Statistics
| Branch: | Revision:

root / block-vmdk.c @ 5fafdf24

History | View | Annotate | Download (25.1 kB)

1
/*
2
 * Block driver for the VMDK format
3
 *
4
 * Copyright (c) 2004 Fabrice Bellard
5
 * Copyright (c) 2005 Filip Navara
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

    
26
#include "vl.h"
27
#include "block_int.h"
28

    
29
#define VMDK3_MAGIC (('C' << 24) | ('O' << 16) | ('W' << 8) | 'D')
30
#define VMDK4_MAGIC (('K' << 24) | ('D' << 16) | ('M' << 8) | 'V')
31

    
32
typedef struct {
33
    uint32_t version;
34
    uint32_t flags;
35
    uint32_t disk_sectors;
36
    uint32_t granularity;
37
    uint32_t l1dir_offset;
38
    uint32_t l1dir_size;
39
    uint32_t file_sectors;
40
    uint32_t cylinders;
41
    uint32_t heads;
42
    uint32_t sectors_per_track;
43
} VMDK3Header;
44

    
45
typedef struct {
46
    uint32_t version;
47
    uint32_t flags;
48
    int64_t capacity;
49
    int64_t granularity;
50
    int64_t desc_offset;
51
    int64_t desc_size;
52
    int32_t num_gtes_per_gte;
53
    int64_t rgd_offset;
54
    int64_t gd_offset;
55
    int64_t grain_offset;
56
    char filler[1];
57
    char check_bytes[4];
58
} __attribute__((packed)) VMDK4Header;
59

    
60
#define L2_CACHE_SIZE 16
61

    
62
typedef struct BDRVVmdkState {
63
    BlockDriverState *hd;
64
    int64_t l1_table_offset;
65
    int64_t l1_backup_table_offset;
66
    uint32_t *l1_table;
67
    uint32_t *l1_backup_table;
68
    unsigned int l1_size;
69
    uint32_t l1_entry_sectors;
70

    
71
    unsigned int l2_size;
72
    uint32_t *l2_cache;
73
    uint32_t l2_cache_offsets[L2_CACHE_SIZE];
74
    uint32_t l2_cache_counts[L2_CACHE_SIZE];
75

    
76
    unsigned int cluster_sectors;
77
    uint32_t parent_cid;
78
    int is_parent;
79
} BDRVVmdkState;
80

    
81
typedef struct VmdkMetaData {
82
    uint32_t offset;
83
    unsigned int l1_index;
84
    unsigned int l2_index;
85
    unsigned int l2_offset;
86
    int valid;
87
} VmdkMetaData;
88

    
89
typedef struct ActiveBDRVState{
90
    BlockDriverState *hd;            // active image handler
91
    uint64_t cluster_offset;         // current write offset
92
}ActiveBDRVState;
93

    
94
static ActiveBDRVState activeBDRV;
95

    
96

    
97
static int vmdk_probe(const uint8_t *buf, int buf_size, const char *filename)
98
{
99
    uint32_t magic;
100

    
101
    if (buf_size < 4)
102
        return 0;
103
    magic = be32_to_cpu(*(uint32_t *)buf);
104
    if (magic == VMDK3_MAGIC ||
105
        magic == VMDK4_MAGIC)
106
        return 100;
107
    else
108
        return 0;
109
}
110

    
111
#define CHECK_CID 1
112

    
113
#define SECTOR_SIZE 512                        
114
#define DESC_SIZE 20*SECTOR_SIZE        // 20 sectors of 512 bytes each
115
#define HEADER_SIZE 512                           // first sector of 512 bytes
116

    
117
static uint32_t vmdk_read_cid(BlockDriverState *bs, int parent)
118
{
119
    BDRVVmdkState *s = bs->opaque;
120
    char desc[DESC_SIZE];
121
    uint32_t cid;
122
    char *p_name, *cid_str;
123
    size_t cid_str_size;
124

    
125
    /* the descriptor offset = 0x200 */
126
    if (bdrv_pread(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
127
        return 0;
128

