Statistics
| Branch: | Revision:

root / block.c @ 4dca4b63

History | View | Annotate | Download (61.6 kB)

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

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

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

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

    
58
BlockDriverState *bdrv_first;
59

    
60
static BlockDriver *first_drv;
61

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
298
    return drv;
299
}
300

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

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

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

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

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

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

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

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

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

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

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

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

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

    
392
        bdrv_delete(bs1);
393

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
658
ro_cleanup:
659

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

    
676
    return ret;
677
}
678

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

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

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

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

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

    
710
    len = bdrv_getlength(bs);
711

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

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

    
718
    return 0;
719
}
720

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
952
    bdrv_get_geometry(bs, &nb_sectors);
953

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1152
void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size)
1153
{
1154
    if (!bs->drv) {
1155
        buf[0] = '\0';
1156
    } else {
1157
        pstrcpy(buf, buf_size, bs->drv->format_name);
1158
    }
1159
}
1160

    
1161
void bdrv_iterate_format(void (*it)(void *opaque, const char *name),
1162
                         void *opaque)
1163
{
1164
    BlockDriver *drv;
1165

    
1166
    for (drv = first_drv; drv != NULL; drv = drv->next) {
1167
        it(opaque, drv->format_name);
1168
    }
1169
}
1170

    
1171
BlockDriverState *bdrv_find(const char *name)
1172
{
1173
    BlockDriverState *bs;
1174

    
1175
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1176
        if (!strcmp(name, bs->device_name))
1177
            return bs;
1178
    }
1179
    return NULL;
1180
}
1181

    
1182
void bdrv_iterate(void (*it)(void *opaque, BlockDriverState *bs), void *opaque)
1183
{
1184
    BlockDriverState *bs;
1185

    
1186
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1187
        it(opaque, bs);
1188
    }
1189
}
1190

    
1191
const char *bdrv_get_device_name(BlockDriverState *bs)
1192
{
1193
    return bs->device_name;
1194
}
1195

    
1196
void bdrv_flush(BlockDriverState *bs)
1197
{
1198
    if (bs->drv && bs->drv->bdrv_flush)
1199
        bs->drv->bdrv_flush(bs);
1200
}
1201

    
1202
void bdrv_flush_all(void)
1203
{
1204
    BlockDriverState *bs;
1205

    
1206
    for (bs = bdrv_first; bs != NULL; bs = bs->next)
1207
        if (bs->drv && !bdrv_is_read_only(bs) && 
1208
            (!bdrv_is_removable(bs) || bdrv_is_inserted(bs)))
1209
            bdrv_flush(bs);
1210
}
1211

    
1212
/*
1213
 * Returns true iff the specified sector is present in the disk image. Drivers
1214
 * not implementing the functionality are assumed to not support backing files,
1215
 * hence all their sectors are reported as allocated.
1216
 *
1217
 * 'pnum' is set to the number of sectors (including and immediately following
1218
 * the specified sector) that are known to be in the same
1219
 * allocated/unallocated state.
1220
 *
1221
 * 'nb_sectors' is the max value 'pnum' should be set to.
1222
 */
1223
int bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
1224
        int *pnum)
1225
{
1226
    int64_t n;
1227
    if (!bs->drv->bdrv_is_allocated) {
1228
        if (sector_num >= bs->total_sectors) {
1229
            *pnum = 0;
1230
            return 0;
1231
        }
1232
        n = bs->total_sectors - sector_num;
1233
        *pnum = (n < nb_sectors) ? (n) : (nb_sectors);
1234
        return 1;
1235
    }
1236
    return bs->drv->bdrv_is_allocated(bs, sector_num, nb_sectors, pnum);
1237
}
1238

    
1239
void bdrv_mon_event(const BlockDriverState *bdrv,
1240
                    BlockMonEventAction action, int is_read)
1241
{
1242
    QObject *data;
1243
    const char *action_str;
1244

    
1245
    switch (action) {
1246
    case BDRV_ACTION_REPORT:
1247
        action_str = "report";
1248
        break;
1249
    case BDRV_ACTION_IGNORE:
1250
        action_str = "ignore";
1251
        break;
1252
    case BDRV_ACTION_STOP:
1253
        action_str = "stop";
1254
        break;
1255
    default:
1256
        abort();
1257
    }
1258

    
1259
    data = qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1260
                              bdrv->device_name,
1261
                              action_str,
1262
                              is_read ? "read" : "write");
1263
    monitor_protocol_event(QEVENT_BLOCK_IO_ERROR, data);
1264

    
1265
    qobject_decref(data);
1266
}
1267

    
1268
static void bdrv_print_dict(QObject *obj, void *opaque)
1269
{
1270
    QDict *bs_dict;
1271
    Monitor *mon = opaque;
1272

    
1273
    bs_dict = qobject_to_qdict(obj);
1274

    
1275
    monitor_printf(mon, "%s: type=%s removable=%d",
1276
                        qdict_get_str(bs_dict, "device"),
1277
                        qdict_get_str(bs_dict, "type"),
1278
                        qdict_get_bool(bs_dict, "removable"));
1279

    
1280
    if (qdict_get_bool(bs_dict, "removable")) {
1281
        monitor_printf(mon, " locked=%d", qdict_get_bool(bs_dict, "locked"));
1282
    }
1283

