root / block / vmdk.c @ 7fa60fa3
History | View | Annotate | Download (36.3 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 "qemu-common.h" |
| 27 |
#include "block_int.h" |
| 28 |
#include "module.h" |
| 29 |
|
| 30 |
#define VMDK3_MAGIC (('C' << 24) | ('O' << 16) | ('W' << 8) | 'D') |
| 31 |
#define VMDK4_MAGIC (('K' << 24) | ('D' << 16) | ('M' << 8) | 'V') |
| 32 |
|
| 33 |
typedef struct { |
| 34 |
uint32_t version; |
| 35 |
uint32_t flags; |
| 36 |
uint32_t disk_sectors; |
| 37 |
uint32_t granularity; |
| 38 |
uint32_t l1dir_offset; |
| 39 |
uint32_t l1dir_size; |
| 40 |
uint32_t file_sectors; |
| 41 |
uint32_t cylinders; |
| 42 |
uint32_t heads; |
| 43 |
uint32_t sectors_per_track; |
| 44 |
} VMDK3Header; |
| 45 |
|
| 46 |
typedef struct { |
| 47 |
uint32_t version; |
| 48 |
uint32_t flags; |
| 49 |
int64_t capacity; |
| 50 |
int64_t granularity; |
| 51 |
int64_t desc_offset; |
| 52 |
int64_t desc_size; |
| 53 |
int32_t num_gtes_per_gte; |
| 54 |
int64_t rgd_offset; |
| 55 |
int64_t gd_offset; |
| 56 |
int64_t grain_offset; |
| 57 |
char filler[1]; |
| 58 |
char check_bytes[4]; |
| 59 |
} __attribute__((packed)) VMDK4Header; |
| 60 |
|
| 61 |
#define L2_CACHE_SIZE 16 |
| 62 |
|
| 63 |
typedef struct VmdkExtent { |
| 64 |
BlockDriverState *file; |
| 65 |
bool flat;
|
| 66 |
int64_t sectors; |
| 67 |
int64_t end_sector; |
| 68 |
int64_t flat_start_offset; |
| 69 |
int64_t l1_table_offset; |
| 70 |
int64_t l1_backup_table_offset; |
| 71 |
uint32_t *l1_table; |
| 72 |
uint32_t *l1_backup_table; |
| 73 |
unsigned int l1_size; |
| 74 |
uint32_t l1_entry_sectors; |
| 75 |
|
| 76 |
unsigned int l2_size; |
| 77 |
uint32_t *l2_cache; |
| 78 |
uint32_t l2_cache_offsets[L2_CACHE_SIZE]; |
| 79 |
uint32_t l2_cache_counts[L2_CACHE_SIZE]; |
| 80 |
|
| 81 |
unsigned int cluster_sectors; |
| 82 |
} VmdkExtent; |
| 83 |
|
| 84 |
typedef struct BDRVVmdkState { |
| 85 |
int desc_offset;
|
| 86 |
bool cid_updated;
|
| 87 |
uint32_t parent_cid; |
| 88 |
int num_extents;
|
| 89 |
/* Extent array with num_extents entries, ascend ordered by address */
|
| 90 |
VmdkExtent *extents; |
| 91 |
} BDRVVmdkState; |
| 92 |
|
| 93 |
typedef struct VmdkMetaData { |
| 94 |
uint32_t offset; |
| 95 |
unsigned int l1_index; |
| 96 |
unsigned int l2_index; |
| 97 |
unsigned int l2_offset; |
| 98 |
int valid;
|
| 99 |
} VmdkMetaData; |
| 100 |
|
| 101 |
static int vmdk_probe(const uint8_t *buf, int buf_size, const char *filename) |
| 102 |
{
|
| 103 |
uint32_t magic; |
| 104 |
|
| 105 |
if (buf_size < 4) |
| 106 |
return 0; |
| 107 |
magic = be32_to_cpu(*(uint32_t *)buf); |
| 108 |
if (magic == VMDK3_MAGIC ||
|
| 109 |
magic == VMDK4_MAGIC) {
|
| 110 |
return 100; |
| 111 |
} else {
|
| 112 |
const char *p = (const char *)buf; |
| 113 |
const char *end = p + buf_size; |
| 114 |
while (p < end) {
|
| 115 |
if (*p == '#') { |
| 116 |
/* skip comment line */
|
| 117 |
while (p < end && *p != '\n') { |
| 118 |
p++; |
| 119 |
} |
| 120 |
p++; |
| 121 |
continue;
|
| 122 |
} |
| 123 |
if (*p == ' ') { |
| 124 |
while (p < end && *p == ' ') { |
| 125 |
p++; |
| 126 |
} |
| 127 |
/* skip '\r' if windows line endings used. */
|
| 128 |
if (p < end && *p == '\r') { |
| 129 |
p++; |
| 130 |
} |
| 131 |
/* only accept blank lines before 'version=' line */
|
| 132 |
if (p == end || *p != '\n') { |
| 133 |
return 0; |
| 134 |
} |
| 135 |
p++; |
| 136 |
continue;
|
| 137 |
} |
| 138 |
if (end - p >= strlen("version=X\n")) { |
| 139 |
if (strncmp("version=1\n", p, strlen("version=1\n")) == 0 || |
| 140 |
strncmp("version=2\n", p, strlen("version=2\n")) == 0) { |
| 141 |
return 100; |
| 142 |
} |
| 143 |
} |
| 144 |
if (end - p >= strlen("version=X\r\n")) { |
| 145 |
if (strncmp("version=1\r\n", p, strlen("version=1\r\n")) == 0 || |
| 146 |
strncmp("version=2\r\n", p, strlen("version=2\r\n")) == 0) { |
| 147 |
return 100; |
| 148 |
} |
| 149 |
} |
| 150 |
return 0; |
| 151 |
} |
| 152 |
return 0; |
| 153 |
} |
| 154 |
} |
| 155 |
|
| 156 |
#define CHECK_CID 1 |
| 157 |
|
| 158 |
#define SECTOR_SIZE 512 |
| 159 |
#define DESC_SIZE 20*SECTOR_SIZE // 20 sectors of 512 bytes each |
| 160 |
#define HEADER_SIZE 512 // first sector of 512 bytes |
| 161 |
|
| 162 |
static void vmdk_free_extents(BlockDriverState *bs) |
| 163 |
{
|
| 164 |
int i;
|
| 165 |
BDRVVmdkState *s = bs->opaque; |
| 166 |
|
| 167 |
for (i = 0; i < s->num_extents; i++) { |
| 168 |
qemu_free(s->extents[i].l1_table); |
| 169 |
qemu_free(s->extents[i].l2_cache); |
| 170 |
qemu_free(s->extents[i].l1_backup_table); |
| 171 |
} |
| 172 |
qemu_free(s->extents); |
| 173 |
} |
| 174 |
|
| 175 |
static uint32_t vmdk_read_cid(BlockDriverState *bs, int parent) |
| 176 |
{
|
| 177 |
char desc[DESC_SIZE];
|
| 178 |
uint32_t cid; |
| 179 |
const char *p_name, *cid_str; |
| 180 |
size_t cid_str_size; |
| 181 |
BDRVVmdkState *s = bs->opaque; |
| 182 |
|
| 183 |
if (bdrv_pread(bs->file, s->desc_offset, desc, DESC_SIZE) != DESC_SIZE) {
|
| 184 |
return 0; |
| 185 |
} |
| 186 |
|
| 187 |
if (parent) {
|
| 188 |
cid_str = "parentCID";
|
| 189 |
cid_str_size = sizeof("parentCID"); |
| 190 |
} else {