    
129
    if (parent) {
130
        cid_str = "parentCID";
131
        cid_str_size = sizeof("parentCID");
132
    } else {
133
        cid_str = "CID";
134
        cid_str_size = sizeof("CID");
135
    }
136

    
137
    if ((p_name = strstr(desc,cid_str)) != 0) {
138
        p_name += cid_str_size;
139
        sscanf(p_name,"%x",&cid);
140
    }
141

    
142
    return cid;
143
}
144

    
145
static int vmdk_write_cid(BlockDriverState *bs, uint32_t cid)
146
{
147
    BDRVVmdkState *s = bs->opaque;
148
    char desc[DESC_SIZE], tmp_desc[DESC_SIZE];
149
    char *p_name, *tmp_str;
150

    
151
    /* the descriptor offset = 0x200 */
152
    if (bdrv_pread(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
153
        return -1;
154

    
155
    tmp_str = strstr(desc,"parentCID");
156
    strcpy(tmp_desc, tmp_str);
157
    if ((p_name = strstr(desc,"CID")) != 0) {
158
        p_name += sizeof("CID");
159
        sprintf(p_name,"%x\n",cid);
160
        strcat(desc,tmp_desc);
161
    }
162

    
163
    if (bdrv_pwrite(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
164
        return -1;
165
    return 0;
166
}
167

    
168
static int vmdk_is_cid_valid(BlockDriverState *bs)
169
{
170
#ifdef CHECK_CID
171
    BDRVVmdkState *s = bs->opaque;
172
    BlockDriverState *p_bs = s->hd->backing_hd;
173
    uint32_t cur_pcid;
174

    
175
    if (p_bs) {
176
        cur_pcid = vmdk_read_cid(p_bs,0);
177
        if (s->parent_cid != cur_pcid)
178
            // CID not valid
179
            return 0;
180
    }
181
#endif
182
    // CID valid
183
    return 1;
184
}
185

    
186
static int vmdk_snapshot_create(const char *filename, const char *backing_file)
187
{
188
    int snp_fd, p_fd;
189
    uint32_t p_cid;
190
    char *p_name, *gd_buf, *rgd_buf;
191
    const char *real_filename, *temp_str;
192
    VMDK4Header header;
193
    uint32_t gde_entries, gd_size;
194
    int64_t gd_offset, rgd_offset, capacity, gt_size;
195
    char p_desc[DESC_SIZE], s_desc[DESC_SIZE], hdr[HEADER_SIZE];
196
    char *desc_template =
197
    "# Disk DescriptorFile\n"
198
    "version=1\n"
199
    "CID=%x\n"
200
    "parentCID=%x\n"
201
    "createType=\"monolithicSparse\"\n"
202
    "parentFileNameHint=\"%s\"\n"
203
    "\n"
204
    "# Extent description\n"
205
    "RW %lu SPARSE \"%s\"\n"
206
    "\n"
207
    "# The Disk Data Base \n"
208
    "#DDB\n"
209
    "\n";
210

    
211
    snp_fd = open(filename, O_RDWR | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, 0644);
212
    if (snp_fd < 0)
213
        return -1;
214
    p_fd = open(backing_file, O_RDONLY | O_BINARY | O_LARGEFILE);
215
    if (p_fd < 0) {
216
        close(snp_fd);
217
        return -1;
218
    }
219

    
220
    /* read the header */
221
    if (lseek(p_fd, 0x0, SEEK_SET) == -1)
222
        goto fail;
223
    if (read(p_fd, hdr, HEADER_SIZE) != HEADER_SIZE)
224
        goto fail;
225

    
226
    /* write the header */
227
    if (lseek(snp_fd, 0x0, SEEK_SET) == -1)
228
        goto fail;
229
    if (write(snp_fd, hdr, HEADER_SIZE) == -1)
230
        goto fail;
231

    
232
    memset(&header, 0, sizeof(header));
233
    memcpy(&header,&hdr[4], sizeof(header)); // skip the VMDK4_MAGIC
234

