector) {
2280
        return -1;
2281
    } else {
2282
        return 0;
2283
    }
2284
}
2285

    
2286
/*
2287
 * Takes a bunch of requests and tries to merge them. Returns the number of
2288
 * requests that remain after merging.
2289
 */
2290
static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
2291
    int num_reqs, MultiwriteCB *mcb)
2292
{
2293
    int i, outidx;
2294

    
2295
    // Sort requests by start sector
2296
    qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
2297

    
2298
    // Check if adjacent requests touch the same clusters. If so, combine them,
2299
    // filling up gaps with zero sectors.
2300
    outidx = 0;
2301
    for (i = 1; i < num_reqs; i++) {
2302
        int merge = 0;
2303
        int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
2304

    
2305
        // This handles the cases that are valid for all block drivers, namely
2306
        // exactly sequential writes and overlapping writes.
2307
        if (reqs[i].sector <= oldreq_last) {
2308
            merge = 1;
2309
        }
2310

    
2311
        // The block driver may decide that it makes sense to combine requests
2312
        // even if there is a gap of some sectors between them. In this case,
2313
        // the gap is filled with zeros (therefore only applicable for yet
2314
        // unused space in format like qcow2).
2315
        if (!merge && bs->drv->bdrv_merge_requests) {
2316
            merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
2317
        }
2318

    
2319
        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
2320
            merge = 0;
2321
        }
2322

    
2323
        if (merge) {
2324
            size_t size;
2325
            QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
2326
            qemu_iovec_init(qiov,
2327
                reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
2328

    
2329
            // Add the first request to the merged one. If the requests are
2330
            // overlapping, drop the last sectors of the first request.
2331
            size = (reqs[i].sector - reqs[outidx].sector) << 9;
2332
            qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
2333

    
2334
            // We might need to add some zeros between the two requests
2335
            if (reqs[i].sector > oldreq_last) {
2336
                size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
2337
                uint8_t *buf = qemu_blockalign(bs, zero_bytes);
2338
                memset(buf, 0, zero_bytes);
2339
                qemu_iovec_add(qiov, buf, zero_bytes);
2340
                mcb->callbacks[i].free_buf = buf;
2341
            }
2342

    
2343
            // Add the second request
2344
            qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
2345

    
2346
            reqs[outidx].nb_sectors = qiov->size >> 9;
2347
            reqs[outidx].qiov = qiov;
2348

    
2349
            mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
2350
        } else {
2351
            outidx++;
2352
            reqs[outidx].sector     = reqs[i].sector;
2353
            reqs[outidx].nb_sectors = reqs[i].nb_sectors;
2354
            reqs[outidx].qiov       = reqs[i].qiov;
2355
        }
2356
    }
2357

    
2358
    return outidx + 1;
2359
}
2360

    
2361
/*
2362
 * Submit multiple AIO write requests at once.
2363
 *
2364
 * On success, the function returns 0 and all requests in the reqs array have
2365
 * been submitted. In error case this function returns -1, and any of the
2366
 * requests may or may not be submitted yet. In particular, this means that the
2367
 * callback will be called for some of the requests, for others it won't. The
2368
 * caller must check the error field of the BlockRequest to wait for the right
2369
 * callbacks (if error != 0, no callback will be called).
2370
 *
2371
 * The implementation may modify the contents of the reqs array, e.g. to merge
2372
 * requests. However, the fields opaque and error are left unmodified as they
2373
 * are used to signal failure for a single request to the caller.
2374
 */
2375
int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
2376
{
2377
    BlockDriverAIOCB *acb;
2378
    MultiwriteCB *mcb;
2379
    int i;
2380

    
2381
    /* don't submit writes if we don't have a medium */
2382
    if (bs->drv == NULL) {
2383
        for (i = 0; i < num_reqs; i++) {
2384
            reqs[i].error = -ENOMEDIUM;
2385
        }
2386
        return -1;
2387
    }
2388

    
2389
    if (num_reqs == 0) {
2390
        return 0;
2391
    }
2392

    
2393
    // Create MultiwriteCB structure
2394
    mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
2395
    mcb->num_requests = 0;
2396
    mcb->num_callbacks = num_reqs;
2397