|
| 191 |
cid_str = "CID";
|
| 192 |
cid_str_size = sizeof("CID"); |
| 193 |
} |
| 194 |
|
| 195 |
if ((p_name = strstr(desc,cid_str)) != NULL) { |
| 196 |
p_name += cid_str_size; |
| 197 |
sscanf(p_name,"%x",&cid);
|
| 198 |
} |
| 199 |
|
| 200 |
return cid;
|
| 201 |
} |
| 202 |
|
| 203 |
static int vmdk_write_cid(BlockDriverState *bs, uint32_t cid) |
| 204 |
{
|
| 205 |
char desc[DESC_SIZE], tmp_desc[DESC_SIZE];
|
| 206 |
char *p_name, *tmp_str;
|
| 207 |
BDRVVmdkState *s = bs->opaque; |
| 208 |
|
| 209 |
memset(desc, 0, sizeof(desc)); |
| 210 |
if (bdrv_pread(bs->file, s->desc_offset, desc, DESC_SIZE) != DESC_SIZE) {
|
| 211 |
return -EIO;
|
| 212 |
} |
| 213 |
|
| 214 |
tmp_str = strstr(desc,"parentCID");
|
| 215 |
pstrcpy(tmp_desc, sizeof(tmp_desc), tmp_str);
|
| 216 |
if ((p_name = strstr(desc,"CID")) != NULL) { |
| 217 |
p_name += sizeof("CID"); |
| 218 |
snprintf(p_name, sizeof(desc) - (p_name - desc), "%x\n", cid); |
| 219 |
pstrcat(desc, sizeof(desc), tmp_desc);
|
| 220 |
} |
| 221 |
|
| 222 |
if (bdrv_pwrite_sync(bs->file, s->desc_offset, desc, DESC_SIZE) < 0) { |
| 223 |
return -EIO;
|
| 224 |
} |
| 225 |
return 0; |
| 226 |
} |
| 227 |
|
| 228 |
static int vmdk_is_cid_valid(BlockDriverState *bs) |
| 229 |
{
|
| 230 |
#ifdef CHECK_CID
|
| 231 |
BDRVVmdkState *s = bs->opaque; |
| 232 |
BlockDriverState *p_bs = bs->backing_hd; |
| 233 |
uint32_t cur_pcid; |
| 234 |
|
| 235 |
if (p_bs) {
|
| 236 |
cur_pcid = vmdk_read_cid(p_bs,0);
|
| 237 |
if (s->parent_cid != cur_pcid)
|
| 238 |
// CID not valid
|
| 239 |
return 0; |
| 240 |
} |
| 241 |
#endif
|
| 242 |
// CID valid
|
| 243 |
return 1; |
| 244 |
} |
| 245 |
|
| 246 |
static int vmdk_snapshot_create(const char *filename, const char *backing_file) |
| 247 |
{
|
| 248 |
int snp_fd, p_fd;
|
| 249 |
int ret;
|
| 250 |
uint32_t p_cid; |
| 251 |
char *p_name, *gd_buf, *rgd_buf;
|
| 252 |
const char *real_filename, *temp_str; |
| 253 |
VMDK4Header header; |
| 254 |
uint32_t gde_entries, gd_size; |
| 255 |
int64_t gd_offset, rgd_offset, capacity, gt_size; |
| 256 |
char p_desc[DESC_SIZE], s_desc[DESC_SIZE], hdr[HEADER_SIZE];
|
| 257 |
static const char desc_template[] = |
| 258 |
"# Disk DescriptorFile\n"
|
| 259 |
"version=1\n"
|
| 260 |
"CID=%x\n"
|
| 261 |
"parentCID=%x\n"
|
| 262 |
"createType=\"monolithicSparse\"\n"
|
| 263 |
"parentFileNameHint=\"%s\"\n"
|
| 264 |
"\n"
|
| 265 |
"# Extent description\n"
|
| 266 |
"RW %u SPARSE \"%s\"\n"
|
| 267 |
"\n"
|
| 268 |
"# The Disk Data Base \n"
|
| 269 |
"#DDB\n"
|
| 270 |
"\n";
|
| 271 |
|
| 272 |
snp_fd = open(filename, O_RDWR | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, 0644);
|
| 273 |
if (snp_fd < 0) |
| 274 |
return -errno;
|
| 275 |
p_fd = open(backing_file, O_RDONLY | O_BINARY | O_LARGEFILE); |
| 276 |
if (p_fd < 0) { |
| 277 |
close(snp_fd); |
| 278 |
return -errno;
|
| 279 |
} |
| 280 |
|
| 281 |
/* read the header */
|
| 282 |
if (lseek(p_fd, 0x0, SEEK_SET) == -1) { |
| 283 |
ret = -errno; |
| 284 |
goto fail;
|
| 285 |
} |
| 286 |
if (read(p_fd, hdr, HEADER_SIZE) != HEADER_SIZE) {
|
| 287 |
ret = -errno; |
| 288 |
goto fail;
|
| 289 |
} |
| 290 |
|
| 291 |
/* write the header */
|
| 292 |
if (lseek(snp_fd, 0x0, SEEK_SET) == -1) { |
| 293 |
ret = -errno; |
| 294 |
goto fail;
|
| 295 |
} |
| 296 |
if (write(snp_fd, hdr, HEADER_SIZE) == -1) { |
| 297 |
ret = -errno; |
| 298 |
goto fail;
|
| 299 |
} |
| 300 |
|
| 301 |
memset(&header, 0, sizeof(header)); |
| 302 |
memcpy(&header,&hdr[4], sizeof(header)); // skip the VMDK4_MAGIC |
| 303 |
|
| 304 |
if (ftruncate(snp_fd, header.grain_offset << 9)) { |
| 305 |
ret = -errno; |
| 306 |
goto fail;
|
| 307 |
} |
| 308 |
/* the descriptor offset = 0x200 */
|
| 309 |
if (lseek(p_fd, 0x200, SEEK_SET) == -1) { |
| 310 |
ret = -errno; |
| 311 |
goto fail;
|
| 312 |
} |
| 313 |
if (read(p_fd, p_desc, DESC_SIZE) != DESC_SIZE) {
|
| 314 |
ret = -errno; |
| 315 |
goto fail;
|
| 316 |
} |
| 317 |
|
| 318 |
if ((p_name = strstr(p_desc,"CID")) != NULL) { |
| 319 |
p_name += sizeof("CID"); |
| 320 |
sscanf(p_name,"%x",&p_cid);
|
| 321 |
} |
| 322 |
|
| 323 |
real_filename = filename; |
| 324 |
if ((temp_str = strrchr(real_filename, '\\')) != NULL) |
| 325 |
real_filename = temp_str + 1;
|
| 326 |
if ((temp_str = strrchr(real_filename, '/')) != NULL) |
| 327 |
real_filename = temp_str + 1;
|
| 328 |
if ((temp_str = strrchr(real_filename, ':')) != NULL) |
| 329 |
real_filename = temp_str + 1;
|
| 330 |
|
| 331 |
snprintf(s_desc, sizeof(s_desc), desc_template, p_cid, p_cid, backing_file,
|
| 332 |
(uint32_t)header.capacity, real_filename); |
| 333 |
|
| 334 |
/* write the descriptor */
|
| 335 |
if (lseek(snp_fd, 0x200, SEEK_SET) == -1) { |
| 336 |
ret = -errno; |
| 337 |
goto fail;
|
| 338 |
} |
| 339 |
if (write(snp_fd, s_desc, strlen(s_desc)) == -1) { |
| 340 |
ret = -errno; |
| 341 |
goto fail;
|
| 342 |
} |
| 343 |
|
| 344 |
gd_offset = header.gd_offset * SECTOR_SIZE; // offset of GD table
|
| 345 |
rgd_offset = header.rgd_offset * SECTOR_SIZE; // offset of RGD table
|
| 346 |
capacity = header.capacity * SECTOR_SIZE; // Extent size
|
