]> git.proxmox.com Git - qemu.git/blame - block-vmdk.c
Don't leak VLANClientState on PCI hot remove
[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;
7ccfb2eb 122 const 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");
363a37d5 156 pstrcpy(tmp_desc, sizeof(tmp_desc), tmp_str);
5f4da8c0
TS
157 if ((p_name = strstr(desc,"CID")) != 0) {
158 p_name += sizeof("CID");
363a37d5
BS
159 snprintf(p_name, sizeof(desc) - (p_name - desc), "%x\n", cid);
160 pstrcat(desc, sizeof(desc), tmp_desc);
5f4da8c0
TS
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];
7ccfb2eb 196 static const char desc_template[] =
5f4da8c0
TS
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"
7ccfb2eb 205 "RW %u SPARSE \"%s\"\n"
5f4da8c0
TS
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
363a37d5
BS
255 snprintf(s_desc, sizeof(s_desc), desc_template, p_cid, p_cid, backing_file,
256 (uint32_t)header.capacity, real_filename);
5f4da8c0
TS
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);
5f4da8c0
TS
279 if (lseek(p_fd, rgd_offset, SEEK_SET) == -1)
280 goto fail_rgd;
281 if (read(p_fd, rgd_buf, gd_size) != gd_size)
282 goto fail_rgd;
283 if (lseek(snp_fd, rgd_offset, SEEK_SET) == -1)
284 goto fail_rgd;
285 if (write(snp_fd, rgd_buf, gd_size) == -1)
286 goto fail_rgd;
287 qemu_free(rgd_buf);
288
289 /* write GD */
290 gd_buf = qemu_malloc(gd_size);
5f4da8c0
TS
291 if (lseek(p_fd, gd_offset, SEEK_SET) == -1)
292 goto fail_gd;
293 if (read(p_fd, gd_buf, gd_size) != gd_size)
294 goto fail_gd;
295 if (lseek(snp_fd, gd_offset, SEEK_SET) == -1)
296 goto fail_gd;
297 if (write(snp_fd, gd_buf, gd_size) == -1)
298 goto fail_gd;
299 qemu_free(gd_buf);
300
301 close(p_fd);
302 close(snp_fd);
303 return 0;
304
305 fail_gd:
306 qemu_free(gd_buf);
3b46e624 307 fail_rgd:
5f4da8c0
TS
308 qemu_free(rgd_buf);
309 fail:
310 close(p_fd);
311 close(snp_fd);
312 return -1;
313}
314
315static void vmdk_parent_close(BlockDriverState *bs)
316{
317 if (bs->backing_hd)
318 bdrv_close(bs->backing_hd);
319}
320
b1d8e52e 321static int parent_open = 0;
5f4da8c0
TS
322static int vmdk_parent_open(BlockDriverState *bs, const char * filename)
323{
324 BDRVVmdkState *s = bs->opaque;
5fafdf24 325 char *p_name;
5f4da8c0
TS
326 char desc[DESC_SIZE];
327 char parent_img_name[1024];
328
329 /* the descriptor offset = 0x200 */
330 if (bdrv_pread(s->hd, 0x200, desc, DESC_SIZE) != DESC_SIZE)
331 return -1;
332
333 if ((p_name = strstr(desc,"parentFileNameHint")) != 0) {
334 char *end_name;
335 struct stat file_buf;
336
337 p_name += sizeof("parentFileNameHint") + 1;
338 if ((end_name = strchr(p_name,'\"')) == 0)
339 return -1;
b34d259a
AZ
340 if ((end_name - p_name) > sizeof (s->hd->backing_file) - 1)
341 return -1;
3b46e624 342
9b8114a8 343 pstrcpy(s->hd->backing_file, end_name - p_name + 1, p_name);
5f4da8c0
TS
344 if (stat(s->hd->backing_file, &file_buf) != 0) {
345 path_combine(parent_img_name, sizeof(parent_img_name),
346 filename, s->hd->backing_file);
347 } else {
363a37d5
BS
348 pstrcpy(parent_img_name, sizeof(parent_img_name),
349 s->hd->backing_file);
5f4da8c0
TS
350 }
351
352 s->hd->backing_hd = bdrv_new("");
353 if (!s->hd->backing_hd) {
354 failure:
355 bdrv_close(s->hd);
ff1afc72 356 return -1;
5f4da8c0 357 }
630530a6
TS
358 parent_open = 1;
359 if (bdrv_open(s->hd->backing_hd, parent_img_name, BDRV_O_RDONLY) < 0)
5f4da8c0 360 goto failure;
630530a6 361 parent_open = 0;
ff1afc72 362 }
5f4da8c0
TS
363
364 return 0;
365}
366
367static int vmdk_open(BlockDriverState *bs, const char *filename, int flags)
368{
369 BDRVVmdkState *s = bs->opaque;
370 uint32_t magic;
371 int l1_size, i, ret;
372
630530a6
TS
373 if (parent_open)
374 // Parent must be opened as RO.
