Statistics
| Branch: | Revision:

root / block / vpc.c @ 69152c09

History | View | Annotate | Download (24.7 kB)

1
/*
2
 * Block driver for Connectix / Microsoft Virtual PC images
3
 *
4
 * Copyright (c) 2005 Alex Beregszaszi
5
 * Copyright (c) 2009 Kevin Wolf <kwolf@suse.de>
6
 *
7
 * Permission is hereby granted, free of charge, to any person obtaining a copy
8
 * of this software and associated documentation files (the "Software"), to deal
9
 * in the Software without restriction, including without limitation the rights
10
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11
 * copies of the Software, and to permit persons to whom the Software is
12
 * furnished to do so, subject to the following conditions:
13
 *
14
 * The above copyright notice and this permission notice shall be included in
15
 * all copies or substantial portions of the Software.
16
 *
17
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23
 * THE SOFTWARE.
24
 */
25
#include "qemu-common.h"
26
#include "block/block_int.h"
27
#include "qemu/module.h"
28
#include "migration/migration.h"
29
#if defined(CONFIG_UUID)
30
#include <uuid/uuid.h>
31
#endif
32

    
33
/**************************************************************/
34

    
35
#define HEADER_SIZE 512
36

    
37
//#define CACHE
38

    
39
enum vhd_type {
40
    VHD_FIXED           = 2,
41
    VHD_DYNAMIC         = 3,
42
    VHD_DIFFERENCING    = 4,
43
};
44

    
45
// Seconds since Jan 1, 2000 0:00:00 (UTC)
46
#define VHD_TIMESTAMP_BASE 946684800
47

    
48
// always big-endian
49
typedef struct vhd_footer {
50
    char        creator[8]; // "conectix"
51
    uint32_t    features;
52
    uint32_t    version;
53

    
54
    // Offset of next header structure, 0xFFFFFFFF if none
55
    uint64_t    data_offset;
56

    
57
    // Seconds since Jan 1, 2000 0:00:00 (UTC)
58
    uint32_t    timestamp;
59

    
60
    char        creator_app[4]; // "vpc "
61
    uint16_t    major;
62
    uint16_t    minor;
63
    char        creator_os[4]; // "Wi2k"
64

    
65
    uint64_t    orig_size;
66
    uint64_t    size;
67

    
68
    uint16_t    cyls;
69
    uint8_t     heads;
70
    uint8_t     secs_per_cyl;
71

    
72
    uint32_t    type;
73

    
74
    // Checksum of the Hard Disk Footer ("one's complement of the sum of all
75
    // the bytes in the footer without the checksum field")
76
    uint32_t    checksum;
77

    
78
    // UUID used to identify a parent hard disk (backing file)
79
    uint8_t     uuid[16];
80

    
81
    uint8_t     in_saved_state;
82
} QEMU_PACKED VHDFooter;
83

    
84
typedef struct vhd_dyndisk_header {
85
    char        magic[8]; // "cxsparse"
86

    
87
    // Offset of next header structure, 0xFFFFFFFF if none
88
    uint64_t    data_offset;
89

    
90
    // Offset of the Block Allocation Table (BAT)
91
    uint64_t    table_offset;
92

    
93
    uint32_t    version;
94
    uint32_t    max_table_entries; // 32bit/entry
95

    
96
    // 2 MB by default, must be a power of two
97
    uint32_t    block_size;
98

    
99
    uint32_t    checksum;
100
    uint8_t     parent_uuid[16];
101
    uint32_t    parent_timestamp;
102
    uint32_t    reserved;
103

    
104
    // Backing file name (in UTF-16)
105
    uint8_t     parent_name[512];
106

    
107
    struct {
108
        uint32_t    platform;
109
        uint32_t    data_space;
110
        uint32_t    data_length;
111
        uint32_t    reserved;
112
        uint64_t    data_offset;
113
    } parent_locator[8];
114
} QEMU_PACKED VHDDynDiskHeader;
115