| 347 |
/*
|
| 348 |
* Each GDE span 32M disk, means:
|
| 349 |
* 512 GTE per GT, each GTE points to grain
|
| 350 |
*/
|
| 351 |
gt_size = (int64_t)header.num_gtes_per_gte * header.granularity * SECTOR_SIZE; |
| 352 |
if (!gt_size) {
|
| 353 |
ret = -EINVAL; |
| 354 |
goto fail;
|
| 355 |
} |
| 356 |
gde_entries = (uint32_t)(capacity / gt_size); // number of gde/rgde
|
| 357 |
gd_size = gde_entries * sizeof(uint32_t);
|
| 358 |
|
| 359 |
/* write RGD */
|
| 360 |
rgd_buf = qemu_malloc(gd_size); |
| 361 |
if (lseek(p_fd, rgd_offset, SEEK_SET) == -1) { |
| 362 |
ret = -errno; |
| 363 |
goto fail_rgd;
|
| 364 |
} |
| 365 |
if (read(p_fd, rgd_buf, gd_size) != gd_size) {
|
| 366 |
ret = -errno; |
| 367 |
goto fail_rgd;
|
| 368 |
} |
| 369 |
if (lseek(snp_fd, rgd_offset, SEEK_SET) == -1) { |
| 370 |
ret = -errno; |
| 371 |
goto fail_rgd;
|
| 372 |
} |
| 373 |
if (write(snp_fd, rgd_buf, gd_size) == -1) { |
| 374 |
ret = -errno; |
| 375 |
goto fail_rgd;
|
| 376 |
} |
| 377 |
|
| 378 |
/* write GD */
|
| 379 |
gd_buf = qemu_malloc(gd_size); |
| 380 |
if (lseek(p_fd, gd_offset, SEEK_SET) == -1) { |
| 381 |
ret = -errno; |
| 382 |
goto fail_gd;
|
| 383 |
} |
| 384 |
if (read(p_fd, gd_buf, gd_size) != gd_size) {
|
| 385 |
ret = -errno; |
| 386 |
goto fail_gd;
|
| 387 |
} |
| 388 |
if (lseek(snp_fd, gd_offset, SEEK_SET) == -1) { |
| 389 |
ret = -errno; |
| 390 |
goto fail_gd;
|
| 391 |
} |
| 392 |
if (write(snp_fd, gd_buf, gd_size) == -1) { |
| 393 |
ret = -errno; |
| 394 |
goto fail_gd;
|
| 395 |
} |
| 396 |
ret = 0;
|
| 397 |
|
| 398 |
fail_gd:
|
| 399 |
qemu_free(gd_buf); |
| 400 |
fail_rgd:
|
| 401 |
qemu_free(rgd_buf); |
| 402 |
fail:
|
| 403 |
close(p_fd); |
| 404 |
close(snp_fd); |
| 405 |
return ret;
|
| 406 |
} |
| 407 |
|
| 408 |
static int vmdk_parent_open(BlockDriverState *bs) |
| 409 |
{
|
| 410 |
char *p_name;
|
| 411 |
char desc[DESC_SIZE + 1]; |
| 412 |
BDRVVmdkState *s = bs->opaque; |
| 413 |
|
| 414 |
desc[DESC_SIZE] = '\0';
|
| 415 |
if (bdrv_pread(bs->file, s->desc_offset, desc, DESC_SIZE) != DESC_SIZE) {
|
| 416 |
return -1; |
| 417 |
} |
| 418 |
|
| 419 |
if ((p_name = strstr(desc,"parentFileNameHint")) != NULL) { |
| 420 |
char *end_name;
|
| 421 |
|
| 422 |
p_name += sizeof("parentFileNameHint") + 1; |
| 423 |
if ((end_name = strchr(p_name,'\"')) == NULL) |
| 424 |
return -1; |
| 425 |
if ((end_name - p_name) > sizeof (bs->backing_file) - 1) |
| 426 |
return -1; |
| 427 |
|
| 428 |
pstrcpy(bs->backing_file, end_name - p_name + 1, p_name);
|
| 429 |
} |
| 430 |
|
| 431 |
return 0; |
| 432 |
} |
| 433 |
|
| 434 |
/* Create and append extent to the extent array. Return the added VmdkExtent
|
| 435 |
* address. return NULL if allocation failed. */
|
| 436 |
static VmdkExtent *vmdk_add_extent(BlockDriverState *bs,
|
| 437 |
BlockDriverState *file, bool flat, int64_t sectors,
|
| 438 |
int64_t l1_offset, int64_t l1_backup_offset, |
| 439 |
uint32_t l1_size, |
| 440 |
int l2_size, unsigned int cluster_sectors) |
| 441 |
{
|
| 442 |
VmdkExtent *extent; |
| 443 |
BDRVVmdkState *s = bs->opaque; |
| 444 |
|
| 445 |
s->extents = qemu_realloc(s->extents, |
| 446 |
(s->num_extents + 1) * sizeof(VmdkExtent)); |
| 447 |
extent = &s->extents[s->num_extents]; |
| 448 |
s->num_extents++; |
| 449 |
|
| 450 |
memset(extent, 0, sizeof(VmdkExtent)); |
| 451 |
extent->file = file; |
| 452 |
extent->flat = flat; |
| 453 |
extent->sectors = sectors; |
| 454 |
extent->l1_table_offset = l1_offset; |
| 455 |
extent->l1_backup_table_offset = l1_backup_offset; |
| 456 |
extent->l1_size = l1_size; |
| 457 |
extent->l1_entry_sectors = l2_size * cluster_sectors; |
| 458 |
extent->l2_size = l2_size; |
| 459 |
extent->cluster_sectors = cluster_sectors; |
| 460 |
|
| 461 |
if (s->num_extents > 1) { |
| 462 |
extent->end_sector = (*(extent - 1)).end_sector + extent->sectors;
|
| 463 |
} else {
|
| 464 |
extent->end_sector = extent->sectors; |
| 465 |
} |
| 466 |
bs->total_sectors = extent->end_sector; |
| 467 |
return extent;
|
| 468 |
} |
| 469 |
|
| 470 |
static int vmdk_init_tables(BlockDriverState *bs, VmdkExtent *extent) |
| 471 |
{
|
| 472 |
int ret;
|
| 473 |
int l1_size, i;
|
| 474 |
|
| 475 |
/* read the L1 table */
|
| 476 |
l1_size = extent->l1_size * sizeof(uint32_t);
|
| 477 |
extent->l1_table = qemu_malloc(l1_size); |
| 478 |
ret = bdrv_pread(extent->file, |
| 479 |
extent->l1_table_offset, |
| 480 |
extent->l1_table, |
| 481 |
l1_size); |
| 482 |
if (ret < 0) { |
| 483 |
goto fail_l1;
|
| 484 |
} |
| 485 |
for (i = 0; i < extent->l1_size; i++) { |
| 486 |
le32_to_cpus(&extent->l1_table[i]); |
| 487 |
} |
| 488 |
|
| 489 |
if (extent->l1_backup_table_offset) {
|
| 490 |
extent->l1_backup_table = qemu_malloc(l1_size); |
| 491 |
ret = bdrv_pread(extent->file, |
| 492 |
extent->l1_backup_table_offset, |
| 493 |
extent->l1_backup_table, |
| 494 |
l1_size); |
| 495 |
if (ret < 0) { |
| 496 |
goto fail_l1b;
|
| 497 |
} |
| 498 |
for (i = 0; i < extent->l1_size; i++) { |
| 499 |
le32_to_cpus(&extent->l1_backup_table[i]); |
| 500 |
} |
| 501 |
} |
| 502 |
|
| 503 |
extent->l2_cache = |
| 504 |