375 flags = BDRV_O_RDONLY;
630530a6 376
b5eff355 377 ret = bdrv_file_open(&s->hd, filename, flags);
5f4da8c0
TS
378 if (ret < 0)
379 return ret;
380 if (bdrv_pread(s->hd, 0, &magic, sizeof(magic)) != sizeof(magic))
ea2384d3 381 goto fail;
5f4da8c0 382
7143c62c 383 magic = be32_to_cpu(magic);
ea2384d3
FB
384 if (magic == VMDK3_MAGIC) {
385 VMDK3Header header;
5f4da8c0
TS
386
387 if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
ea2384d3
FB
388 goto fail;
389 s->cluster_sectors = le32_to_cpu(header.granularity);
390 s->l2_size = 1 << 9;
391 s->l1_size = 1 << 6;
392 bs->total_sectors = le32_to_cpu(header.disk_sectors);
ff1afc72
FB
393 s->l1_table_offset = le32_to_cpu(header.l1dir_offset) << 9;
394 s->l1_backup_table_offset = 0;
ea2384d3
FB
395 s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
396 } else if (magic == VMDK4_MAGIC) {
397 VMDK4Header header;
5f4da8c0
TS
398
399 if (bdrv_pread(s->hd, sizeof(magic), &header, sizeof(header)) != sizeof(header))
ea2384d3 400 goto fail;
bd6ea3c8
FB
401 bs->total_sectors = le64_to_cpu(header.capacity);
402 s->cluster_sectors = le64_to_cpu(header.granularity);
ea2384d3
FB
403 s->l2_size = le32_to_cpu(header.num_gtes_per_gte);
404 s->l1_entry_sectors = s->l2_size * s->cluster_sectors;
405 if (s->l1_entry_sectors <= 0)
406 goto fail;
5fafdf24 407 s->l1_size = (bs->total_sectors + s->l1_entry_sectors - 1)
ea2384d3 408 / s->l1_entry_sectors;
ff1afc72
FB
409 s->l1_table_offset = le64_to_cpu(header.rgd_offset) << 9;
410 s->l1_backup_table_offset = le64_to_cpu(header.gd_offset) << 9;
5f4da8c0 411
630530a6
TS
412 if (parent_open)
413 s->is_parent = 1;
414 else
415 s->is_parent = 0;
416
5f4da8c0
TS
417 // try to open parent images, if exist
418 if (vmdk_parent_open(bs, filename) != 0)
419 goto fail;
420 // write the CID once after the image creation
421 s->parent_cid = vmdk_read_cid(bs,1);
ea2384d3
FB
422 } else {
423 goto fail;
424 }
5f4da8c0 425
ea2384d3
FB
426 /* read the L1 table */
427 l1_size = s->l1_size * sizeof(uint32_t);
428 s->l1_table = qemu_malloc(l1_size);
5f4da8c0 429 if (bdrv_pread(s->hd, s->l1_table_offset, s->l1_table, l1_size) != l1_size)
ea2384d3
FB
430 goto fail;
431 for(i = 0; i < s->l1_size; i++) {
432 le32_to_cpus(&s->l1_table[i]);
433 }
434
ff1afc72
FB
435 if (s->l1_backup_table_offset) {
436 s->l1_backup_table = qemu_malloc(l1_size);
5f4da8c0 437 if (bdrv_pread(s->hd, s->l1_backup_table_offset, s->l1_backup_table, l1_size) != l1_size)
ff1afc72
FB
438 goto fail;
439 for(i = 0; i < s->l1_size; i++) {
440 le32_to_cpus(&s->l1_backup_table[i]);
441 }
442 }
443
ea2384d3 444 s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint32_t));
ea2384d3
FB
445 return 0;
446 fail:
ff1afc72 447 qemu_free(s->l1_backup_table);
ea2384d3
FB
448 qemu_free(s->l1_table);
449 qemu_free(s->l2_cache);
5f4da8c0 450 bdrv_delete(s->hd);
ea2384d3
FB
451 return -1;
452}
453
630530a6
TS
454static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
455 uint64_t offset, int allocate);
5f4da8c0
TS
456
457static int get_whole_cluster(BlockDriverState *bs, uint64_t cluster_offset,
458 uint64_t offset, int allocate)
459{
460 uint64_t parent_cluster_offset;
461 BDRVVmdkState *s = bs->opaque;
462 uint8_t whole_grain[s->cluster_sectors*512]; // 128 sectors * 512 bytes each = grain size 64KB
463
464 // we will be here if it's first write on non-exist grain(cluster).
