Statistics
| Branch: | Revision:

root / block / vpc.c @ 76abe407

History | View | Annotate | Download (24.8 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
            error_setg(errp, "invalid VPC image");
194
            ret = -EINVAL;
195
            goto fail;
196
        }
197
        disk_type = VHD_FIXED;
198
    }
199

    
200
    checksum = be32_to_cpu(footer->checksum);
201
    footer->checksum = 0;
202
    if (vpc_checksum(s->footer_buf, HEADER_SIZE) != checksum)
203
        fprintf(stderr, "block-vpc: The header checksum of '%s' is "
204
            "incorrect.\n", bs->filename);
205

    
206
    /* Write 'checksum' back to footer, or else will leave it with zero. */
207
    footer->checksum = be32_to_cpu(checksum);
208

    
209
    // The visible size of a image in Virtual PC depends on the geometry
210
    // rather than on the size stored in the footer (the size in the footer
211
    // is too large usually)
212
    bs->total_sectors = (int64_t)
213
        be16_to_cpu(footer->cyls) * footer->heads * footer->secs_per_cyl;
214

    
215
    /* images created with disk2vhd report a far higher virtual size
216
     * than expected with the cyls * heads * sectors_per_cyl formula.
217
     * use the footer->size instead if the image was created with
218
     * disk2vhd.
219
     */
220
    if (!strncmp(footer->creator_app, "d2v", 4)) {
221
        bs->total_sectors = be64_to_cpu(footer->size) / BDRV_SECTOR_SIZE;
222
    }
223

    
224
    /* Allow a maximum disk size of approximately 2 TB */
225
    if (bs->total_sectors >= 65535LL * 255 * 255) {
226
        ret = -EFBIG;
227
        goto fail;
228
    }
229

    
230
    if (disk_type == VHD_DYNAMIC) {
231
        ret = bdrv_pread(bs->file, be64_to_cpu(footer->data_offset), buf,
232
                         HEADER_SIZE);
233
        if (ret < 0) {
234
            goto fail;
235
        }
236

    
237
        dyndisk_header = (VHDDynDiskHeader *) buf;
238

    
239
        if (strncmp(dyndisk_header->magic, "cxsparse", 8)) {
240
            ret = -EINVAL;
241
            goto fail;
242
        }
243

    
244
        s->block_size = be32_to_cpu(dyndisk_header->block_size);
245
        s->bitmap_size = ((s->block_size / (8 * 512)) + 511) & ~511;
246

    
247
        s->max_table_entries = be32_to_cpu(dyndisk_header->max_table_entries);
248
        s->pagetable = g_malloc(s->max_table_entries * 4);
249

    
250
        s->bat_offset = be64_to_cpu(dyndisk_header->table_offset);
251

    
252
        ret = bdrv_pread(bs->file, s->bat_offset, s->pagetable,
253
                         s->max_table_entries * 4);
254
        if (ret < 0) {
255
            goto fail;
256
        }
257

    
258
        s->free_data_block_offset =
259
            (s->bat_offset + (s->max_table_entries * 4) + 511) & ~511;
260

    
261
        for (i = 0; i < s->max_table_entries; i++) {
262
            be32_to_cpus(&s->pagetable[i]);
263
            if (s->pagetable[i] != 0xFFFFFFFF) {
264
                int64_t next = (512 * (int64_t) s->pagetable[i]) +
265
                    s->bitmap_size + s->block_size;
266

    
267
                if (next > s->free_data_block_offset) {
268
                    s->free_data_block_offset = next;
269
                }
270
            }
271
        }
272

    
273
        if (s->free_data_block_offset > bdrv_getlength(bs->file)) {
274
            error_setg(errp, "block-vpc: free_data_block_offset points after "
275
                             "the end of file. The image has been truncated.");
276
            ret = -EINVAL;
277
            goto fail;
278
        }
279

    
280
        s->last_bitmap_offset = (int64_t) -1;
281

    
282
#ifdef CACHE
283
        s->pageentry_u8 = g_malloc(512);
284
        s->pageentry_u32 = s->pageentry_u8;
285
        s->pageentry_u16 = s->pageentry_u8;
286
        s->last_pagetable = -1;
287
#endif
288
    }
289

    
290
    qemu_co_mutex_init(&s->lock);
291

    
292
    /* Disable migration when VHD images are used */
293
    error_set(&s->migration_blocker,
294
              QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
295
              "vpc", bs->device_name, "live migration");
296
    migrate_add_blocker(s->migration_blocker);
297