    
1284
    if (qdict_haskey(bs_dict, "inserted")) {
1285
        QDict *qdict = qobject_to_qdict(qdict_get(bs_dict, "inserted"));
1286

    
1287
        monitor_printf(mon, " file=");
1288
        monitor_print_filename(mon, qdict_get_str(qdict, "file"));
1289
        if (qdict_haskey(qdict, "backing_file")) {
1290
            monitor_printf(mon, " backing_file=");
1291
            monitor_print_filename(mon, qdict_get_str(qdict, "backing_file"));
1292
        }
1293
        monitor_printf(mon, " ro=%d drv=%s encrypted=%d",
1294
                            qdict_get_bool(qdict, "ro"),
1295
                            qdict_get_str(qdict, "drv"),
1296
                            qdict_get_bool(qdict, "encrypted"));
1297
    } else {
1298
        monitor_printf(mon, " [not inserted]");
1299
    }
1300

    
1301
    monitor_printf(mon, "\n");
1302
}
1303

    
1304
void bdrv_info_print(Monitor *mon, const QObject *data)
1305
{
1306
    qlist_iter(qobject_to_qlist(data), bdrv_print_dict, mon);
1307
}
1308

    
1309
/**
1310
 * bdrv_info(): Block devices information
1311
 *
1312
 * Each block device information is stored in a QDict and the
1313
 * returned QObject is a QList of all devices.
1314
 *
1315
 * The QDict contains the following:
1316
 *
1317
 * - "device": device name
1318
 * - "type": device type
1319
 * - "removable": true if the device is removable, false otherwise
1320
 * - "locked": true if the device is locked, false otherwise
1321
 * - "inserted": only present if the device is inserted, it is a QDict
1322
 *    containing the following:
1323
 *          - "file": device file name
1324
 *          - "ro": true if read-only, false otherwise
1325
 *          - "drv": driver format name
1326
 *          - "backing_file": backing file name if one is used
1327
 *          - "encrypted": true if encrypted, false otherwise
1328
 *
1329
 * Example:
1330
 *
1331
 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1332
 *     "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1333
 *   { "device": "floppy0", "type": "floppy", "removable": true,
1334
 *     "locked": false } ]
1335
 */
1336
void bdrv_info(Monitor *mon, QObject **ret_data)
1337
{
1338
    QList *bs_list;
1339
    BlockDriverState *bs;
1340

    
1341
    bs_list = qlist_new();
1342

    
1343
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1344
        QObject *bs_obj;
1345
        const char *type = "unknown";
1346

    
1347
        switch(bs->type) {
1348
        case BDRV_TYPE_HD:
1349
            type = "hd";
1350
            break;
1351
        case BDRV_TYPE_CDROM:
1352
            type = "cdrom";
1353
            break;
1354
        case BDRV_TYPE_FLOPPY:
1355
            type = "floppy";
1356
            break;
1357
        }
1358

    
1359
        bs_obj = qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1360
                                    "'removable': %i, 'locked': %i }",
1361
                                    bs->device_name, type, bs->removable,
1362
                                    bs->locked);
1363

    
1364
        if (bs->drv) {
1365
            QObject *obj;
1366
            QDict *bs_dict = qobject_to_qdict(bs_obj);
1367

    
1368
            obj = qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1369
                                     "'encrypted': %i }",
1370
                                     bs->filename, bs->read_only,
1371
                                     bs->drv->format_name,
1372
                                     bdrv_is_encrypted(bs));
1373
            if (bs->backing_file[0] != '\0') {
1374
                QDict *qdict = qobject_to_qdict(obj);
1375
                qdict_put(qdict, "backing_file",
1376
                          qstring_from_str(bs->backing_file));
1377
            }
1378

    
1379
            qdict_put_obj(bs_dict, "inserted", obj);
1380
        }
1381
        qlist_append_obj(bs_list, bs_obj);
1382
    }
1383

    
1384
    *ret_data = QOBJECT(bs_list);
1385
}
1386

    
1387
static void bdrv_stats_iter(QObject *data, void *opaque)
1388
{
1389
    QDict *qdict;
1390
    Monitor *mon = opaque;
1391

    
1392
    qdict = qobject_to_qdict(data);
1393
    monitor_printf(mon, "%s:", qdict_get_str(qdict, "device"));
1394

    
1395
    qdict = qobject_to_qdict(qdict_get(qdict, "stats"));
1396
    monitor_printf(mon, " rd_bytes=%" PRId64
1397
                        " wr_bytes=%" PRId64
1398
                        " rd_operations=%" PRId64
1399
                        " wr_operations=%" PRId64
1400
                        "\n",
1401
                        qdict_get_int(qdict, "rd_bytes"),
1402
                        qdict_get_int(qdict, "wr_bytes"),
1403
                        qdict_get_int(qdict, "rd_operations"),
1404
                        qdict_get_int(qdict, "wr_operations"));
1405
}
1406

    
1407
void bdrv_stats_print(Monitor *mon, const QObject *data)
1408
{
1409
    qlist_iter(qobject_to_qlist(data), bdrv_stats_iter, mon);
1410
}
1411