465 // try to read from parent image, if exist
466 if (s->hd->backing_hd) {
467 BDRVVmdkState *ps = s->hd->backing_hd->opaque;
468
469 if (!vmdk_is_cid_valid(bs))
470 return -1;
5f4da8c0 471
630530a6
TS
472 parent_cluster_offset = get_cluster_offset(s->hd->backing_hd, NULL, offset, allocate);
473
474 if (parent_cluster_offset) {
475 BDRVVmdkState *act_s = activeBDRV.hd->opaque;
476
477 if (bdrv_pread(ps->hd, parent_cluster_offset, whole_grain, ps->cluster_sectors*512) != ps->cluster_sectors*512)
478 return -1;
479
480 //Write grain only into the active image
481 if (bdrv_pwrite(act_s->hd, activeBDRV.cluster_offset << 9, whole_grain, sizeof(whole_grain)) != sizeof(whole_grain))
482 return -1;
483 }
484 }
485 return 0;
486}
487
488static int vmdk_L2update(BlockDriverState *bs, VmdkMetaData *m_data)
489{
490 BDRVVmdkState *s = bs->opaque;
491
492 /* update L2 table */
493 if (bdrv_pwrite(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
494 &(m_data->offset), sizeof(m_data->offset)) != sizeof(m_data->offset))
495 return -1;
496 /* update backup L2 table */
497 if (s->l1_backup_table_offset != 0) {
498 m_data->l2_offset = s->l1_backup_table[m_data->l1_index];
499 if (bdrv_pwrite(s->hd, ((int64_t)m_data->l2_offset * 512) + (m_data->l2_index * sizeof(m_data->offset)),
500 &(m_data->offset), sizeof(m_data->offset)) != sizeof(m_data->offset))
5f4da8c0
TS
501 return -1;
502 }
630530a6 503
5f4da8c0
TS
504 return 0;
505}
506
630530a6 507static uint64_t get_cluster_offset(BlockDriverState *bs, VmdkMetaData *m_data,
ff1afc72 508 uint64_t offset, int allocate)
ea2384d3
FB
509{
510 BDRVVmdkState *s = bs->opaque;
511 unsigned int l1_index, l2_offset, l2_index;
512 int min_index, i, j;
630530a6 513 uint32_t min_count, *l2_table, tmp = 0;
ea2384d3 514 uint64_t cluster_offset;
630530a6
TS
515
516 if (m_data)
517 m_data->valid = 0;
518
ea2384d3
FB
519 l1_index = (offset >> 9) / s->l1_entry_sectors;
520 if (l1_index >= s->l1_size)
521 return 0;
522 l2_offset = s->l1_table[l1_index];
523 if (!l2_offset)
524 return 0;
ea2384d3
FB
525 for(i = 0; i < L2_CACHE_SIZE; i++) {
526 if (l2_offset == s->l2_cache_offsets[i]) {
527 /* increment the hit count */
528 if (++s->l2_cache_counts[i] == 0xffffffff) {
529 for(j = 0; j < L2_CACHE_SIZE; j++) {
530 s->l2_cache_counts[j] >>= 1;
531 }
532 }
533 l2_table = s->l2_cache + (i * s->l2_size);
534 goto found;
535 }
536 }
537 /* not found: load a new entry in the least used one */
538 min_index = 0;
539 min_count = 0xffffffff;
540 for(i = 0; i < L2_CACHE_SIZE; i++) {
541 if (s->l2_cache_counts[i] < min_count) {
542 min_count = s->l2_cache_counts[i];
543 min_index = i;
544 }
545 }
546 l2_table = s->l2_cache + (min_index * s->l2_size);
5fafdf24 547 if (bdrv_pread(s->hd, (int64_t)l2_offset * 512, l2_table, s->l2_size * sizeof(uint32_t)) !=
5f4da8c0 548 s->l2_size * sizeof(uint32_t))
ea2384d3 549 return 0;
5f4da8c0 550
ea2384d3
FB
551 s->l2_cache_offsets[min_index] = l2_offset;
552 s->l2_cache_counts[min_index] = 1;
553 found:
554 l2_index = ((offset >> 9) / s->cluster_sectors) % s->l2_size;
555 cluster_offset = le32_to_cpu(l2_table[l2_index]);
630530a6 556
ff1afc72
FB
557 if (!cluster_offset) {
558 if (!allocate)
559 return 0;
630530a6
TS
560 // Avoid the L2 tables update for the images that have snapshots.
