Statistics
| Branch: | Revision:

root / block.c @ 15dc2697

History | View | Annotate | Download (59.9 kB)

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

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

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

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

    
58
BlockDriverState *bdrv_first;
59

    
60
static BlockDriver *first_drv;
61

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
298
    return drv;
299
}
300

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

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

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

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

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

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

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

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

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

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

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

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

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

    
391
        bdrv_delete(bs1);
392

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

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

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

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

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

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

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

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

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

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

    
453
    bs->read_only = (flags & BDRV_O_RDWR) == 0;
454

    
455
    /*
456
     * Clear flags that are internal to the block layer before opening the
457
     * image.
458
     */
459
    open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT | BDRV_O_NO_BACKING);
460

    
461
    /*
462
     * Snapshots should be writeable.
463
     *
464
     * XXX(hch): and what is the point of a snapshot during a read-only open?
465
     */
466
    if (!(flags & BDRV_O_FILE) && bs->is_temporary) {
467
        open_flags |= BDRV_O_RDWR;
468
    }
469

    
470
    ret = drv->bdrv_open(bs, filename, open_flags);
471
    if (ret < 0) {
472
        goto free_and_fail;
473
    }
474

    
475
    if (drv->bdrv_getlength) {
476
        bs->total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
477
    }
478
#ifndef _WIN32
479
    if (bs->is_temporary) {
480
        unlink(filename);
481
    }
482
#endif
483
    if ((flags & BDRV_O_NO_BACKING) == 0 && bs->backing_file[0] != '\0') {
484
        /* if there is a backing file, use it */
485
        BlockDriver *back_drv = NULL;
486
        bs->backing_hd = bdrv_new("");
487
        path_combine(backing_filename, sizeof(backing_filename),
488
                     filename, bs->backing_file);
489
        if (bs->backing_format[0] != '\0')
490
            back_drv = bdrv_find_format(bs->backing_format);
491
        ret = bdrv_open2(bs->backing_hd, backing_filename, open_flags,
492
                         back_drv);
493
        bs->backing_hd->read_only =  (open_flags & BDRV_O_RDWR) == 0;
494
        if (ret < 0) {
495
            bdrv_close(bs);
496
            return ret;
497
        }
498
    }
499

    
500
    if (!bdrv_key_required(bs)) {
501
        /* call the change callback */
502
        bs->media_changed = 1;
503
        if (bs->change_cb)
504
            bs->change_cb(bs->change_opaque);
505
    }
506
    return 0;
507

    
508
free_and_fail:
509
    qemu_free(bs->opaque);
510
    bs->opaque = NULL;
511
    bs->drv = NULL;
512
unlink_and_fail:
513
    if (bs->is_temporary)
514
        unlink(filename);
515
    return ret;
516
}
517

    
518
void bdrv_close(BlockDriverState *bs)
519
{
520
    if (bs->drv) {
521
        if (bs->backing_hd)
522
            bdrv_delete(bs->backing_hd);
523
        bs->drv->bdrv_close(bs);
524
        qemu_free(bs->opaque);
525
#ifdef _WIN32
526
        if (bs->is_temporary) {
527
            unlink(bs->filename);
528
        }
529
#endif
530
        bs->opaque = NULL;
531
        bs->drv = NULL;
532

    
533
        /* call the change callback */
534
        bs->media_changed = 1;
535
        if (bs->change_cb)
536
            bs->change_cb(bs->change_opaque);
537
    }
538
}
539

    
540
void bdrv_delete(BlockDriverState *bs)
541
{
542
    BlockDriverState **pbs;
543

    
544
    pbs = &bdrv_first;
545
    while (*pbs != bs && *pbs != NULL)
546
        pbs = &(*pbs)->next;
547
    if (*pbs == bs)
548
        *pbs = bs->next;
549

    
550
    bdrv_close(bs);
551
    qemu_free(bs);
552
}
553

    
554
/*
555
 * Run consistency checks on an image
556
 *
557
 * Returns the number of errors or -errno when an internal error occurs
558
 */
559
int bdrv_check(BlockDriverState *bs)
560
{
561
    if (bs->drv->bdrv_check == NULL) {
562
        return -ENOTSUP;
563
    }
564

    
565
    return bs->drv->bdrv_check(bs);
566
}
567

    
568
/* commit COW file into the raw image */
569
int bdrv_commit(BlockDriverState *bs)
570
{
571
    BlockDriver *drv = bs->drv;
572
    int64_t i, total_sectors;
573
    int n, j;
574
    int ret = 0;
575
    unsigned char sector[512];
576

    
577
    if (!drv)
578
        return -ENOMEDIUM;
579

    
580
    if (bs->read_only) {
581
        return -EACCES;
582
    }
583

    
584
    if (!bs->backing_hd) {
585
        return -ENOTSUP;
586
    }
587

    
588
    total_sectors = bdrv_getlength(bs) >> BDRV_SECTOR_BITS;
589
    for (i = 0; i < total_sectors;) {
590
        if (drv->bdrv_is_allocated(bs, i, 65536, &n)) {
591
            for(j = 0; j < n; j++) {
592
                if (bdrv_read(bs, i, sector, 1) != 0) {
593
                    return -EIO;
594
                }
595

    
596
                if (bdrv_write(bs->backing_hd, i, sector, 1) != 0) {
597
                    return -EIO;
598
                }
599
                i++;
600
            }
601
        } else {
602
            i += n;
603
        }
604
    }
605

    
606
    if (drv->bdrv_make_empty) {
607
        ret = drv->bdrv_make_empty(bs);
608
        bdrv_flush(bs);
609
    }
610

    
611
    /*
612
     * Make sure all data we wrote to the backing device is actually
613
     * stable on disk.
614
     */
615
    if (bs->backing_hd)
616
        bdrv_flush(bs->backing_hd);
617
    return ret;
618
}
619

    
620
/*
621
 * Return values:
622
 * 0        - success
623
 * -EINVAL  - backing format specified, but no file
624
 * -ENOSPC  - can't update the backing file because no space is left in the
625
 *            image file header
626
 * -ENOTSUP - format driver doesn't support changing the backing file
627
 */
628
int bdrv_change_backing_file(BlockDriverState *bs,
629
    const char *backing_file, const char *backing_fmt)
630
{
631
    BlockDriver *drv = bs->drv;
632

    
633
    if (drv->bdrv_change_backing_file != NULL) {
634
        return drv->bdrv_change_backing_file(bs, backing_file, backing_fmt);
635
    } else {
636
        return -ENOTSUP;
637
    }
638
}
639

    
640
static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
641
                                   size_t size)
642
{
643
    int64_t len;
644

    
645
    if (!bdrv_is_inserted(bs))
646
        return -ENOMEDIUM;
647

    
648
    if (bs->growable)
649
        return 0;
650

    
651
    len = bdrv_getlength(bs);
652

    
653
    if (offset < 0)
654
        return -EIO;
655

    
656
    if ((offset > len) || (len - offset < size))
657
        return -EIO;
658

    
659
    return 0;
660
}
661

    
662
static int bdrv_check_request(BlockDriverState *bs, int64_t sector_num,
663
                              int nb_sectors)
664
{
665
    return bdrv_check_byte_request(bs, sector_num * 512, nb_sectors * 512);
666
}
667

