]> git.proxmox.com Git - qemu.git/blame - block-vmdk.c
Fix wrwim masking (Luis Pureza)
[qemu.git] / block-vmdk.c
CommitLineData
ea2384d3
FB
1/*
2 * Block driver for the VMDK format
5fafdf24 3 *
ea2384d3 4 * Copyright (c) 2004 Fabrice Bellard
ff1afc72 5 * Copyright (c) 2005 Filip Navara
5fafdf24 6 *
ea2384d3
FB
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 */
5f4da8c0 25
faf07963 26#include "qemu-common.h"
ea2384d3
FB
27#include "block_int.h"
28
ea2384d3
FB
29#define VMDK3_MAGIC (('C' << 24) | ('O' << 16) | ('W' << 8) | 'D')
30#define VMDK4_MAGIC (('K' << 24) | ('D' << 16) | ('M' << 8) | 'V')
31
32typedef 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
45typedef 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];
ff1afc72 58} __attribute__((packed)) VMDK4Header;
ea2384d3
FB
59
60#define L2_CACHE_SIZE 16
61
62typedef struct BDRVVmdkState {
5f4da8c0 63 BlockDriverState *hd;
ea2384d3 64 int64_t l1_table_offset;
ff1afc72 65 int64_t l1_backup_table_offset;
ea2384d3 66 uint32_t *l1_table;
ff1afc72 67 uint32_t *l1_backup_table;
ea2384d3
FB
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;
5f4da8c0 77 uint32_t parent_cid;
630530a6 78 int is_parent;
ea2384d3
FB
79} BDRVVmdkState;
80
630530a6
TS
81typedef 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
89typedef struct ActiveBDRVState{
90 BlockDriverState *hd; // active image handler
91 uint64_t cluster_offset; // current write offset
92}ActiveBDRVState;
93
94static ActiveBDRVState activeBDRV;
95
96
ea2384d3
FB
97static 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
5f4da8c0
TS
111#define CHECK_CID 1
112
3b46e624 113#define SECTOR_SIZE 512
5f4da8c0 114#define DESC_SIZE 20*SECTOR_SIZE // 20 sectors of 512 bytes each
5fafdf24 115#define HEADER_SIZE 512 // first sector of 512 bytes
5f4da8c0
TS
116
117static uint32_t vmdk_read_cid(BlockDriverState *bs, int parent)
ea2384d3
FB
118{
119 BDRVVmdkState *s = bs->opaque;
5f4da8c0
TS
120 char desc[DESC_SIZE];
121 uint32_t cid;
5fafdf24 122 char *p_name, *cid_str;
5f4da8c0
TS
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
145static 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
168static 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
186static int vmdk_snapshot_create(const char *filename, const char *backing_file)
187{
188 int snp_fd, p_fd;
189 uint32_t p_cid;
5fafdf24 190 char *p_name, *gd_buf, *rgd_buf;
5f4da8c0
TS
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;
ea2384d3 263
5f4da8c0
TS
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;
5fafdf24 274 gde_entries = (uint32_t)(capacity / gt_size); // number of gde/rgde
5f4da8c0
TS
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);
3b46e624 311 fail_rgd:
5f4da8c0
TS
312 qemu_free(rgd_buf);
313 fail:
314 close(p_fd);
315 close(snp_fd);
316 return -1;
317}
318
319static void vmdk_parent_close(BlockDriverState *bs)
320{
321 if (bs->backing_hd)
322 bdrv_close(bs->backing_hd);
323}
324
630530a6 325int parent_open = 0;
5f4da8c0
TS
326static int vmdk_parent_open(BlockDriverState *bs, const char * filename)
327{
328 BDRVVmdkState *s = bs->opaque;
5fafdf24 329 char *p_name;
5f4da8c0
TS
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;
b34d259a
AZ
344 if ((end_name - p_name) > sizeof (s->hd->backing_file) - 1)
345 return -1;
3b46e624 346
5f4da8c0
TS
347 strncpy(s->hd->backing_file, p_name, end_name - p_name);
348 if (stat(s->hd->backing_file, &file_buf) != 0) {
349 path_combine(parent_img_name, sizeof(parent_img_name),
350 filename, s->hd->backing_file);
351 } else {
352 strcpy(parent_img_name, s->hd->backing_file);
353 }
354
355 s->hd->backing_hd = bdrv_new("");
356 if (!s->hd->backing_hd) {
357 failure:
358 bdrv_close(s->hd);
ff1afc72 359 return -1;
5f4da8c0 360 }
630530a6
TS
361 parent_open = 1;
362 if (bdrv_open(s->hd->backing_hd, parent_img_name, BDRV_O_RDONLY) < 0)
5f4da8c0 363 goto failure;
630530a6 364 parent_open = 0;
ff1afc72 365 }
5f4da8c0
TS
366
367 return 0;
368}
369
370static int vmdk_open(BlockDriverState *bs, const char *filename, int flags)
371{
372 BDRVVmdkState *s = bs->opaque;
373 uint32_t magic;
374 int l1_size, i, ret;
375
630530a6
TS
376 if (parent_open)
377 // Parent must be opened as RO.