qemu_malloc(extent->l2_size * L2_CACHE_SIZE * sizeof(uint32_t));
|
| 505 |
return 0; |
| 506 |
fail_l1b:
|
| 507 |
qemu_free(extent->l1_backup_table); |
| 508 |
fail_l1:
|
| 509 |
qemu_free(extent->l1_table); |
| 510 |
return ret;
|
| 511 |
} |
| 512 |
|
| 513 |
static int vmdk_open_vmdk3(BlockDriverState *bs, int flags) |
| 514 |
{
|
| 515 |
int ret;
|
| 516 |
uint32_t magic; |
| 517 |
VMDK3Header header; |
| 518 |
BDRVVmdkState *s = bs->opaque; |
| 519 |
VmdkExtent *extent; |
| 520 |
|
| 521 |
s->desc_offset = 0x200;
|
| 522 |
ret = bdrv_pread(bs->file, sizeof(magic), &header, sizeof(header)); |
| 523 |
if (ret < 0) { |
| 524 |
goto fail;
|
| 525 |
} |
| 526 |
extent = vmdk_add_extent(bs, |
| 527 |
bs->file, false,
|
| 528 |
le32_to_cpu(header.disk_sectors), |
| 529 |
le32_to_cpu(header.l1dir_offset) << 9,
|
| 530 |
0, 1 << 6, 1 << 9, |
| 531 |
le32_to_cpu(header.granularity)); |
| 532 |
ret = vmdk_init_tables(bs, extent); |
| 533 |
if (ret) {
|
| 534 |
/* vmdk_init_tables cleans up on fail, so only free allocation of
|
| 535 |
* vmdk_add_extent here. */
|
| 536 |
goto fail;
|
| 537 |
} |
| 538 |
return 0; |
| 539 |
fail:
|
| 540 |
vmdk_free_extents(bs); |
| 541 |
return ret;
|
| 542 |
} |
| 543 |
|
| 544 |
static int vmdk_open_vmdk4(BlockDriverState *bs, int flags) |
| 545 |
{
|
| 546 |
int ret;
|
| 547 |
uint32_t magic; |
| 548 |
uint32_t l1_size, l1_entry_sectors; |
| 549 |
VMDK4Header header; |
| 550 |
BDRVVmdkState *s = bs->opaque; |
| 551 |
VmdkExtent *extent; |
| 552 |
|
| 553 |
s->desc_offset = 0x200;
|
| 554 |
ret = bdrv_pread(bs->file, sizeof(magic), &header, sizeof(header)); |
| 555 |
if (ret < 0) { |
| 556 |
goto fail;
|
| 557 |
} |
| 558 |
l1_entry_sectors = le32_to_cpu(header.num_gtes_per_gte) |
| 559 |
* le64_to_cpu(header.granularity); |
| 560 |
l1_size = (le64_to_cpu(header.capacity) + l1_entry_sectors - 1)
|
| 561 |
/ l1_entry_sectors; |
| 562 |
extent = vmdk_add_extent(bs, bs->file, false,
|
| 563 |
le64_to_cpu(header.capacity), |
| 564 |
le64_to_cpu(header.gd_offset) << 9,
|
| 565 |
le64_to_cpu(header.rgd_offset) << 9,
|
| 566 |
l1_size, |
| 567 |
le32_to_cpu(header.num_gtes_per_gte), |
| 568 |
le64_to_cpu(header.granularity)); |
| 569 |
if (extent->l1_entry_sectors <= 0) { |
| 570 |
ret = -EINVAL; |
| 571 |
goto fail;
|
| 572 |
} |
| 573 |
/* try to open parent images, if exist */
|
| 574 |
ret = vmdk_parent_open(bs); |
| 575 |
if (ret) {
|
| 576 |
goto fail;
|
| 577 |
} |
| 578 |
s->parent_cid = vmdk_read_cid(bs, 1);
|
| 579 |
ret = vmdk_init_tables(bs, extent); |
| 580 |
if (ret) {
|
| 581 |
goto fail;
|
| 582 |
} |
| 583 |
return 0; |
| 584 |
fail:
|
| 585 |
vmdk_free_extents(bs); |
| 586 |
return ret;
|
| 587 |
} |
| 588 |
|
| 589 |
/* find an option value out of descriptor file */
|
| 590 |
static int vmdk_parse_description(const char *desc, const char *opt_name, |
| 591 |
char *buf, int buf_size) |
| 592 |
{
|
| 593 |
char *opt_pos, *opt_end;
|
| 594 |
const char *end = desc + strlen(desc); |
| 595 |
|
| 596 |
opt_pos = strstr(desc, opt_name); |
| 597 |
if (!opt_pos) {
|
| 598 |
return -1; |
| 599 |
} |
| 600 |
/* Skip "=\"" following opt_name */
|
| 601 |
opt_pos += strlen(opt_name) + 2;
|
| 602 |
if (opt_pos >= end) {
|
| 603 |
return -1; |
| 604 |
} |
| 605 |
opt_end = opt_pos; |
| 606 |
while (opt_end < end && *opt_end != '"') { |
| 607 |
opt_end++; |
| 608 |
} |
| 609 |
if (opt_end == end || buf_size < opt_end - opt_pos + 1) { |
| 610 |
return -1; |
| 611 |
} |
| 612 |
pstrcpy(buf, opt_end - opt_pos + 1, opt_pos);
|
| 613 |
return 0; |
| 614 |
} |
| 615 |
|
| 616 |
static int vmdk_parse_extents(const char *desc, BlockDriverState *bs, |
| 617 |
const char *desc_file_path) |
| 618 |
{
|
| 619 |
int ret;
|
| 620 |
char access[11]; |
| 621 |
char type[11]; |
| 622 |
char fname[512]; |
| 623 |
const char *p = desc; |
| 624 |
int64_t sectors = 0;
|
| 625 |
int64_t flat_offset; |
| 626 |
|
| 627 |
while (*p) {
|
| 628 |
/* parse extent line:
|
| 629 |
* RW [size in sectors] FLAT "file-name.vmdk" OFFSET
|
| 630 |
* or
|
| 631 |
* RW [size in sectors] SPARSE "file-name.vmdk"
|
| 632 |
*/
|
| 633 |
flat_offset = -1;
|
| 634 |
ret = sscanf(p, "%10s %" SCNd64 " %10s %511s %" SCNd64, |
| 635 |
access, §ors, type, fname, &flat_offset); |
| 636 |
if (ret < 4 || strcmp(access, "RW")) { |
| 637 |
goto next_line;
|
| 638 |
} else if (!strcmp(type, "FLAT")) { |
| 639 |
if (ret != 5 || flat_offset < 0) { |
| 640 |
return -EINVAL;
|
| 641 |
} |
| 642 |
} else if (ret != 4) { |
| 643 |
return -EINVAL;
|
| 644 |
} |
| 645 |
|
| 646 |
/* trim the quotation marks around */
|
| 647 |
if (fname[0] == '"') { |
| 648 |
memmove(fname, fname + 1, strlen(fname));
|
| 649 |
if (strlen(fname) <= 1 || fname[strlen(fname) - 1] != '"') { |
| 650 |
return -EINVAL;
|
| 651 |
} |
| 652 |
fname[strlen(fname) - 1] = '\0'; |
| 653 |
} |
| 654 |
if (sectors <= 0 || |
| 655 |
(strcmp(type, "FLAT") && strcmp(type, "SPARSE")) || |
| 656 |
(strcmp(access, "RW"))) {
|
| 657 |
goto next_line;
|
| 658 |
} |
| 659 |
|
| 660 |
/* save to extents array */
|
| 661 |
if (!strcmp(type, "FLAT")) { |
| 662 |
/* FLAT extent */
|
| 663 |