    
668
/* return < 0 if error. See bdrv_write() for the return codes */
669
int bdrv_read(BlockDriverState *bs, int64_t sector_num,
670
              uint8_t *buf, int nb_sectors)
671
{
672
    BlockDriver *drv = bs->drv;
673

    
674
    if (!drv)
675
        return -ENOMEDIUM;
676
    if (bdrv_check_request(bs, sector_num, nb_sectors))
677
        return -EIO;
678

    
679
    return drv->bdrv_read(bs, sector_num, buf, nb_sectors);
680
}
681

    
682
static void set_dirty_bitmap(BlockDriverState *bs, int64_t sector_num,
683
                             int nb_sectors, int dirty)
684
{
685
    int64_t start, end;
686
    unsigned long val, idx, bit;
687

    
688
    start = sector_num / BDRV_SECTORS_PER_DIRTY_CHUNK;
689
    end = (sector_num + nb_sectors - 1) / BDRV_SECTORS_PER_DIRTY_CHUNK;
690

    
691
    for (; start <= end; start++) {
692
        idx = start / (sizeof(unsigned long) * 8);
693
        bit = start % (sizeof(unsigned long) * 8);
694
        val = bs->dirty_bitmap[idx];
695
        if (dirty) {
696
            if (!(val & (1 << bit))) {
697
                bs->dirty_count++;
698
                val |= 1 << bit;
699
            }
700
        } else {
701
            if (val & (1 << bit)) {
702
                bs->dirty_count--;
703
                val &= ~(1 << bit);
704
            }
705
        }
706
        bs->dirty_bitmap[idx] = val;
707
    }
708
}
709

    
710
/* Return < 0 if error. Important errors are:
711
  -EIO         generic I/O error (may happen for all errors)
712
  -ENOMEDIUM   No media inserted.
713
  -EINVAL      Invalid sector number or nb_sectors
714
  -EACCES      Trying to write a read-only device
715
*/
716
int bdrv_write(BlockDriverState *bs, int64_t sector_num,
717
               const uint8_t *buf, int nb_sectors)
718
{
719
    BlockDriver *drv = bs->drv;
720
    if (!bs->drv)
721
        return -ENOMEDIUM;
722
    if (bs->read_only)
723
        return -EACCES;
724
    if (bdrv_check_request(bs, sector_num, nb_sectors))
725
        return -EIO;
726

    
727
    if (bs->dirty_bitmap) {
728
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
729
    }
730

    
731
    return drv->bdrv_write(bs, sector_num, buf, nb_sectors);
732
}
733

    
734
int bdrv_pread(BlockDriverState *bs, int64_t offset,
735
               void *buf, int count1)
736
{
737
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
738
    int len, nb_sectors, count;
739
    int64_t sector_num;
740
    int ret;
741

    
742
    count = count1;
743
    /* first read to align to sector start */
744
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
745
    if (len > count)
746
        len = count;
747
    sector_num = offset >> BDRV_SECTOR_BITS;
748
    if (len > 0) {
749
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
750
            return ret;
751
        memcpy(buf, tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), len);
752
        count -= len;
753
        if (count == 0)
754
            return count1;
755
        sector_num++;
756
        buf += len;
757
    }
758

    
759
    /* read the sectors "in place" */
760
    nb_sectors = count >> BDRV_SECTOR_BITS;
761
    if (nb_sectors > 0) {
762
        if ((ret = bdrv_read(bs, sector_num, buf, nb_sectors)) < 0)
763
            return ret;
764
        sector_num += nb_sectors;
765
        len = nb_sectors << BDRV_SECTOR_BITS;
766
        buf += len;
767
        count -= len;
768
    }
769

    
770
    /* add data from the last sector */
771
    if (count > 0) {
772
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
773
            return ret;
774
        memcpy(buf, tmp_buf, count);
775
    }
776
    return count1;
777
}
778

    
779
int bdrv_pwrite(BlockDriverState *bs, int64_t offset,
780
                const void *buf, int count1)
781
{
782
    uint8_t tmp_buf[BDRV_SECTOR_SIZE];
783
    int len, nb_sectors, count;
784
    int64_t sector_num;
785
    int ret;
786

    
787
    count = count1;
788
    /* first write to align to sector start */
789
    len = (BDRV_SECTOR_SIZE - offset) & (BDRV_SECTOR_SIZE - 1);
790
    if (len > count)
791
        len = count;
792
    sector_num = offset >> BDRV_SECTOR_BITS;
793
    if (len > 0) {
794
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
795
            return ret;
796
        memcpy(tmp_buf + (offset & (BDRV_SECTOR_SIZE - 1)), buf, len);
797
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
798
            return ret;
799
        count -= len;
800
        if (count == 0)
801
            return count1;
802
        sector_num++;
803
        buf += len;
804
    }
805

    
806
    /* write the sectors "in place" */
807
    nb_sectors = count >> BDRV_SECTOR_BITS;
808
    if (nb_sectors > 0) {
809
        if ((ret = bdrv_write(bs, sector_num, buf, nb_sectors)) < 0)
810
            return ret;
811
        sector_num += nb_sectors;
812
        len = nb_sectors << BDRV_SECTOR_BITS;
813
        buf += len;
814
        count -= len;
815
    }
816

    
817
    /* add data from the last sector */
818
    if (count > 0) {
819
        if ((ret = bdrv_read(bs, sector_num, tmp_buf, 1)) < 0)
820
            return ret;
821
        memcpy(tmp_buf, buf, count);
822
        if ((ret = bdrv_write(bs, sector_num, tmp_buf, 1)) < 0)
823
            return ret;
824
    }
825
    return count1;
826
}
827

    
828
/**
829
 * Truncate file to 'offset' bytes (needed only for file protocols)
830
 */
831
int bdrv_truncate(BlockDriverState *bs, int64_t offset)
832
{
833
    BlockDriver *drv = bs->drv;
834
    if (!drv)
835
        return -ENOMEDIUM;
836
    if (!drv->bdrv_truncate)
837
        return -ENOTSUP;
838
    if (bs->read_only)
839
        return -EACCES;
840
    return drv->bdrv_truncate(bs, offset);
841
}
842

    
843
/**
844
 * Length of a file in bytes. Return < 0 if error or unknown.
845
 */
846
int64_t bdrv_getlength(BlockDriverState *bs)
847
{
848
    BlockDriver *drv = bs->drv;
849
    if (!drv)
850
        return -ENOMEDIUM;
851
    if (!drv->bdrv_getlength) {
852
        /* legacy mode */
853
        return bs->total_sectors * BDRV_SECTOR_SIZE;
854
    }
855
    return drv->bdrv_getlength(bs);
856
}
857

    
858
/* return 0 as number of sectors if no device present or error */
859
void bdrv_get_geometry(BlockDriverState *bs, uint64_t *nb_sectors_ptr)
860
{
861
    int64_t length;
862
    length = bdrv_getlength(bs);
863
    if (length < 0)
864
        length = 0;
865
    else
866
        length = length >> BDRV_SECTOR_BITS;
867
    *nb_sectors_ptr = length;
868
}
869

    
870
struct partition {
871
        uint8_t boot_ind;           /* 0x80 - active */
872
        uint8_t head;               /* starting head */
873
        uint8_t sector;             /* starting sector */
874
        uint8_t cyl;                /* starting cylinder */
875
        uint8_t sys_ind;            /* What partition type */
876
        uint8_t end_head;           /* end head */
877
        uint8_t end_sector;         /* end sector */
878
        uint8_t end_cyl;            /* end cylinder */
879
        uint32_t start_sect;        /* starting sector counting from 0 */
880
        uint32_t nr_sects;          /* nr of sectors in partition */
881
} __attribute__((packed));
882