    
235
    ftruncate(snp_fd, header.grain_offset << 9);
236
    /* the descriptor offset = 0x200 */
237
    if (lseek(p_fd, 0x200, SEEK_SET) == -1)
238
        goto fail;
239
    if (read(p_fd, p_desc, DESC_SIZE) != DESC_SIZE)
240
        goto fail;
241

    
242
    if ((p_name = strstr(p_desc,"CID")) != 0) {
243
        p_name += sizeof("CID");
244
        sscanf(p_name,"%x",&p_cid);
245
    }
246

    
247
    real_filename = filename;
248
    if ((temp_str = strrchr(real_filename, '\\')) != NULL)
249
        real_filename = temp_str + 1;
250
    if ((temp_str = strrchr(real_filename, '/')) != NULL)
251
        real_filename = temp_str + 1;
252
    if ((temp_str = strrchr(real_filename, ':')) != NULL)
253
        real_filename = temp_str + 1;
254

    
255
    sprintf(s_desc, desc_template, p_cid, p_cid, backing_file
256
            , (uint32_t)header.capacity, real_filename);
257

    
258
    /* write the descriptor */
259
    if (lseek(snp_fd, 0x200, SEEK_SET) == -1)
260
        goto fail;
261
    if (write(snp_fd, s_desc, strlen(s_desc)) == -1)
262
        goto fail;
263

    
264
    gd_offset = header.gd_offset * SECTOR_SIZE;     // offset of GD table
265
    rgd_offset = header.rgd_offset * SECTOR_SIZE;   // offset of RGD table
266
    capacity = header.capacity * SECTOR_SIZE;       // Extent size
267
    /*
268
     * Each GDE span 32M disk, means:
269
     * 512 GTE per GT, each GTE points to grain
270
     */
271
    gt_size = (int64_t)header.num_gtes_per_gte * header.granularity * SECTOR_SIZE;
272
    if (!gt_size)
273
        goto fail;
274
    gde_entries = (uint32_t)(capacity / gt_size);  // number of gde/rgde
275
    gd_size = gde_entries * sizeof(uint32_t);
276

    
277
    /* write RGD */
278
    rgd_buf = qemu_malloc(gd_size);
279
    if (!rgd_buf)
280
        goto fail;
281
    if (lseek(p_fd, rgd_offset, SEEK_SET) == -1)
282
        goto fail_rgd;
283
    if (read(p_fd, rgd_buf, gd_size) != gd_size)
284
        goto fail_rgd;
285
    if (lseek(snp_fd, rgd_offset, SEEK_SET) == -1)
286
        goto fail_rgd;
287
    if (write(snp_fd, rgd_buf, gd_size) == -1)
288
        goto fail_rgd;
289
    qemu_free(rgd_buf);
290

    
291
    /* write GD */
292
    gd_buf = qemu_malloc(gd_size);
293
    if (!gd_buf)
294
        goto fail_rgd;
295
    if (lseek(p_fd, gd_offset, SEEK_SET) == -1)
296
        goto fail_gd;
297
    if (read(p_fd, gd_buf, gd_size) != gd_size)
298
        goto fail_gd;
299
    if (lseek(snp_fd, gd_offset, SEEK_SET) == -1)
300
        goto fail_gd;
301
    if (write(snp_fd, gd_buf, gd_size) == -1)
302
        goto fail_gd;
303
    qemu_free(gd_buf);
304

    
305
    close(p_fd);
306
    close(snp_fd);
307
    return 0;
308

    
309
    fail_gd:
310
    qemu_free(gd_buf);
311
    fail_rgd:  
312
    qemu_free(rgd_buf);
313
    fail:
314
    close(p_fd);
315
    close(snp_fd);
316
    return -1;
317
}
318

    
319
static void vmdk_parent_close(BlockDriverState *bs)
320
{
321
    if (bs->backing_hd)
322
        bdrv_close(bs->backing_hd);
323
}
324

    
325
int parent_open = 0;
326
static int vmdk_parent_open(BlockDriverState *bs, const char * filename)
327
{
328
    BDRVVmdkState *s = bs->opaque;
329
    char *p_name;
330
    char desc[DESC_SIZE];
331
    char parent_img_name[1024];
332