    
116
typedef struct BDRVVPCState {
117
    CoMutex lock;
118
    uint8_t footer_buf[HEADER_SIZE];
119
    uint64_t free_data_block_offset;
120
    int max_table_entries;
121
    uint32_t *pagetable;
122
    uint64_t bat_offset;
123
    uint64_t last_bitmap_offset;
124

    
125
    uint32_t block_size;
126
    uint32_t bitmap_size;
127

    
128
#ifdef CACHE
129
    uint8_t *pageentry_u8;
130
    uint32_t *pageentry_u32;
131
    uint16_t *pageentry_u16;
132

    
133
    uint64_t last_bitmap;
134
#endif
135

    
136
    Error *migration_blocker;
137
} BDRVVPCState;
138

    
139
static uint32_t vpc_checksum(uint8_t* buf, size_t size)
140
{
141
    uint32_t res = 0;
142
    int i;
143

    
144
    for (i = 0; i < size; i++)
145
        res += buf[i];
146

    
147
    return ~res;
148
}
149

    
150

    
151
static int vpc_probe(const uint8_t *buf, int buf_size, const char *filename)
152
{
153
    if (buf_size >= 8 && !strncmp((char *)buf, "conectix", 8))
154
        return 100;
155
    return 0;
156
}
157

    
158
static int vpc_open(BlockDriverState *bs, QDict *options, int flags,
159
                    Error **errp)
160
{
161
    BDRVVPCState *s = bs->opaque;
162
    int i;
163
    VHDFooter *footer;
164
    VHDDynDiskHeader *dyndisk_header;
165
    uint8_t buf[HEADER_SIZE];
166
    uint32_t checksum;
167
    int disk_type = VHD_DYNAMIC;
168
    int ret;
169

    
170
    ret = bdrv_pread(bs->file, 0, s->footer_buf, HEADER_SIZE);
171
    if (ret < 0) {
172
        goto fail;
173
    }
174

    
175
    footer = (VHDFooter *) s->footer_buf;
176
    if (strncmp(footer->creator, "conectix", 8)) {
177
        int64_t offset = bdrv_getlength(bs->file);
178
        if (offset < 0) {
179
            ret = offset;
180
            goto fail;
181
        } else if (offset < HEADER_SIZE) {
182
            ret = -EINVAL;
183
            goto fail;
184
        }
185

    
186
        /* If a fixed disk, the footer is found only at the end of the file */
187
        ret = bdrv_pread(bs->file, offset-HEADER_SIZE, s->footer_buf,
188
                         HEADER_SIZE);
189
        if (ret < 0) {
190
            goto fail;
191
        }
192
        if (strncmp(footer->creator, "conectix", 8)) {
193
            ret = -EMEDIUMTYPE;
194
            goto fail;
195
        }
196
        disk_type = VHD_FIXED;
197
    }
198

    
199
    checksum = be32_to_cpu(footer->checksum);
200
    footer->checksum = 0;
201
    if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum)
202
        fprintf(stderr, "block-vpc: The header checksum of '%s' is "
203
            "incorrect.\n", bs->filename);
204

    
205
    /* Write 'checksum' back to footer, or else will leave it with zero. */
206
    footer->checksum = be32_to_cpu(checksum);
207

    
208
    // The visible size of a image in Virtual PC depends on the geometry
209
    // rather than on the size stored in the footer (the size in the footer
210
    // is too large usually)
211
    bs->total_sectors = (int64_t)
212
        be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl;
213

    
214
    /* images created with disk2vhd report a far higher virtual size
215
     * than expected with the cyls * heads * sectors_per_cyl formula.
216
     * use the footer->size instead if the image was created with
217
     * disk2vhd.
218
     */
219
    if (!strncmp(footer->creator_app, "d2v", 4)) {
220
        bs->total_sectors = be64_to_cpu(footer->size) / BDRV_SECTOR_SIZE;
221
    }
222