    
883
/* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
884
static int guess_disk_lchs(BlockDriverState *bs,
885
                           int *pcylinders, int *pheads, int *psectors)
886
{
887
    uint8_t buf[512];
888
    int ret, i, heads, sectors, cylinders;
889
    struct partition *p;
890
    uint32_t nr_sects;
891
    uint64_t nb_sectors;
892

    
893
    bdrv_get_geometry(bs, &nb_sectors);
894

    
895
    ret = bdrv_read(bs, 0, buf, 1);
896
    if (ret < 0)
897
        return -1;
898
    /* test msdos magic */
899
    if (buf[510] != 0x55 || buf[511] != 0xaa)
900
        return -1;
901
    for(i = 0; i < 4; i++) {
902
        p = ((struct partition *)(buf + 0x1be)) + i;
903
        nr_sects = le32_to_cpu(p->nr_sects);
904
        if (nr_sects && p->end_head) {
905
            /* We make the assumption that the partition terminates on
906
               a cylinder boundary */
907
            heads = p->end_head + 1;
908
            sectors = p->end_sector & 63;
909
            if (sectors == 0)
910
                continue;
911
            cylinders = nb_sectors / (heads * sectors);
912
            if (cylinders < 1 || cylinders > 16383)
913
                continue;
914
            *pheads = heads;
915
            *psectors = sectors;
916
            *pcylinders = cylinders;
917
#if 0
918
            printf("guessed geometry: LCHS=%d %d %d\n",
919
                   cylinders, heads, sectors);
920
#endif
921
            return 0;
922
        }
923
    }
924
    return -1;
925
}
926

    
927
void bdrv_guess_geometry(BlockDriverState *bs, int *pcyls, int *pheads, int *psecs)
928
{
929
    int translation, lba_detected = 0;
930
    int cylinders, heads, secs;
931
    uint64_t nb_sectors;
932

    
933
    /* if a geometry hint is available, use it */
934
    bdrv_get_geometry(bs, &nb_sectors);
935
    bdrv_get_geometry_hint(bs, &cylinders, &heads, &secs);
936
    translation = bdrv_get_translation_hint(bs);
937
    if (cylinders != 0) {
938
        *pcyls = cylinders;
939
        *pheads = heads;
940
        *psecs = secs;
941
    } else {
942
        if (guess_disk_lchs(bs, &cylinders, &heads, &secs) == 0) {
943
            if (heads > 16) {
944
                /* if heads > 16, it means that a BIOS LBA
945
                   translation was active, so the default
946
                   hardware geometry is OK */
947
                lba_detected = 1;
948
                goto default_geometry;
949
            } else {
950
                *pcyls = cylinders;
951
                *pheads = heads;
952
                *psecs = secs;
953
                /* disable any translation to be in sync with
954
                   the logical geometry */
955
                if (translation == BIOS_ATA_TRANSLATION_AUTO) {
956
                    bdrv_set_translation_hint(bs,
957
                                              BIOS_ATA_TRANSLATION_NONE);
958
                }
959
            }
960
        } else {
961
        default_geometry:
962
            /* if no geometry, use a standard physical disk geometry */
963
            cylinders = nb_sectors / (16 * 63);
964

    
965
            if (cylinders > 16383)
966
                cylinders = 16383;
967
            else if (cylinders < 2)
968
                cylinders = 2;
969
            *pcyls = cylinders;
970
            *pheads = 16;
971
            *psecs = 63;
972
            if ((lba_detected == 1) && (translation == BIOS_ATA_TRANSLATION_AUTO)) {
973
                if ((*pcyls * *pheads) <= 131072) {
974
                    bdrv_set_translation_hint(bs,
975
                                              BIOS_ATA_TRANSLATION_LARGE);
976
                } else {
977
                    bdrv_set_translation_hint(bs,
978
                                              BIOS_ATA_TRANSLATION_LBA);
979
                }
980
            }
981
        }
982
        bdrv_set_geometry_hint(bs, *pcyls, *pheads, *psecs);
983
    }
984
}
985

    
986
void bdrv_set_geometry_hint(BlockDriverState *bs,
987
                            int cyls, int heads, int secs)
988
{
989
    bs->cyls = cyls;
990
    bs->heads = heads;
991
    bs->secs = secs;
992
}
993

    
994
void bdrv_set_type_hint(BlockDriverState *bs, int type)
995
{
996
    bs->type = type;
997
    bs->removable = ((type == BDRV_TYPE_CDROM ||
998
                      type == BDRV_TYPE_FLOPPY));
999
}
1000

    
1001
void bdrv_set_translation_hint(BlockDriverState *bs, int translation)
1002
{
1003
    bs->translation = translation;
1004
}
1005

    
1006
void bdrv_get_geometry_hint(BlockDriverState *bs,
1007
                            int *pcyls, int *pheads, int *psecs)
1008
{
1009
    *pcyls = bs->cyls;
1010
    *pheads = bs->heads;
1011
    *psecs = bs->secs;
1012
}
1013

    
1014
int bdrv_get_type_hint(BlockDriverState *bs)
1015
{
1016
    return bs->type;
1017
}
1018

    
1019
int bdrv_get_translation_hint(BlockDriverState *bs)
1020
{
1021
    return bs->translation;
1022
}
1023

    
1024
int bdrv_is_removable(BlockDriverState *bs)
1025
{
1026
    return bs->removable;
1027
}
1028

    
1029
int bdrv_is_read_only(BlockDriverState *bs)
1030
{
1031
    return bs->read_only;
1032
}
1033

    
1034
int bdrv_is_sg(BlockDriverState *bs)
1035
{
1036
    return bs->sg;
1037
}
1038

    
1039
int bdrv_enable_write_cache(BlockDriverState *bs)
1040
{
1041
    return bs->enable_write_cache;
1042
}
1043

    
1044
/* XXX: no longer used */
1045
void bdrv_set_change_cb(BlockDriverState *bs,
1046
                        void (*change_cb)(void *opaque), void *opaque)
1047
{
1048
    bs->change_cb = change_cb;
1049
    bs->change_opaque = opaque;
1050
}
1051

    
1052
int bdrv_is_encrypted(BlockDriverState *bs)
1053
{
1054
    if (bs->backing_hd && bs->backing_hd->encrypted)
1055
        return 1;
1056
    return bs->encrypted;
1057
}
1058

    
1059
int bdrv_key_required(BlockDriverState *bs)
1060
{
1061
    BlockDriverState *backing_hd = bs->backing_hd;
1062

    
1063
    if (backing_hd && backing_hd->encrypted && !backing_hd->valid_key)
1064
        return 1;
1065
    return (bs->encrypted && !bs->valid_key);
1066
}
1067

