]>
Commit | Line | Data |
---|---|---|
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/block_int.h" | |
28 | #include "qemu/module.h" | |
29 | #include "migration/migration.h" | |
30 | #include <zlib.h> | |
31 | ||
32 | #define VMDK3_MAGIC (('C' << 24) | ('O' << 16) | ('W' << 8) | 'D') | |
33 | #define VMDK4_MAGIC (('K' << 24) | ('D' << 16) | ('M' << 8) | 'V') | |
34 | #define VMDK4_COMPRESSION_DEFLATE 1 | |
35 | #define VMDK4_FLAG_RGD (1 << 1) | |
36 | #define VMDK4_FLAG_COMPRESS (1 << 16) | |
37 | #define VMDK4_FLAG_MARKER (1 << 17) | |
38 | #define VMDK4_GD_AT_END 0xffffffffffffffffULL | |
39 | ||
40 | ||
41 | /* VMDK internal error codes */ | |
42 | #define VMDK_OK 0 | |
43 | #define VMDK_ERROR (-1) | |
44 | /* Cluster not allocated */ | |
45 | #define VMDK_UNALLOC (-2) | |
46 | #define VMDK_ZEROED (-3) | |
47 | ||
48 | typedef struct { | |
49 | uint32_t version; | |
50 | uint32_t flags; | |
51 | uint32_t disk_sectors; | |
52 | uint32_t granularity; | |
53 | uint32_t l1dir_offset; | |
54 | uint32_t l1dir_size; | |
55 | uint32_t file_sectors; | |
56 | uint32_t cylinders; | |
57 | uint32_t heads; | |
58 | uint32_t sectors_per_track; | |
59 | } VMDK3Header; | |
60 | ||
61 | typedef struct { | |
62 | uint32_t version; | |
63 | uint32_t flags; | |
64 | int64_t capacity; | |
65 | int64_t granularity; | |
66 | int64_t desc_offset; | |
67 | int64_t desc_size; | |
68 | int32_t num_gtes_per_gte; | |
69 | int64_t rgd_offset; | |
70 | int64_t gd_offset; | |
71 | int64_t grain_offset; | |
72 | char filler[1]; | |
73 | char check_bytes[4]; | |
74 | uint16_t compressAlgorithm; | |
75 | } QEMU_PACKED VMDK4Header; | |
76 | ||
77 | #define L2_CACHE_SIZE 16 | |
78 | ||
79 | typedef struct VmdkExtent { | |
80 | BlockDriverState *file; | |
81 | bool flat; | |
82 | bool compressed; | |
83 | bool has_marker; | |
84 | int64_t sectors; | |
85 | int64_t end_sector; | |
86 | int64_t flat_start_offset; | |
87 | int64_t l1_table_offset; | |
88 | int64_t l1_backup_table_offset; | |
89 | uint32_t *l1_table; | |
90 | uint32_t *l1_backup_table; | |
91 | unsigned int l1_size; | |
92 | uint32_t l1_entry_sectors; | |
93 | ||
94 | unsigned int l2_size; | |
95 | uint32_t *l2_cache; | |
96 | uint32_t l2_cache_offsets[L2_CACHE_SIZE]; | |
97 | uint32_t l2_cache_counts[L2_CACHE_SIZE]; | |
98 | ||
99 | unsigned int cluster_sectors; | |
100 | } VmdkExtent; | |
101 | ||
102 | typedef struct BDRVVmdkState { | |
103 | CoMutex lock; | |
104 | int desc_offset; | |
105 | bool cid_updated; | |
106 | uint32_t parent_cid; | |
107 | int num_extents; | |
108 | /* Extent array with num_extents entries, ascend ordered by address */ | |
109 | VmdkExtent *extents; | |
110 | Error *migration_blocker; | |
111 | } BDRVVmdkState; | |
112 | ||
113 | typedef struct VmdkMetaData { | |
114 | uint32_t offset; | |
115 | unsigned int l1_index; | |
116 | unsigned int l2_index; | |
117 | unsigned int l2_offset; | |
118 | int valid; | |
119 | } VmdkMetaData; | |
120 | ||
121 | typedef struct VmdkGrainMarker { | |
122 | uint64_t lba; | |
123 | uint32_t size; | |
124 | uint8_t data[0]; | |
125 | } VmdkGrainMarker; | |
126 | ||
127 | enum { | |