378 flags = BDRV_O_RDONLY;
630530a6 379
b5eff355 380 ret = bdrv_file_open(&s->hd, filename, flags);
5f4da8c0
TS
381 if (ret < 0)
382 return ret;
383 if (bdrv_pread(s->hd, 0, &magic, sizeof(magic)) != sizeof(magic))
ea2384d3 384 goto fail;
5f4da8c0 385
7143c62c 386 magic = be32_to_cpu(magic);
ea2384d3
FB
387 if (magic == VMDK3_MAGIC) {
388 VMDK3Header header;
5f4da8c0
TS
389
390 if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
ea2384d3
FB
391 goto fail;
392 s->cluster_sectors = le32_to_cpu(header.granularity);
393 s->l2_size = 1 << 9;
394 s->l1_size = 1 << 6;
395 bs->total_sectors = le32_to_cpu(header.disk_sectors);
ff1afc72
FB
396 s->l1_table_offset = le32_to_cpu(header.l1dir_offset) << 9;
397 s->l1_backup_table_offset = 0;
ea2384d3
FB
398 s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
399 } else if (magic == VMDK4_MAGIC) {
400 VMDK4Header header;
5f4da8c0
TS
401
402 if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
ea2384d3 403 goto fail;
bd6ea3c8
FB
404 bs->total_sectors = le64_to_cpu(header.capacity);
405 s->cluster_sectors = le64_to_cpu(header.granularity);
ea2384d3
FB
406 s->l2_size = le32_to_cpu(header.num_gtes_per_gte);
407 s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
408 if (s->l1_entry_sectors <= 0)
409 goto fail;
5fafdf24 410 s->l1_size = (bs->total_sectors + s->l1_entry_sectors - 1)
ea2384d3 411 / s->l1_entry_sectors;
ff1afc72
FB
412 s->l1_table_offset = le64_to_cpu(header.rgd_offset) << 9;
413 s->l1_backup_table_offset = le64_to_cpu(header.gd_offset) << 9;
5f4da8c0 414
630530a6
TS
415 if (parent_open)
416 s->is_parent = 1;
417 else
418 s->is_parent = 0;
419
5f4da8c0
TS
420 // try to open parent images, if exist
421 if (vmdk_parent_open(bs, filename) != 0)
422 goto fail;
423 // write the CID once after the image creation
424 s->parent_cid = vmdk_read_cid(bs,1);
ea2384d3
FB
425 } else {
426 goto fail;
427 }
5f4da8c0 428
ea2384d3
FB
429 /* read the L1 table */
430 l1_size = s->l1_size * sizeof(uint32_t);
431 s->l1_table = qemu_malloc(l1_size);
432 if (!s->l1_table)
433 goto fail;
5f4da8c0 434 if (bdrv_pread(s->hd, s->l1_table_offset, s->l1_table, l1_size) != l1_size)
ea2384d3
FB
435 goto fail;
436 for(i = 0; i < s->l1_size; i++) {
437 le32_to_cpus(&s->l1_table[i]);
438 }
439
ff1afc72
FB
440 if (s->l1_backup_table_offset) {
441 s->l1_backup_table = qemu_malloc(l1_size);
442 if (!s->l1_backup_table)
443 goto fail;
5f4da8c0 444 if (bdrv_pread(s->hd, s->l1_backup_table_offset, s->l1_backup_table, l1_size) != l1_size)
ff1afc72
FB
445 goto fail;
446 for(i = 0; i < s->l1_size; i++) {
447 le32_to_cpus(&s->l1_backup_table[i]);
448 }
449 }
450
ea2384d3
FB
451 s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint32_t));
452 if (!s->l2_cache)
453 goto fail;
ea2384d3
FB
454 return 0;
455 fail:
ff1afc72 456 qemu_free(s->l1_backup_table);
ea2384d3
FB
457 qemu_free(s->l1_table);
458 qemu_free(s->l2_cache);
5f4da8c0 459 bdrv_delete(s->hd);
ea2384d3
FB
460 return -1;
461}
462
630530a6
TS
463static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