    
2398
    for (i = 0; i < num_reqs; i++) {
2399
        mcb->callbacks[i].cb = reqs[i].cb;
2400
        mcb->callbacks[i].opaque = reqs[i].opaque;
2401
    }
2402

    
2403
    // Check for mergable requests
2404
    num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
2405

    
2406
    trace_bdrv_aio_multiwrite(mcb, mcb->num_callbacks, num_reqs);
2407

    
2408
    /*
2409
     * Run the aio requests. As soon as one request can't be submitted
2410
     * successfully, fail all requests that are not yet submitted (we must
2411
     * return failure for all requests anyway)
2412
     *
2413
     * num_requests cannot be set to the right value immediately: If
2414
     * bdrv_aio_writev fails for some request, num_requests would be too high
2415
     * and therefore multiwrite_cb() would never recognize the multiwrite
2416
     * request as completed. We also cannot use the loop variable i to set it
2417
     * when the first request fails because the callback may already have been
2418
     * called for previously submitted requests. Thus, num_requests must be
2419
     * incremented for each request that is submitted.
2420
     *
2421
     * The problem that callbacks may be called early also means that we need
2422
     * to take care that num_requests doesn't become 0 before all requests are
2423
     * submitted - multiwrite_cb() would consider the multiwrite request
2424
     * completed. A dummy request that is "completed" by a manual call to
2425
     * multiwrite_cb() takes care of this.
2426
     */
2427
    mcb->num_requests = 1;
2428

    
2429
    // Run the aio requests
2430
    for (i = 0; i < num_reqs; i++) {
2431
        mcb->num_requests++;
2432
        acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
2433
            reqs[i].nb_sectors, multiwrite_cb, mcb);
2434

    
2435
        if (acb == NULL) {
2436
            // We can only fail the whole thing if no request has been
2437
            // submitted yet. Otherwise we'll wait for the submitted AIOs to
2438
            // complete and report the error in the callback.
2439
            if (i == 0) {
2440
                trace_bdrv_aio_multiwrite_earlyfail(mcb);
2441
                goto fail;
2442
            } else {
2443
                trace_bdrv_aio_multiwrite_latefail(mcb, i);
2444
                multiwrite_cb(mcb, -EIO);
2445
                break;
2446
            }
2447
        }
2448
    }
2449

    
2450
    /* Complete the dummy request */
2451
    multiwrite_cb(mcb, 0);
2452

    
2453
    return 0;
2454

    
2455
fail:
2456
    for (i = 0; i < mcb->num_callbacks; i++) {
2457
        reqs[i].error = -EIO;
2458
    }
2459
    qemu_free(mcb);
2460
    return -1;
2461
}
2462

    
2463
BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
2464
        BlockDriverCompletionFunc *cb, void *opaque)
2465
{
2466
    BlockDriver *drv = bs->drv;
2467

    
2468
    trace_bdrv_aio_flush(bs, opaque);
2469

    
2470
    if (bs->open_flags & BDRV_O_NO_FLUSH) {
2471
        return bdrv_aio_noop_em(bs, cb, opaque);
2472
    }
2473

    
2474
    if (!drv)
2475
        return NULL;
2476
    return drv->bdrv_aio_flush(bs, cb, opaque);
2477
}
2478

    
2479
void bdrv_aio_cancel(BlockDriverAIOCB *acb)
2480
{
2481
    acb->pool->cancel(acb);
2482
}
2483

    
2484

    
2485
/**************************************************************/
2486
/* async block device emulation */
2487

    
2488
typedef struct BlockDriverAIOCBSync {
2489
    BlockDriverAIOCB common;
2490
    QEMUBH *bh;
2491
    int ret;
2492
    /* vector translation state */
2493
    QEMUIOVector *qiov;
2494
    uint8_t *bounce;
2495
    int is_write;
2496
} BlockDriverAIOCBSync;
2497

    
2498
static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
2499
{
2500
    BlockDriverAIOCBSync *acb =
2501
        container_of(blockacb, BlockDriverAIOCBSync, common);
2502
    qemu_bh_delete(acb->bh);
2503
    acb->bh = NULL;
2504
    qemu_aio_release(acb);
2505
}
2506