561 if (!s->is_parent) {
49140a77 562 cluster_offset = bdrv_getlength(s->hd);
630530a6
TS
563 bdrv_truncate(s->hd, cluster_offset + (s->cluster_sectors << 9));
564
565 cluster_offset >>= 9;
566 tmp = cpu_to_le32(cluster_offset);
567 l2_table[l2_index] = tmp;
568 // Save the active image state
569 activeBDRV.cluster_offset = cluster_offset;
570 activeBDRV.hd = bs;
ff1afc72 571 }
630530a6
TS
572 /* First of all we write grain itself, to avoid race condition
573 * that may to corrupt the image.
574 * This problem may occur because of insufficient space on host disk
575 * or inappropriate VM shutdown.
576 */
5f4da8c0
TS
577 if (get_whole_cluster(bs, cluster_offset, offset, allocate) == -1)
578 return 0;
630530a6
TS
579
580 if (m_data) {
581 m_data->offset = tmp;
582 m_data->l1_index = l1_index;
583 m_data->l2_index = l2_index;
584 m_data->l2_offset = l2_offset;
585 m_data->valid = 1;
586 }
ff1afc72 587 }
ea2384d3
FB
588 cluster_offset <<= 9;
589 return cluster_offset;
590}
591
5fafdf24 592static int vmdk_is_allocated(BlockDriverState *bs, int64_t sector_num,
ea2384d3
FB
593 int nb_sectors, int *pnum)
594{
595 BDRVVmdkState *s = bs->opaque;
596 int index_in_cluster, n;
597 uint64_t cluster_offset;
598
630530a6 599 cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
ea2384d3
FB
600 index_in_cluster = sector_num % s->cluster_sectors;
601 n = s->cluster_sectors - index_in_cluster;
602 if (n > nb_sectors)
603 n = nb_sectors;
604 *pnum = n;
605 return (cluster_offset != 0);
606}
607
5fafdf24 608static int vmdk_read(BlockDriverState *bs, int64_t sector_num,
ea2384d3
FB
609 uint8_t *buf, int nb_sectors)
610{
611 BDRVVmdkState *s = bs->opaque;
5f4da8c0 612 int index_in_cluster, n, ret;
ea2384d3 613 uint64_t cluster_offset;
5f4da8c0 614
ea2384d3 615 while (nb_sectors > 0) {
630530a6 616 cluster_offset = get_cluster_offset(bs, NULL, sector_num << 9, 0);
ea2384d3
FB
617 index_in_cluster = sector_num % s->cluster_sectors;
618 n = s->cluster_sectors - index_in_cluster;
619 if (n > nb_sectors)
620 n = nb_sectors;
621 if (!cluster_offset) {
5f4da8c0
TS
622 // try to read from parent image, if exist
623 if (s->hd->backing_hd) {
624 if (!vmdk_is_cid_valid(bs))
625 return -1;
626 ret = bdrv_read(s->hd->backing_hd, sector_num, buf, n);
627 if (ret < 0)
628 return -1;
629 } else {
630 memset(buf, 0, 512 * n);
631 }
ea2384d3 632 } else {
5f4da8c0 633 if(bdrv_pread(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
ea2384d3
FB
634 return -1;
635 }
636 nb_sectors -= n;
637 sector_num += n;
638 buf += n * 512;
639 }
640 return 0;
641}
642
5fafdf24 643static int vmdk_write(BlockDriverState *bs, int64_t sector_num,
ea2384d3
FB
644 const uint8_t *buf, int nb_sectors)
645{
ff1afc72 646 BDRVVmdkState *s = bs->opaque;
630530a6 647 VmdkMetaData m_data;
5f4da8c0 648 int index_in_cluster, n;
ff1afc72 649 uint64_t cluster_offset;
5f4da8c0 650 static int cid_update = 0;