    
223
    /* Allow a maximum disk size of approximately 2 TB */
224
    if (bs->total_sectors >= 65535LL * 255 * 255) {
225
        ret = -EFBIG;
226
        goto fail;
227
    }
228

    
229
    if (disk_type == VHD_DYNAMIC) {
230
        ret = bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf,
231
                         HEADER_SIZE);
232
        if (ret < 0) {
233
            goto fail;
234
        }
235

    
236
        dyndisk_header = (VHDDynDiskHeader *) buf;
237

    
238
        if (strncmp(dyndisk_header->magic, "cxsparse", 8)) {
239
            ret = -EINVAL;
240
            goto fail;
241
        }
242

    
243
        s->block_size = be32_to_cpu(dyndisk_header->block_size);
244
        s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511;
245

    
246
        s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries);
247
        s->pagetable = g_malloc(s->max_table_entries * 4);
248

    
249
        s->bat_offset = be64_to_cpu(dyndisk_header->table_offset);
250

    
251
        ret = bdrv_pread(bs->file, s->bat_offset, s->pagetable,
252
                         s->max_table_entries * 4);
253
        if (ret < 0) {
254
            goto fail;
255
        }
256

    
257
        s->free_data_block_offset =
258
            (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511;
259

    
260
        for (i = 0; i < s->max_table_entries; i++) {
261
            be32_to_cpus(&s->pagetable[i]);
262
            if (s->pagetable[i] != 0xFFFFFFFF) {
263
                int64_t next = (512 * (int64_t) s->pagetable[i]) +
264
                    s->bitmap_size + s->block_size;
265

    
266
                if (next > s->free_data_block_offset) {
267
                    s->free_data_block_offset = next;
268
                }
269
            }
270
        }
271

    
272
        if (s->free_data_block_offset > bdrv_getlength(bs->file)) {
273
            error_setg(errp, "block-vpc: free_data_block_offset points after "
274
                             "the end of file. The image has been truncated.");
275
            ret = -EINVAL;
276
            goto fail;
277
        }
278

    
279
        s->last_bitmap_offset = (int64_t) -1;
280

    
281
#ifdef CACHE
282
        s->pageentry_u8 = g_malloc(512);
283
        s->pageentry_u32 = s->pageentry_u8;
284
        s->pageentry_u16 = s->pageentry_u8;
285
        s->last_pagetable = -1;
286
#endif
287
    }
288

    
289
    qemu_co_mutex_init(&s->lock);
290

    
291
    /* Disable migration when VHD images are used */
292
    error_set(&s->migration_blocker,
293
              QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
294
              "vpc", bs->device_name, "live migration");
295
    migrate_add_blocker(s->migration_blocker);
296

    
297
    return 0;
298

    
299
fail:
300
    g_free(s->pagetable);
301
#ifdef CACHE
302
    g_free(s->pageentry_u8);
303
#endif
304
    return ret;
305
}
306

    
307
static int vpc_reopen_prepare(BDRVReopenState *state,
308
                              BlockReopenQueue *queue, Error **errp)
309
{
310
    return 0;
311
}
312

    
313
/*
314
 * Returns the absolute byte offset of the given sector in the image file.
315
 * If the sector is not allocated, -1 is returned instead.
316
 *
317
 * The parameter write must be 1 if the offset will be used for a write
318
 * operation (the block bitmaps is updated then), 0 otherwise.
319
 */
320
static inline int64_t get_sector_offset(BlockDriverState *bs,
321
    int64_t sector_num, int write)
322
{
323
    BDRVVPCState *s = bs->opaque;
324
    uint64_t offset = sector_num * 512;
325
    uint64_t bitmap_offset, block_offset;
326
    uint32_t pagetable_index, pageentry_index;
327

    
328
    pagetable_index = offset / s->block_size;
329
    pageentry_index = (offset % s->block_size) / 512;
330