    
2507
static AIOPool bdrv_em_aio_pool = {
2508
    .aiocb_size         = sizeof(BlockDriverAIOCBSync),
2509
    .cancel             = bdrv_aio_cancel_em,
2510
};
2511

    
2512
static void bdrv_aio_bh_cb(void *opaque)
2513
{
2514
    BlockDriverAIOCBSync *acb = opaque;
2515

    
2516
    if (!acb->is_write)
2517
        qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
2518
    qemu_vfree(acb->bounce);
2519
    acb->common.cb(acb->common.opaque, acb->ret);
2520
    qemu_bh_delete(acb->bh);
2521
    acb->bh = NULL;
2522
    qemu_aio_release(acb);
2523
}
2524

    
2525
static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
2526
                                            int64_t sector_num,
2527
                                            QEMUIOVector *qiov,
2528
                                            int nb_sectors,
2529
                                            BlockDriverCompletionFunc *cb,
2530
                                            void *opaque,
2531
                                            int is_write)
2532

    
2533
{
2534
    BlockDriverAIOCBSync *acb;
2535

    
2536
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2537
    acb->is_write = is_write;
2538
    acb->qiov = qiov;
2539
    acb->bounce = qemu_blockalign(bs, qiov->size);
2540

    
2541
    if (!acb->bh)
2542
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2543

    
2544
    if (is_write) {
2545
        qemu_iovec_to_buffer(acb->qiov, acb->bounce);
2546
        acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
2547
    } else {
2548
        acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
2549
    }
2550

    
2551
    qemu_bh_schedule(acb->bh);
2552

    
2553
    return &acb->common;
2554
}
2555

    
2556
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
2557
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2558
        BlockDriverCompletionFunc *cb, void *opaque)
2559
{
2560
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
2561
}
2562

    
2563
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
2564
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
2565
        BlockDriverCompletionFunc *cb, void *opaque)
2566
{
2567
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
2568
}
2569

    
2570
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2571
        BlockDriverCompletionFunc *cb, void *opaque)
2572
{
2573
    BlockDriverAIOCBSync *acb;
2574

    
2575
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2576
    acb->is_write = 1; /* don't bounce in the completion hadler */
2577
    acb->qiov = NULL;
2578
    acb->bounce = NULL;
2579
    acb->ret = 0;
2580

    
2581
    if (!acb->bh)
2582
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2583

    
2584
    bdrv_flush(bs);
2585
    qemu_bh_schedule(acb->bh);
2586
    return &acb->common;
2587
}
2588

    
2589
static BlockDriverAIOCB *bdrv_aio_noop_em(BlockDriverState *bs,
2590
        BlockDriverCompletionFunc *cb, void *opaque)
2591
{
2592
    BlockDriverAIOCBSync *acb;
2593

    
2594
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2595
    acb->is_write = 1; /* don't bounce in the completion handler */
2596
    acb->qiov = NULL;
2597
    acb->bounce = NULL;
2598
    acb->ret = 0;
2599

    
2600
    if (!acb->bh) {
2601
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2602
    }
2603

    
2604
    qemu_bh_schedule(acb->bh);
2605
    return &acb->common;
2606
}
2607

    
2608
/**************************************************************/
2609
/* sync block device emulation */
2610

    
2611
static void bdrv_rw_em_cb(void *opaque, int ret)
2612
{
2613
    *(int *)opaque = ret;
2614
}
2615

    
2616
#define NOT_DONE 0x7fffffff
2617

    
2618
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2619
                        uint8_t *buf, int nb_sectors)
2620
{
2621
    int async_ret;
2622
    BlockDriverAIOCB *acb;
2623
    struct iovec iov;
2624
    QEMUIOVector qiov;
2625

    
2626
    async_context_push();
2627

    
2628
    async_ret = NOT_DONE;
2629
    iov.iov_base = (void *)buf;
2630
    iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2631
    qemu_iovec_init_external(&qiov, &iov, 1);
2632
    acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2633
        bdrv_rw_em_cb, &async_ret);
2634
    if (acb == NULL) {
2635
        async_ret = -1;
2636
        goto fail;
2637
    }
2638

    
2639
    while (async_ret == NOT_DONE) {
2640
        qemu_aio_wait();
2641
    }
2642