464 uint64_t offset, int allocate);
5f4da8c0
TS
465
466static int get_whole_cluster(BlockDriverState *bs, uint64_t cluster_offset,
467 uint64_t offset, int allocate)
468{
469 uint64_t parent_cluster_offset;
470 BDRVVmdkState *s = bs->opaque;
471 uint8_t whole_grain[s->cluster_sectors*512]; // 128 sectors * 512 bytes each = grain size 64KB
472
473 // we will be here if it's first write on non-exist grain(cluster).
474 // try to read from parent image, if exist
475 if (s->hd->backing_hd) {
476 BDRVVmdkState *ps = s->hd->backing_hd->opaque;
477
478 if (!vmdk_is_cid_valid(bs))
479 return -1;
5f4da8c0 480
630530a6
TS
481 parent_cluster_offset = get_cluster_offset(s->hd->backing_hd, NULL, offset, allocate);
482
483 if (parent_cluster_offset) {
484 BDRVVmdkState *act_s = activeBDRV.hd->opaque;
485
486 if (bdrv_pread(ps->hd, parent_cluster_offset, whole_grain, ps->cluster_sectors*512) != ps->cluster_sectors*512)
487 return -1;
488
489 //Write grain only into the active image
490 if (bdrv_pwrite(act_s->hd, activeBDRV.cluster_offset << 9, whole_grain, sizeof(whole_grain)) != sizeof(whole_grain))
491 return -1;
492 }
493 }
494 return 0;
495}
496
497static int vmdk_L2update(BlockDriverState *bs, VmdkMetaData *m_data)
498{
499 BDRVVmdkState *s = bs->opaque;
500
501 /* update L2 table */
502 if (bdrv_pwrite(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
503 &(m_data->offset), sizeof(m_data->offset)) != sizeof(m_data->offset))
504 return -1;
505 /* update backup L2 table */
506 if (s->l1_backup_table_offset != 0) {
507 m_data->l2_offset = s->l1_backup_table[m_data->l1_index];
508 if (bdrv_pwrite(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
509 &(m_data->offset), sizeof(m_data->offset)) != sizeof(m_data->offset))
5f4da8c0
TS
510 return -1;
511 }
630530a6 512
5f4da8c0
TS
513 return 0;
514}
515
630530a6 516static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
ff1afc72 517 uint64_t offset, int allocate)
ea2384d3
FB
518{
519 BDRVVmdkState *s = bs->opaque;
520 unsigned int l1_index, l2_offset, l2_index;
521 int min_index, i, j;
630530a6 522 uint32_t min_count, *l2_table, tmp = 0;
ea2384d3 523 uint64_t cluster_offset;
630530a6
TS
524
525 if (m_data)
526 m_data->valid = 0;
527
ea2384d3
FB
528 l1_index = (offset >> 9) / s->l1_entry_sectors;
529 if (l1_index >= s->l1_size)
530 return 0;
531 l2_offset = s->l1_table[l1_index];
532 if (!l2_offset)
533 return 0;
ea2384d3
FB
534 for(i = 0; i < L2_CACHE_SIZE; i++) {
535 if (l2_offset == s->l2_cache_offsets[i]) {
536 /* increment the hit count */
537 if (++s->l2_cache_counts[i] == 0xffffffff) {
538 for(j = 0; j < L2_CACHE_SIZE; j++) {
539 s->l2_cache_counts[j] >>= 1;
540 }
541 }
542 l2_table = s->l2_cache + (i * s->l2_size);
543 goto found;
544 }
545 }
546 /* not found: load a new entry in the least used one */
547 min_index = 0;
548 min_count = 0xffffffff;
549 for(i = 0; i < L2_CACHE_SIZE; i++) {
550 if (s->l2_cache_counts[i] < min_count) {
551 min_count = s->l2_cache_counts[i];