    
333
    /* the descriptor offset = 0x200 */
334
    if (bdrv_pread(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
335
        return -1;
336

    
337
    if ((p_name = strstr(desc,"parentFileNameHint")) != 0) {
338
        char *end_name;
339
        struct stat file_buf;
340

    
341
        p_name += sizeof("parentFileNameHint") + 1;
342
        if ((end_name = strchr(p_name,'\"')) == 0)
343
            return -1;
344
               
345
        strncpy(s->hd->backing_file, p_name, end_name - p_name);
346
        if (stat(s->hd->backing_file, &file_buf) != 0) {
347
            path_combine(parent_img_name, sizeof(parent_img_name),
348
                         filename, s->hd->backing_file);
349
        } else {
350
            strcpy(parent_img_name, s->hd->backing_file);
351
        }
352

    
353
        s->hd->backing_hd = bdrv_new("");
354
        if (!s->hd->backing_hd) {
355
            failure:
356
            bdrv_close(s->hd);
357
            return -1;
358
        }
359
        parent_open = 1;
360
        if (bdrv_open(s->hd->backing_hd, parent_img_name, BDRV_O_RDONLY) < 0)
361
            goto failure;
362
        parent_open = 0;
363
    }
364

    
365
    return 0;
366
}
367

    
368
static int vmdk_open(BlockDriverState *bs, const char *filename, int flags)
369
{
370
    BDRVVmdkState *s = bs->opaque;
371
    uint32_t magic;
372
    int l1_size, i, ret;
373

    
374
    if (parent_open)
375
        // Parent must be opened as RO.
376
        flags = BDRV_O_RDONLY;
377
    fprintf(stderr, "(VMDK) image open: flags=0x%x filename=%s\n", flags, bs->filename);
378

    
379
    ret = bdrv_file_open(&s->hd, filename, flags);
380
    if (ret < 0)
381
        return ret;
382
    if (bdrv_pread(s->hd, 0, &magic, sizeof(magic)) != sizeof(magic))
383
        goto fail;
384

    
385
    magic = be32_to_cpu(magic);
386
    if (magic == VMDK3_MAGIC) {
387
        VMDK3Header header;
388

    
389
        if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
390
            goto fail;
391
        s->cluster_sectors = le32_to_cpu(header.granularity);
392
        s->l2_size = 1 << 9;
393
        s->l1_size = 1 << 6;
394
        bs->total_sectors = le32_to_cpu(header.disk_sectors);
395
        s->l1_table_offset = le32_to_cpu(header.l1dir_offset) << 9;
396
        s->l1_backup_table_offset = 0;
397
        s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
398
    } else if (magic == VMDK4_MAGIC) {
399
        VMDK4Header header;
400

    
401
        if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
402
            goto fail;
403
        bs->total_sectors = le64_to_cpu(header.capacity);
404
        s->cluster_sectors = le64_to_cpu(header.granularity);
405
        s->l2_size = le32_to_cpu(header.num_gtes_per_gte);
406
        s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
407
        if (s->l1_entry_sectors <= 0)
408
            goto fail;
409
        s->l1_size = (bs->total_sectors + s->l1_entry_sectors - 1)
410
            / s->l1_entry_sectors;
411
        s->l1_table_offset = le64_to_cpu(header.rgd_offset) << 9;
412
        s->l1_backup_table_offset = le64_to_cpu(header.gd_offset) << 9;
413

    
414
        if (parent_open)
415
            s->is_parent = 1;
416
        else
417
            s->is_parent = 0;
418

    
419
        // try to open parent images, if exist
420
        if (vmdk_parent_open(bs, filename) != 0)
421
            goto fail;
422
        // write the CID once after the image creation
423
        s->parent_cid = vmdk_read_cid(bs,1);
424
    } else {
425
        goto fail;
426
    }
427