    
1068
int bdrv_set_key(BlockDriverState *bs, const char *key)
1069
{
1070
    int ret;
1071
    if (bs->backing_hd && bs->backing_hd->encrypted) {
1072
        ret = bdrv_set_key(bs->backing_hd, key);
1073
        if (ret < 0)
1074
            return ret;
1075
        if (!bs->encrypted)
1076
            return 0;
1077
    }
1078
    if (!bs->encrypted || !bs->drv || !bs->drv->bdrv_set_key)
1079
        return -1;
1080
    ret = bs->drv->bdrv_set_key(bs, key);
1081
    if (ret < 0) {
1082
        bs->valid_key = 0;
1083
    } else if (!bs->valid_key) {
1084
        bs->valid_key = 1;
1085
        /* call the change callback now, we skipped it on open */
1086
        bs->media_changed = 1;
1087
        if (bs->change_cb)
1088
            bs->change_cb(bs->change_opaque);
1089
    }
1090
    return ret;
1091
}
1092

    
1093
void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1094
{
1095
    if (!bs->drv) {
1096
        buf[0] = '\0';
1097
    } else {
1098
        pstrcpy(buf, buf_size, bs->drv->format_name);
1099
    }
1100
}
1101

    
1102
void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1103
                         void *opaque)
1104
{
1105
    BlockDriver *drv;
1106

    
1107
    for (drv = first_drv; drv != NULL; drv = drv->next) {
1108
        it(opaque, drv->format_name);
1109
    }
1110
}
1111

    
1112
BlockDriverState *bdrv_find(const char *name)
1113
{
1114
    BlockDriverState *bs;
1115

    
1116
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1117
        if (!strcmp(name, bs->device_name))
1118
            return bs;
1119
    }
1120
    return NULL;
1121
}
1122

    
1123
void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1124
{
1125
    BlockDriverState *bs;
1126

    
1127
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1128
        it(opaque, bs);
1129
    }
1130
}
1131

    
1132
const char *bdrv_get_device_name(BlockDriverState *bs)
1133
{
1134
    return bs->device_name;
1135
}
1136

    
1137
void bdrv_flush(BlockDriverState *bs)
1138
{
1139
    if (bs->drv && bs->drv->bdrv_flush)
1140
        bs->drv->bdrv_flush(bs);
1141
}
1142

    
1143
void bdrv_flush_all(void)
1144
{
1145
    BlockDriverState *bs;
1146

    
1147
    for (bs = bdrv_first; bs != NULL; bs = bs->next)
1148
        if (bs->drv && !bdrv_is_read_only(bs) && 
1149
            (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)))
1150
            bdrv_flush(bs);
1151
}
1152

    
1153
/*
1154
 * Returns true iff the specified sector is present in the disk image. Drivers
1155
 * not implementing the functionality are assumed to not support backing files,
1156
 * hence all their sectors are reported as allocated.
1157
 *
1158
 * 'pnum' is set to the number of sectors (including and immediately following
1159
 * the specified sector) that are known to be in the same
1160
 * allocated/unallocated state.
1161
 *
1162
 * 'nb_sectors' is the max value 'pnum' should be set to.
1163
 */
1164
int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1165
        int *pnum)
1166
{
1167
    int64_t n;
1168
    if (!bs->drv->bdrv_is_allocated) {
1169
        if (sector_num >= bs->total_sectors) {
1170
            *pnum = 0;
1171
            return 0;
1172
        }
1173
        n = bs->total_sectors - sector_num;
1174
        *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1175
        return 1;
1176
    }
1177
    return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1178
}
1179

    
1180
void bdrv_mon_event(const BlockDriverState *bdrv,
1181
                    BlockMonEventAction action, int is_read)
1182
{
1183
    QObject *data;
1184
    const char *action_str;
1185

    
1186
    switch (action) {
1187
    case BDRV_ACTION_REPORT:
1188
        action_str = "report";
1189
        break;
1190
    case BDRV_ACTION_IGNORE:
1191
        action_str = "ignore";
1192
        break;
1193
    case BDRV_ACTION_STOP:
1194
        action_str = "stop";
1195
        break;
1196
    default:
1197
        abort();
1198
    }
1199

    
1200
    data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1201
                              bdrv->device_name,
1202
                              action_str,
1203
                              is_read ? "read" : "write");
1204
    monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1205

    
1206
    qobject_decref(data);
1207
}
1208

    
1209
static void bdrv_print_dict(QObject *obj, void *opaque)
1210
{
1211
    QDict *bs_dict;
1212
    Monitor *mon = opaque;
1213

    
1214
    bs_dict = qobject_to_qdict(obj);
1215

    
1216
    monitor_printf(mon, "%s: type=%s removable=%d",
1217
                        qdict_get_str(bs_dict, "device"),
1218
                        qdict_get_str(bs_dict, "type"),
1219
                        qdict_get_bool(bs_dict, "removable"));
1220

    
1221
    if (qdict_get_bool(bs_dict, "removable")) {
1222
        monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1223
    }
1224

    
1225
    if (qdict_haskey(bs_dict, "inserted")) {
1226
        QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1227

    
1228
        monitor_printf(mon, " file=");
1229
        monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1230
        if (qdict_haskey(qdict, "backing_file")) {
1231
            monitor_printf(mon, " backing_file=");
1232
            monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1233
        }
1234
        monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1235
                            qdict_get_bool(qdict, "ro"),
1236
                            qdict_get_str(qdict, "drv"),
1237
                            qdict_get_bool(qdict, "encrypted"));
1238
    } else {
1239
        monitor_printf(mon, " [not inserted]");
1240
    }
1241

    
1242
    monitor_printf(mon, "\n");
1243
}
1244

    
1245
void bdrv_info_print(Monitor *mon, const QObject *data)
1246
{
1247
    qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1248
}
1249

    
1250
/**
1251
 * bdrv_info(): Block devices information
1252
 *
1253
 * Each block device information is stored in a QDict and the
1254
 * returned QObject is a QList of all devices.
1255
 *
1256
 * The QDict contains the following:
1257
 *
1258
 * - "device": device name
1259
 * - "type": device type
1260
 * - "removable": true if the device is removable, false otherwise
1261
 * - "locked": true if the device is locked, false otherwise
1262
 * - "inserted": only present if the device is inserted, it is a QDict
1263
 *    containing the following:
1264
 *          - "file": device file name
1265
 *          - "ro": true if read-only, false otherwise
1266
 *          - "drv": driver format name
1267
 *          - "backing_file": backing file name if one is used
1268
 *          - "encrypted": true if encrypted, false otherwise
1269
 *
1270
 * Example:
1271
 *
1272
 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1273
 *     "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1274
 *   { "device": "floppy0", "type": "floppy", "removable": true,
1275
 *     "locked": false } ]
1276
 */
1277
void bdrv_info(Monitor *mon, QObject **ret_data)
1278
{
1279
    QList *bs_list;
1280
    BlockDriverState *bs;
1281

    
1282
    bs_list = qlist_new();
1283

    
1284
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1285
        QObject *bs_obj;
1286
        const char *type = "unknown";
1287

    
1288
        switch(bs->type) {
1289
        case BDRV_TYPE_HD:
1290
            type = "hd";
1291
            break;
1292
        case BDRV_TYPE_CDROM:
1293
            type = "cdrom";
1294
            break;
1295
        case BDRV_TYPE_FLOPPY:
1296
            type = "floppy";
1297
            break;
1298
        }
1299

    
1300
        bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1301
                                    "'removable': %i, 'locked': %i }",
1302
                                    bs->device_name, type, bs->removable,
1303
                                    bs->locked);
1304
        assert(bs_obj != NULL);
1305

    
1306
        if (bs->drv) {
1307
            QObject *obj;
1308
            QDict *bs_dict = qobject_to_qdict(bs_obj);
1309