    
298
    return 0;
299

    
300
fail:
301
    g_free(s->pagetable);
302
#ifdef CACHE
303
    g_free(s->pageentry_u8);
304
#endif
305
    return ret;
306
}
307

    
308
static int vpc_reopen_prepare(BDRVReopenState *state,
309
                              BlockReopenQueue *queue, Error **errp)
310
{
311
    return 0;
312
}
313

    
314
/*
315
 * Returns the absolute byte offset of the given sector in the image file.
316
 * If the sector is not allocated, -1 is returned instead.
317
 *
318
 * The parameter write must be 1 if the offset will be used for a write
319
 * operation (the block bitmaps is updated then), 0 otherwise.
320
 */
321
static inline int64_t get_sector_offset(BlockDriverState *bs,
322
    int64_t sector_num, int write)
323
{
324
    BDRVVPCState *s = bs->opaque;
325
    uint64_t offset = sector_num * 512;
326
    uint64_t bitmap_offset, block_offset;
327
    uint32_t pagetable_index, pageentry_index;
328

    
329
    pagetable_index = offset / s->block_size;
330
    pageentry_index = (offset % s->block_size) / 512;
331

    
332
    if (pagetable_index >= s->max_table_entries || s->pagetable[pagetable_index] == 0xffffffff)
333
        return -1; // not allocated
334

    
335
    bitmap_offset = 512 * (uint64_t) s->pagetable[pagetable_index];
336
    block_offset = bitmap_offset + s->bitmap_size + (512 * pageentry_index);
337

    
338
    // We must ensure that we don't write to any sectors which are marked as
339
    // unused in the bitmap. We get away with setting all bits in the block
340
    // bitmap each time we write to a new block. This might cause Virtual PC to
341
    // miss sparse read optimization, but it's not a problem in terms of
342
    // correctness.
343
    if (write && (s->last_bitmap_offset != bitmap_offset)) {
344
        uint8_t bitmap[s->bitmap_size];
345

    
346
        s->last_bitmap_offset = bitmap_offset;
347
        memset(bitmap, 0xff, s->bitmap_size);
348
        bdrv_pwrite_sync(bs->file, bitmap_offset, bitmap, s->bitmap_size);
349
    }
350

    
351
//    printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
352
//        sector_num, pagetable_index, pageentry_index,
353
//        bitmap_offset, block_offset);
354

    
355
// disabled by reason
356
#if 0
357
#ifdef CACHE
358
    if (bitmap_offset != s->last_bitmap)
359
    {
360
        lseek(s->fd, bitmap_offset, SEEK_SET);
361

362
        s->last_bitmap = bitmap_offset;
363

364
        // Scary! Bitmap is stored as big endian 32bit entries,
365
        // while we used to look it up byte by byte
366
        read(s->fd, s->pageentry_u8, 512);
367
        for (i = 0; i < 128; i++)
368
            be32_to_cpus(&s->pageentry_u32[i]);
369
    }
370

371
    if ((s->pageentry_u8[pageentry_index / 8] >> (pageentry_index % 8)) & 1)
372
        return -1;
373
#else
374
    lseek(s->fd, bitmap_offset + (pageentry_index / 8), SEEK_SET);
375

    
376
    read(s->fd, &bitmap_entry, 1);
377

    
378
    if ((bitmap_entry >> (pageentry_index % 8)) & 1)
379
        return -1; // not allocated
380
#endif
381
#endif
382

    
383
    return block_offset;
384
}
385

    
386
/*
387
 * Writes the footer to the end of the image file. This is needed when the
388
 * file grows as it overwrites the old footer
389
 *
390
 * Returns 0 on success and < 0 on error
391
 */
392
static int rewrite_footer(BlockDriverState* bs)
393
{
394
    int ret;
395
    BDRVVPCState *s = bs->opaque;
396
    int64_t offset = s->free_data_block_offset;
397

    
398
    ret = bdrv_pwrite_sync(bs->file, offset, s->footer_buf, HEADER_SIZE);
399
    if (ret < 0)
400
        return ret;
401

    
402
    return 0;
403
}
404

    
405
/*
406
 * Allocates a new block. This involves writing a new footer and updating
407
 * the Block Allocation Table to use the space at the old end of the image
408
 * file (overwriting the old footer)
409
 *
410
 * Returns the sectors' offset in the image file on success and < 0 on error
411
 */
412
static int64_t alloc_block(BlockDriverState* bs, int64_t sector_num)
413
{
414
    BDRVVPCState *s = bs->opaque;
415
    int64_t bat_offset;
416
    uint32_t index, bat_value;
417
    int ret;
418
    uint8_t bitmap[s->bitmap_size];
419