    
1412
/**
1413
 * bdrv_info_stats(): show block device statistics
1414
 *
1415
 * Each device statistic information is stored in a QDict and
1416
 * the returned QObject is a QList of all devices.
1417
 *
1418
 * The QDict contains the following:
1419
 *
1420
 * - "device": device name
1421
 * - "stats": A QDict with the statistics information, it contains:
1422
 *     - "rd_bytes": bytes read
1423
 *     - "wr_bytes": bytes written
1424
 *     - "rd_operations": read operations
1425
 *     - "wr_operations": write operations
1426
 * 
1427
 * Example:
1428
 *
1429
 * [ { "device": "ide0-hd0",
1430
 *               "stats": { "rd_bytes": 512,
1431
 *                          "wr_bytes": 0,
1432
 *                          "rd_operations": 1,
1433
 *                          "wr_operations": 0 } },
1434
 *   { "device": "ide1-cd0",
1435
 *               "stats": { "rd_bytes": 0,
1436
 *                          "wr_bytes": 0,
1437
 *                          "rd_operations": 0,
1438
 *                          "wr_operations": 0 } } ]
1439
 */
1440
void bdrv_info_stats(Monitor *mon, QObject **ret_data)
1441
{
1442
    QObject *obj;
1443
    QList *devices;
1444
    BlockDriverState *bs;
1445

    
1446
    devices = qlist_new();
1447

    
1448
    for (bs = bdrv_first; bs != NULL; bs = bs->next) {
1449
        obj = qobject_from_jsonf("{ 'device': %s, 'stats': {"
1450
                                 "'rd_bytes': %" PRId64 ","
1451
                                 "'wr_bytes': %" PRId64 ","
1452
                                 "'rd_operations': %" PRId64 ","
1453
                                 "'wr_operations': %" PRId64
1454
                                 "} }",
1455
                                 bs->device_name,
1456
                                 bs->rd_bytes, bs->wr_bytes,
1457
                                 bs->rd_ops, bs->wr_ops);
1458
        qlist_append_obj(devices, obj);
1459
    }
1460

    
1461
    *ret_data = QOBJECT(devices);
1462
}
1463

    
1464
const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
1465
{
1466
    if (bs->backing_hd && bs->backing_hd->encrypted)
1467
        return bs->backing_file;
1468
    else if (bs->encrypted)
1469
        return bs->filename;
1470
    else
1471
        return NULL;
1472
}
1473

    
1474
void bdrv_get_backing_filename(BlockDriverState *bs,
1475
                               char *filename, int filename_size)
1476
{
1477
    if (!bs->backing_file) {
1478
        pstrcpy(filename, filename_size, "");
1479
    } else {
1480
        pstrcpy(filename, filename_size, bs->backing_file);
1481
    }
1482
}
1483

    
1484
int bdrv_write_compressed(BlockDriverState *bs, int64_t sector_num,
1485
                          const uint8_t *buf, int nb_sectors)
1486
{
1487
    BlockDriver *drv = bs->drv;
1488
    if (!drv)
1489
        return -ENOMEDIUM;
1490
    if (!drv->bdrv_write_compressed)
1491
        return -ENOTSUP;
1492
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1493
        return -EIO;
1494

    
1495
    if (bs->dirty_bitmap) {
1496
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1497
    }
1498

    
1499
    return drv->bdrv_write_compressed(bs, sector_num, buf, nb_sectors);
1500
}
1501

    
1502
int bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
1503
{
1504
    BlockDriver *drv = bs->drv;
1505
    if (!drv)
1506
        return -ENOMEDIUM;
1507
    if (!drv->bdrv_get_info)
1508
        return -ENOTSUP;
1509
    memset(bdi, 0, sizeof(*bdi));
1510
    return drv->bdrv_get_info(bs, bdi);
1511
}
1512

    
1513
int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
1514
                      int64_t pos, int size)
1515
{
1516
    BlockDriver *drv = bs->drv;
1517
    if (!drv)
1518
        return -ENOMEDIUM;
1519
    if (!drv->bdrv_save_vmstate)
1520
        return -ENOTSUP;
1521
    return drv->bdrv_save_vmstate(bs, buf, pos, size);
1522
}
1523

    
1524
int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
1525
                      int64_t pos, int size)
1526
{
1527
    BlockDriver *drv = bs->drv;
1528
    if (!drv)
1529
        return -ENOMEDIUM;
1530
    if (!drv->bdrv_load_vmstate)
1531
        return -ENOTSUP;
1532
    return drv->bdrv_load_vmstate(bs, buf, pos, size);
1533
}
1534

    
1535
/**************************************************************/
1536
/* handling of snapshots */
1537

    
1538
int bdrv_snapshot_create(BlockDriverState *bs,
1539
                         QEMUSnapshotInfo *sn_info)
1540
{
1541
    BlockDriver *drv = bs->drv;
1542
    if (!drv)
1543
        return -ENOMEDIUM;
1544
    if (!drv->bdrv_snapshot_create)
1545
        return -ENOTSUP;
1546
    return drv->bdrv_snapshot_create(bs, sn_info);
1547
}
1548