    
1310
            obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1311
                                     "'encrypted': %i }",
1312
                                     bs->filename, bs->read_only,
1313
                                     bs->drv->format_name,
1314
                                     bdrv_is_encrypted(bs));
1315
            assert(obj != NULL);
1316
            if (bs->backing_file[0] != '\0') {
1317
                QDict *qdict = qobject_to_qdict(obj);
1318
                qdict_put(qdict, "backing_file",
1319
                          qstring_from_str(bs->backing_file));
1320
            }
1321

    
1322
            qdict_put_obj(bs_dict, "inserted", obj);
1323
        }
1324
        qlist_append_obj(bs_list, bs_obj);
1325
    }
1326

    
1327
    *ret_data = QOBJECT(bs_list);
1328
}
1329

    
1330
static void bdrv_stats_iter(QObject *data, void *opaque)
1331
{
1332
    QDict *qdict;
1333
    Monitor *mon = opaque;
1334

    
1335
    qdict = qobject_to_qdict(data);
1336
    monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1337

    
1338
    qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1339
    monitor_printf(mon, " rd_bytes=%" PRId64
1340
                        " wr_bytes=%" PRId64
1341
                        " rd_operations=%" PRId64
1342
                        " wr_operations=%" PRId64
1343
                        "\n",
1344
                        qdict_get_int(qdict, "rd_bytes"),
1345
                        qdict_get_int(qdict, "wr_bytes"),
1346
                        qdict_get_int(qdict, "rd_operations"),
1347
                        qdict_get_int(qdict, "wr_operations"));
1348
}
1349

    
1350
void bdrv_stats_print(Monitor *mon, const QObject *data)
1351
{
1352
    qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1353
}
1354

    
1355
/**
1356
 * bdrv_info_stats(): show block device statistics
1357
 *
1358
 * Each device statistic information is stored in a QDict and
1359
 * the returned QObject is a QList of all devices.
1360
 *
1361
 * The QDict contains the following:
1362
 *
1363
 * - "device": device name
1364
 * - "stats": A QDict with the statistics information, it contains:
1365
 *     - "rd_bytes": bytes read
1366
 *     - "wr_bytes": bytes written
1367
 *     - "rd_operations": read operations
1368
 *     - "wr_operations": write operations
1369
 * 
1370
 * Example:
1371
 *
1372
 * [ { "device": "ide0-hd0",
1373
 *               "stats": { "rd_bytes": 512,
1374
 *                          "wr_bytes": 0,
1375
 *                          "rd_operations": 1,
1376
 *                          "wr_operations": 0 } },
1377
 *   { "device": "ide1-cd0",
1378
 *               "stats": { "rd_bytes": 0,
1379
 *                          "wr_bytes": 0,
1380
 *                          "rd_operations": 0,
1381
 *                          "wr_operations": 0 } } ]
1382
 */
1383
void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1384
{
1385
    QObject *obj;
1386
    QList *devices;
1387
    BlockDriverState *bs;
1388

    
1389
    devices = qlist_new();
1390

    
1391
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1392
        obj = qobject_from_jsonf("{ 'device': %s, 'stats': {"
1393
                                 "'rd_bytes': %" PRId64 ","
1394
                                 "'wr_bytes': %" PRId64 ","
1395
                                 "'rd_operations': %" PRId64 ","
1396
                                 "'wr_operations': %" PRId64
1397
                                 "} }",
1398
                                 bs->device_name,
1399
                                 bs->rd_bytes, bs->wr_bytes,
1400
                                 bs->rd_ops, bs->wr_ops);
1401
        assert(obj != NULL);
1402
        qlist_append_obj(devices, obj);
1403
    }
1404

    
1405
    *ret_data = QOBJECT(devices);
1406
}
1407

    
1408
const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1409
{
1410
    if (bs->backing_hd && bs->backing_hd->encrypted)
1411
        return bs->backing_file;
1412
    else if (bs->encrypted)
1413
        return bs->filename;
1414
    else
1415
        return NULL;
1416
}
1417

    
1418
void bdrv_get_backing_filename(BlockDriverState *bs,
1419
                               char *filename, int filename_size)
1420
{
1421
    if (!bs->backing_file) {
1422
        pstrcpy(filename, filename_size, "");
1423
    } else {
1424
        pstrcpy(filename, filename_size, bs->backing_file);
1425
    }
1426
}
1427

    
1428
int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1429
                          const uint8_t *buf, int nb_sectors)
1430
{
1431
    BlockDriver *drv = bs->drv;
1432
    if (!drv)
1433
        return -ENOMEDIUM;
1434
    if (!drv->bdrv_write_compressed)
1435
        return -ENOTSUP;
1436
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1437
        return -EIO;
1438

    
1439
    if (bs->dirty_bitmap) {
1440
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1441
    }
1442

    
1443
    return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1444
}
1445

    
1446
int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1447
{
1448
    BlockDriver *drv = bs->drv;
1449
    if (!drv)
1450
        return -ENOMEDIUM;
1451
    if (!drv->bdrv_get_info)
1452
        return -ENOTSUP;
1453
    memset(bdi, 0, sizeof(*bdi));
1454
    return drv->bdrv_get_info(bs, bdi);
1455
}
1456

    
1457
int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1458
                      int64_t pos, int size)
1459
{
1460
    BlockDriver *drv = bs->drv;
1461
    if (!drv)
1462
        return -ENOMEDIUM;
1463
    if (!drv->bdrv_save_vmstate)
1464
        return -ENOTSUP;
1465
    return drv->bdrv_save_vmstate(bs, buf, pos, size);
1466
}
1467

    
1468
int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1469
                      int64_t pos, int size)
1470
{
1471
    BlockDriver *drv = bs->drv;
1472
    if (!drv)
1473
        return -ENOMEDIUM;
1474
    if (!drv->bdrv_load_vmstate)
1475
        return -ENOTSUP;
1476
    return drv->bdrv_load_vmstate(bs, buf, pos, size);
1477
}
1478

    
1479
/**************************************************************/
1480
/* handling of snapshots */
1481

    
1482
int bdrv_snapshot_create(BlockDriverState *bs,
1483
                         QEMUSnapshotInfo *sn_info)
1484
{
1485
    BlockDriver *drv = bs->drv;
1486
    if (!drv)
1487
        return -ENOMEDIUM;
1488
    if (!drv->bdrv_snapshot_create)
1489
        return -ENOTSUP;
1490
    return drv->bdrv_snapshot_create(bs, sn_info);
1491
}
1492

    
1493
int bdrv_snapshot_goto(BlockDriverState *bs,
1494
                       const char *snapshot_id)
1495
{
1496
    BlockDriver *drv = bs->drv;
1497
    if (!drv)
1498
        return -ENOMEDIUM;
1499
    if (!drv->bdrv_snapshot_goto)
1500
        return -ENOTSUP;
1501
    return drv->bdrv_snapshot_goto(bs, snapshot_id);
1502
}
1503

    
1504
int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1505
{
1506
    BlockDriver *drv = bs->drv;
1507
    if (!drv)
1508
        return -ENOMEDIUM;
1509
    if (!drv->bdrv_snapshot_delete)
1510
        return -ENOTSUP;
1511
    return drv->bdrv_snapshot_delete(bs, snapshot_id);
1512
}
1513