128 | MARKER_END_OF_STREAM = 0, | |
129 | MARKER_GRAIN_TABLE = 1, | |
130 | MARKER_GRAIN_DIRECTORY = 2, | |
131 | MARKER_FOOTER = 3, | |
132 | }; | |
133 | ||
134 | static int vmdk_probe(const uint8_t *buf, int buf_size, const char *filename) | |
135 | { | |
136 | uint32_t magic; | |
137 | ||
138 | if (buf_size < 4) { | |
139 | return 0; | |
140 | } | |
141 | magic = be32_to_cpu(*(uint32_t *)buf); | |
142 | if (magic == VMDK3_MAGIC || | |
143 | magic == VMDK4_MAGIC) { | |
144 | return 100; | |
145 | } else { | |
146 | const char *p = (const char *)buf; | |
147 | const char *end = p + buf_size; | |
148 | while (p < end) { | |
149 | if (*p == '#') { | |
150 | /* skip comment line */ | |
151 | while (p < end && *p != '\n') { | |
152 | p++; | |
153 | } | |
154 | p++; | |
155 | continue; | |
156 | } | |
157 | if (*p == ' ') { | |
158 | while (p < end && *p == ' ') { | |
159 | p++; | |
160 | } | |
161 | /* skip '\r' if windows line endings used. */ | |
162 | if (p < end && *p == '\r') { | |
163 | p++; | |
164 | } | |
165 | /* only accept blank lines before 'version=' line */ | |
166 | if (p == end || *p != '\n') { | |
167 | return 0; | |
168 | } | |
169 | p++; | |
170 | continue; | |
171 | } | |
172 | if (end - p >= strlen("version=X\n")) { | |
173 | if (strncmp("version=1\n", p, strlen("version=1\n")) == 0 || | |
174 | strncmp("version=2\n", p, strlen("version=2\n")) == 0) { | |
175 | return 100; | |
176 | } | |
177 | } | |
178 | if (end - p >= strlen("version=X\r\n")) { | |
179 | if (strncmp("version=1\r\n", p, strlen("version=1\r\n")) == 0 || | |
180 | strncmp("version=2\r\n", p, strlen("version=2\r\n")) == 0) { | |
181 | return 100; | |
182 | } | |
183 | } | |
184 | return 0; | |
185 | } | |
186 | return 0; | |
187 | } | |
188 | } | |
189 | ||
190 | #define CHECK_CID 1 | |
191 | ||
192 | #define SECTOR_SIZE 512 | |
193 | #define DESC_SIZE (20 * SECTOR_SIZE) /* 20 sectors of 512 bytes each */ | |
194 | #define BUF_SIZE 4096 | |
195 | #define HEADER_SIZE 512 /* first sector of 512 bytes */ | |
196 | ||
197 | static void vmdk_free_extents(BlockDriverState *bs) | |
198 | { | |
199 | int i; | |
200 | BDRVVmdkState *s = bs->opaque; | |
201 | VmdkExtent *e; | |
202 | ||
203 | for (i = 0; i < s->num_extents; i++) { | |
204 | e = &s->extents[i]; | |
205 | g_free(e->l1_table); | |
206 | g_free(e->l2_cache); | |
207 | g_free(e->l1_backup_table); | |
208 | if (e->file != bs->file) { | |
209 | bdrv_delete(e->file); | |
210 | } | |
211 | } | |
212 | g_free(s->extents); | |
213 | } | |
214 | ||
215 | static void vmdk_free_last_extent(BlockDriverState *bs) | |
216 | { | |
217 | BDRVVmdkState *s = bs->opaque; | |
218 | ||
219 | if (s->num_extents == 0) { | |
220 | return; | |
221 | } | |
222 | s->num_extents--; | |
223 | s->extents = g_realloc(s->extents, s->num_extents * sizeof(VmdkExtent)); | |
224 | } | |
225 | ||
226 | static uint32_t vmdk_read_cid(BlockDriverState *bs, int parent) | |
227 | { | |
228 | char desc[DESC_SIZE]; | |
229 | uint32_t cid = 0xffffffff; | |
230 | const char *p_name, *cid_str; | |
231 | size_t cid_str_size; | |
232 | BDRVVmdkState *s = bs->opaque; | |
233 | int ret; | |
234 | ||