    
331
    if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff)
332
        return -1; // not allocated
333

    
334
    bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index];
335
    block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index);
336

    
337
    // We must ensure that we don't write to any sectors which are marked as
338
    // unused in the bitmap. We get away with setting all bits in the block
339
    // bitmap each time we write to a new block. This might cause Virtual PC to
340
    // miss sparse read optimization, but it's not a problem in terms of
341
    // correctness.
342
    if (write && (s->last_bitmap_offset != bitmap_offset)) {
343
        uint8_t bitmap[s->bitmap_size];
344

    
345
        s->last_bitmap_offset = bitmap_offset;
346
        memset(bitmap, 0xff, s->bitmap_size);
347
        bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size);
348
    }
349

    
350
//    printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
351
//        sector_num, pagetable_index, pageentry_index,
352
//        bitmap_offset, block_offset);
353

    
354
// disabled by reason
355
#if 0
356
#ifdef CACHE
357
    if (bitmap_offset != s->last_bitmap)
358
    {
359
        lseek(s->fd, bitmap_offset, SEEK_SET);
360

361
        s->last_bitmap = bitmap_offset;
362

363
        // Scary! Bitmap is stored as big endian 32bit entries,
364
        // while we used to look it up byte by byte
365
        read(s->fd, s->pageentry_u8, 512);
366
        for (i = 0; i < 128; i++)
367
            be32_to_cpus(&s->pageentry_u32[i]);
368
    }
369

370
    if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1)
371
        return -1;
372
#else
373
    lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET);
374

    
375
    read(s->fd, &bitmap_entry, 1);
376

    
377
    if ((bitmap_entry >> (pageentry_index % 8)) & 1)
378
        return -1; // not allocated
379
#endif
380
#endif
381

    
382
    return block_offset;
383
}
384

    
385
/*
386
 * Writes the footer to the end of the image file. This is needed when the
387
 * file grows as it overwrites the old footer
388
 *
389
 * Returns 0 on success and < 0 on error
390
 */
391
static int rewrite_footer(BlockDriverState* bs)
392
{
393
    int ret;
394
    BDRVVPCState *s = bs->opaque;
395
    int64_t offset = s->free_data_block_offset;
396

    
397
    ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE);
398
    if (ret < 0)
399
        return ret;
400

    
401
    return 0;
402
}
403

    
404
/*
405
 * Allocates a new block. This involves writing a new footer and updating
406
 * the Block Allocation Table to use the space at the old end of the image
407
 * file (overwriting the old footer)
408
 *
409
 * Returns the sectors' offset in the image file on success and < 0 on error
410
 */
411
static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num)
412
{
413
    BDRVVPCState *s = bs->opaque;
414
    int64_t bat_offset;
415
    uint32_t index, bat_value;
416
    int ret;
417
    uint8_t bitmap[s->bitmap_size];
418

    
419
    // Check if sector_num is valid
420
    if ((sector_num < 0) || (sector_num > bs->total_sectors))
421
        return -1;
422

    
423
    // Write entry into in-memory BAT
424
    index = (sector_num * 512) / s->block_size;
425
    if (s->pagetable[index] != 0xFFFFFFFF)
426
        return -1;
427

    
428
    s->pagetable[index] = s->free_data_block_offset / 512;
429

    
430
    // Initialize the block's bitmap
431
    memset(bitmap, 0xff, s->bitmap_size);
432
    ret = bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap,
433
        s->bitmap_size);
434
    if (ret < 0) {
435
        return ret;
436
    }
437

    
438
    // Write new footer (the old one will be overwritten)
439
    s->free_data_block_offset += s->block_size + s->bitmap_size;
440
    ret = rewrite_footer(bs);
441
    if (ret < 0)
442
        goto fail;
443