552 min_index = i;
553 }
554 }
555 l2_table = s->l2_cache + (min_index * s->l2_size);
5fafdf24 556 if (bdrv_pread(s->hd, (int64_t)l2_offset * 512, l2_table, s->l2_size * sizeof(uint32_t)) !=
5f4da8c0 557 s->l2_size * sizeof(uint32_t))
ea2384d3 558 return 0;
5f4da8c0 559
ea2384d3
FB
560 s->l2_cache_offsets[min_index] = l2_offset;
561 s->l2_cache_counts[min_index] = 1;
562 found:
563 l2_index = ((offset >> 9) / s->cluster_sectors) % s->l2_size;
564 cluster_offset = le32_to_cpu(l2_table[l2_index]);
630530a6 565
ff1afc72
FB
566 if (!cluster_offset) {
567 if (!allocate)
568 return 0;
630530a6
TS
569 // Avoid the L2 tables update for the images that have snapshots.
570 if (!s->is_parent) {
49140a77 571 cluster_offset = bdrv_getlength(s->hd);
630530a6
TS
572 bdrv_truncate(s->hd, cluster_offset + (s->cluster_sectors << 9));
573
574 cluster_offset >>= 9;
575 tmp = cpu_to_le32(cluster_offset);
576 l2_table[l2_index] = tmp;
577 // Save the active image state
578 activeBDRV.cluster_offset = cluster_offset;
579 activeBDRV.hd = bs;
ff1afc72 580 }
630530a6
TS
581 /* First of all we write grain itself, to avoid race condition
582 * that may to corrupt the image.
583 * This problem may occur because of insufficient space on host disk
584 * or inappropriate VM shutdown.
585 */
5f4da8c0
TS
586 if (get_whole_cluster(bs, cluster_offset, offset, allocate) == -1)
587 return 0;
630530a6
TS
588
589 if (m_data) {
590 m_data->offset = tmp;
591 m_data->l1_index = l1_index;
592 m_data->l2_index = l2_index;
593 m_data->l2_offset = l2_offset;
594 m_data->valid = 1;
595 }
ff1afc72 596 }
ea2384d3
FB
597 cluster_offset <<= 9;
598 return cluster_offset;
599}
600
5fafdf24 601static int vmdk_is_allocated(BlockDriverState *bs, int64_t sector_num,
ea2384d3
FB
602 int nb_sectors, int *pnum)
603{
604 BDRVVmdkState *s = bs->opaque;
605 int index_in_cluster, n;
606 uint64_t cluster_offset;
607
630530a6 608 cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
ea2384d3
FB
609 index_in_cluster = sector_num % s->cluster_sectors;
610 n = s->cluster_sectors - index_in_cluster;
611 if (n > nb_sectors)
612 n = nb_sectors;
613 *pnum = n;
614 return (cluster_offset != 0);
615}
616
5fafdf24 617static int vmdk_read(BlockDriverState *bs, int64_t sector_num,
ea2384d3
FB
618 uint8_t *buf, int nb_sectors)
619{
620 BDRVVmdkState *s = bs->opaque;
5f4da8c0 621 int index_in_cluster, n, ret;
ea2384d3 622 uint64_t cluster_offset;
5f4da8c0 623
ea2384d3 624 while (nb_sectors > 0) {
630530a6 625 cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
ea2384d3
FB
626 index_in_cluster = sector_num % s->cluster_sectors;
627 n = s->cluster_sectors - index_in_cluster;
628 if (n > nb_sectors)
629 n = nb_sectors;
630 if (!cluster_offset) {
5f4da8c0
TS
631 // try to read from parent image, if exist
632 if (s->hd->backing_hd) {
633 if (!vmdk_is_cid_valid(bs))
634 return -1;
635 ret = bdrv_read(s->hd->backing_hd, sector_num, buf, n);
636 if (ret < 0)
637 return -1;
638 } else {
639 memset(buf, 0, 512 * n);
640 }
ea2384d3 641 } else {