ff1afc72 651
630530a6
TS
652 if (sector_num > bs->total_sectors) {
653 fprintf(stderr,
92868412
JM
654 "(VMDK) Wrong offset: sector_num=0x%" PRIx64
655 " total_sectors=0x%" PRIx64 "\n",
630530a6
TS
656 sector_num, bs->total_sectors);
657 return -1;
658 }
659
ff1afc72
FB
660 while (nb_sectors > 0) {
661 index_in_cluster = sector_num & (s->cluster_sectors - 1);
662 n = s->cluster_sectors - index_in_cluster;
663 if (n > nb_sectors)
664 n = nb_sectors;
630530a6 665 cluster_offset = get_cluster_offset(bs, &m_data, sector_num << 9, 1);
ff1afc72
FB
666 if (!cluster_offset)
667 return -1;
630530a6 668
5f4da8c0 669 if (bdrv_pwrite(s->hd, cluster_offset + index_in_cluster * 512, buf, n * 512) != n * 512)
ff1afc72 670 return -1;
630530a6
TS
671 if (m_data.valid) {
672 /* update L2 tables */
673 if (vmdk_L2update(bs, &m_data) == -1)
674 return -1;
675 }
ff1afc72
FB
676 nb_sectors -= n;
677 sector_num += n;
678 buf += n * 512;
5f4da8c0
TS
679
680 // update CID on the first write every time the virtual disk is opened
681 if (!cid_update) {
682 vmdk_write_cid(bs, time(NULL));
683 cid_update++;
684 }
ff1afc72
FB
685 }
686 return 0;
ea2384d3
FB
687}
688
8979b227
FB
689static int vmdk_create(const char *filename, int64_t total_size,
690 const char *backing_file, int flags)
691{
692 int fd, i;
693 VMDK4Header header;
694 uint32_t tmp, magic, grains, gd_size, gt_size, gt_count;
7ccfb2eb 695 static const char desc_template[] =
8979b227
FB
696 "# Disk DescriptorFile\n"
697 "version=1\n"
698 "CID=%x\n"
699 "parentCID=ffffffff\n"
700 "createType=\"monolithicSparse\"\n"
701 "\n"
702 "# Extent description\n"
7fd6d9fc 703 "RW %" PRId64 " SPARSE \"%s\"\n"
8979b227
FB
704 "\n"
705 "# The Disk Data Base \n"
706 "#DDB\n"
707 "\n"
ec36ba14 708 "ddb.virtualHWVersion = \"%d\"\n"
7fd6d9fc 709 "ddb.geometry.cylinders = \"%" PRId64 "\"\n"
8979b227
FB
710 "ddb.geometry.heads = \"16\"\n"
711 "ddb.geometry.sectors = \"63\"\n"
712 "ddb.adapterType = \"ide\"\n";
713 char desc[1024];
714 const char *real_filename, *temp_str;
715
716 /* XXX: add support for backing file */
5f4da8c0
TS
717 if (backing_file) {
718 return vmdk_snapshot_create(filename, backing_file);
719 }
8979b227
FB
720
721 fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE,
722 0644);
723 if (fd < 0)
724 return -1;
725 magic = cpu_to_be32(VMDK4_MAGIC);
726 memset(&header, 0, sizeof(header));
727 header.version = cpu_to_le32(1);
728 header.flags = cpu_to_le32(3); /* ?? */
729 header.capacity = cpu_to_le64(total_size);
730 header.granularity = cpu_to_le64(128);
731 header.num_gtes_per_gte = cpu_to_le32(512);
732
733 grains = (total_size + header.granularity - 1) / header.granularity;
734 gt_size = ((header.num_gtes_per_gte * sizeof(uint32_t)) + 511) >> 9;
735 gt_count = (grains + header.num_gtes_per_gte - 1) / header.num_gtes_per_gte;
736 gd_size = (gt_count * sizeof(uint32_t) + 511) >> 9;
737