    
444
    // Write BAT entry to disk
445
    bat_offset = s->bat_offset + (4 * index);
446
    bat_value = be32_to_cpu(s->pagetable[index]);
447
    ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4);
448
    if (ret < 0)
449
        goto fail;
450

    
451
    return get_sector_offset(bs, sector_num, 0);
452

    
453
fail:
454
    s->free_data_block_offset -= (s->block_size + s->bitmap_size);
455
    return -1;
456
}
457

    
458
static int vpc_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
459
{
460
    BDRVVPCState *s = (BDRVVPCState *)bs->opaque;
461
    VHDFooter *footer = (VHDFooter *) s->footer_buf;
462

    
463
    if (cpu_to_be32(footer->type) != VHD_FIXED) {
464
        bdi->cluster_size = s->block_size;
465
    }
466

    
467
    bdi->unallocated_blocks_are_zero = true;
468
    return 0;
469
}
470

    
471
static int vpc_read(BlockDriverState *bs, int64_t sector_num,
472
                    uint8_t *buf, int nb_sectors)
473
{
474
    BDRVVPCState *s = bs->opaque;
475
    int ret;
476
    int64_t offset;
477
    int64_t sectors, sectors_per_block;
478
    VHDFooter *footer = (VHDFooter *) s->footer_buf;
479

    
480
    if (cpu_to_be32(footer->type) == VHD_FIXED) {
481
        return bdrv_read(bs->file, sector_num, buf, nb_sectors);
482
    }
483
    while (nb_sectors > 0) {
484
        offset = get_sector_offset(bs, sector_num, 0);
485

    
486
        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
487
        sectors = sectors_per_block - (sector_num % sectors_per_block);
488
        if (sectors > nb_sectors) {
489
            sectors = nb_sectors;
490
        }
491

    
492
        if (offset == -1) {
493
            memset(buf, 0, sectors * BDRV_SECTOR_SIZE);
494
        } else {
495
            ret = bdrv_pread(bs->file, offset, buf,
496
                sectors * BDRV_SECTOR_SIZE);
497
            if (ret != sectors * BDRV_SECTOR_SIZE) {
498
                return -1;
499
            }
500
        }
501

    
502
        nb_sectors -= sectors;
503
        sector_num += sectors;
504
        buf += sectors * BDRV_SECTOR_SIZE;
505
    }
506
    return 0;
507
}
508

    
509
static coroutine_fn int vpc_co_read(BlockDriverState *bs, int64_t sector_num,
510
                                    uint8_t *buf, int nb_sectors)
511
{
512
    int ret;
513
    BDRVVPCState *s = bs->opaque;
514
    qemu_co_mutex_lock(&s->lock);
515
    ret = vpc_read(bs, sector_num, buf, nb_sectors);
516
    qemu_co_mutex_unlock(&s->lock);
517
    return ret;
518
}
519

    
520
static int vpc_write(BlockDriverState *bs, int64_t sector_num,
521
    const uint8_t *buf, int nb_sectors)
522
{
523
    BDRVVPCState *s = bs->opaque;
524
    int64_t offset;
525
    int64_t sectors, sectors_per_block;
526
    int ret;
527
    VHDFooter *footer =  (VHDFooter *) s->footer_buf;
528

    
529
    if (cpu_to_be32(footer->type) == VHD_FIXED) {
530
        return bdrv_write(bs->file, sector_num, buf, nb_sectors);
531
    }
532
    while (nb_sectors > 0) {
533
        offset = get_sector_offset(bs, sector_num, 1);
534

    
535
        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
536
        sectors = sectors_per_block - (sector_num % sectors_per_block);
537
        if (sectors > nb_sectors) {
538
            sectors = nb_sectors;
539
        }
540

    
541
        if (offset == -1) {
542
            offset = alloc_block(bs, sector_num);
543
            if (offset < 0)
544
                return -1;
545
        }
546

    
547
        ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE);
548
        if (ret != sectors * BDRV_SECTOR_SIZE) {
549
            return -1;
550
        }
551