char extent_path[PATH_MAX];
|
| 664 |
BlockDriverState *extent_file; |
| 665 |
VmdkExtent *extent; |
| 666 |
|
| 667 |
path_combine(extent_path, sizeof(extent_path),
|
| 668 |
desc_file_path, fname); |
| 669 |
ret = bdrv_file_open(&extent_file, extent_path, bs->open_flags); |
| 670 |
if (ret) {
|
| 671 |
return ret;
|
| 672 |
} |
| 673 |
extent = vmdk_add_extent(bs, extent_file, true, sectors,
|
| 674 |
0, 0, 0, 0, sectors); |
| 675 |
extent->flat_start_offset = flat_offset; |
| 676 |
} else {
|
| 677 |
/* SPARSE extent, not supported for now */
|
| 678 |
fprintf(stderr, |
| 679 |
"VMDK: Not supported extent type \"%s\""".\n", type); |
| 680 |
return -ENOTSUP;
|
| 681 |
} |
| 682 |
next_line:
|
| 683 |
/* move to next line */
|
| 684 |
while (*p && *p != '\n') { |
| 685 |
p++; |
| 686 |
} |
| 687 |
p++; |
| 688 |
} |
| 689 |
return 0; |
| 690 |
} |
| 691 |
|
| 692 |
static int vmdk_open_desc_file(BlockDriverState *bs, int flags) |
| 693 |
{
|
| 694 |
int ret;
|
| 695 |
char buf[2048]; |
| 696 |
char ct[128]; |
| 697 |
BDRVVmdkState *s = bs->opaque; |
| 698 |
|
| 699 |
ret = bdrv_pread(bs->file, 0, buf, sizeof(buf)); |
| 700 |
if (ret < 0) { |
| 701 |
return ret;
|
| 702 |
} |
| 703 |
buf[2047] = '\0'; |
| 704 |
if (vmdk_parse_description(buf, "createType", ct, sizeof(ct))) { |
| 705 |
return -EINVAL;
|
| 706 |
} |
| 707 |
if (strcmp(ct, "monolithicFlat")) { |
| 708 |
fprintf(stderr, |
| 709 |
"VMDK: Not supported image type \"%s\""".\n", ct); |
| 710 |
return -ENOTSUP;
|
| 711 |
} |
| 712 |
s->desc_offset = 0;
|
| 713 |
ret = vmdk_parse_extents(buf, bs, bs->file->filename); |
| 714 |
if (ret) {
|
| 715 |
return ret;
|
| 716 |
} |
| 717 |
|
| 718 |
/* try to open parent images, if exist */
|
| 719 |
if (vmdk_parent_open(bs)) {
|
| 720 |
qemu_free(s->extents); |
| 721 |
return -EINVAL;
|
| 722 |
} |
| 723 |
s->parent_cid = vmdk_read_cid(bs, 1);
|
| 724 |
return 0; |
| 725 |
} |
| 726 |
|
| 727 |
static int vmdk_open(BlockDriverState *bs, int flags) |
| 728 |
{
|
| 729 |
uint32_t magic; |
| 730 |
|
| 731 |
if (bdrv_pread(bs->file, 0, &magic, sizeof(magic)) != sizeof(magic)) { |
| 732 |
return -EIO;
|
| 733 |
} |
| 734 |
|
| 735 |
magic = be32_to_cpu(magic); |
| 736 |
if (magic == VMDK3_MAGIC) {
|
| 737 |
return vmdk_open_vmdk3(bs, flags);
|
| 738 |
} else if (magic == VMDK4_MAGIC) { |
| 739 |
return vmdk_open_vmdk4(bs, flags);
|
| 740 |
} else {
|
| 741 |
return vmdk_open_desc_file(bs, flags);
|
| 742 |
} |
| 743 |
} |
| 744 |
|
| 745 |
static int get_whole_cluster(BlockDriverState *bs, |
| 746 |
VmdkExtent *extent, |
| 747 |
uint64_t cluster_offset, |
| 748 |
uint64_t offset, |
| 749 |
bool allocate)
|
| 750 |
{
|
| 751 |
/* 128 sectors * 512 bytes each = grain size 64KB */
|
| 752 |
uint8_t whole_grain[extent->cluster_sectors * 512];
|
| 753 |
|
| 754 |
/* we will be here if it's first write on non-exist grain(cluster).
|
| 755 |
* try to read from parent image, if exist */
|
| 756 |
if (bs->backing_hd) {
|
| 757 |
int ret;
|
| 758 |
|
| 759 |
if (!vmdk_is_cid_valid(bs))
|
| 760 |
return -1; |
| 761 |
|
| 762 |
/* floor offset to cluster */
|
| 763 |
offset -= offset % (extent->cluster_sectors * 512);
|
| 764 |
ret = bdrv_read(bs->backing_hd, offset >> 9, whole_grain,
|
| 765 |
extent->cluster_sectors); |
| 766 |
if (ret < 0) { |
| 767 |
return -1; |
| 768 |
} |
| 769 |
|
| 770 |
/* Write grain only into the active image */
|
| 771 |
ret = bdrv_write(extent->file, cluster_offset, whole_grain, |
| 772 |
extent->cluster_sectors); |
| 773 |
if (ret < 0) { |
| 774 |
return -1; |
| 775 |
} |
| 776 |
} |
| 777 |
return 0; |
| 778 |
} |
| 779 |
|
| 780 |
static int vmdk_L2update(VmdkExtent *extent, VmdkMetaData *m_data) |
| 781 |
{
|
| 782 |
/* update L2 table */
|
| 783 |
if (bdrv_pwrite_sync(
|
| 784 |
extent->file, |
| 785 |
((int64_t)m_data->l2_offset * 512)
|
| 786 |
+ (m_data->l2_index * sizeof(m_data->offset)),
|
| 787 |
&(m_data->offset), |
| 788 |
sizeof(m_data->offset)
|
| 789 |
) < 0) {
|
| 790 |
return -1; |
| 791 |
} |
| 792 |
/* update backup L2 table */
|
| 793 |
if (extent->l1_backup_table_offset != 0) { |
| 794 |
m_data->l2_offset = extent->l1_backup_table[m_data->l1_index]; |
| 795 |
if (bdrv_pwrite_sync(
|
| 796 |
extent->file, |
| 797 |
((int64_t)m_data->l2_offset * 512)
|
| 798 |
+ (m_data->l2_index * sizeof(m_data->offset)),
|
| 799 |
&(m_data->offset), sizeof(m_data->offset)
|
| 800 |
) < 0) {
|
| 801 |
return -1; |
| 802 |
} |
| 803 |
} |
| 804 |
|
| 805 |
return 0; |
| 806 |
} |
| 807 |
|
| 808 |
static int get_cluster_offset(BlockDriverState *bs, |
| 809 |
VmdkExtent *extent, |
| 810 |
VmdkMetaData *m_data, |
| 811 |
uint64_t offset, |
| 812 |
int allocate,
|
| 813 |
uint64_t *cluster_offset) |
| 814 |
{
|
| 815 |
unsigned int l1_index, l2_offset, l2_index; |
| 816 |
int min_index, i, j;
|
| 817 |
uint32_t min_count, *l2_table, tmp = 0;
|
| 818 |
|
| 819 |
if (m_data)
|
| 820 |
m_data->valid = 0;
|
| 821 |
if (extent->flat) {
|
| 822 |
*cluster_offset = extent->flat_start_offset; |
| 823 |
return 0; |
| 824 |
} |
| 825 |
|
| 826 |
l1_index = (offset >> 9) / extent->l1_entry_sectors;