    
1549
int bdrv_snapshot_goto(BlockDriverState *bs,
1550
                       const char *snapshot_id)
1551
{
1552
    BlockDriver *drv = bs->drv;
1553
    if (!drv)
1554
        return -ENOMEDIUM;
1555
    if (!drv->bdrv_snapshot_goto)
1556
        return -ENOTSUP;
1557
    return drv->bdrv_snapshot_goto(bs, snapshot_id);
1558
}
1559

    
1560
int bdrv_snapshot_delete(BlockDriverState *bs, const char *snapshot_id)
1561
{
1562
    BlockDriver *drv = bs->drv;
1563
    if (!drv)
1564
        return -ENOMEDIUM;
1565
    if (!drv->bdrv_snapshot_delete)
1566
        return -ENOTSUP;
1567
    return drv->bdrv_snapshot_delete(bs, snapshot_id);
1568
}
1569

    
1570
int bdrv_snapshot_list(BlockDriverState *bs,
1571
                       QEMUSnapshotInfo **psn_info)
1572
{
1573
    BlockDriver *drv = bs->drv;
1574
    if (!drv)
1575
        return -ENOMEDIUM;
1576
    if (!drv->bdrv_snapshot_list)
1577
        return -ENOTSUP;
1578
    return drv->bdrv_snapshot_list(bs, psn_info);
1579
}
1580

    
1581
#define NB_SUFFIXES 4
1582

    
1583
char *get_human_readable_size(char *buf, int buf_size, int64_t size)
1584
{
1585
    static const char suffixes[NB_SUFFIXES] = "KMGT";
1586
    int64_t base;
1587
    int i;
1588

    
1589
    if (size <= 999) {
1590
        snprintf(buf, buf_size, "%" PRId64, size);
1591
    } else {
1592
        base = 1024;
1593
        for(i = 0; i < NB_SUFFIXES; i++) {
1594
            if (size < (10 * base)) {
1595
                snprintf(buf, buf_size, "%0.1f%c",
1596
                         (double)size / base,
1597
                         suffixes[i]);
1598
                break;
1599
            } else if (size < (1000 * base) || i == (NB_SUFFIXES - 1)) {
1600
                snprintf(buf, buf_size, "%" PRId64 "%c",
1601
                         ((size + (base >> 1)) / base),
1602
                         suffixes[i]);
1603
                break;
1604
            }
1605
            base = base * 1024;
1606
        }
1607
    }
1608
    return buf;
1609
}
1610

    
1611
char *bdrv_snapshot_dump(char *buf, int buf_size, QEMUSnapshotInfo *sn)
1612
{
1613
    char buf1[128], date_buf[128], clock_buf[128];
1614
#ifdef _WIN32
1615
    struct tm *ptm;
1616
#else
1617
    struct tm tm;
1618
#endif
1619
    time_t ti;
1620
    int64_t secs;
1621

    
1622
    if (!sn) {
1623
        snprintf(buf, buf_size,
1624
                 "%-10s%-20s%7s%20s%15s",
1625
                 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1626
    } else {
1627
        ti = sn->date_sec;
1628
#ifdef _WIN32
1629
        ptm = localtime(&ti);
1630
        strftime(date_buf, sizeof(date_buf),
1631
                 "%Y-%m-%d %H:%M:%S", ptm);
1632
#else
1633
        localtime_r(&ti, &tm);
1634
        strftime(date_buf, sizeof(date_buf),
1635
                 "%Y-%m-%d %H:%M:%S", &tm);
1636
#endif
1637
        secs = sn->vm_clock_nsec / 1000000000;
1638
        snprintf(clock_buf, sizeof(clock_buf),
1639
                 "%02d:%02d:%02d.%03d",
1640
                 (int)(secs / 3600),
1641
                 (int)((secs / 60) % 60),
1642
                 (int)(secs % 60),
1643
                 (int)((sn->vm_clock_nsec / 1000000) % 1000));
1644
        snprintf(buf, buf_size,
1645
                 "%-10s%-20s%7s%20s%15s",
1646
                 sn->id_str, sn->name,
1647
                 get_human_readable_size(buf1, sizeof(buf1), sn->vm_state_size),
1648
                 date_buf,
1649
                 clock_buf);
1650
    }
1651
    return buf;
1652
}
1653

    
1654

    
1655
/**************************************************************/
1656
/* async I/Os */
1657

    
1658
BlockDriverAIOCB *bdrv_aio_readv(BlockDriverState *bs, int64_t sector_num,
1659
                                 QEMUIOVector *qiov, int nb_sectors,
1660
                                 BlockDriverCompletionFunc *cb, void *opaque)
1661
{
1662
    BlockDriver *drv = bs->drv;
1663
    BlockDriverAIOCB *ret;
1664

    
1665
    if (!drv)
1666
        return NULL;
1667
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1668
        return NULL;
1669

    
1670
    ret = drv->bdrv_aio_readv(bs, sector_num, qiov, nb_sectors,
1671
                              cb, opaque);
1672

    
1673
    if (ret) {
1674
        /* Update stats even though technically transfer has not happened. */
1675
        bs->rd_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1676
        bs->rd_ops ++;
1677
    }
1678