    
2643

    
2644
fail:
2645
    async_context_pop();
2646
    return async_ret;
2647
}
2648

    
2649
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2650
                         const uint8_t *buf, int nb_sectors)
2651
{
2652
    int async_ret;
2653
    BlockDriverAIOCB *acb;
2654
    struct iovec iov;
2655
    QEMUIOVector qiov;
2656

    
2657
    async_context_push();
2658

    
2659
    async_ret = NOT_DONE;
2660
    iov.iov_base = (void *)buf;
2661
    iov.iov_len = nb_sectors * BDRV_SECTOR_SIZE;
2662
    qemu_iovec_init_external(&qiov, &iov, 1);
2663
    acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2664
        bdrv_rw_em_cb, &async_ret);
2665
    if (acb == NULL) {
2666
        async_ret = -1;
2667
        goto fail;
2668
    }
2669
    while (async_ret == NOT_DONE) {
2670
        qemu_aio_wait();
2671
    }
2672

    
2673
fail:
2674
    async_context_pop();
2675
    return async_ret;
2676
}
2677

    
2678
void bdrv_init(void)
2679
{
2680
    module_call_init(MODULE_INIT_BLOCK);
2681
}
2682

    
2683
void bdrv_init_with_whitelist(void)
2684
{
2685
    use_bdrv_whitelist = 1;
2686
    bdrv_init();
2687
}
2688

    
2689
void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2690
                   BlockDriverCompletionFunc *cb, void *opaque)
2691
{
2692
    BlockDriverAIOCB *acb;
2693

    
2694
    if (pool->free_aiocb) {
2695
        acb = pool->free_aiocb;
2696
        pool->free_aiocb = acb->next;
2697
    } else {
2698
        acb = qemu_mallocz(pool->aiocb_size);
2699
        acb->pool = pool;
2700
    }
2701
    acb->bs = bs;
2702
    acb->cb = cb;
2703
    acb->opaque = opaque;
2704
    return acb;
2705
}
2706

    
2707
void qemu_aio_release(void *p)
2708
{
2709
    BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2710
    AIOPool *pool = acb->pool;
2711
    acb->next = pool->free_aiocb;
2712
    pool->free_aiocb = acb;
2713
}
2714

    
2715
/**************************************************************/
2716
/* removable device support */
2717

    
2718
/**
2719
 * Return TRUE if the media is present
2720
 */
2721
int bdrv_is_inserted(BlockDriverState *bs)
2722
{
2723
    BlockDriver *drv = bs->drv;
2724
    int ret;
2725
    if (!drv)
2726
        return 0;
2727
    if (!drv->bdrv_is_inserted)
2728
        return !bs->tray_open;
2729
    ret = drv->bdrv_is_inserted(bs);
2730
    return ret;
2731
}
2732

    
2733
/**
2734
 * Return TRUE if the media changed since the last call to this
2735
 * function. It is currently only used for floppy disks
2736
 */
2737
int bdrv_media_changed(BlockDriverState *bs)
2738
{
2739
    BlockDriver *drv = bs->drv;
2740
    int ret;
2741

    
2742
    if (!drv || !drv->bdrv_media_changed)
2743
        ret = -ENOTSUP;
2744
    else
2745
        ret = drv->bdrv_media_changed(bs);
2746
    if (ret == -ENOTSUP)
2747
        ret = bs->media_changed;
2748
    bs->media_changed = 0;
2749
    return ret;
2750
}
2751

    
2752
/**
2753
 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2754
 */
2755
int bdrv_eject(BlockDriverState *bs, int eject_flag)
2756
{
2757
    BlockDriver *drv = bs->drv;
2758
    int ret;
2759

    
2760
    if (bs->locked) {
2761
        return -EBUSY;
2762
    }
2763

    
2764
    if (!drv || !drv->bdrv_eject) {
2765
        ret = -ENOTSUP;
2766
    } else {
2767
        ret = drv->bdrv_eject(bs, eject_flag);
2768
    }
2769
    if (ret == -ENOTSUP) {
2770
        ret = 0;
2771
    }
2772
    if (ret >= 0) {
2773
        bs->tray_open = eject_flag;
2774
    }
2775