|
| 827 |
if (l1_index >= extent->l1_size) {
|
| 828 |
return -1; |
| 829 |
} |
| 830 |
l2_offset = extent->l1_table[l1_index]; |
| 831 |
if (!l2_offset) {
|
| 832 |
return -1; |
| 833 |
} |
| 834 |
for (i = 0; i < L2_CACHE_SIZE; i++) { |
| 835 |
if (l2_offset == extent->l2_cache_offsets[i]) {
|
| 836 |
/* increment the hit count */
|
| 837 |
if (++extent->l2_cache_counts[i] == 0xffffffff) { |
| 838 |
for (j = 0; j < L2_CACHE_SIZE; j++) { |
| 839 |
extent->l2_cache_counts[j] >>= 1;
|
| 840 |
} |
| 841 |
} |
| 842 |
l2_table = extent->l2_cache + (i * extent->l2_size); |
| 843 |
goto found;
|
| 844 |
} |
| 845 |
} |
| 846 |
/* not found: load a new entry in the least used one */
|
| 847 |
min_index = 0;
|
| 848 |
min_count = 0xffffffff;
|
| 849 |
for (i = 0; i < L2_CACHE_SIZE; i++) { |
| 850 |
if (extent->l2_cache_counts[i] < min_count) {
|
| 851 |
min_count = extent->l2_cache_counts[i]; |
| 852 |
min_index = i; |
| 853 |
} |
| 854 |
} |
| 855 |
l2_table = extent->l2_cache + (min_index * extent->l2_size); |
| 856 |
if (bdrv_pread(
|
| 857 |
extent->file, |
| 858 |
(int64_t)l2_offset * 512,
|
| 859 |
l2_table, |
| 860 |
extent->l2_size * sizeof(uint32_t)
|
| 861 |
) != extent->l2_size * sizeof(uint32_t)) {
|
| 862 |
return -1; |
| 863 |
} |
| 864 |
|
| 865 |
extent->l2_cache_offsets[min_index] = l2_offset; |
| 866 |
extent->l2_cache_counts[min_index] = 1;
|
| 867 |
found:
|
| 868 |
l2_index = ((offset >> 9) / extent->cluster_sectors) % extent->l2_size;
|
| 869 |
*cluster_offset = le32_to_cpu(l2_table[l2_index]); |
| 870 |
|
| 871 |
if (!*cluster_offset) {
|
| 872 |
if (!allocate) {
|
| 873 |
return -1; |
| 874 |
} |
| 875 |
|
| 876 |
// Avoid the L2 tables update for the images that have snapshots.
|
| 877 |
*cluster_offset = bdrv_getlength(extent->file); |
| 878 |
bdrv_truncate( |
| 879 |
extent->file, |
| 880 |
*cluster_offset + (extent->cluster_sectors << 9)
|
| 881 |
); |
| 882 |
|
| 883 |
*cluster_offset >>= 9;
|
| 884 |
tmp = cpu_to_le32(*cluster_offset); |
| 885 |
l2_table[l2_index] = tmp; |
| 886 |
|
| 887 |
/* First of all we write grain itself, to avoid race condition
|
| 888 |
* that may to corrupt the image.
|
| 889 |
* This problem may occur because of insufficient space on host disk
|
| 890 |
* or inappropriate VM shutdown.
|
| 891 |
*/
|
| 892 |
if (get_whole_cluster(
|
| 893 |
bs, extent, *cluster_offset, offset, allocate) == -1)
|
| 894 |
return -1; |
| 895 |
|
| 896 |
if (m_data) {
|
| 897 |
m_data->offset = tmp; |
| 898 |
m_data->l1_index = l1_index; |
| 899 |
m_data->l2_index = l2_index; |
| 900 |
m_data->l2_offset = l2_offset; |
| 901 |
m_data->valid = 1;
|
| 902 |
} |
| 903 |
} |
| 904 |
*cluster_offset <<= 9;
|
| 905 |
return 0; |
| 906 |
} |
| 907 |
|
| 908 |
static VmdkExtent *find_extent(BDRVVmdkState *s,
|
| 909 |
int64_t sector_num, VmdkExtent *start_hint) |
| 910 |
{
|
| 911 |
VmdkExtent *extent = start_hint; |
| 912 |
|
| 913 |
if (!extent) {
|
| 914 |
extent = &s->extents[0];
|
| 915 |
} |
| 916 |
while (extent < &s->extents[s->num_extents]) {
|
| 917 |
if (sector_num < extent->end_sector) {
|
| 918 |
return extent;
|
| 919 |
} |
| 920 |
extent++; |
| 921 |
} |
| 922 |
return NULL; |
| 923 |
} |
| 924 |
|
| 925 |
static int vmdk_is_allocated(BlockDriverState *bs, int64_t sector_num, |
| 926 |
int nb_sectors, int *pnum) |
| 927 |
{
|
| 928 |
BDRVVmdkState *s = bs->opaque; |
| 929 |
int64_t index_in_cluster, n, ret; |
| 930 |
uint64_t offset; |
| 931 |
VmdkExtent *extent; |
| 932 |
|
| 933 |
extent = find_extent(s, sector_num, NULL);
|
| 934 |
if (!extent) {
|
| 935 |
return 0; |
| 936 |
} |
| 937 |
ret = get_cluster_offset(bs, extent, NULL,
|
| 938 |
sector_num * 512, 0, &offset); |
| 939 |
/* get_cluster_offset returning 0 means success */
|
| 940 |
ret = !ret; |
| 941 |
|
| 942 |
index_in_cluster = sector_num % extent->cluster_sectors; |
| 943 |
n = extent->cluster_sectors - index_in_cluster; |
| 944 |
if (n > nb_sectors)
|
| 945 |
n = nb_sectors; |
| 946 |
*pnum = n; |
| 947 |
return ret;
|
| 948 |
} |
| 949 |
|
| 950 |
static int vmdk_read(BlockDriverState *bs, int64_t sector_num, |
| 951 |
uint8_t *buf, int nb_sectors)
|
| 952 |
{
|
| 953 |
BDRVVmdkState *s = bs->opaque; |
| 954 |
int ret;
|
| 955 |
uint64_t n, index_in_cluster; |
| 956 |
VmdkExtent *extent = NULL;
|
| 957 |
uint64_t cluster_offset; |
| 958 |
|
| 959 |
while (nb_sectors > 0) { |
| 960 |
extent = find_extent(s, sector_num, extent); |
| 961 |
if (!extent) {
|
| 962 |
return -EIO;
|
| 963 |
} |
| 964 |
ret = get_cluster_offset( |
| 965 |
bs, extent, NULL,
|
| 966 |
sector_num << 9, 0, &cluster_offset); |
| 967 |
index_in_cluster = sector_num % extent->cluster_sectors; |
| 968 |
n = extent->cluster_sectors - index_in_cluster; |
| 969 |
if (n > nb_sectors)
|
| 970 |
n = nb_sectors; |
| 971 |
if (ret) {
|
| 972 |
/* if not allocated, try to read from parent image, if exist */
|
| 973 |
if (bs->backing_hd) {
|
| 974 |
if (!vmdk_is_cid_valid(bs))
|
| 975 |
return -EINVAL;
|
| 976 |
ret = bdrv_read(bs->backing_hd, sector_num, buf, n); |
| 977 |
if (ret < 0) |
| 978 |
return ret;
|
| 979 |
} else {