    
1679
    return ret;
1680
}
1681

    
1682
BlockDriverAIOCB *bdrv_aio_writev(BlockDriverState *bs, int64_t sector_num,
1683
                                  QEMUIOVector *qiov, int nb_sectors,
1684
                                  BlockDriverCompletionFunc *cb, void *opaque)
1685
{
1686
    BlockDriver *drv = bs->drv;
1687
    BlockDriverAIOCB *ret;
1688

    
1689
    if (!drv)
1690
        return NULL;
1691
    if (bs->read_only)
1692
        return NULL;
1693
    if (bdrv_check_request(bs, sector_num, nb_sectors))
1694
        return NULL;
1695

    
1696
    if (bs->dirty_bitmap) {
1697
        set_dirty_bitmap(bs, sector_num, nb_sectors, 1);
1698
    }
1699

    
1700
    ret = drv->bdrv_aio_writev(bs, sector_num, qiov, nb_sectors,
1701
                               cb, opaque);
1702

    
1703
    if (ret) {
1704
        /* Update stats even though technically transfer has not happened. */
1705
        bs->wr_bytes += (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
1706
        bs->wr_ops ++;
1707
    }
1708

    
1709
    return ret;
1710
}
1711

    
1712

    
1713
typedef struct MultiwriteCB {
1714
    int error;
1715
    int num_requests;
1716
    int num_callbacks;
1717
    struct {
1718
        BlockDriverCompletionFunc *cb;
1719
        void *opaque;
1720
        QEMUIOVector *free_qiov;
1721
        void *free_buf;
1722
    } callbacks[];
1723
} MultiwriteCB;
1724

    
1725
static void multiwrite_user_cb(MultiwriteCB *mcb)
1726
{
1727
    int i;
1728

    
1729
    for (i = 0; i < mcb->num_callbacks; i++) {
1730
        mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
1731
        qemu_free(mcb->callbacks[i].free_qiov);
1732
        qemu_vfree(mcb->callbacks[i].free_buf);
1733
    }
1734
}
1735

    
1736
static void multiwrite_cb(void *opaque, int ret)
1737
{
1738
    MultiwriteCB *mcb = opaque;
1739

    
1740
    if (ret < 0) {
1741
        mcb->error = ret;
1742
        multiwrite_user_cb(mcb);
1743
    }
1744

    
1745
    mcb->num_requests--;
1746
    if (mcb->num_requests == 0) {
1747
        if (mcb->error == 0) {
1748
            multiwrite_user_cb(mcb);
1749
        }
1750
        qemu_free(mcb);
1751
    }
1752
}
1753

    
1754
static int multiwrite_req_compare(const void *a, const void *b)
1755
{
1756
    return (((BlockRequest*) a)->sector - ((BlockRequest*) b)->sector);
1757
}
1758

    
1759
/*
1760
 * Takes a bunch of requests and tries to merge them. Returns the number of
1761
 * requests that remain after merging.
1762
 */
1763
static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
1764
    int num_reqs, MultiwriteCB *mcb)
1765
{
1766
    int i, outidx;
1767

    
1768
    // Sort requests by start sector
1769
    qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
1770

    
1771
    // Check if adjacent requests touch the same clusters. If so, combine them,
1772
    // filling up gaps with zero sectors.
1773
    outidx = 0;
1774
    for (i = 1; i < num_reqs; i++) {
1775
        int merge = 0;
1776
        int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
1777

    
1778
        // This handles the cases that are valid for all block drivers, namely
1779
        // exactly sequential writes and overlapping writes.
1780
        if (reqs[i].sector <= oldreq_last) {
1781
            merge = 1;
1782
        }
1783

    
1784
        // The block driver may decide that it makes sense to combine requests
1785
        // even if there is a gap of some sectors between them. In this case,
1786
        // the gap is filled with zeros (therefore only applicable for yet
1787
        // unused space in format like qcow2).
1788
        if (!merge && bs->drv->bdrv_merge_requests) {
1789
            merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
1790
        }
1791

    
1792
        if (reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1 > IOV_MAX) {
1793
            merge = 0;
1794
        }
1795

    
1796
        if (merge) {
1797
            size_t size;
1798
            QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
1799
            qemu_iovec_init(qiov,
1800
                reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
1801

    
1802
            // Add the first request to the merged one. If the requests are
1803
            // overlapping, drop the last sectors of the first request.
1804
            size = (reqs[i].sector - reqs[outidx].sector) << 9;
1805
            qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
1806

    
1807
            // We might need to add some zeros between the two requests
1808
            if (reqs[i].sector > oldreq_last) {
1809
                size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
1810
                uint8_t *buf = qemu_blockalign(bs, zero_bytes);
1811
                memset(buf, 0, zero_bytes);
1812
                qemu_iovec_add(qiov, buf, zero_bytes);
1813
                mcb->callbacks[i].free_buf = buf;
1814
            }
1815

    
1816
            // Add the second request
1817
            qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
1818

    
1819
            reqs[outidx].nb_sectors += reqs[i].nb_sectors;
1820
            reqs[outidx].qiov = qiov;
1821

    
1822
            mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
1823
        } else {
1824
            outidx++;
1825
            reqs[outidx].sector     = reqs[i].sector;
1826
            reqs[outidx].nb_sectors = reqs[i].nb_sectors;
1827
            reqs[outidx].qiov       = reqs[i].qiov;
1828
        }
1829
    }
1830