    
428
    /* read the L1 table */
429
    l1_size = s->l1_size * sizeof(uint32_t);
430
    s->l1_table = qemu_malloc(l1_size);
431
    if (!s->l1_table)
432
        goto fail;
433
    if (bdrv_pread(s->hd, s->l1_table_offset, s->l1_table, l1_size) != l1_size)
434
        goto fail;
435
    for(i = 0; i < s->l1_size; i++) {
436
        le32_to_cpus(&s->l1_table[i]);
437
    }
438

    
439
    if (s->l1_backup_table_offset) {
440
        s->l1_backup_table = qemu_malloc(l1_size);
441
        if (!s->l1_backup_table)
442
            goto fail;
443
        if (bdrv_pread(s->hd, s->l1_backup_table_offset, s->l1_backup_table, l1_size) != l1_size)
444
            goto fail;
445
        for(i = 0; i < s->l1_size; i++) {
446
            le32_to_cpus(&s->l1_backup_table[i]);
447
        }
448
    }
449

    
450
    s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint32_t));
451
    if (!s->l2_cache)
452
        goto fail;
453
    return 0;
454
 fail:
455
    qemu_free(s->l1_backup_table);
456
    qemu_free(s->l1_table);
457
    qemu_free(s->l2_cache);
458
    bdrv_delete(s->hd);
459
    return -1;
460
}
461

    
462
static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
463
                                   uint64_t offset, int allocate);
464

    
465
static int get_whole_cluster(BlockDriverState *bs, uint64_t cluster_offset,
466
                             uint64_t offset, int allocate)
467
{
468
    uint64_t parent_cluster_offset;
469
    BDRVVmdkState *s = bs->opaque;
470
    uint8_t  whole_grain[s->cluster_sectors*512];        // 128 sectors * 512 bytes each = grain size 64KB
471

    
472
    // we will be here if it's first write on non-exist grain(cluster).
473
    // try to read from parent image, if exist
474
    if (s->hd->backing_hd) {
475
        BDRVVmdkState *ps = s->hd->backing_hd->opaque;
476

    
477
        if (!vmdk_is_cid_valid(bs))
478
            return -1;
479

    
480
        parent_cluster_offset = get_cluster_offset(s->hd->backing_hd, NULL, offset, allocate);
481

    
482
        if (parent_cluster_offset) {
483
            BDRVVmdkState *act_s = activeBDRV.hd->opaque;
484

    
485
            if (bdrv_pread(ps->hd, parent_cluster_offset, whole_grain, ps->cluster_sectors*512) != ps->cluster_sectors*512)
486
                return -1;
487

    
488
            //Write grain only into the active image
489
            if (bdrv_pwrite(act_s->hd, activeBDRV.cluster_offset << 9, whole_grain, sizeof(whole_grain)) != sizeof(whole_grain))
490
                return -1;
491
        }
492
    }
493
    return 0;
494
}
495

    
496
static int vmdk_L2update(BlockDriverState *bs, VmdkMetaData *m_data)
497
{
498
    BDRVVmdkState *s = bs->opaque;
499

    
500
    /* update L2 table */
501
    if (bdrv_pwrite(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
502
                    &(m_data->offset), sizeof(m_data->offset)) != sizeof(m_data->offset))
503
        return -1;
504
    /* update backup L2 table */
505
    if (s->l1_backup_table_offset != 0) {
506
        m_data->l2_offset = s->l1_backup_table[m_data->l1_index];
507
        if (bdrv_pwrite(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
508
                        &(m_data->offset), sizeof(m_data->offset)) != sizeof(m_data->offset))
509
            return -1;
510
    }
511

    
512
    return 0;
513
}
514

    
515
static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
516
                                   uint64_t offset, int allocate)
517
{
518
    BDRVVmdkState *s = bs->opaque;
519
    unsigned int l1_index, l2_offset, l2_index;
520
    int min_index, i, j;
521
    uint32_t min_count, *l2_table, tmp = 0;
522
    uint64_t cluster_offset;
523