235 | ret = bdrv_pread(bs->file, s->desc_offset, desc, DESC_SIZE); | |
236 | if (ret < 0) { | |
237 | return 0; | |
238 | } | |
239 | ||
240 | if (parent) { | |
241 | cid_str = "parentCID"; | |
242 | cid_str_size = sizeof("parentCID"); | |
243 | } else { | |
244 | cid_str = "CID"; | |
245 | cid_str_size = sizeof("CID"); | |
246 | } | |
247 | ||
248 | desc[DESC_SIZE - 1] = '\0'; | |
249 | p_name = strstr(desc, cid_str); | |
250 | if (p_name != NULL) { | |
251 | p_name += cid_str_size; | |
252 | sscanf(p_name, "%x", &cid); | |
253 | } | |
254 | ||
255 | return cid; | |
256 | } | |
257 | ||
258 | static int vmdk_write_cid(BlockDriverState *bs, uint32_t cid) | |
259 | { | |
260 | char desc[DESC_SIZE], tmp_desc[DESC_SIZE]; | |
261 | char *p_name, *tmp_str; | |
262 | BDRVVmdkState *s = bs->opaque; | |
263 | int ret; | |
264 | ||
265 | ret = bdrv_pread(bs->file, s->desc_offset, desc, DESC_SIZE); | |
266 | if (ret < 0) { | |
267 | return ret; | |
268 | } | |
269 | ||
270 | desc[DESC_SIZE - 1] = '\0'; | |
271 | tmp_str = strstr(desc, "parentCID"); | |
272 | if (tmp_str == NULL) { | |
273 | return -EINVAL; | |
274 | } | |
275 | ||
276 | pstrcpy(tmp_desc, sizeof(tmp_desc), tmp_str); | |
277 | p_name = strstr(desc, "CID"); | |
278 | if (p_name != NULL) { | |
279 | p_name += sizeof("CID"); | |
280 | snprintf(p_name, sizeof(desc) - (p_name - desc), "%x\n", cid); | |
281 | pstrcat(desc, sizeof(desc), tmp_desc); | |
282 | } | |
283 | ||
284 | ret = bdrv_pwrite_sync(bs->file, s->desc_offset, desc, DESC_SIZE); | |
285 | if (ret < 0) { | |
286 | return ret; | |
287 | } | |
288 | ||
289 | return 0; | |
290 | } | |
291 | ||
292 | static int vmdk_is_cid_valid(BlockDriverState *bs) | |
293 | { | |
294 | #ifdef CHECK_CID | |
295 | BDRVVmdkState *s = bs->opaque; | |
296 | BlockDriverState *p_bs = bs->backing_hd; | |
297 | uint32_t cur_pcid; | |
298 | ||
299 | if (p_bs) { | |
300 | cur_pcid = vmdk_read_cid(p_bs, 0); | |
301 | if (s->parent_cid != cur_pcid) { | |
302 | /* CID not valid */ | |
303 | return 0; | |
304 | } | |
305 | } | |
306 | #endif | |
307 | /* CID valid */ | |
308 | return 1; | |
309 | } | |
310 | ||
311 | /* Queue extents, if any, for reopen() */ | |
312 | static int vmdk_reopen_prepare(BDRVReopenState *state, | |
313 | BlockReopenQueue *queue, Error **errp) | |
314 | { | |
315 | BDRVVmdkState *s; | |
316 | int ret = -1; | |
317 | int i; | |
318 | VmdkExtent *e; | |
319 | ||
320 | assert(state != NULL); | |
321 | assert(state->bs != NULL); | |
322 | ||
323 | if (queue == NULL) { | |
324 | error_set(errp, ERROR_CLASS_GENERIC_ERROR, | |
325 | "No reopen queue for VMDK extents"); | |
326 | goto exit; | |
327 | } | |
328 | ||
329 | s = state->bs->opaque; | |
330 | ||
331 | assert(s != NULL); | |
332 | ||
333 | for (i = 0; i < s->num_extents; i++) { | |
334 | e = &s->extents[i]; | |
335 | if (e->file != state->bs->file) { | |
336 | bdrv_reopen_queue(queue, e->file, state->flags); | |
337 | } | |
338 | } | |
339 | ret = 0; | |
340 | ||
341 | exit: | |
342 | return ret; | |
343 | } | |
344 | ||
345 | static int vmdk_parent_open(BlockDriverState *bs) | |
346 | { | |
347 | char *p_name; | |