|
| 980 |
memset(buf, 0, 512 * n); |
| 981 |
} |
| 982 |
} else {
|
| 983 |
ret = bdrv_pread(extent->file, |
| 984 |
cluster_offset + index_in_cluster * 512,
|
| 985 |
buf, n * 512);
|
| 986 |
if (ret < 0) { |
| 987 |
return ret;
|
| 988 |
} |
| 989 |
} |
| 990 |
nb_sectors -= n; |
| 991 |
sector_num += n; |
| 992 |
buf += n * 512;
|
| 993 |
} |
| 994 |
return 0; |
| 995 |
} |
| 996 |
|
| 997 |
static int vmdk_write(BlockDriverState *bs, int64_t sector_num, |
| 998 |
const uint8_t *buf, int nb_sectors) |
| 999 |
{
|
| 1000 |
BDRVVmdkState *s = bs->opaque; |
| 1001 |
VmdkExtent *extent = NULL;
|
| 1002 |
int n, ret;
|
| 1003 |
int64_t index_in_cluster; |
| 1004 |
uint64_t cluster_offset; |
| 1005 |
VmdkMetaData m_data; |
| 1006 |
|
| 1007 |
if (sector_num > bs->total_sectors) {
|
| 1008 |
fprintf(stderr, |
| 1009 |
"(VMDK) Wrong offset: sector_num=0x%" PRIx64
|
| 1010 |
" total_sectors=0x%" PRIx64 "\n", |
| 1011 |
sector_num, bs->total_sectors); |
| 1012 |
return -EIO;
|
| 1013 |
} |
| 1014 |
|
| 1015 |
while (nb_sectors > 0) { |
| 1016 |
extent = find_extent(s, sector_num, extent); |
| 1017 |
if (!extent) {
|
| 1018 |
return -EIO;
|
| 1019 |
} |
| 1020 |
ret = get_cluster_offset( |
| 1021 |
bs, |
| 1022 |
extent, |
| 1023 |
&m_data, |
| 1024 |
sector_num << 9, 1, |
| 1025 |
&cluster_offset); |
| 1026 |
if (ret) {
|
| 1027 |
return -EINVAL;
|
| 1028 |
} |
| 1029 |
index_in_cluster = sector_num % extent->cluster_sectors; |
| 1030 |
n = extent->cluster_sectors - index_in_cluster; |
| 1031 |
if (n > nb_sectors) {
|
| 1032 |
n = nb_sectors; |
| 1033 |
} |
| 1034 |
|
| 1035 |
ret = bdrv_pwrite(extent->file, |
| 1036 |
cluster_offset + index_in_cluster * 512,
|
| 1037 |
buf, |
| 1038 |
n * 512);
|
| 1039 |
if (ret < 0) { |
| 1040 |
return ret;
|
| 1041 |
} |
| 1042 |
if (m_data.valid) {
|
| 1043 |
/* update L2 tables */
|
| 1044 |
if (vmdk_L2update(extent, &m_data) == -1) { |
| 1045 |
return -EIO;
|
| 1046 |
} |
| 1047 |
} |
| 1048 |
nb_sectors -= n; |
| 1049 |
sector_num += n; |
| 1050 |
buf += n * 512;
|
| 1051 |
|
| 1052 |
// update CID on the first write every time the virtual disk is opened
|
| 1053 |
if (!s->cid_updated) {
|
| 1054 |
vmdk_write_cid(bs, time(NULL));
|
| 1055 |
s->cid_updated = true;
|
| 1056 |
} |
| 1057 |
} |
| 1058 |
return 0; |
| 1059 |
} |
| 1060 |
|
| 1061 |
static int vmdk_create(const char *filename, QEMUOptionParameter *options) |
| 1062 |
{
|
| 1063 |
int fd, i;
|
| 1064 |
VMDK4Header header; |
| 1065 |
uint32_t tmp, magic, grains, gd_size, gt_size, gt_count; |
| 1066 |
static const char desc_template[] = |
| 1067 |
"# Disk DescriptorFile\n"
|
| 1068 |
"version=1\n"
|
| 1069 |
"CID=%x\n"
|
| 1070 |
"parentCID=ffffffff\n"
|
| 1071 |
"createType=\"monolithicSparse\"\n"
|
| 1072 |
"\n"
|
| 1073 |
"# Extent description\n"
|
| 1074 |
"RW %" PRId64 " SPARSE \"%s\"\n" |
| 1075 |
"\n"
|
| 1076 |
"# The Disk Data Base \n"
|
| 1077 |
"#DDB\n"
|
| 1078 |
"\n"
|
| 1079 |
"ddb.virtualHWVersion = \"%d\"\n"
|
| 1080 |
"ddb.geometry.cylinders = \"%" PRId64 "\"\n" |
| 1081 |
"ddb.geometry.heads = \"16\"\n"
|
| 1082 |
"ddb.geometry.sectors = \"63\"\n"
|
| 1083 |
"ddb.adapterType = \"ide\"\n";
|
| 1084 |
char desc[1024]; |
| 1085 |
const char *real_filename, *temp_str; |
| 1086 |
int64_t total_size = 0;
|
| 1087 |
const char *backing_file = NULL; |
| 1088 |
int flags = 0; |
| 1089 |
int ret;
|
| 1090 |
|
| 1091 |
// Read out options
|
| 1092 |
while (options && options->name) {
|
| 1093 |
if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
|
| 1094 |
total_size = options->value.n / 512;
|
| 1095 |
} else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) { |
| 1096 |
backing_file = options->value.s; |
| 1097 |
} else if (!strcmp(options->name, BLOCK_OPT_COMPAT6)) { |
| 1098 |
flags |= options->value.n ? BLOCK_FLAG_COMPAT6: 0;
|
| 1099 |
} |
| 1100 |
options++; |
| 1101 |
} |
| 1102 |
|
| 1103 |
/* XXX: add support for backing file */
|
| 1104 |
if (backing_file) {
|
| 1105 |
return vmdk_snapshot_create(filename, backing_file);
|
| 1106 |
} |
| 1107 |
|
| 1108 |
fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, |
| 1109 |
0644);
|
| 1110 |
if (fd < 0) |
| 1111 |
return -errno;
|
| 1112 |
magic = cpu_to_be32(VMDK4_MAGIC); |
| 1113 |
memset(&header, 0, sizeof(header)); |
| 1114 |
header.version = 1;
|
| 1115 |
header.flags = 3; /* ?? */ |
| 1116 |
header.capacity = total_size; |
| 1117 |
header.granularity = 128;
|
| 1118 |
header.num_gtes_per_gte = 512;
|
| 1119 |
|
| 1120 |
grains = (total_size + header.granularity - 1) / header.granularity;
|
| 1121 |
gt_size = ((header.num_gtes_per_gte * sizeof(uint32_t)) + 511) >> 9; |
| 1122 |
gt_count = (grains + header.num_gtes_per_gte - 1) / header.num_gtes_per_gte;
|
| 1123 |
gd_size = (gt_count * sizeof(uint32_t) + 511) >> 9; |
| 1124 |
|
| 1125 |
header.desc_offset = 1;
|
| 1126 |
header.desc_size = 20;
|
| 1127 |
header.rgd_offset = header.desc_offset + header.desc_size; |
| 1128 |
header.gd_offset = header.rgd_offset + gd_size + (gt_size * gt_count); |
| 1129 |
header.grain_offset = |
| 1130 |
((header.gd_offset + gd_size + (gt_size * gt_count) + |