    
1514
int bdrv_snapshot_list(BlockDriverState *bs,
1515
                       QEMUSnapshotInfo **psn_info)
1516
{
1517
    BlockDriver *drv = bs->drv;
1518
    if (!drv)
1519
        return -ENOMEDIUM;
1520
    if (!drv->bdrv_snapshot_list)
1521
        return -ENOTSUP;
1522
    return drv->bdrv_snapshot_list(bs, psn_info);
1523
}
1524

    
1525
#define NB_SUFFIXES 4
1526

    
1527
char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1528
{
1529
    static const char suffixes[NB_SUFFIXES] = "KMGT";
1530
    int64_t base;
1531
    int i;
1532

    
1533
    if (size <= 999) {
1534
        snprintf(buf, buf_size, "%" PRId64, size);
1535
    } else {
1536
        base = 1024;
1537
        for(i = 0; i < NB_SUFFIXES; i++) {
1538
            if (size < (10 * base)) {
1539
                snprintf(buf, buf_size, "%0.1f%c",
1540
                         (double)size / base,
1541
                         suffixes[i]);
1542
                break;
1543
            } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1544
                snprintf(buf, buf_size, "%" PRId64 "%c",
1545
                         ((size + (base >> 1)) / base),
1546
                         suffixes[i]);
1547
                break;
1548
            }
1549
            base = base * 1024;
1550
        }
1551
    }
1552
    return buf;
1553
}
1554

    
1555
char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1556
{
1557
    char buf1[128], date_buf[128], clock_buf[128];
1558
#ifdef _WIN32
1559
    struct tm *ptm;
1560
#else
1561
    struct tm tm;
1562
#endif
1563
    time_t ti;
1564
    int64_t secs;
1565

    
1566
    if (!sn) {
1567
        snprintf(buf, buf_size,
1568
                 "%-10s%-20s%7s%20s%15s",
1569
                 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1570
    } else {
1571
        ti = sn->date_sec;
1572
#ifdef _WIN32
1573
        ptm = localtime(&ti);
1574
        strftime(date_buf, sizeof(date_buf),
1575
                 "%Y-%m-%d %H:%M:%S", ptm);
1576
#else
1577
        localtime_r(&ti, &tm);
1578
        strftime(date_buf, sizeof(date_buf),
1579
                 "%Y-%m-%d %H:%M:%S", &tm);
1580
#endif
1581
        secs = sn->vm_clock_nsec / 1000000000;
1582
        snprintf(clock_buf, sizeof(clock_buf),
1583
                 "%02d:%02d:%02d.%03d",
1584
                 (int)(secs / 3600),
1585
                 (int)((secs / 60) % 60),
1586
                 (int)(secs % 60),
1587
                 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1588
        snprintf(buf, buf_size,
1589
                 "%-10s%-20s%7s%20s%15s",
1590
                 sn->id_str, sn->name,
1591
                 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1592
                 date_buf,
1593
                 clock_buf);
1594
    }
1595
    return buf;
1596
}
1597

    
1598

    
1599
/**************************************************************/
1600
/* async I/Os */
1601

    
1602
BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1603
                                 QEMUIOVector *qiov, int nb_sectors,
1604
                                 BlockDriverCompletionFunc *cb, void *opaque)
1605
{
1606
    BlockDriver *drv = bs->drv;
1607
    BlockDriverAIOCB *ret;
1608

    
1609
    if (!drv)
1610
        return NULL;
1611
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1612
        return NULL;
1613

    
1614
    ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1615
                              cb, opaque);
1616

    
1617
    if (ret) {
1618
        /* Update stats even though technically transfer has not happened. */
1619
        bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1620
        bs->rd_ops ++;
1621
    }
1622

    
1623
    return ret;
1624
}
1625

    
1626
BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1627
                                  QEMUIOVector *qiov, int nb_sectors,
1628
                                  BlockDriverCompletionFunc *cb, void *opaque)
1629
{
1630
    BlockDriver *drv = bs->drv;
1631
    BlockDriverAIOCB *ret;
1632

    
1633
    if (!drv)
1634
        return NULL;
1635
    if (bs->read_only)
1636
        return NULL;
1637
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1638
        return NULL;
1639

    
1640
    if (bs->dirty_bitmap) {
1641
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1642
    }
1643

    
1644
    ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1645
                               cb, opaque);
1646

    
1647
    if (ret) {
1648
        /* Update stats even though technically transfer has not happened. */
1649
        bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1650
        bs->wr_ops ++;
1651
    }
1652

    
1653
    return ret;
1654
}
1655

    
1656

    
1657
typedef struct MultiwriteCB {
1658
    int error;
1659
    int num_requests;
1660
    int num_callbacks;
1661
    struct {
1662
        BlockDriverCompletionFunc *cb;
1663
        void *opaque;
1664
        QEMUIOVector *free_qiov;
1665
        void *free_buf;
1666
    } callbacks[];
1667
} MultiwriteCB;
1668

    
1669
static void multiwrite_user_cb(MultiwriteCB *mcb)
1670
{
1671
    int i;
1672

    
1673
    for (i = 0; i < mcb->num_callbacks; i++) {
1674
        mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1675
        qemu_free(mcb->callbacks[i].free_qiov);
1676
        qemu_vfree(mcb->callbacks[i].free_buf);
1677
    }
1678
}
1679

    
1680
static void multiwrite_cb(void *opaque, int ret)
1681
{
1682
    MultiwriteCB *mcb = opaque;
1683

    
1684
    if (ret < 0) {
1685
        mcb->error = ret;
1686
        multiwrite_user_cb(mcb);
1687
    }
1688

    
1689
    mcb->num_requests--;
1690
    if (mcb->num_requests == 0) {
1691
        if (mcb->error == 0) {
1692
            multiwrite_user_cb(mcb);
1693
        }
1694
        qemu_free(mcb);
1695
    }
1696
}
1697

    
1698
static int multiwrite_req_compare(const void *a, const void *b)
1699
{
1700
    return (((BlockRequest*) a)->sector - ((BlockRequest*) b)->sector);
1701
}
1702

    
1703
/*
1704
 * Takes a bunch of requests and tries to merge them. Returns the number of
1705
 * requests that remain after merging.
1706
 */
1707
static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1708
    int num_reqs, MultiwriteCB *mcb)
1709
{
1710
    int i, outidx;
1711

    
1712
    // Sort requests by start sector
1713
    qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1714

    
1715
    // Check if adjacent requests touch the same clusters. If so, combine them,
1716
    // filling up gaps with zero sectors.
1717
    outidx = 0;
1718
    for (i = 1; i < num_reqs; i++) {
1719
        int merge = 0;
1720
        int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1721

    
1722
        // This handles the cases that are valid for all block drivers, namely
1723
        // exactly sequential writes and overlapping writes.
1724
        if (reqs[i].sector <= oldreq_last) {
1725
            merge = 1;
1726
        }
1727

    
1728
        // The block driver may decide that it makes sense to combine requests
1729
        // even if there is a gap of some sectors between them. In this case,
1730
        // the gap is filled with zeros (therefore only applicable for yet
1731
        // unused space in format like qcow2).
1732
        if (!merge && bs->drv->bdrv_merge_requests) {
1733
            merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1734
        }
1735

    
1736
        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
1737
            merge = 0;
1738
        }
1739