348 | char desc[DESC_SIZE + 1]; | |
349 | BDRVVmdkState *s = bs->opaque; | |
350 | int ret; | |
351 | ||
352 | desc[DESC_SIZE] = '\0'; | |
353 | ret = bdrv_pread(bs->file, s->desc_offset, desc, DESC_SIZE); | |
354 | if (ret < 0) { | |
355 | return ret; | |
356 | } | |
357 | ||
358 | p_name = strstr(desc, "parentFileNameHint"); | |
359 | if (p_name != NULL) { | |
360 | char *end_name; | |
361 | ||
362 | p_name += sizeof("parentFileNameHint") + 1; | |
363 | end_name = strchr(p_name, '\"'); | |
364 | if (end_name == NULL) { | |
365 | return -EINVAL; | |
366 | } | |
367 | if ((end_name - p_name) > sizeof(bs->backing_file) - 1) { | |
368 | return -EINVAL; | |
369 | } | |
370 | ||
371 | pstrcpy(bs->backing_file, end_name - p_name + 1, p_name); | |
372 | } | |
373 | ||
374 | return 0; | |
375 | } | |
376 | ||
377 | /* Create and append extent to the extent array. Return the added VmdkExtent | |
378 | * address. return NULL if allocation failed. */ | |
379 | static VmdkExtent *vmdk_add_extent(BlockDriverState *bs, | |
380 | BlockDriverState *file, bool flat, int64_t sectors, | |
381 | int64_t l1_offset, int64_t l1_backup_offset, | |
382 | uint32_t l1_size, | |
383 | int l2_size, unsigned int cluster_sectors) | |
384 | { | |
385 | VmdkExtent *extent; | |
386 | BDRVVmdkState *s = bs->opaque; | |
387 | ||
388 | s->extents = g_realloc(s->extents, | |
389 | (s->num_extents + 1) * sizeof(VmdkExtent)); | |
390 | extent = &s->extents[s->num_extents]; | |
391 | s->num_extents++; | |
392 | ||
393 | memset(extent, 0, sizeof(VmdkExtent)); | |
394 | extent->file = file; | |
395 | extent->flat = flat; | |
396 | extent->sectors = sectors; | |
397 | extent->l1_table_offset = l1_offset; | |
398 | extent->l1_backup_table_offset = l1_backup_offset; | |
399 | extent->l1_size = l1_size; | |
400 | extent->l1_entry_sectors = l2_size * cluster_sectors; | |
401 | extent->l2_size = l2_size; | |
402 | extent->cluster_sectors = cluster_sectors; | |
403 | ||
404 | if (s->num_extents > 1) { | |
405 | extent->end_sector = (*(extent - 1)).end_sector + extent->sectors; | |
406 | } else { | |
407 | extent->end_sector = extent->sectors; | |
408 | } | |
409 | bs->total_sectors = extent->end_sector; | |
410 | return extent; | |
411 | } | |
412 | ||
413 | static int vmdk_init_tables(BlockDriverState *bs, VmdkExtent *extent) | |
414 | { | |
415 | int ret; | |
416 | int l1_size, i; | |
417 | ||
418 | /* read the L1 table */ | |
419 | l1_size = extent->l1_size * sizeof(uint32_t); | |
420 | extent->l1_table = g_malloc(l1_size); | |
421 | ret = bdrv_pread(extent->file, | |
422 | extent->l1_table_offset, | |
423 | extent->l1_table, | |
424 | l1_size); | |
425 | if (ret < 0) { | |
426 | goto fail_l1; | |
427 | } | |
428 | for (i = 0; i < extent->l1_size; i++) { | |
429 | le32_to_cpus(&extent->l1_table[i]); | |
430 | } | |
431 | ||
432 | if (extent->l1_backup_table_offset) { | |
433 | extent->l1_backup_table = g_malloc(l1_size); | |
434 | ret = bdrv_pread(extent->file, | |
435 | extent->l1_backup_table_offset, | |
436 | extent->l1_backup_table, | |
437 | l1_size); | |
438 | if (ret < 0) { | |
439 | goto fail_l1b; | |
440 | } | |