| 1131 |
header.granularity - 1) / header.granularity) *
|
| 1132 |
header.granularity; |
| 1133 |
|
| 1134 |
/* swap endianness for all header fields */
|
| 1135 |
header.version = cpu_to_le32(header.version); |
| 1136 |
header.flags = cpu_to_le32(header.flags); |
| 1137 |
header.capacity = cpu_to_le64(header.capacity); |
| 1138 |
header.granularity = cpu_to_le64(header.granularity); |
| 1139 |
header.num_gtes_per_gte = cpu_to_le32(header.num_gtes_per_gte); |
| 1140 |
header.desc_offset = cpu_to_le64(header.desc_offset); |
| 1141 |
header.desc_size = cpu_to_le64(header.desc_size); |
| 1142 |
header.rgd_offset = cpu_to_le64(header.rgd_offset); |
| 1143 |
header.gd_offset = cpu_to_le64(header.gd_offset); |
| 1144 |
header.grain_offset = cpu_to_le64(header.grain_offset); |
| 1145 |
|
| 1146 |
header.check_bytes[0] = 0xa; |
| 1147 |
header.check_bytes[1] = 0x20; |
| 1148 |
header.check_bytes[2] = 0xd; |
| 1149 |
header.check_bytes[3] = 0xa; |
| 1150 |
|
| 1151 |
/* write all the data */
|
| 1152 |
ret = qemu_write_full(fd, &magic, sizeof(magic));
|
| 1153 |
if (ret != sizeof(magic)) { |
| 1154 |
ret = -errno; |
| 1155 |
goto exit;
|
| 1156 |
} |
| 1157 |
ret = qemu_write_full(fd, &header, sizeof(header));
|
| 1158 |
if (ret != sizeof(header)) { |
| 1159 |
ret = -errno; |
| 1160 |
goto exit;
|
| 1161 |
} |
| 1162 |
|
| 1163 |
ret = ftruncate(fd, le64_to_cpu(header.grain_offset) << 9);
|
| 1164 |
if (ret < 0) { |
| 1165 |
ret = -errno; |
| 1166 |
goto exit;
|
| 1167 |
} |
| 1168 |
|
| 1169 |
/* write grain directory */
|
| 1170 |
lseek(fd, le64_to_cpu(header.rgd_offset) << 9, SEEK_SET);
|
| 1171 |
for (i = 0, tmp = le64_to_cpu(header.rgd_offset) + gd_size; |
| 1172 |
i < gt_count; i++, tmp += gt_size) {
|
| 1173 |
ret = qemu_write_full(fd, &tmp, sizeof(tmp));
|
| 1174 |
if (ret != sizeof(tmp)) { |
| 1175 |
ret = -errno; |
| 1176 |
goto exit;
|
| 1177 |
} |
| 1178 |
} |
| 1179 |
|
| 1180 |
/* write backup grain directory */
|
| 1181 |
lseek(fd, le64_to_cpu(header.gd_offset) << 9, SEEK_SET);
|
| 1182 |
for (i = 0, tmp = le64_to_cpu(header.gd_offset) + gd_size; |
| 1183 |
i < gt_count; i++, tmp += gt_size) {
|
| 1184 |
ret = qemu_write_full(fd, &tmp, sizeof(tmp));
|
| 1185 |
if (ret != sizeof(tmp)) { |
| 1186 |
ret = -errno; |
| 1187 |
goto exit;
|
| 1188 |
} |
| 1189 |
} |
| 1190 |
|
| 1191 |
/* compose the descriptor */
|
| 1192 |
real_filename = filename; |
| 1193 |
if ((temp_str = strrchr(real_filename, '\\')) != NULL) |
| 1194 |
real_filename = temp_str + 1;
|
| 1195 |
if ((temp_str = strrchr(real_filename, '/')) != NULL) |
| 1196 |
real_filename = temp_str + 1;
|
| 1197 |
if ((temp_str = strrchr(real_filename, ':')) != NULL) |
| 1198 |
real_filename = temp_str + 1;
|
| 1199 |
snprintf(desc, sizeof(desc), desc_template, (unsigned int)time(NULL), |
| 1200 |
total_size, real_filename, |
| 1201 |
(flags & BLOCK_FLAG_COMPAT6 ? 6 : 4), |
| 1202 |
total_size / (int64_t)(63 * 16)); |
| 1203 |
|
| 1204 |
/* write the descriptor */
|
| 1205 |
lseek(fd, le64_to_cpu(header.desc_offset) << 9, SEEK_SET);
|
| 1206 |
ret = qemu_write_full(fd, desc, strlen(desc)); |
| 1207 |
if (ret != strlen(desc)) {
|
| 1208 |
ret = -errno; |
| 1209 |
goto exit;
|
| 1210 |
} |
| 1211 |
|
| 1212 |
ret = 0;
|
| 1213 |
exit:
|
| 1214 |
close(fd); |
| 1215 |
return ret;
|
| 1216 |
} |
| 1217 |
|
| 1218 |
static void vmdk_close(BlockDriverState *bs) |
| 1219 |
{
|
| 1220 |
vmdk_free_extents(bs); |
| 1221 |
} |
| 1222 |
|
| 1223 |
static int vmdk_flush(BlockDriverState *bs) |
| 1224 |
{
|
| 1225 |
int i, ret, err;
|
| 1226 |
BDRVVmdkState *s = bs->opaque; |
| 1227 |
|
| 1228 |
ret = bdrv_flush(bs->file); |
| 1229 |
for (i = 0; i < s->num_extents; i++) { |
| 1230 |
err = bdrv_flush(s->extents[i].file); |
| 1231 |
if (err < 0) { |
| 1232 |
ret = err; |
| 1233 |
} |
| 1234 |
} |
| 1235 |
return ret;
|
| 1236 |
} |
| 1237 |
|
| 1238 |
|
| 1239 |
static QEMUOptionParameter vmdk_create_options[] = {
|
| 1240 |
{
|
| 1241 |
.name = BLOCK_OPT_SIZE, |
| 1242 |
.type = OPT_SIZE, |
| 1243 |
.help = "Virtual disk size"
|
| 1244 |
}, |
| 1245 |
{
|
| 1246 |
.name = BLOCK_OPT_BACKING_FILE, |
| 1247 |
.type = OPT_STRING, |
| 1248 |
.help = "File name of a base image"
|
| 1249 |
}, |
| 1250 |
{
|
| 1251 |
.name = BLOCK_OPT_COMPAT6, |
| 1252 |
.type = OPT_FLAG, |
| 1253 |
.help = "VMDK version 6 image"
|
| 1254 |
}, |
| 1255 |
{ NULL }
|
| 1256 |
}; |
| 1257 |
|
| 1258 |
static BlockDriver bdrv_vmdk = {
|
| 1259 |
.format_name = "vmdk",
|
| 1260 |
.instance_size = sizeof(BDRVVmdkState),
|
| 1261 |
.bdrv_probe = vmdk_probe, |
| 1262 |
.bdrv_open = vmdk_open, |
| 1263 |
.bdrv_read = vmdk_read, |
| 1264 |
.bdrv_write = vmdk_write, |
| 1265 |
.bdrv_close = vmdk_close, |
| 1266 |
.bdrv_create = vmdk_create, |
| 1267 |
.bdrv_flush = vmdk_flush, |
| 1268 |
.bdrv_is_allocated = vmdk_is_allocated, |
| 1269 |
|
| 1270 |
.create_options = vmdk_create_options, |
| 1271 |
}; |
| 1272 |
|
| 1273 |
static void bdrv_vmdk_init(void) |
| 1274 |
{
|
| 1275 |
bdrv_register(&bdrv_vmdk); |
| 1276 |
} |
| 1277 |
|
| 1278 |
block_init(bdrv_vmdk_init); |