    
524
    if (m_data)
525
        m_data->valid = 0;
526

    
527
    l1_index = (offset >> 9) / s->l1_entry_sectors;
528
    if (l1_index >= s->l1_size)
529
        return 0;
530
    l2_offset = s->l1_table[l1_index];
531
    if (!l2_offset)
532
        return 0;
533
    for(i = 0; i < L2_CACHE_SIZE; i++) {
534
        if (l2_offset == s->l2_cache_offsets[i]) {
535
            /* increment the hit count */
536
            if (++s->l2_cache_counts[i] == 0xffffffff) {
537
                for(j = 0; j < L2_CACHE_SIZE; j++) {
538
                    s->l2_cache_counts[j] >>= 1;
539
                }
540
            }
541
            l2_table = s->l2_cache + (i * s->l2_size);
542
            goto found;
543
        }
544
    }
545
    /* not found: load a new entry in the least used one */
546
    min_index = 0;
547
    min_count = 0xffffffff;
548
    for(i = 0; i < L2_CACHE_SIZE; i++) {
549
        if (s->l2_cache_counts[i] < min_count) {
550
            min_count = s->l2_cache_counts[i];
551
            min_index = i;
552
        }
553
    }
554
    l2_table = s->l2_cache + (min_index * s->l2_size);
555
    if (bdrv_pread(s->hd, (int64_t)l2_offset * 512, l2_table, s->l2_size * sizeof(uint32_t)) !=
556
                                                                        s->l2_size * sizeof(uint32_t))
557
        return 0;
558

    
559
    s->l2_cache_offsets[min_index] = l2_offset;
560
    s->l2_cache_counts[min_index] = 1;
561
 found:
562
    l2_index = ((offset >> 9) / s->cluster_sectors) % s->l2_size;
563
    cluster_offset = le32_to_cpu(l2_table[l2_index]);
564

    
565
    if (!cluster_offset) {
566
        if (!allocate)
567
            return 0;
568
        // Avoid the L2 tables update for the images that have snapshots.
569
        if (!s->is_parent) {
570
            cluster_offset = bdrv_getlength(s->hd);
571
            bdrv_truncate(s->hd, cluster_offset + (s->cluster_sectors << 9));
572

    
573
            cluster_offset >>= 9;
574
            tmp = cpu_to_le32(cluster_offset);
575
            l2_table[l2_index] = tmp;
576
            // Save the active image state
577
            activeBDRV.cluster_offset = cluster_offset;
578
            activeBDRV.hd = bs;
579
        }
580
        /* First of all we write grain itself, to avoid race condition
581
         * that may to corrupt the image.
582
         * This problem may occur because of insufficient space on host disk
583
         * or inappropriate VM shutdown.
584
         */
585
        if (get_whole_cluster(bs, cluster_offset, offset, allocate) == -1)
586
            return 0;
587

    
588
        if (m_data) {
589
            m_data->offset = tmp;
590
            m_data->l1_index = l1_index;
591
            m_data->l2_index = l2_index;
592
            m_data->l2_offset = l2_offset;
593
            m_data->valid = 1;
594
        }
595
    }
596
    cluster_offset <<= 9;
597
    return cluster_offset;
598
}
599

    
600
static int vmdk_is_allocated(BlockDriverState *bs, int64_t sector_num,
601
                             int nb_sectors, int *pnum)
602
{
603
    BDRVVmdkState *s = bs->opaque;
604
    int index_in_cluster, n;
605
    uint64_t cluster_offset;
606

    
607
    cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
608
    index_in_cluster = sector_num % s->cluster_sectors;
609
    n = s->cluster_sectors - index_in_cluster;
610
    if (n > nb_sectors)
611
        n = nb_sectors;
612
    *pnum = n;
613
    return (cluster_offset != 0);
614
}
615

    
616
static int vmdk_read(BlockDriverState *bs, int64_t sector_num,
617
                    uint8_t *buf, int nb_sectors)
618
{
619
    BDRVVmdkState *s = bs->opaque;
620
    int index_in_cluster, n, ret;
621
    uint64_t cluster_offset;
622