    
552
        nb_sectors -= sectors;
553
        sector_num += sectors;
554
        buf += sectors * BDRV_SECTOR_SIZE;
555
    }
556

    
557
    return 0;
558
}
559

    
560
static coroutine_fn int vpc_co_write(BlockDriverState *bs, int64_t sector_num,
561
                                     const uint8_t *buf, int nb_sectors)
562
{
563
    int ret;
564
    BDRVVPCState *s = bs->opaque;
565
    qemu_co_mutex_lock(&s->lock);
566
    ret = vpc_write(bs, sector_num, buf, nb_sectors);
567
    qemu_co_mutex_unlock(&s->lock);
568
    return ret;
569
}
570

    
571
/*
572
 * Calculates the number of cylinders, heads and sectors per cylinder
573
 * based on a given number of sectors. This is the algorithm described
574
 * in the VHD specification.
575
 *
576
 * Note that the geometry doesn't always exactly match total_sectors but
577
 * may round it down.
578
 *
579
 * Returns 0 on success, -EFBIG if the size is larger than ~2 TB. Override
580
 * the hardware EIDE and ATA-2 limit of 16 heads (max disk size of 127 GB)
581
 * and instead allow up to 255 heads.
582
 */
583
static int calculate_geometry(int64_t total_sectors, uint16_t* cyls,
584
    uint8_t* heads, uint8_t* secs_per_cyl)
585
{
586
    uint32_t cyls_times_heads;
587

    
588
    /* Allow a maximum disk size of approximately 2 TB */
589
    if (total_sectors > 65535LL * 255 * 255) {
590
        return -EFBIG;
591
    }
592

    
593
    if (total_sectors > 65535 * 16 * 63) {
594
        *secs_per_cyl = 255;
595
        if (total_sectors > 65535 * 16 * 255) {
596
            *heads = 255;
597
        } else {
598
            *heads = 16;
599
        }
600
        cyls_times_heads = total_sectors / *secs_per_cyl;
601
    } else {
602
        *secs_per_cyl = 17;
603
        cyls_times_heads = total_sectors / *secs_per_cyl;
604
        *heads = (cyls_times_heads + 1023) / 1024;
605

    
606
        if (*heads < 4)
607
            *heads = 4;
608

    
609
        if (cyls_times_heads >= (*heads * 1024) || *heads > 16) {
610
            *secs_per_cyl = 31;
611
            *heads = 16;
612
            cyls_times_heads = total_sectors / *secs_per_cyl;
613
        }
614

    
615
        if (cyls_times_heads >= (*heads * 1024)) {
616
            *secs_per_cyl = 63;
617
            *heads = 16;
618
            cyls_times_heads = total_sectors / *secs_per_cyl;
619
        }
620
    }
621

    
622
    *cyls = cyls_times_heads / *heads;
623

    
624
    return 0;
625
}
626

    
627
static int create_dynamic_disk(int fd, uint8_t *buf, int64_t total_sectors)
628
{
629
    VHDDynDiskHeader *dyndisk_header =
630
        (VHDDynDiskHeader *) buf;
631
    size_t block_size, num_bat_entries;
632
    int i;
633
    int ret = -EIO;
634

    
635
    // Write the footer (twice: at the beginning and at the end)
636
    block_size = 0x200000;
637
    num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);
638

    
639
    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
640
        goto fail;
641
    }
642

    
643
    if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0) {
644
        goto fail;
645
    }
646
    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
647
        goto fail;
648
    }
649

    
650
    // Write the initial BAT
651
    if (lseek(fd, 3 * 512, SEEK_SET) < 0) {
652
        goto fail;
653
    }
654

    
655
    memset(buf, 0xFF, 512);
656
    for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) {
657
        if (write(fd, buf, 512) != 512) {
658
            goto fail;
659
        }
660
    }
661