    
420
    // Check if sector_num is valid
421
    if ((sector_num < 0) || (sector_num > bs->total_sectors))
422
        return -1;
423

    
424
    // Write entry into in-memory BAT
425
    index = (sector_num * 512) / s->block_size;
426
    if (s->pagetable[index] != 0xFFFFFFFF)
427
        return -1;
428

    
429
    s->pagetable[index] = s->free_data_block_offset / 512;
430

    
431
    // Initialize the block's bitmap
432
    memset(bitmap, 0xff, s->bitmap_size);
433
    ret = bdrv_pwrite_sync(bs->file, s->free_data_block_offset, bitmap,
434
        s->bitmap_size);
435
    if (ret < 0) {
436
        return ret;
437
    }
438

    
439
    // Write new footer (the old one will be overwritten)
440
    s->free_data_block_offset += s->block_size + s->bitmap_size;
441
    ret = rewrite_footer(bs);
442
    if (ret < 0)
443
        goto fail;
444

    
445
    // Write BAT entry to disk
446
    bat_offset = s->bat_offset + (4 * index);
447
    bat_value = be32_to_cpu(s->pagetable[index]);
448
    ret = bdrv_pwrite_sync(bs->file, bat_offset, &bat_value, 4);
449
    if (ret < 0)
450
        goto fail;
451

    
452
    return get_sector_offset(bs, sector_num, 0);
453

    
454
fail:
455
    s->free_data_block_offset -= (s->block_size + s->bitmap_size);
456
    return -1;
457
}
458

    
459
static int vpc_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
460
{
461
    BDRVVPCState *s = (BDRVVPCState *)bs->opaque;
462
    VHDFooter *footer = (VHDFooter *) s->footer_buf;
463

    
464
    if (cpu_to_be32(footer->type) != VHD_FIXED) {
465
        bdi->cluster_size = s->block_size;
466
    }
467

    
468
    bdi->unallocated_blocks_are_zero = true;
469
    return 0;
470
}
471

    
472
static int vpc_read(BlockDriverState *bs, int64_t sector_num,
473
                    uint8_t *buf, int nb_sectors)
474
{
475
    BDRVVPCState *s = bs->opaque;
476
    int ret;
477
    int64_t offset;
478
    int64_t sectors, sectors_per_block;
479
    VHDFooter *footer = (VHDFooter *) s->footer_buf;
480

    
481
    if (cpu_to_be32(footer->type) == VHD_FIXED) {
482
        return bdrv_read(bs->file, sector_num, buf, nb_sectors);
483
    }
484
    while (nb_sectors > 0) {
485
        offset = get_sector_offset(bs, sector_num, 0);
486

    
487
        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
488
        sectors = sectors_per_block - (sector_num % sectors_per_block);
489
        if (sectors > nb_sectors) {
490
            sectors = nb_sectors;
491
        }
492

    
493
        if (offset == -1) {
494
            memset(buf, 0, sectors * BDRV_SECTOR_SIZE);
495
        } else {
496
            ret = bdrv_pread(bs->file, offset, buf,
497
                sectors * BDRV_SECTOR_SIZE);
498
            if (ret != sectors * BDRV_SECTOR_SIZE) {
499
                return -1;
500
            }
501
        }
502

    
503
        nb_sectors -= sectors;
504
        sector_num += sectors;
505
        buf += sectors * BDRV_SECTOR_SIZE;
506
    }
507
    return 0;
508
}
509

    
510
static coroutine_fn int vpc_co_read(BlockDriverState *bs, int64_t sector_num,
511
                                    uint8_t *buf, int nb_sectors)
512
{
513
    int ret;
514
    BDRVVPCState *s = bs->opaque;
515
    qemu_co_mutex_lock(&s->lock);
516
    ret = vpc_read(bs, sector_num, buf, nb_sectors);
517
    qemu_co_mutex_unlock(&s->lock);
518
    return ret;
519
}
520

    
521
static int vpc_write(BlockDriverState *bs, int64_t sector_num,
522
    const uint8_t *buf, int nb_sectors)
523
{
524
    BDRVVPCState *s = bs->opaque;
525
    int64_t offset;
526
    int64_t sectors, sectors_per_block;
527
    int ret;
528
    VHDFooter *footer =  (VHDFooter *) s->footer_buf;
529