    
623
    while (nb_sectors > 0) {
624
        cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
625
        index_in_cluster = sector_num % s->cluster_sectors;
626
        n = s->cluster_sectors - index_in_cluster;
627
        if (n > nb_sectors)
628
            n = nb_sectors;
629
        if (!cluster_offset) {
630
            // try to read from parent image, if exist
631
            if (s->hd->backing_hd) {
632
                if (!vmdk_is_cid_valid(bs))
633
                    return -1;
634
                ret = bdrv_read(s->hd->backing_hd, sector_num, buf, n);
635
                if (ret < 0)
636
                    return -1;
637
            } else {
638
                memset(buf, 0, 512 * n);
639
            }
640
        } else {
641
            if(bdrv_pread(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
642
                return -1;
643
        }
644
        nb_sectors -= n;
645
        sector_num += n;
646
        buf += n * 512;
647
    }
648
    return 0;
649
}
650

    
651
static int vmdk_write(BlockDriverState *bs, int64_t sector_num,
652
                     const uint8_t *buf, int nb_sectors)
653
{
654
    BDRVVmdkState *s = bs->opaque;
655
    VmdkMetaData m_data;
656
    int index_in_cluster, n;
657
    uint64_t cluster_offset;
658
    static int cid_update = 0;
659

    
660
    if (sector_num > bs->total_sectors) {
661
        fprintf(stderr,
662
                "(VMDK) Wrong offset: sector_num=0x%llx total_sectors=0x%llx\n",
663
                sector_num, bs->total_sectors);
664
        return -1;
665
    }
666

    
667
    while (nb_sectors > 0) {
668
        index_in_cluster = sector_num & (s->cluster_sectors - 1);
669
        n = s->cluster_sectors - index_in_cluster;
670
        if (n > nb_sectors)
671
            n = nb_sectors;
672
        cluster_offset = get_cluster_offset(bs, &m_data, sector_num << 9, 1);
673
        if (!cluster_offset)
674
            return -1;
675

    
676
        if (bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
677
            return -1;
678
        if (m_data.valid) {
679
            /* update L2 tables */
680
            if (vmdk_L2update(bs, &m_data) == -1)
681
                return -1;
682
        }
683
        nb_sectors -= n;
684
        sector_num += n;
685
        buf += n * 512;
686

    
687
        // update CID on the first write every time the virtual disk is opened
688
        if (!cid_update) {
689
            vmdk_write_cid(bs, time(NULL));
690
            cid_update++;
691
        }
692
    }
693
    return 0;
694
}
695

    
696
static int vmdk_create(const char *filename, int64_t total_size,
697
                       const char *backing_file, int flags)
698
{
699
    int fd, i;
700
    VMDK4Header header;
701
    uint32_t tmp, magic, grains, gd_size, gt_size, gt_count;
702
    char *desc_template =
703
        "# Disk DescriptorFile\n"
704
        "version=1\n"
705
        "CID=%x\n"
706
        "parentCID=ffffffff\n"
707
        "createType=\"monolithicSparse\"\n"
708
        "\n"
709
        "# Extent description\n"
710
        "RW %lu SPARSE \"%s\"\n"
711
        "\n"
712
        "# The Disk Data Base \n"
713
        "#DDB\n"
714
        "\n"
715
        "ddb.virtualHWVersion = \"4\"\n"
716
        "ddb.geometry.cylinders = \"%lu\"\n"
717
        "ddb.geometry.heads = \"16\"\n"
718
        "ddb.geometry.sectors = \"63\"\n"
719
        "ddb.adapterType = \"ide\"\n";
720
    char desc[1024];
721
    const char *real_filename, *temp_str;
722

    
723
    /* XXX: add support for backing file */
724
    if (backing_file) {
725
        return vmdk_snapshot_create(filename, backing_file);
726
    }
727

    
728
    fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE,
729
              0644);
730
    if (fd < 0)
731
        return -1;
732
    magic = cpu_to_be32(VMDK4_MAGIC);
733
    memset(&header, 0, sizeof(header));
734
    header.version = cpu_to_le32(1);
735
    header.flags = cpu_to_le32(3); /* ?? */
736
    header.capacity = cpu_to_le64(total_size);
737
    header.granularity = cpu_to_le64(128);
738
    header.num_gtes_per_gte = cpu_to_le32(512);
739