    
1831
    return outidx + 1;
1832
}
1833

    
1834
/*
1835
 * Submit multiple AIO write requests at once.
1836
 *
1837
 * On success, the function returns 0 and all requests in the reqs array have
1838
 * been submitted. In error case this function returns -1, and any of the
1839
 * requests may or may not be submitted yet. In particular, this means that the
1840
 * callback will be called for some of the requests, for others it won't. The
1841
 * caller must check the error field of the BlockRequest to wait for the right
1842
 * callbacks (if error != 0, no callback will be called).
1843
 *
1844
 * The implementation may modify the contents of the reqs array, e.g. to merge
1845
 * requests. However, the fields opaque and error are left unmodified as they
1846
 * are used to signal failure for a single request to the caller.
1847
 */
1848
int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
1849
{
1850
    BlockDriverAIOCB *acb;
1851
    MultiwriteCB *mcb;
1852
    int i;
1853

    
1854
    if (num_reqs == 0) {
1855
        return 0;
1856
    }
1857

    
1858
    // Create MultiwriteCB structure
1859
    mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
1860
    mcb->num_requests = 0;
1861
    mcb->num_callbacks = num_reqs;
1862

    
1863
    for (i = 0; i < num_reqs; i++) {
1864
        mcb->callbacks[i].cb = reqs[i].cb;
1865
        mcb->callbacks[i].opaque = reqs[i].opaque;
1866
    }
1867

    
1868
    // Check for mergable requests
1869
    num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
1870

    
1871
    // Run the aio requests
1872
    for (i = 0; i < num_reqs; i++) {
1873
        acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
1874
            reqs[i].nb_sectors, multiwrite_cb, mcb);
1875

    
1876
        if (acb == NULL) {
1877
            // We can only fail the whole thing if no request has been
1878
            // submitted yet. Otherwise we'll wait for the submitted AIOs to
1879
            // complete and report the error in the callback.
1880
            if (mcb->num_requests == 0) {
1881
                reqs[i].error = EIO;
1882
                goto fail;
1883
            } else {
1884
                mcb->error = EIO;
1885
                break;
1886
            }
1887
        } else {
1888
            mcb->num_requests++;
1889
        }
1890
    }
1891

    
1892
    return 0;
1893

    
1894
fail:
1895
    free(mcb);
1896
    return -1;
1897
}
1898

    
1899
BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
1900
        BlockDriverCompletionFunc *cb, void *opaque)
1901
{
1902
    BlockDriver *drv = bs->drv;
1903

    
1904
    if (!drv)
1905
        return NULL;
1906
    return drv->bdrv_aio_flush(bs, cb, opaque);
1907
}
1908

    
1909
void bdrv_aio_cancel(BlockDriverAIOCB *acb)
1910
{
1911
    acb->pool->cancel(acb);
1912
}
1913

    
1914

    
1915
/**************************************************************/
1916
/* async block device emulation */
1917

    
1918
typedef struct BlockDriverAIOCBSync {
1919
    BlockDriverAIOCB common;
1920
    QEMUBH *bh;
1921
    int ret;
1922
    /* vector translation state */
1923
    QEMUIOVector *qiov;
1924
    uint8_t *bounce;
1925
    int is_write;
1926
} BlockDriverAIOCBSync;
1927

    
1928
static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
1929
{
1930
    BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
1931
    qemu_bh_delete(acb->bh);
1932
    acb->bh = NULL;
1933
    qemu_aio_release(acb);
1934
}
1935

    
1936
static AIOPool bdrv_em_aio_pool = {
1937
    .aiocb_size         = sizeof(BlockDriverAIOCBSync),
1938
    .cancel             = bdrv_aio_cancel_em,
1939
};
1940

    
1941
static void bdrv_aio_bh_cb(void *opaque)
1942
{
1943
    BlockDriverAIOCBSync *acb = opaque;
1944

    
1945
    if (!acb->is_write)
1946
        qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
1947
    qemu_vfree(acb->bounce);
1948
    acb->common.cb(acb->common.opaque, acb->ret);
1949
    qemu_bh_delete(acb->bh);
1950
    acb->bh = NULL;
1951
    qemu_aio_release(acb);
1952
}
1953

    
1954
static BlockDriverAIOCB *bdrv_aio_rw_vector(BlockDriverState *bs,
1955
                                            int64_t sector_num,
1956
                                            QEMUIOVector *qiov,
1957
                                            int nb_sectors,
1958
                                            BlockDriverCompletionFunc *cb,
1959
                                            void *opaque,
1960
                                            int is_write)
1961

    
1962
{
1963
    BlockDriverAIOCBSync *acb;
1964

    
1965
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
1966
    acb->is_write = is_write;
1967
    acb->qiov = qiov;
1968
    acb->bounce = qemu_blockalign(bs, qiov->size);
1969

    
1970
    if (!acb->bh)
1971
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
1972

    
1973
    if (is_write) {
1974
        qemu_iovec_to_buffer(acb->qiov, acb->bounce);
1975
        acb->ret = bdrv_write(bs, sector_num, acb->bounce, nb_sectors);
1976
    } else {
1977
        acb->ret = bdrv_read(bs, sector_num, acb->bounce, nb_sectors);
1978
    }
1979