441 | for (i = 0; i < extent->l1_size; i++) { | |
442 | le32_to_cpus(&extent->l1_backup_table[i]); | |
443 | } | |
444 | } | |
445 | ||
446 | extent->l2_cache = | |
447 | g_malloc(extent->l2_size * L2_CACHE_SIZE * sizeof(uint32_t)); | |
448 | return 0; | |
449 | fail_l1b: | |
450 | g_free(extent->l1_backup_table); | |
451 | fail_l1: | |
452 | g_free(extent->l1_table); | |
453 | return ret; | |
454 | } | |
455 | ||
456 | static int vmdk_open_vmdk3(BlockDriverState *bs, | |
457 | BlockDriverState *file, | |
458 | int flags) | |
459 | { | |
460 | int ret; | |
461 | uint32_t magic; | |
462 | VMDK3Header header; | |
463 | VmdkExtent *extent; | |
464 | ||
465 | ret = bdrv_pread(file, sizeof(magic), &header, sizeof(header)); | |
466 | if (ret < 0) { | |
467 | return ret; | |
468 | } | |
469 | extent = vmdk_add_extent(bs, | |
470 | bs->file, false, | |
471 | le32_to_cpu(header.disk_sectors), | |
472 | le32_to_cpu(header.l1dir_offset) << 9, | |
473 | 0, 1 << 6, 1 << 9, | |
474 | le32_to_cpu(header.granularity)); | |
475 | ret = vmdk_init_tables(bs, extent); | |
476 | if (ret) { | |
477 | /* free extent allocated by vmdk_add_extent */ | |
478 | vmdk_free_last_extent(bs); | |
479 | } | |
480 | return ret; | |
481 | } | |
482 | ||
483 | static int vmdk_open_desc_file(BlockDriverState *bs, int flags, | |
484 | int64_t desc_offset); | |
485 | ||
486 | static int vmdk_open_vmdk4(BlockDriverState *bs, | |
487 | BlockDriverState *file, | |
488 | int flags) | |
489 | { | |
490 | int ret; | |
491 | uint32_t magic; | |
492 | uint32_t l1_size, l1_entry_sectors; | |
493 | VMDK4Header header; | |
494 | VmdkExtent *extent; | |
495 | int64_t l1_backup_offset = 0; | |
496 | ||
497 | ret = bdrv_pread(file, sizeof(magic), &header, sizeof(header)); | |
498 | if (ret < 0) { | |
499 | return ret; | |
500 | } | |
501 | if (header.capacity == 0 && header.desc_offset) { | |
502 | return vmdk_open_desc_file(bs, flags, header.desc_offset << 9); | |
503 | } | |
504 | ||
505 | if (le64_to_cpu(header.gd_offset) == VMDK4_GD_AT_END) { | |
506 | /* | |
507 | * The footer takes precedence over the header, so read it in. The | |
508 | * footer starts at offset -1024 from the end: One sector for the | |
509 | * footer, and another one for the end-of-stream marker. | |
510 | */ | |
511 | struct { | |
512 | struct { | |
513 | uint64_t val; | |
514 | uint32_t size; | |
515 | uint32_t type; | |
516 | uint8_t pad[512 - 16]; | |
517 | } QEMU_PACKED footer_marker; | |
518 | ||
519 | uint32_t magic; | |
520 | VMDK4Header header; | |
521 | uint8_t pad[512 - 4 - sizeof(VMDK4Header)]; | |
522 | ||
523 | struct { | |
524 | uint64_t val; | |
525 | uint32_t size; | |
526 | uint32_t type; | |
527 | uint8_t pad[512 - 16]; | |
528 | } QEMU_PACKED eos_marker; | |
529 | } QEMU_PACKED footer; | |
530 | ||
531 | ret = bdrv_pread(file, | |
532 | bs->file->total_sectors * 512 - 1536, | |
533 | &footer, sizeof(footer)); | |
534 | if (ret < 0) { | |
535 | return ret; | |
536 | } | |
537 | ||
538 | /* Some sanity checks for the footer */ | |
539 | if (be32_to_cpu(footer.magic) != VMDK4_MAGIC || | |
540 | le32_to_cpu(footer.footer_marker.size) != 0 || | |