    
740
    grains = (total_size + header.granularity - 1) / header.granularity;
741
    gt_size = ((header.num_gtes_per_gte * sizeof(uint32_t)) + 511) >> 9;
742
    gt_count = (grains + header.num_gtes_per_gte - 1) / header.num_gtes_per_gte;
743
    gd_size = (gt_count * sizeof(uint32_t) + 511) >> 9;
744

    
745
    header.desc_offset = 1;
746
    header.desc_size = 20;
747
    header.rgd_offset = header.desc_offset + header.desc_size;
748
    header.gd_offset = header.rgd_offset + gd_size + (gt_size * gt_count);
749
    header.grain_offset =
750
       ((header.gd_offset + gd_size + (gt_size * gt_count) +
751
         header.granularity - 1) / header.granularity) *
752
        header.granularity;
753

    
754
    header.desc_offset = cpu_to_le64(header.desc_offset);
755
    header.desc_size = cpu_to_le64(header.desc_size);
756
    header.rgd_offset = cpu_to_le64(header.rgd_offset);
757
    header.gd_offset = cpu_to_le64(header.gd_offset);
758
    header.grain_offset = cpu_to_le64(header.grain_offset);
759

    
760
    header.check_bytes[0] = 0xa;
761
    header.check_bytes[1] = 0x20;
762
    header.check_bytes[2] = 0xd;
763
    header.check_bytes[3] = 0xa;
764
   
765
    /* write all the data */   
766
    write(fd, &magic, sizeof(magic));
767
    write(fd, &header, sizeof(header));
768

    
769
    ftruncate(fd, header.grain_offset << 9);
770

    
771
    /* write grain directory */
772
    lseek(fd, le64_to_cpu(header.rgd_offset) << 9, SEEK_SET);
773
    for (i = 0, tmp = header.rgd_offset + gd_size;
774
         i < gt_count; i++, tmp += gt_size)
775
        write(fd, &tmp, sizeof(tmp));
776
  
777
    /* write backup grain directory */
778
    lseek(fd, le64_to_cpu(header.gd_offset) << 9, SEEK_SET);
779
    for (i = 0, tmp = header.gd_offset + gd_size;
780
         i < gt_count; i++, tmp += gt_size)
781
        write(fd, &tmp, sizeof(tmp));
782

    
783
    /* compose the descriptor */
784
    real_filename = filename;
785
    if ((temp_str = strrchr(real_filename, '\\')) != NULL)
786
        real_filename = temp_str + 1;
787
    if ((temp_str = strrchr(real_filename, '/')) != NULL)
788
        real_filename = temp_str + 1;
789
    if ((temp_str = strrchr(real_filename, ':')) != NULL)
790
        real_filename = temp_str + 1;
791
    sprintf(desc, desc_template, time(NULL), (unsigned long)total_size,
792
            real_filename, total_size / (63 * 16));
793

    
794
    /* write the descriptor */
795
    lseek(fd, le64_to_cpu(header.desc_offset) << 9, SEEK_SET);
796
    write(fd, desc, strlen(desc));
797

    
798
    close(fd);
799
    return 0;
800
}
801

    
802
static void vmdk_close(BlockDriverState *bs)
803
{
804
    BDRVVmdkState *s = bs->opaque;
805

    
806
    qemu_free(s->l1_table);
807
    qemu_free(s->l2_cache);
808
    bdrv_delete(s->hd);
809
    // try to close parent image, if exist
810
    vmdk_parent_close(s->hd);
811
}
812

    
813
static void vmdk_flush(BlockDriverState *bs)
814
{
815
    BDRVVmdkState *s = bs->opaque;
816
    bdrv_flush(s->hd);
817
}
818

    
819
BlockDriver bdrv_vmdk = {
820
    "vmdk",
821
    sizeof(BDRVVmdkState),
822
    vmdk_probe,
823
    vmdk_open,
824
    vmdk_read,
825
    vmdk_write,
826
    vmdk_close,
827
    vmdk_create,
828
    vmdk_flush,
829
    vmdk_is_allocated,
830
};