    
1980
    qemu_bh_schedule(acb->bh);
1981

    
1982
    return &acb->common;
1983
}
1984

    
1985
static BlockDriverAIOCB *bdrv_aio_readv_em(BlockDriverState *bs,
1986
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1987
        BlockDriverCompletionFunc *cb, void *opaque)
1988
{
1989
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 0);
1990
}
1991

    
1992
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
1993
        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
1994
        BlockDriverCompletionFunc *cb, void *opaque)
1995
{
1996
    return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
1997
}
1998

    
1999
static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
2000
        BlockDriverCompletionFunc *cb, void *opaque)
2001
{
2002
    BlockDriverAIOCBSync *acb;
2003

    
2004
    acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
2005
    acb->is_write = 1; /* don't bounce in the completion hadler */
2006
    acb->qiov = NULL;
2007
    acb->bounce = NULL;
2008
    acb->ret = 0;
2009

    
2010
    if (!acb->bh)
2011
        acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
2012

    
2013
    bdrv_flush(bs);
2014
    qemu_bh_schedule(acb->bh);
2015
    return &acb->common;
2016
}
2017

    
2018
/**************************************************************/
2019
/* sync block device emulation */
2020

    
2021
static void bdrv_rw_em_cb(void *opaque, int ret)
2022
{
2023
    *(int *)opaque = ret;
2024
}
2025

    
2026
#define NOT_DONE 0x7fffffff
2027

    
2028
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
2029
                        uint8_t *buf, int nb_sectors)
2030
{
2031
    int async_ret;
2032
    BlockDriverAIOCB *acb;
2033
    struct iovec iov;
2034
    QEMUIOVector qiov;
2035

    
2036
    async_context_push();
2037

    
2038
    async_ret = NOT_DONE;
2039
    iov.iov_base = (void *)buf;
2040
    iov.iov_len = nb_sectors * 512;
2041
    qemu_iovec_init_external(&qiov, &iov, 1);
2042
    acb = bdrv_aio_readv(bs, sector_num, &qiov, nb_sectors,
2043
        bdrv_rw_em_cb, &async_ret);
2044
    if (acb == NULL) {
2045
        async_ret = -1;
2046
        goto fail;
2047
    }
2048

    
2049
    while (async_ret == NOT_DONE) {
2050
        qemu_aio_wait();
2051
    }
2052

    
2053

    
2054
fail:
2055
    async_context_pop();
2056
    return async_ret;
2057
}
2058

    
2059
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
2060
                         const uint8_t *buf, int nb_sectors)
2061
{
2062
    int async_ret;
2063
    BlockDriverAIOCB *acb;
2064
    struct iovec iov;
2065
    QEMUIOVector qiov;
2066

    
2067
    async_context_push();
2068

    
2069
    async_ret = NOT_DONE;
2070
    iov.iov_base = (void *)buf;
2071
    iov.iov_len = nb_sectors * 512;
2072
    qemu_iovec_init_external(&qiov, &iov, 1);
2073
    acb = bdrv_aio_writev(bs, sector_num, &qiov, nb_sectors,
2074
        bdrv_rw_em_cb, &async_ret);
2075
    if (acb == NULL) {
2076
        async_ret = -1;
2077
        goto fail;
2078
    }
2079
    while (async_ret == NOT_DONE) {
2080
        qemu_aio_wait();
2081
    }
2082

    
2083
fail:
2084
    async_context_pop();
2085
    return async_ret;
2086
}
2087

    
2088
void bdrv_init(void)
2089
{
2090
    module_call_init(MODULE_INIT_BLOCK);
2091
}
2092

    
2093
void bdrv_init_with_whitelist(void)
2094
{
2095
    use_bdrv_whitelist = 1;
2096
    bdrv_init();
2097
}
2098

    
2099
void *qemu_aio_get(AIOPool *pool, BlockDriverState *bs,
2100
                   BlockDriverCompletionFunc *cb, void *opaque)
2101
{
2102
    BlockDriverAIOCB *acb;
2103

    
2104
    if (pool->free_aiocb) {
2105
        acb = pool->free_aiocb;
2106
        pool->free_aiocb = acb->next;
2107
    } else {
2108
        acb = qemu_mallocz(pool->aiocb_size);
2109
        acb->pool = pool;
2110
    }
2111
    acb->bs = bs;
2112
    acb->cb = cb;
2113
    acb->opaque = opaque;
2114
    return acb;
2115
}
2116

    
2117
void qemu_aio_release(void *p)
2118
{
2119
    BlockDriverAIOCB *acb = (BlockDriverAIOCB *)p;
2120
    AIOPool *pool = acb->pool;
2121
    acb->next = pool->free_aiocb;
2122
    pool->free_aiocb = acb;
2123
}
2124

    
2125
/**************************************************************/
2126
/* removable device support */
2127

    
2128
/**
2129
 * Return TRUE if the media is present
2130
 */
2131
int bdrv_is_inserted(BlockDriverState *bs)
2132
{
2133
    BlockDriver *drv = bs->drv;
2134
    int ret;
2135
    if (!drv)
2136
        return 0;
2137
    if (!drv->bdrv_is_inserted)
2138
        return 1;
2139
    ret = drv->bdrv_is_inserted(bs);
2140
    return ret;
2141
}
2142