541 | le32_to_cpu(footer.footer_marker.type) != MARKER_FOOTER || | |
542 | le64_to_cpu(footer.eos_marker.val) != 0 || | |
543 | le32_to_cpu(footer.eos_marker.size) != 0 || | |
544 | le32_to_cpu(footer.eos_marker.type) != MARKER_END_OF_STREAM) | |
545 | { | |
546 | return -EINVAL; | |
547 | } | |
548 | ||
549 | header = footer.header; | |
550 | } | |
551 | ||
552 | l1_entry_sectors = le32_to_cpu(header.num_gtes_per_gte) | |
553 | * le64_to_cpu(header.granularity); | |
554 | if (l1_entry_sectors == 0) { | |
555 | return -EINVAL; | |
556 | } | |
557 | l1_size = (le64_to_cpu(header.capacity) + l1_entry_sectors - 1) | |
558 | / l1_entry_sectors; | |
559 | if (le32_to_cpu(header.flags) & VMDK4_FLAG_RGD) { | |
560 | l1_backup_offset = le64_to_cpu(header.rgd_offset) << 9; | |
561 | } | |
562 | extent = vmdk_add_extent(bs, file, false, | |
563 | le64_to_cpu(header.capacity), | |
564 | le64_to_cpu(header.gd_offset) << 9, | |
565 | l1_backup_offset, | |
566 | l1_size, | |
567 | le32_to_cpu(header.num_gtes_per_gte), | |
568 | le64_to_cpu(header.granularity)); | |
569 | extent->compressed = | |
570 | le16_to_cpu(header.compressAlgorithm) == VMDK4_COMPRESSION_DEFLATE; | |
571 | extent->has_marker = le32_to_cpu(header.flags) & VMDK4_FLAG_MARKER; | |
572 | ret = vmdk_init_tables(bs, extent); | |
573 | if (ret) { | |
574 | /* free extent allocated by vmdk_add_extent */ | |
575 | vmdk_free_last_extent(bs); | |
576 | } | |
577 | return ret; | |
578 | } | |
579 | ||
580 | /* find an option value out of descriptor file */ | |
581 | static int vmdk_parse_description(const char *desc, const char *opt_name, | |
582 | char *buf, int buf_size) | |
583 | { | |
584 | char *opt_pos, *opt_end; | |
585 | const char *end = desc + strlen(desc); | |
586 | ||
587 | opt_pos = strstr(desc, opt_name); | |
588 | if (!opt_pos) { | |
589 | return VMDK_ERROR; | |
590 | } | |
591 | /* Skip "=\"" following opt_name */ | |
592 | opt_pos += strlen(opt_name) + 2; | |
593 | if (opt_pos >= end) { | |
594 | return VMDK_ERROR; | |
595 | } | |
596 | opt_end = opt_pos; | |
597 | while (opt_end < end && *opt_end != '"') { | |
598 | opt_end++; | |
599 | } | |
600 | if (opt_end == end || buf_size < opt_end - opt_pos + 1) { | |
601 | return VMDK_ERROR; | |
602 | } | |
603 | pstrcpy(buf, opt_end - opt_pos + 1, opt_pos); | |
604 | return VMDK_OK; | |
605 | } | |
606 | ||
607 | /* Open an extent file and append to bs array */ | |
608 | static int vmdk_open_sparse(BlockDriverState *bs, | |
609 | BlockDriverState *file, | |
610 | int flags) | |
611 | { | |
612 | uint32_t magic; | |
613 | ||
614 | if (bdrv_pread(file, 0, &magic, sizeof(magic)) != sizeof(magic)) { | |
615 | return -EIO; | |
616 | } | |
617 | ||
618 | magic = be32_to_cpu(magic); | |
619 | switch (magic) { | |
620 | case VMDK3_MAGIC: | |
621 | return vmdk_open_vmdk3(bs, file, flags); | |
622 | break; | |
623 | case VMDK4_MAGIC: | |
624 | return vmdk_open_vmdk4(bs, file, flags); | |
625 | break; | |
626 | default: | |
627 | return -EMEDIUMTYPE; | |
628 | break; | |
629 | } | |
630 | } | |
631 | ||
632 | static int vmdk_parse_extents(const char *desc, BlockDriverState *bs, | |