    
2776
    return ret;
2777
}
2778

    
2779
int bdrv_is_locked(BlockDriverState *bs)
2780
{
2781
    return bs->locked;
2782
}
2783

    
2784
/**
2785
 * Lock or unlock the media (if it is locked, the user won't be able
2786
 * to eject it manually).
2787
 */
2788
void bdrv_set_locked(BlockDriverState *bs, int locked)
2789
{
2790
    BlockDriver *drv = bs->drv;
2791

    
2792
    trace_bdrv_set_locked(bs, locked);
2793

    
2794
    bs->locked = locked;
2795
    if (drv && drv->bdrv_set_locked) {
2796
        drv->bdrv_set_locked(bs, locked);
2797
    }
2798
}
2799

    
2800
/* needed for generic scsi interface */
2801

    
2802
int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2803
{
2804
    BlockDriver *drv = bs->drv;
2805

    
2806
    if (drv && drv->bdrv_ioctl)
2807
        return drv->bdrv_ioctl(bs, req, buf);
2808
    return -ENOTSUP;
2809
}
2810

    
2811
BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2812
        unsigned long int req, void *buf,
2813
        BlockDriverCompletionFunc *cb, void *opaque)
2814
{
2815
    BlockDriver *drv = bs->drv;
2816

    
2817
    if (drv && drv->bdrv_aio_ioctl)
2818
        return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2819
    return NULL;
2820
}
2821

    
2822

    
2823

    
2824
void *qemu_blockalign(BlockDriverState *bs, size_t size)
2825
{
2826
    return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2827
}
2828

    
2829
void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2830
{
2831
    int64_t bitmap_size;
2832

    
2833
    bs->dirty_count = 0;
2834
    if (enable) {
2835
        if (!bs->dirty_bitmap) {
2836
            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2837
                    BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2838
            bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2839

    
2840
            bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2841
        }
2842
    } else {
2843
        if (bs->dirty_bitmap) {
2844
            qemu_free(bs->dirty_bitmap);
2845
            bs->dirty_bitmap = NULL;
2846
        }
2847
    }
2848
}
2849

    
2850
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2851
{
2852
    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2853

    
2854
    if (bs->dirty_bitmap &&
2855
        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2856
        return !!(bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2857
            (1UL << (chunk % (sizeof(unsigned long) * 8))));
2858
    } else {
2859
        return 0;
2860
    }
2861
}
2862

    
2863
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2864
                      int nr_sectors)
2865
{
2866
    set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2867
}
2868

    
2869
int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2870
{
2871
    return bs->dirty_count;
2872
}
2873

    
2874
void bdrv_set_in_use(BlockDriverState *bs, int in_use)
2875
{
2876
    assert(bs->in_use != in_use);
2877
    bs->in_use = in_use;
2878
}
2879

    
2880
int bdrv_in_use(BlockDriverState *bs)
2881
{
2882
    return bs->in_use;
2883
}
2884

    
2885
int bdrv_img_create(const char *filename, const char *fmt,
2886
                    const char *base_filename, const char *base_fmt,
2887
                    char *options, uint64_t img_size, int flags)
2888
{
2889
    QEMUOptionParameter *param = NULL, *create_options = NULL;
2890
    QEMUOptionParameter *backing_fmt, *backing_file;
2891
    BlockDriverState *bs = NULL;
2892
    BlockDriver *drv, *proto_drv;
2893
    BlockDriver *backing_drv = NULL;
2894
    int ret = 0;
2895

    
2896
    /* Find driver and parse its options */
2897
    drv = bdrv_find_format(fmt);
2898
    if (!drv) {
2899
        error_report("Unknown file format '%s'", fmt);
2900
        ret = -EINVAL;
2901
        goto out;
2902
    }
2903

    
2904
    proto_drv = bdrv_find_protocol(filename);
2905
    if (!proto_drv) {
2906
        error_report("Unknown protocol '%s'", filename);
2907
        ret = -EINVAL;
2908
        goto out;
2909
    }
2910

    
2911
    create_options = append_option_parameters(create_options,
2912
                                              drv->create_options);
2913
    create_options = append_option_parameters(create_options,
2914
                                              proto_drv->create_options);
2915