5f4da8c0 642 if(bdrv_pread(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
ea2384d3
FB
643 return -1;
644 }
645 nb_sectors -= n;
646 sector_num += n;
647 buf += n * 512;
648 }
649 return 0;
650}
651
5fafdf24 652static int vmdk_write(BlockDriverState *bs, int64_t sector_num,
ea2384d3
FB
653 const uint8_t *buf, int nb_sectors)
654{
ff1afc72 655 BDRVVmdkState *s = bs->opaque;
630530a6 656 VmdkMetaData m_data;
5f4da8c0 657 int index_in_cluster, n;
ff1afc72 658 uint64_t cluster_offset;
5f4da8c0 659 static int cid_update = 0;
ff1afc72 660
630530a6
TS
661 if (sector_num > bs->total_sectors) {
662 fprintf(stderr,
92868412
JM
663 "(VMDK) Wrong offset: sector_num=0x%" PRIx64
664 " total_sectors=0x%" PRIx64 "\n",
630530a6
TS
665 sector_num, bs->total_sectors);
666 return -1;
667 }
668
ff1afc72
FB
669 while (nb_sectors > 0) {
670 index_in_cluster = sector_num & (s->cluster_sectors - 1);
671 n = s->cluster_sectors - index_in_cluster;
672 if (n > nb_sectors)
673 n = nb_sectors;
630530a6 674 cluster_offset = get_cluster_offset(bs, &m_data, sector_num << 9, 1);
ff1afc72
FB
675 if (!cluster_offset)
676 return -1;
630530a6 677
5f4da8c0 678 if (bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
ff1afc72 679 return -1;
630530a6
TS
680 if (m_data.valid) {
681 /* update L2 tables */
682 if (vmdk_L2update(bs, &m_data) == -1)
683 return -1;
684 }
ff1afc72
FB
685 nb_sectors -= n;
686 sector_num += n;
687 buf += n * 512;
5f4da8c0
TS
688
689 // update CID on the first write every time the virtual disk is opened
690 if (!cid_update) {
691 vmdk_write_cid(bs, time(NULL));
692 cid_update++;
693 }
ff1afc72
FB
694 }
695 return 0;
ea2384d3
FB
696}
697
8979b227
FB
698static int vmdk_create(const char *filename, int64_t total_size,
699 const char *backing_file, int flags)
700{
701 int fd, i;
702 VMDK4Header header;
703 uint32_t tmp, magic, grains, gd_size, gt_size, gt_count;
704 char *desc_template =
705 "# Disk DescriptorFile\n"
706 "version=1\n"
707 "CID=%x\n"
708 "parentCID=ffffffff\n"
709 "createType=\"monolithicSparse\"\n"
710 "\n"
711 "# Extent description\n"
712 "RW %lu SPARSE \"%s\"\n"
713 "\n"
714 "# The Disk Data Base \n"
715 "#DDB\n"
716 "\n"
ec36ba14 717 "ddb.virtualHWVersion = \"%d\"\n"
8979b227
FB
718 "ddb.geometry.cylinders = \"%lu\"\n"
719 "ddb.geometry.heads = \"16\"\n"
720 "ddb.geometry.sectors = \"63\"\n"
721 "ddb.adapterType = \"ide\"\n";
722 char desc[1024];
723 const char *real_filename, *temp_str;
724
725 /* XXX: add support for backing file */
5f4da8c0
TS
726 if (backing_file) {
727 return vmdk_snapshot_create(filename, backing_file);
728 }
8979b227
FB
729
730 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE,
731 0644);
732 if (fd < 0)
733 return -1;
734 magic = cpu_to_be32(VMDK4_MAGIC);
735 memset(&header, 0, sizeof(header));
736 header.version = cpu_to_le32(1);
737 header.flags = cpu_to_le32(3); /* ?? */
738 header.capacity = cpu_to_le64(total_size);
739 header.granularity = cpu_to_le64(128);
740 header.num_gtes_per_gte = cpu_to_le32(512);
741