633 | const char *desc_file_path) | |
634 | { | |
635 | int ret; | |
636 | char access[11]; | |
637 | char type[11]; | |
638 | char fname[512]; | |
639 | const char *p = desc; | |
640 | int64_t sectors = 0; | |
641 | int64_t flat_offset; | |
642 | char extent_path[PATH_MAX]; | |
643 | BlockDriverState *extent_file; | |
644 | ||
645 | while (*p) { | |
646 | /* parse extent line: | |
647 | * RW [size in sectors] FLAT "file-name.vmdk" OFFSET | |
648 | * or | |
649 | * RW [size in sectors] SPARSE "file-name.vmdk" | |
650 | */ | |
651 | flat_offset = -1; | |
652 | ret = sscanf(p, "%10s %" SCNd64 " %10s \"%511[^\n\r\"]\" %" SCNd64, | |
653 | access, §ors, type, fname, &flat_offset); | |
654 | if (ret < 4 || strcmp(access, "RW")) { | |
655 | goto next_line; | |
656 | } else if (!strcmp(type, "FLAT")) { | |
657 | if (ret != 5 || flat_offset < 0) { | |
658 | return -EINVAL; | |
659 | } | |
660 | } else if (ret != 4) { | |
661 | return -EINVAL; | |
662 | } | |
663 | ||
664 | if (sectors <= 0 || | |
665 | (strcmp(type, "FLAT") && strcmp(type, "SPARSE")) || | |
666 | (strcmp(access, "RW"))) { | |
667 | goto next_line; | |
668 | } | |
669 | ||
670 | path_combine(extent_path, sizeof(extent_path), | |
671 | desc_file_path, fname); | |
672 | ret = bdrv_file_open(&extent_file, extent_path, NULL, bs->open_flags); | |
673 | if (ret) { | |
674 | return ret; | |
675 | } | |
676 | ||
677 | /* save to extents array */ | |
678 | if (!strcmp(type, "FLAT")) { | |
679 | /* FLAT extent */ | |
680 | VmdkExtent *extent; | |
681 | ||
682 | extent = vmdk_add_extent(bs, extent_file, true, sectors, | |
683 | 0, 0, 0, 0, sectors); | |
684 | extent->flat_start_offset = flat_offset << 9; | |
685 | } else if (!strcmp(type, "SPARSE")) { | |
686 | /* SPARSE extent */ | |
687 | ret = vmdk_open_sparse(bs, extent_file, bs->open_flags); | |
688 | if (ret) { | |
689 | bdrv_delete(extent_file); | |
690 | return ret; | |
691 | } | |
692 | } else { | |
693 | fprintf(stderr, | |
694 | "VMDK: Not supported extent type \"%s\""".\n", type); | |
695 | return -ENOTSUP; | |
696 | } | |
697 | next_line: | |
698 | /* move to next line */ | |
699 | while (*p && *p != '\n') { | |
700 | p++; | |
701 | } | |
702 | p++; | |
703 | } | |
704 | return 0; | |
705 | } | |
706 | ||
707 | static int vmdk_open_desc_file(BlockDriverState *bs, int flags, | |
708 | int64_t desc_offset) | |
709 | { | |
710 | int ret; | |
711 | char buf[2048]; | |
712 | char ct[128]; | |
713 | BDRVVmdkState *s = bs->opaque; | |
714 | ||
715 | ret = bdrv_pread(bs->file, desc_offset, buf, sizeof(buf)); | |
716 | if (ret < 0) { | |
717 | return ret; | |
718 | } | |
719 | buf[2047] = '\0'; | |
720 | if (vmdk_parse_description(buf, "createType", ct, sizeof(ct))) { | |
721 | return -EMEDIUMTYPE; | |
722 | } | |
723 | if (strcmp(ct, "monolithicFlat") && | |
724 | strcmp(ct, "twoGbMaxExtentSparse") && | |
725 | strcmp(ct, "twoGbMaxExtentFlat")) { | |
726 | fprintf(stderr, | |
727 | "VMDK: Not supported image type \"%s\""".\n", ct); | |
728 | return -ENOTSUP; | |
729 | } | |
730 | s->desc_offset = 0; | |
731 | return vmdk_parse_extents(buf, bs, bs->file->filename); | |
732 | } | |
733 | ||