    
2916
    /* Create parameter list with default values */
2917
    param = parse_option_parameters("", create_options, param);
2918

    
2919
    set_option_parameter_int(param, BLOCK_OPT_SIZE, img_size);
2920

    
2921
    /* Parse -o options */
2922
    if (options) {
2923
        param = parse_option_parameters(options, create_options, param);
2924
        if (param == NULL) {
2925
            error_report("Invalid options for file format '%s'.", fmt);
2926
            ret = -EINVAL;
2927
            goto out;
2928
        }
2929
    }
2930

    
2931
    if (base_filename) {
2932
        if (set_option_parameter(param, BLOCK_OPT_BACKING_FILE,
2933
                                 base_filename)) {
2934
            error_report("Backing file not supported for file format '%s'",
2935
                         fmt);
2936
            ret = -EINVAL;
2937
            goto out;
2938
        }
2939
    }
2940

    
2941
    if (base_fmt) {
2942
        if (set_option_parameter(param, BLOCK_OPT_BACKING_FMT, base_fmt)) {
2943
            error_report("Backing file format not supported for file "
2944
                         "format '%s'", fmt);
2945
            ret = -EINVAL;
2946
            goto out;
2947
        }
2948
    }
2949

    
2950
    backing_file = get_option_parameter(param, BLOCK_OPT_BACKING_FILE);
2951
    if (backing_file && backing_file->value.s) {
2952
        if (!strcmp(filename, backing_file->value.s)) {
2953
            error_report("Error: Trying to create an image with the "
2954
                         "same filename as the backing file");
2955
            ret = -EINVAL;
2956
            goto out;
2957
        }
2958
    }
2959

    
2960
    backing_fmt = get_option_parameter(param, BLOCK_OPT_BACKING_FMT);
2961
    if (backing_fmt && backing_fmt->value.s) {
2962
        backing_drv = bdrv_find_format(backing_fmt->value.s);
2963
        if (!backing_drv) {
2964
            error_report("Unknown backing file format '%s'",
2965
                         backing_fmt->value.s);
2966
            ret = -EINVAL;
2967
            goto out;
2968
        }
2969
    }
2970

    
2971
    // The size for the image must always be specified, with one exception:
2972
    // If we are using a backing file, we can obtain the size from there
2973
    if (get_option_parameter(param, BLOCK_OPT_SIZE)->value.n == -1) {
2974
        if (backing_file && backing_file->value.s) {
2975
            uint64_t size;
2976
            char buf[32];
2977

    
2978
            bs = bdrv_new("");
2979

    
2980
            ret = bdrv_open(bs, backing_file->value.s, flags, backing_drv);
2981
            if (ret < 0) {
2982
                error_report("Could not open '%s'", backing_file->value.s);
2983
                goto out;
2984
            }
2985
            bdrv_get_geometry(bs, &size);
2986
            size *= 512;
2987

    
2988
            snprintf(buf, sizeof(buf), "%" PRId64, size);
2989
            set_option_parameter(param, BLOCK_OPT_SIZE, buf);
2990
        } else {
2991
            error_report("Image creation needs a size parameter");
2992
            ret = -EINVAL;
2993
            goto out;
2994
        }
2995
    }
2996

    
2997
    printf("Formatting '%s', fmt=%s ", filename, fmt);
2998
    print_option_parameters(param);
2999
    puts("");
3000

    
3001
    ret = bdrv_create(drv, filename, param);
3002

    
3003
    if (ret < 0) {
3004
        if (ret == -ENOTSUP) {
3005
            error_report("Formatting or formatting option not supported for "
3006
                         "file format '%s'", fmt);
3007
        } else if (ret == -EFBIG) {
3008
            error_report("The image size is too large for file format '%s'",
3009
                         fmt);
3010
        } else {
3011
            error_report("%s: error while creating %s: %s", filename, fmt,
3012
                         strerror(-ret));
3013
        }
3014
    }
3015

    
3016
out:
3017
    free_option_parameters(create_options);
3018
    free_option_parameters(param);
3019

    
3020
    if (bs) {
3021
        bdrv_delete(bs);
3022
    }
3023

    
3024
    return ret;
3025
}