742 grains = (total_size + header.granularity - 1) / header.granularity;
743 gt_size = ((header.num_gtes_per_gte * sizeof(uint32_t)) + 511) >> 9;
744 gt_count = (grains + header.num_gtes_per_gte - 1) / header.num_gtes_per_gte;
745 gd_size = (gt_count * sizeof(uint32_t) + 511) >> 9;
746
747 header.desc_offset = 1;
748 header.desc_size = 20;
749 header.rgd_offset = header.desc_offset + header.desc_size;
750 header.gd_offset = header.rgd_offset + gd_size + (gt_size * gt_count);
751 header.grain_offset =
752 ((header.gd_offset + gd_size + (gt_size * gt_count) +
753 header.granularity - 1) / header.granularity) *
754 header.granularity;
755
756 header.desc_offset = cpu_to_le64(header.desc_offset);
757 header.desc_size = cpu_to_le64(header.desc_size);
758 header.rgd_offset = cpu_to_le64(header.rgd_offset);
759 header.gd_offset = cpu_to_le64(header.gd_offset);
760 header.grain_offset = cpu_to_le64(header.grain_offset);
761
762 header.check_bytes[0] = 0xa;
763 header.check_bytes[1] = 0x20;
764 header.check_bytes[2] = 0xd;
765 header.check_bytes[3] = 0xa;
3b46e624
TS
766
767 /* write all the data */
8979b227
FB
768 write(fd, &magic, sizeof(magic));
769 write(fd, &header, sizeof(header));
770
771 ftruncate(fd, header.grain_offset << 9);
772
773 /* write grain directory */
774 lseek(fd, le64_to_cpu(header.rgd_offset) << 9, SEEK_SET);
775 for (i = 0, tmp = header.rgd_offset + gd_size;
776 i < gt_count; i++, tmp += gt_size)
777 write(fd, &tmp, sizeof(tmp));
3b46e624 778
8979b227
FB
779 /* write backup grain directory */
780 lseek(fd, le64_to_cpu(header.gd_offset) << 9, SEEK_SET);
781 for (i = 0, tmp = header.gd_offset + gd_size;
782 i < gt_count; i++, tmp += gt_size)
783 write(fd, &tmp, sizeof(tmp));
784
785 /* compose the descriptor */
786 real_filename = filename;
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 if ((temp_str = strrchr(real_filename, ':')) != NULL)
792 real_filename = temp_str + 1;
793 sprintf(desc, desc_template, time(NULL), (unsigned long)total_size,
ec36ba14 794 real_filename, (flags & BLOCK_FLAG_COMPAT6 ? 6 : 4), total_size / (63 * 16));
8979b227
FB
795
796 /* write the descriptor */
797 lseek(fd, le64_to_cpu(header.desc_offset) << 9, SEEK_SET);
798 write(fd, desc, strlen(desc));
799
800 close(fd);
801 return 0;
802}
803
e2731add 804static void vmdk_close(BlockDriverState *bs)
ea2384d3
FB
805{
806 BDRVVmdkState *s = bs->opaque;
5f4da8c0 807
ea2384d3
FB
808 qemu_free(s->l1_table);
809 qemu_free(s->l2_cache);
5f4da8c0
TS
810 // try to close parent image, if exist
811 vmdk_parent_close(s->hd);
5cbdd273 812 bdrv_delete(s->hd);
ea2384d3
FB
813}
814
7a6cba61
PB
815static void vmdk_flush(BlockDriverState *bs)
816{
817 BDRVVmdkState *s = bs->opaque;
5f4da8c0 818 bdrv_flush(s->hd);
7a6cba61
PB
819}
820
ea2384d3
FB
821BlockDriver bdrv_vmdk = {
822 "vmdk",
823 sizeof(BDRVVmdkState),
824 vmdk_probe,
825 vmdk_open,
826 vmdk_read,
827 vmdk_write,
828 vmdk_close,
8979b227 829 vmdk_create,
7a6cba61 830 vmdk_flush,
ea2384d3
FB
831 vmdk_is_allocated,
832};