    
530
    if (cpu_to_be32(footer->type) == VHD_FIXED) {
531
        return bdrv_write(bs->file, sector_num, buf, nb_sectors);
532
    }
533
    while (nb_sectors > 0) {
534
        offset = get_sector_offset(bs, sector_num, 1);
535

    
536
        sectors_per_block = s->block_size >> BDRV_SECTOR_BITS;
537
        sectors = sectors_per_block - (sector_num % sectors_per_block);
538
        if (sectors > nb_sectors) {
539
            sectors = nb_sectors;
540
        }
541

    
542
        if (offset == -1) {
543
            offset = alloc_block(bs, sector_num);
544
            if (offset < 0)
545
                return -1;
546
        }
547

    
548
        ret = bdrv_pwrite(bs->file, offset, buf, sectors * BDRV_SECTOR_SIZE);
549
        if (ret != sectors * BDRV_SECTOR_SIZE) {
550
            return -1;
551
        }
552

    
553
        nb_sectors -= sectors;
554
        sector_num += sectors;
555
        buf += sectors * BDRV_SECTOR_SIZE;
556
    }
557

    
558
    return 0;
559
}
560

    
561
static coroutine_fn int vpc_co_write(BlockDriverState *bs, int64_t sector_num,
562
                                     const uint8_t *buf, int nb_sectors)
563
{
564
    int ret;
565
    BDRVVPCState *s = bs->opaque;
566
    qemu_co_mutex_lock(&s->lock);
567
    ret = vpc_write(bs, sector_num, buf, nb_sectors);
568
    qemu_co_mutex_unlock(&s->lock);
569
    return ret;
570
}
571

    
572
/*
573
 * Calculates the number of cylinders, heads and sectors per cylinder
574
 * based on a given number of sectors. This is the algorithm described
575
 * in the VHD specification.
576
 *
577
 * Note that the geometry doesn't always exactly match total_sectors but
578
 * may round it down.
579
 *
580
 * Returns 0 on success, -EFBIG if the size is larger than ~2 TB. Override
581
 * the hardware EIDE and ATA-2 limit of 16 heads (max disk size of 127 GB)
582
 * and instead allow up to 255 heads.
583
 */
584
static int calculate_geometry(int64_t total_sectors, uint16_t* cyls,
585
    uint8_t* heads, uint8_t* secs_per_cyl)
586
{
587
    uint32_t cyls_times_heads;
588

    
589
    /* Allow a maximum disk size of approximately 2 TB */
590
    if (total_sectors > 65535LL * 255 * 255) {
591
        return -EFBIG;
592
    }
593

    
594
    if (total_sectors > 65535 * 16 * 63) {
595
        *secs_per_cyl = 255;
596
        if (total_sectors > 65535 * 16 * 255) {
597
            *heads = 255;
598
        } else {
599
            *heads = 16;
600
        }
601
        cyls_times_heads = total_sectors / *secs_per_cyl;
602
    } else {
603
        *secs_per_cyl = 17;
604
        cyls_times_heads = total_sectors / *secs_per_cyl;
605
        *heads = (cyls_times_heads + 1023) / 1024;
606

    
607
        if (*heads < 4)
608
            *heads = 4;
609

    
610
        if (cyls_times_heads >= (*heads * 1024) || *heads > 16) {
611
            *secs_per_cyl = 31;
612
            *heads = 16;
613
            cyls_times_heads = total_sectors / *secs_per_cyl;
614
        }
615

    
616
        if (cyls_times_heads >= (*heads * 1024)) {
617
            *secs_per_cyl = 63;
618
            *heads = 16;
619
            cyls_times_heads = total_sectors / *secs_per_cyl;
620
        }
621
    }
622

    
623
    *cyls = cyls_times_heads / *heads;
624

    
625
    return 0;
626
}
627

    
628
static int create_dynamic_disk(int fd, uint8_t *buf, int64_t total_sectors)
629
{
630
    VHDDynDiskHeader *dyndisk_header =
631
        (VHDDynDiskHeader *) buf;
632
    size_t block_size, num_bat_entries;
633
    int i;
634
    int ret = -EIO;
635

    
636
    // Write the footer (twice: at the beginning and at the end)
637
    block_size = 0x200000;
638
    num_bat_entries = (total_sectors + block_size / 512) / (block_size / 512);
639

    
640
    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
641
        goto fail;
642
    }
643