    
1740
        if (merge) {
1741
            size_t size;
1742
            QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1743
            qemu_iovec_init(qiov,
1744
                reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1745

    
1746
            // Add the first request to the merged one. If the requests are
1747
            // overlapping, drop the last sectors of the first request.
1748
            size = (reqs[i].sector - reqs[outidx].sector) << 9;
1749
            qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1750

    
1751
            // We might need to add some zeros between the two requests
1752
            if (reqs[i].sector > oldreq_last) {
1753
                size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1754
                uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1755
                memset(buf, 0, zero_bytes);
1756
                qemu_iovec_add(qiov, buf, zero_bytes);
1757
                mcb->callbacks[i].free_buf = buf;
1758
            }
1759

    
1760
            // Add the second request
1761
            qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
1762

    
1763
            reqs[outidx].nb_sectors += reqs[i].nb_sectors;
1764
            reqs[outidx].qiov = qiov;
1765

    
1766
            mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
1767
        } else {
1768
            outidx++;
1769
            reqs[outidx].sector     = reqs[i].sector;
1770
            reqs[outidx].nb_sectors = reqs[i].nb_sectors;
1771
            reqs[outidx].qiov       = reqs[i].qiov;
1772
        }
1773
    }
1774

    
1775
    return outidx + 1;
1776
}
1777

    
1778
/*
1779
 * Submit multiple AIO write requests at once.
1780
 *
1781
 * On success, the function returns 0 and all requests in the reqs array have
1782
 * been submitted. In error case this function returns -1, and any of the
1783
 * requests may or may not be submitted yet. In particular, this means that the
1784
 * callback will be called for some of the requests, for others it won't. The
1785
 * caller must check the error field of the BlockRequest to wait for the right
1786
 * callbacks (if error != 0, no callback will be called).
1787
 *
1788
 * The implementation may modify the contents of the reqs array, e.g. to merge
1789
 * requests. However, the fields opaque and error are left unmodified as they
1790
 * are used to signal failure for a single request to the caller.
1791
 */
1792
int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
1793
{
1794
    BlockDriverAIOCB *acb;
1795
    MultiwriteCB *mcb;
1796
    int i;
1797

    
1798
    if (num_reqs == 0) {
1799
        return 0;
1800
    }
1801

    
1802
    // Create MultiwriteCB structure
1803
    mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
1804
    mcb->num_requests = 0;
1805
    mcb->num_callbacks = num_reqs;
1806

    
1807
    for (i = 0; i < num_reqs; i++) {
1808
        mcb->callbacks[i].cb = reqs[i].cb;
1809
        mcb->callbacks[i].opaque = reqs[i].opaque;
1810
    }
1811

    
1812
    // Check for mergable requests
1813
    num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
1814

    
1815
    // Run the aio requests
1816
    for (i = 0; i < num_reqs; i++) {
1817
        acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
1818
            reqs[i].nb_sectors, multiwrite_cb, mcb);
1819

    
1820
        if (acb == NULL) {
1821
            // We can only fail the whole thing if no request has been
1822
            // submitted yet. Otherwise we'll wait for the submitted AIOs to
1823
            // complete and report the error in the callback.
1824
            if (mcb->num_requests == 0) {
1825
                reqs[i].error = EIO;
1826
                goto fail;
1827
            } else {
1828
                mcb->error = EIO;
1829
                break;
1830
            }
1831
        } else {
1832
            mcb->num_requests++;
1833
        }
1834
    }
1835

    
1836
    return 0;
1837

    
1838
fail:
1839
    free(mcb);
1840
    return -1;
1841
}
1842

    
1843
BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
1844
        BlockDriverCompletionFunc *cb, void *opaque)
1845
{
1846
    BlockDriver *drv = bs->drv;
1847

    
1848
    if (!drv)
1849
        return NULL;
1850
    return drv->bdrv_aio_flush(bs, cb, opaque);
1851
}
1852

    
1853
void bdrv_aio_cancel(BlockDriverAIOCB *acb)
1854
{
1855
    acb->pool->cancel(acb);
1856
}
1857

    
1858

    
1859
/**************************************************************/
1860
/* async block device emulation */
1861

    
1862
typedef struct BlockDriverAIOCBSync {
1863
    BlockDriverAIOCB common;
1864
    QEMUBH *bh;
1865
    int ret;
1866
    /* vector translation state */
1867
    QEMUIOVector *qiov;
1868
    uint8_t *bounce;
1869
    int is_write;
1870
} BlockDriverAIOCBSync;
1871

    
1872
static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
1873
{
1874
    BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
1875
    qemu_bh_delete(acb->bh);
1876
    acb->bh = NULL;
1877
    qemu_aio_release(acb);
1878
}
1879

    
1880
static AIOPool bdrv_em_aio_pool = {
1881
    .aiocb_size         = sizeof(BlockDriverAIOCBSync),
1882
    .cancel             = bdrv_aio_cancel_em,
1883
};
1884

    
1885
static void bdrv_aio_bh_cb(void *opaque)
1886
{
1887
    BlockDriverAIOCBSync *acb = opaque;
1888

    
1889
    if (!acb->is_write)
1890
        qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
1891
    qemu_vfree(acb->bounce);
1892
    acb->common.cb(acb->common.opaque, acb->ret);
1893
    qemu_bh_delete(acb->bh);
1894
    acb->bh = NULL;
1895
    qemu_aio_release(acb);
1896
}
1897

    
1898
static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
1899
                                            int64_t sector_num,
1900
                                            QEMUIOVector *qiov,
1901
                                            int nb_sectors,
1902
                                            BlockDriverCompletionFunc *cb,
1903
                                            void *opaque,
1904
                                            int is_write)
1905

    
1906
{
1907
    BlockDriverAIOCBSync *acb;
1908

    
1909
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1910
    acb->is_write = is_write;
1911
    acb->qiov = qiov;
1912
    acb->bounce = qemu_blockalign(bs, qiov->size);
1913

    
1914
    if (!acb->bh)
1915
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1916

    
1917
    if (is_write) {
1918
        qemu_iovec_to_buffer(acb->qiov, acb->bounce);
1919
        acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
1920
    } else {
1921
        acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
1922
    }
1923

    
1924
    qemu_bh_schedule(acb->bh);
1925

    
1926
    return &acb->common;
1927
}
1928

    
1929
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
1930
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1931
        BlockDriverCompletionFunc *cb, void *opaque)
1932
{
1933
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
1934
}
1935

    
1936
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
1937
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1938
        BlockDriverCompletionFunc *cb, void *opaque)
1939
{
1940
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
1941
}
1942

    
1943
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
1944
        BlockDriverCompletionFunc *cb, void *opaque)
1945
{
1946
    BlockDriverAIOCBSync *acb;
1947

    
1948
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1949
    acb->is_write = 1; /* don't bounce in the completion hadler */
1950
    acb->qiov = NULL;
1951
    acb->bounce = NULL;
1952
    acb->ret = 0;
1953

    
1954
    if (!acb->bh)
1955
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1956

    
1957
    bdrv_flush(bs);
1958
    qemu_bh_schedule(acb->bh);
1959
    return &acb->common;
1960
}
1961

    
1962
/**************************************************************/
1963
/* sync block device emulation */
1964

    
1965
static void bdrv_rw_em_cb(void *opaque, int ret)
1966
{
1967
    *(int *)opaque = ret;
1968
}
1969