    
662
    // Prepare the Dynamic Disk Header
663
    memset(buf, 0, 1024);
664

    
665
    memcpy(dyndisk_header->magic, "cxsparse", 8);
666

    
667
    /*
668
     * Note: The spec is actually wrong here for data_offset, it says
669
     * 0xFFFFFFFF, but MS tools expect all 64 bits to be set.
670
     */
671
    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
672
    dyndisk_header->table_offset = be64_to_cpu(3 * 512);
673
    dyndisk_header->version = be32_to_cpu(0x00010000);
674
    dyndisk_header->block_size = be32_to_cpu(block_size);
675
    dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries);
676

    
677
    dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024));
678

    
679
    // Write the header
680
    if (lseek(fd, 512, SEEK_SET) < 0) {
681
        goto fail;
682
    }
683

    
684
    if (write(fd, buf, 1024) != 1024) {
685
        goto fail;
686
    }
687
    ret = 0;
688

    
689
 fail:
690
    return ret;
691
}
692

    
693
static int create_fixed_disk(int fd, uint8_t *buf, int64_t total_size)
694
{
695
    int ret = -EIO;
696

    
697
    /* Add footer to total size */
698
    total_size += 512;
699
    if (ftruncate(fd, total_size) != 0) {
700
        ret = -errno;
701
        goto fail;
702
    }
703
    if (lseek(fd, -512, SEEK_END) < 0) {
704
        goto fail;
705
    }
706
    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
707
        goto fail;
708
    }
709

    
710
    ret = 0;
711

    
712
 fail:
713
    return ret;
714
}
715

    
716
static int vpc_create(const char *filename, QEMUOptionParameter *options,
717
                      Error **errp)
718
{
719
    uint8_t buf[1024];
720
    VHDFooter *footer = (VHDFooter *) buf;
721
    QEMUOptionParameter *disk_type_param;
722
    int fd, i;
723
    uint16_t cyls = 0;
724
    uint8_t heads = 0;
725
    uint8_t secs_per_cyl = 0;
726
    int64_t total_sectors;
727
    int64_t total_size;
728
    int disk_type;
729
    int ret = -EIO;
730

    
731
    /* Read out options */
732
    total_size = get_option_parameter(options, BLOCK_OPT_SIZE)->value.n;
733

    
734
    disk_type_param = get_option_parameter(options, BLOCK_OPT_SUBFMT);
735
    if (disk_type_param && disk_type_param->value.s) {
736
        if (!strcmp(disk_type_param->value.s, "dynamic")) {
737
            disk_type = VHD_DYNAMIC;
738
        } else if (!strcmp(disk_type_param->value.s, "fixed")) {
739
            disk_type = VHD_FIXED;
740
        } else {
741
            return -EINVAL;
742
        }
743
    } else {
744
        disk_type = VHD_DYNAMIC;
745
    }
746

    
747
    /* Create the file */
748
    fd = qemu_open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
749
    if (fd < 0) {
750
        return -EIO;
751
    }
752

    
753
    /*
754
     * Calculate matching total_size and geometry. Increase the number of
755
     * sectors requested until we get enough (or fail). This ensures that
756
     * qemu-img convert doesn't truncate images, but rather rounds up.
757
     */
758
    total_sectors = total_size / BDRV_SECTOR_SIZE;
759
    for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) {
760
        if (calculate_geometry(total_sectors + i, &cyls, &heads,
761
                               &secs_per_cyl))
762
        {
763
            ret = -EFBIG;
764
            goto fail;
765
        }
766
    }
767

    
768
    total_sectors = (int64_t) cyls * heads * secs_per_cyl;
769

    
770
    /* Prepare the Hard Disk Footer */
771
    memset(buf, 0, 1024);
772

    
773
    memcpy(footer->creator, "conectix", 8);
774
    /* TODO Check if "qemu" creator_app is ok for VPC */
775
    memcpy(footer->creator_app, "qemu", 4);
776
    memcpy(footer->creator_os, "Wi2k", 4);
777