    
644
    if (lseek(fd, 1536 + ((num_bat_entries * 4 + 511) & ~511), SEEK_SET) < 0) {
645
        goto fail;
646
    }
647
    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
648
        goto fail;
649
    }
650

    
651
    // Write the initial BAT
652
    if (lseek(fd, 3 * 512, SEEK_SET) < 0) {
653
        goto fail;
654
    }
655

    
656
    memset(buf, 0xFF, 512);
657
    for (i = 0; i < (num_bat_entries * 4 + 511) / 512; i++) {
658
        if (write(fd, buf, 512) != 512) {
659
            goto fail;
660
        }
661
    }
662

    
663
    // Prepare the Dynamic Disk Header
664
    memset(buf, 0, 1024);
665

    
666
    memcpy(dyndisk_header->magic, "cxsparse", 8);
667

    
668
    /*
669
     * Note: The spec is actually wrong here for data_offset, it says
670
     * 0xFFFFFFFF, but MS tools expect all 64 bits to be set.
671
     */
672
    dyndisk_header->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
673
    dyndisk_header->table_offset = be64_to_cpu(3 * 512);
674
    dyndisk_header->version = be32_to_cpu(0x00010000);
675
    dyndisk_header->block_size = be32_to_cpu(block_size);
676
    dyndisk_header->max_table_entries = be32_to_cpu(num_bat_entries);
677

    
678
    dyndisk_header->checksum = be32_to_cpu(vpc_checksum(buf, 1024));
679

    
680
    // Write the header
681
    if (lseek(fd, 512, SEEK_SET) < 0) {
682
        goto fail;
683
    }
684

    
685
    if (write(fd, buf, 1024) != 1024) {
686
        goto fail;
687
    }
688
    ret = 0;
689

    
690
 fail:
691
    return ret;
692
}
693

    
694
static int create_fixed_disk(int fd, uint8_t *buf, int64_t total_size)
695
{
696
    int ret = -EIO;
697

    
698
    /* Add footer to total size */
699
    total_size += 512;
700
    if (ftruncate(fd, total_size) != 0) {
701
        ret = -errno;
702
        goto fail;
703
    }
704
    if (lseek(fd, -512, SEEK_END) < 0) {
705
        goto fail;
706
    }
707
    if (write(fd, buf, HEADER_SIZE) != HEADER_SIZE) {
708
        goto fail;
709
    }
710

    
711
    ret = 0;
712

    
713
 fail:
714
    return ret;
715
}
716

    
717
static int vpc_create(const char *filename, QEMUOptionParameter *options,
718
                      Error **errp)
719
{
720
    uint8_t buf[1024];
721
    VHDFooter *footer = (VHDFooter *) buf;
722
    QEMUOptionParameter *disk_type_param;
723
    int fd, i;
724
    uint16_t cyls = 0;
725
    uint8_t heads = 0;
726
    uint8_t secs_per_cyl = 0;
727
    int64_t total_sectors;
728
    int64_t total_size;
729
    int disk_type;
730
    int ret = -EIO;
731

    
732
    /* Read out options */
733
    total_size = get_option_parameter(options, BLOCK_OPT_SIZE)->value.n;
734

    
735
    disk_type_param = get_option_parameter(options, BLOCK_OPT_SUBFMT);
736
    if (disk_type_param && disk_type_param->value.s) {
737
        if (!strcmp(disk_type_param->value.s, "dynamic")) {
738
            disk_type = VHD_DYNAMIC;
739
        } else if (!strcmp(disk_type_param->value.s, "fixed")) {
740
            disk_type = VHD_FIXED;
741
        } else {
742
            return -EINVAL;
743
        }
744
    } else {
745
        disk_type = VHD_DYNAMIC;
746
    }
747

    
748
    /* Create the file */
749
    fd = qemu_open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644);
750
    if (fd < 0) {
751
        return -EIO;
752
    }
753

    
754
    /*
755
     * Calculate matching total_size and geometry. Increase the number of
756
     * sectors requested until we get enough (or fail). This ensures that
757
     * qemu-img convert doesn't truncate images, but rather rounds up.
758
     */
759
    total_sectors = total_size / BDRV_SECTOR_SIZE;
760
    for (i = 0; total_sectors > (int64_t)cyls * heads * secs_per_cyl; i++) {
761
        if (calculate_geometry(total_sectors + i, &cyls, &heads,
762
                               &secs_per_cyl))
763
        {
764
            ret = -EFBIG;
765
            goto fail;
766
        }
767
    }
768