    
2143
/**
2144
 * Return TRUE if the media changed since the last call to this
2145
 * function. It is currently only used for floppy disks
2146
 */
2147
int bdrv_media_changed(BlockDriverState *bs)
2148
{
2149
    BlockDriver *drv = bs->drv;
2150
    int ret;
2151

    
2152
    if (!drv || !drv->bdrv_media_changed)
2153
        ret = -ENOTSUP;
2154
    else
2155
        ret = drv->bdrv_media_changed(bs);
2156
    if (ret == -ENOTSUP)
2157
        ret = bs->media_changed;
2158
    bs->media_changed = 0;
2159
    return ret;
2160
}
2161

    
2162
/**
2163
 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2164
 */
2165
int bdrv_eject(BlockDriverState *bs, int eject_flag)
2166
{
2167
    BlockDriver *drv = bs->drv;
2168
    int ret;
2169

    
2170
    if (bs->locked) {
2171
        return -EBUSY;
2172
    }
2173

    
2174
    if (!drv || !drv->bdrv_eject) {
2175
        ret = -ENOTSUP;
2176
    } else {
2177
        ret = drv->bdrv_eject(bs, eject_flag);
2178
    }
2179
    if (ret == -ENOTSUP) {
2180
        if (eject_flag)
2181
            bdrv_close(bs);
2182
        ret = 0;
2183
    }
2184

    
2185
    return ret;
2186
}
2187

    
2188
int bdrv_is_locked(BlockDriverState *bs)
2189
{
2190
    return bs->locked;
2191
}
2192

    
2193
/**
2194
 * Lock or unlock the media (if it is locked, the user won't be able
2195
 * to eject it manually).
2196
 */
2197
void bdrv_set_locked(BlockDriverState *bs, int locked)
2198
{
2199
    BlockDriver *drv = bs->drv;
2200

    
2201
    bs->locked = locked;
2202
    if (drv && drv->bdrv_set_locked) {
2203
        drv->bdrv_set_locked(bs, locked);
2204
    }
2205
}
2206

    
2207
/* needed for generic scsi interface */
2208

    
2209
int bdrv_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
2210
{
2211
    BlockDriver *drv = bs->drv;
2212

    
2213
    if (drv && drv->bdrv_ioctl)
2214
        return drv->bdrv_ioctl(bs, req, buf);
2215
    return -ENOTSUP;
2216
}
2217

    
2218
BlockDriverAIOCB *bdrv_aio_ioctl(BlockDriverState *bs,
2219
        unsigned long int req, void *buf,
2220
        BlockDriverCompletionFunc *cb, void *opaque)
2221
{
2222
    BlockDriver *drv = bs->drv;
2223

    
2224
    if (drv && drv->bdrv_aio_ioctl)
2225
        return drv->bdrv_aio_ioctl(bs, req, buf, cb, opaque);
2226
    return NULL;
2227
}
2228

    
2229

    
2230

    
2231
void *qemu_blockalign(BlockDriverState *bs, size_t size)
2232
{
2233
    return qemu_memalign((bs && bs->buffer_alignment) ? bs->buffer_alignment : 512, size);
2234
}
2235

    
2236
void bdrv_set_dirty_tracking(BlockDriverState *bs, int enable)
2237
{
2238
    int64_t bitmap_size;
2239

    
2240
    bs->dirty_count = 0;
2241
    if (enable) {
2242
        if (!bs->dirty_bitmap) {
2243
            bitmap_size = (bdrv_getlength(bs) >> BDRV_SECTOR_BITS) +
2244
                    BDRV_SECTORS_PER_DIRTY_CHUNK * 8 - 1;
2245
            bitmap_size /= BDRV_SECTORS_PER_DIRTY_CHUNK * 8;
2246

    
2247
            bs->dirty_bitmap = qemu_mallocz(bitmap_size);
2248
        }
2249
    } else {
2250
        if (bs->dirty_bitmap) {
2251
            qemu_free(bs->dirty_bitmap);
2252
            bs->dirty_bitmap = NULL;
2253
        }
2254
    }
2255
}
2256

    
2257
int bdrv_get_dirty(BlockDriverState *bs, int64_t sector)
2258
{
2259
    int64_t chunk = sector / (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK;
2260

    
2261
    if (bs->dirty_bitmap &&
2262
        (sector << BDRV_SECTOR_BITS) < bdrv_getlength(bs)) {
2263
        return bs->dirty_bitmap[chunk / (sizeof(unsigned long) * 8)] &
2264
            (1 << (chunk % (sizeof(unsigned long) * 8)));
2265
    } else {
2266
        return 0;
2267
    }
2268
}
2269

    
2270
void bdrv_reset_dirty(BlockDriverState *bs, int64_t cur_sector,
2271
                      int nr_sectors)
2272
{
2273
    set_dirty_bitmap(bs, cur_sector, nr_sectors, 0);
2274
}
2275

    
2276
int64_t bdrv_get_dirty_count(BlockDriverState *bs)
2277
{
2278
    return bs->dirty_count;
2279
}