    
778
    footer->features = be32_to_cpu(0x02);
779
    footer->version = be32_to_cpu(0x00010000);
780
    if (disk_type == VHD_DYNAMIC) {
781
        footer->data_offset = be64_to_cpu(HEADER_SIZE);
782
    } else {
783
        footer->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
784
    }
785
    footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE);
786

    
787
    /* Version of Virtual PC 2007 */
788
    footer->major = be16_to_cpu(0x0005);
789
    footer->minor = be16_to_cpu(0x0003);
790
    if (disk_type == VHD_DYNAMIC) {
791
        footer->orig_size = be64_to_cpu(total_sectors * 512);
792
        footer->size = be64_to_cpu(total_sectors * 512);
793
    } else {
794
        footer->orig_size = be64_to_cpu(total_size);
795
        footer->size = be64_to_cpu(total_size);
796
    }
797
    footer->cyls = be16_to_cpu(cyls);
798
    footer->heads = heads;
799
    footer->secs_per_cyl = secs_per_cyl;
800

    
801
    footer->type = be32_to_cpu(disk_type);
802

    
803
#if defined(CONFIG_UUID)
804
    uuid_generate(footer->uuid);
805
#endif
806

    
807
    footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE));
808

    
809
    if (disk_type == VHD_DYNAMIC) {
810
        ret = create_dynamic_disk(fd, buf, total_sectors);
811
    } else {
812
        ret = create_fixed_disk(fd, buf, total_size);
813
    }
814

    
815
 fail:
816
    qemu_close(fd);
817
    return ret;
818
}
819

    
820
static int vpc_has_zero_init(BlockDriverState *bs)
821
{
822
    BDRVVPCState *s = bs->opaque;
823
    VHDFooter *footer =  (VHDFooter *) s->footer_buf;
824

    
825
    if (cpu_to_be32(footer->type) == VHD_FIXED) {
826
        return bdrv_has_zero_init(bs->file);
827
    } else {
828
        return 1;
829
    }
830
}
831

    
832
static void vpc_close(BlockDriverState *bs)
833
{
834
    BDRVVPCState *s = bs->opaque;
835
    g_free(s->pagetable);
836
#ifdef CACHE
837
    g_free(s->pageentry_u8);
838
#endif
839

    
840
    migrate_del_blocker(s->migration_blocker);
841
    error_free(s->migration_blocker);
842
}
843

    
844
static QEMUOptionParameter vpc_create_options[] = {
845
    {
846
        .name = BLOCK_OPT_SIZE,
847
        .type = OPT_SIZE,
848
        .help = "Virtual disk size"
849
    },
850
    {
851
        .name = BLOCK_OPT_SUBFMT,
852
        .type = OPT_STRING,
853
        .help =
854
            "Type of virtual hard disk format. Supported formats are "
855
            "{dynamic (default) | fixed} "
856
    },
857
    { NULL }
858
};
859

    
860
static BlockDriver bdrv_vpc = {
861
    .format_name    = "vpc",
862
    .instance_size  = sizeof(BDRVVPCState),
863

    
864
    .bdrv_probe             = vpc_probe,
865
    .bdrv_open              = vpc_open,
866
    .bdrv_close             = vpc_close,
867
    .bdrv_reopen_prepare    = vpc_reopen_prepare,
868
    .bdrv_create            = vpc_create,
869

    
870
    .bdrv_read              = vpc_co_read,
871
    .bdrv_write             = vpc_co_write,
872

    
873
    .bdrv_get_info          = vpc_get_info,
874

    
875
    .create_options         = vpc_create_options,
876
    .bdrv_has_zero_init     = vpc_has_zero_init,
877
};
878

    
879
static void bdrv_vpc_init(void)
880
{
881
    bdrv_register(&bdrv_vpc);
882
}
883

    
884
block_init(bdrv_vpc_init);