    
769
    total_sectors = (int64_t) cyls * heads * secs_per_cyl;
770

    
771
    /* Prepare the Hard Disk Footer */
772
    memset(buf, 0, 1024);
773

    
774
    memcpy(footer->creator, "conectix", 8);
775
    /* TODO Check if "qemu" creator_app is ok for VPC */
776
    memcpy(footer->creator_app, "qemu", 4);
777
    memcpy(footer->creator_os, "Wi2k", 4);
778

    
779
    footer->features = be32_to_cpu(0x02);
780
    footer->version = be32_to_cpu(0x00010000);
781
    if (disk_type == VHD_DYNAMIC) {
782
        footer->data_offset = be64_to_cpu(HEADER_SIZE);
783
    } else {
784
        footer->data_offset = be64_to_cpu(0xFFFFFFFFFFFFFFFFULL);
785
    }
786
    footer->timestamp = be32_to_cpu(time(NULL) - VHD_TIMESTAMP_BASE);
787

    
788
    /* Version of Virtual PC 2007 */
789
    footer->major = be16_to_cpu(0x0005);
790
    footer->minor = be16_to_cpu(0x0003);
791
    if (disk_type == VHD_DYNAMIC) {
792
        footer->orig_size = be64_to_cpu(total_sectors * 512);
793
        footer->size = be64_to_cpu(total_sectors * 512);
794
    } else {
795
        footer->orig_size = be64_to_cpu(total_size);
796
        footer->size = be64_to_cpu(total_size);
797
    }
798
    footer->cyls = be16_to_cpu(cyls);
799
    footer->heads = heads;
800
    footer->secs_per_cyl = secs_per_cyl;
801

    
802
    footer->type = be32_to_cpu(disk_type);
803

    
804
#if defined(CONFIG_UUID)
805
    uuid_generate(footer->uuid);
806
#endif
807

    
808
    footer->checksum = be32_to_cpu(vpc_checksum(buf, HEADER_SIZE));
809

    
810
    if (disk_type == VHD_DYNAMIC) {
811
        ret = create_dynamic_disk(fd, buf, total_sectors);
812
    } else {
813
        ret = create_fixed_disk(fd, buf, total_size);
814
    }
815

    
816
 fail:
817
    qemu_close(fd);
818
    return ret;
819
}
820

    
821
static int vpc_has_zero_init(BlockDriverState *bs)
822
{
823
    BDRVVPCState *s = bs->opaque;
824
    VHDFooter *footer =  (VHDFooter *) s->footer_buf;
825

    
826
    if (cpu_to_be32(footer->type) == VHD_FIXED) {
827
        return bdrv_has_zero_init(bs->file);
828
    } else {
829
        return 1;
830
    }
831
}
832

    
833
static void vpc_close(BlockDriverState *bs)
834
{
835
    BDRVVPCState *s = bs->opaque;
836
    g_free(s->pagetable);
837
#ifdef CACHE
838
    g_free(s->pageentry_u8);
839
#endif
840

    
841
    migrate_del_blocker(s->migration_blocker);
842
    error_free(s->migration_blocker);
843
}
844

    
845
static QEMUOptionParameter vpc_create_options[] = {
846
    {
847
        .name = BLOCK_OPT_SIZE,
848
        .type = OPT_SIZE,
849
        .help = "Virtual disk size"
850
    },
851
    {
852
        .name = BLOCK_OPT_SUBFMT,
853
        .type = OPT_STRING,
854
        .help =
855
            "Type of virtual hard disk format. Supported formats are "
856
            "{dynamic (default) | fixed} "
857
    },
858
    { NULL }
859
};
860

    
861
static BlockDriver bdrv_vpc = {
862
    .format_name    = "vpc",
863
    .instance_size  = sizeof(BDRVVPCState),
864

    
865
    .bdrv_probe             = vpc_probe,
866
    .bdrv_open              = vpc_open,
867
    .bdrv_close             = vpc_close,
868
    .bdrv_reopen_prepare    = vpc_reopen_prepare,
869
    .bdrv_create            = vpc_create,
870

    
871
    .bdrv_read              = vpc_co_read,
872
    .bdrv_write             = vpc_co_write,
873

    
874
    .bdrv_get_info          = vpc_get_info,
875

    
876
    .create_options         = vpc_create_options,
877
    .bdrv_has_zero_init     = vpc_has_zero_init,
878
};
879

    
880
static void bdrv_vpc_init(void)
881
{
882
    bdrv_register(&bdrv_vpc);
883
}
884

    
885
block_init(bdrv_vpc_init);