738 header.desc_offset = 1;
739 header.desc_size = 20;
740 header.rgd_offset = header.desc_offset + header.desc_size;
741 header.gd_offset = header.rgd_offset + gd_size + (gt_size * gt_count);
742 header.grain_offset =
743 ((header.gd_offset + gd_size + (gt_size * gt_count) +
744 header.granularity - 1) / header.granularity) *
745 header.granularity;
746
747 header.desc_offset = cpu_to_le64(header.desc_offset);
748 header.desc_size = cpu_to_le64(header.desc_size);
749 header.rgd_offset = cpu_to_le64(header.rgd_offset);
750 header.gd_offset = cpu_to_le64(header.gd_offset);
751 header.grain_offset = cpu_to_le64(header.grain_offset);
752
753 header.check_bytes[0] = 0xa;
754 header.check_bytes[1] = 0x20;
755 header.check_bytes[2] = 0xd;
756 header.check_bytes[3] = 0xa;
3b46e624
TS
757
758 /* write all the data */
8979b227
FB
759 write(fd, &magic, sizeof(magic));
760 write(fd, &header, sizeof(header));
761
762 ftruncate(fd, header.grain_offset << 9);
763
764 /* write grain directory */
765 lseek(fd, le64_to_cpu(header.rgd_offset) << 9, SEEK_SET);
766 for (i = 0, tmp = header.rgd_offset + gd_size;
767 i < gt_count; i++, tmp += gt_size)
768 write(fd, &tmp, sizeof(tmp));
3b46e624 769
8979b227
FB
770 /* write backup grain directory */
771 lseek(fd, le64_to_cpu(header.gd_offset) << 9, SEEK_SET);
772 for (i = 0, tmp = header.gd_offset + gd_size;
773 i < gt_count; i++, tmp += gt_size)
774 write(fd, &tmp, sizeof(tmp));
775
776 /* compose the descriptor */
777 real_filename = filename;
778 if ((temp_str = strrchr(real_filename, '\\')) != NULL)
779 real_filename = temp_str + 1;
780 if ((temp_str = strrchr(real_filename, '/')) != NULL)
781 real_filename = temp_str + 1;
782 if ((temp_str = strrchr(real_filename, ':')) != NULL)
783 real_filename = temp_str + 1;
7ccfb2eb 784 snprintf(desc, sizeof(desc), desc_template, (unsigned int)time(NULL),
7fd6d9fc
BS
785 total_size, real_filename,
786 (flags & BLOCK_FLAG_COMPAT6 ? 6 : 4),
787 total_size / (int64_t)(63 * 16));
8979b227
FB
788
789 /* write the descriptor */
790 lseek(fd, le64_to_cpu(header.desc_offset) << 9, SEEK_SET);
791 write(fd, desc, strlen(desc));
792
793 close(fd);
794 return 0;
795}
796
e2731add 797static void vmdk_close(BlockDriverState *bs)
ea2384d3
FB
798{
799 BDRVVmdkState *s = bs->opaque;
5f4da8c0 800
ea2384d3
FB
801 qemu_free(s->l1_table);
802 qemu_free(s->l2_cache);
5f4da8c0
TS
803 // try to close parent image, if exist
804 vmdk_parent_close(s->hd);
5cbdd273 805 bdrv_delete(s->hd);
ea2384d3
FB
806}
807
7a6cba61
PB
808static void vmdk_flush(BlockDriverState *bs)
809{
810 BDRVVmdkState *s = bs->opaque;
5f4da8c0 811 bdrv_flush(s->hd);
7a6cba61
PB
812}
813
ea2384d3
FB
814BlockDriver bdrv_vmdk = {
815 "vmdk",
816 sizeof(BDRVVmdkState),
817 vmdk_probe,
818 vmdk_open,
819 vmdk_read,
820 vmdk_write,
821 vmdk_close,
8979b227 822 vmdk_create,
7a6cba61 823 vmdk_flush,
ea2384d3
FB
824 vmdk_is_allocated,
825};