    
1970
#define NOT_DONE 0x7fffffff
1971

    
1972
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
1973
                        uint8_t *buf, int nb_sectors)
1974
{
1975
    int async_ret;
1976
    BlockDriverAIOCB *acb;
1977
    struct iovec iov;
1978
    QEMUIOVector qiov;
1979

    
1980
    async_context_push();
1981

    
1982
    async_ret = NOT_DONE;
1983
    iov.iov_base = (void *)buf;
1984
    iov.iov_len = nb_sectors * 512;
1985
    qemu_iovec_init_external(&qiov, &iov, 1);
1986
    acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
1987
        bdrv_rw_em_cb, &async_ret);
1988
    if (acb == NULL) {
1989
        async_ret = -1;
1990
        goto fail;
1991
    }
1992

    
1993
    while (async_ret == NOT_DONE) {
1994
        qemu_aio_wait();
1995
    }
1996

    
1997

    
1998
fail:
1999
    async_context_pop();
2000
    return async_ret;
2001
}
2002

    
2003
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2004
                         const uint8_t *buf, int nb_sectors)
2005
{
2006
    int async_ret;
2007
    BlockDriverAIOCB *acb;
2008
    struct iovec iov;
2009
    QEMUIOVector qiov;
2010

    
2011
    async_context_push();
2012

    
2013
    async_ret = NOT_DONE;
2014
    iov.iov_base = (void *)buf;
2015
    iov.iov_len = nb_sectors * 512;
2016
    qemu_iovec_init_external(&qiov, &iov, 1);
2017
    acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2018
        bdrv_rw_em_cb, &async_ret);
2019
    if (acb == NULL) {
2020
        async_ret = -1;
2021
        goto fail;
2022
    }
2023
    while (async_ret == NOT_DONE) {
2024
        qemu_aio_wait();
2025
    }
2026

    
2027
fail:
2028
    async_context_pop();
2029
    return async_ret;
2030
}
2031

    
2032
void bdrv_init(void)
2033
{
2034
    module_call_init(MODULE_INIT_BLOCK);
2035
}
2036

    
2037
void bdrv_init_with_whitelist(void)
2038
{
2039
    use_bdrv_whitelist = 1;
2040
    bdrv_init();
2041
}
2042

    
2043
void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2044
                   BlockDriverCompletionFunc *cb, void *opaque)
2045
{
2046
    BlockDriverAIOCB *acb;
2047

    
2048
    if (pool->free_aiocb) {
2049
        acb = pool->free_aiocb;
2050
        pool->free_aiocb = acb->next;
2051
    } else {
2052
        acb = qemu_mallocz(pool->aiocb_size);
2053
        acb->pool = pool;
2054
    }
2055
    acb->bs = bs;
2056
    acb->cb = cb;
2057
    acb->opaque = opaque;
2058
    return acb;
2059
}
2060

    
2061
void qemu_aio_release(void *p)
2062
{
2063
    BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2064
    AIOPool *pool = acb->pool;
2065
    acb->next = pool->free_aiocb;
2066
    pool->free_aiocb = acb;
2067
}
2068

    
2069
/**************************************************************/
2070
/* removable device support */
2071

    
2072
/**
2073
 * Return TRUE if the media is present
2074
 */
2075
int bdrv_is_inserted(BlockDriverState *bs)
2076
{
2077
    BlockDriver *drv = bs->drv;
2078
    int ret;
2079
    if (!drv)
2080
        return 0;
2081
    if (!drv->bdrv_is_inserted)
2082
        return 1;
2083
    ret = drv->bdrv_is_inserted(bs);
2084
    return ret;
2085
}
2086

    
2087
/**
2088
 * Return TRUE if the media changed since the last call to this
2089
 * function. It is currently only used for floppy disks
2090
 */
2091
int bdrv_media_changed(BlockDriverState *bs)
2092
{
2093
    BlockDriver *drv = bs->drv;
2094
    int ret;
2095

    
2096
    if (!drv || !drv->bdrv_media_changed)
2097
        ret = -ENOTSUP;
2098
    else
2099
        ret = drv->bdrv_media_changed(bs);
2100
    if (ret == -ENOTSUP)
2101
        ret = bs->media_changed;
2102
    bs->media_changed = 0;
2103
    return ret;
2104
}
2105

    
2106
/**
2107
 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2108
 */
2109
int bdrv_eject(BlockDriverState *bs, int eject_flag)
2110
{
2111
    BlockDriver *drv = bs->drv;
2112
    int ret;
2113

    
2114
    if (bs->locked) {
2115
        return -EBUSY;
2116
    }
2117

    
2118
    if (!drv || !drv->bdrv_eject) {
2119
        ret = -ENOTSUP;
2120
    } else {
2121
        ret = drv->bdrv_eject(bs, eject_flag);
2122
    }
2123
    if (ret == -ENOTSUP) {
2124
        if (eject_flag)
2125
            bdrv_close(bs);
2126
        ret = 0;
2127
    }
2128

    
2129
    return ret;
2130
}
2131

    
2132
int bdrv_is_locked(BlockDriverState *bs)
2133
{
2134
    return bs->locked;
2135
}
2136

    
2137
/**
2138
 * Lock or unlock the media (if it is locked, the user won't be able
2139
 * to eject it manually).
2140
 */
2141
void bdrv_set_locked(BlockDriverState *bs, int locked)
2142
{
2143
    BlockDriver *drv = bs->drv;
2144

    
2145
    bs->locked = locked;
2146
    if (drv && drv->bdrv_set_locked) {
2147
        drv->bdrv_set_locked(bs, locked);
2148
    }
2149
}
2150

    
2151
/* needed for generic scsi interface */
2152

    
2153
int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2154
{
2155
    BlockDriver *drv = bs->drv;
2156

    
2157
    if (drv && drv->bdrv_ioctl)
2158
        return drv->bdrv_ioctl(bs, req, buf);
2159
    return -ENOTSUP;
2160
}
2161

    
2162
BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2163
        unsigned long int req, void *buf,
2164
        BlockDriverCompletionFunc *cb, void *opaque)
2165
{
2166
    BlockDriver *drv = bs->drv;
2167

    
2168
    if (drv && drv->bdrv_aio_ioctl)
2169
        return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2170
    return NULL;
2171
}
2172

    
2173

    
2174

    
2175
void *qemu_blockalign(BlockDriverState *bs, size_t size)
2176
{
2177
    return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2178
}
2179

    
2180
void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2181
{
2182
    int64_t bitmap_size;
2183

    
2184
    bs->dirty_count = 0;
2185
    if (enable) {
2186
        if (!bs->dirty_bitmap) {
2187
            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2188
                    BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2189
            bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2190

    
2191
            bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2192
        }
2193
    } else {
2194
        if (bs->dirty_bitmap) {
2195
            qemu_free(bs->dirty_bitmap);
2196
            bs->dirty_bitmap = NULL;
2197
        }
2198
    }
2199
}
2200

    
2201
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2202
{
2203
    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2204

    
2205
    if (bs->dirty_bitmap &&
2206
        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2207
        return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2208
            (1 << (chunk % (sizeof(unsigned long) * 8)));
2209
    } else {
2210
        return 0;
2211
    }
2212
}
2213

    
2214
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2215
                      int nr_sectors)
2216
{
2217
    set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2218
}
2219

    
2220
int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2221
{
2222
    return bs->dirty_count;
2223
}