734 | static int vmdk_open(BlockDriverState *bs, QDict *options, int flags) | |
735 | { | |
736 | int ret; | |
737 | BDRVVmdkState *s = bs->opaque; | |
738 | ||
739 | if (vmdk_open_sparse(bs, bs->file, flags) == 0) { | |
740 | s->desc_offset = 0x200; | |
741 | } else { | |
742 | ret = vmdk_open_desc_file(bs, flags, 0); | |
743 | if (ret) { | |
744 | goto fail; | |
745 | } | |
746 | } | |
747 | /* try to open parent images, if exist */ | |
748 | ret = vmdk_parent_open(bs); | |
749 | if (ret) { | |
750 | goto fail; | |
751 | } | |
752 | s->parent_cid = vmdk_read_cid(bs, 1); | |
753 | qemu_co_mutex_init(&s->lock); | |
754 | ||
755 | /* Disable migration when VMDK images are used */ | |
756 | error_set(&s->migration_blocker, | |
757 | QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED, | |
758 | "vmdk", bs->device_name, "live migration"); | |
759 | migrate_add_blocker(s->migration_blocker); | |
760 | ||
761 | return 0; | |
762 | ||
763 | fail: | |
764 | vmdk_free_extents(bs); | |
765 | return ret; | |
766 | } | |
767 | ||
768 | static int get_whole_cluster(BlockDriverState *bs, | |
769 | VmdkExtent *extent, | |
770 | uint64_t cluster_offset, | |
771 | uint64_t offset, | |
772 | bool allocate) | |
773 | { | |
774 | /* 128 sectors * 512 bytes each = grain size 64KB */ | |
775 | uint8_t whole_grain[extent->cluster_sectors * 512]; | |
776 | ||
777 | /* we will be here if it's first write on non-exist grain(cluster). | |
778 | * try to read from parent image, if exist */ | |
779 | if (bs->backing_hd) { | |
780 | int ret; | |
781 | ||
782 | if (!vmdk_is_cid_valid(bs)) { | |
783 | return VMDK_ERROR; | |
784 | } | |
785 | ||
786 | /* floor offset to cluster */ | |
787 | offset -= offset % (extent->cluster_sectors * 512); | |
788 | ret = bdrv_read(bs->backing_hd, offset >> 9, whole_grain, | |
789 | extent->cluster_sectors); | |
790 | if (ret < 0) { | |
791 | return VMDK_ERROR; | |
792 | } | |
793 | ||
794 | /* Write grain only into the active image */ | |
795 | ret = bdrv_write(extent->file, cluster_offset, whole_grain, | |
796 | extent->cluster_sectors); | |
797 | if (ret < 0) { | |
798 | return VMDK_ERROR; | |
799 | } | |
800 | } | |
801 | return VMDK_OK; | |
802 | } | |
803 | ||
804 | static int vmdk_L2update(VmdkExtent *extent, VmdkMetaData *m_data) | |
805 | { | |
806 | /* update L2 table */ | |
807 | if (bdrv_pwrite_sync( | |
808 | extent->file, | |
809 | ((int64_t)m_data->l2_offset * 512) | |
810 | + (m_data->l2_index * sizeof(m_data->offset)), | |
811 | &(m_data->offset), | |
812 | sizeof(m_data->offset) | |
813 | ) < 0) { | |
814 | return VMDK_ERROR; | |
815 | } | |
816 | /* update backup L2 table */ | |
817 | if (extent->l1_backup_table_offset != 0) { | |
818 | m_data->l2_offset = extent->l1_backup_table[m_data->l1_index]; | |
819 | if (bdrv_pwrite_sync( | |
820 | extent->file, | |
821 | ((int64_t)m_data->l2_offset * 512) | |
822 | + (m_data->l2_index * sizeof(m_data->offset)), | |
823 | &(m_data->offset), sizeof(m_data->offset) | |
824 | ) < 0) { | |
825 | return VMDK_ERROR; | |
826 | } | |
827 | } | |
828 | ||
829 | return VMDK_OK; | |
830 | } | |
831 | ||
832 | static int get_cluster_offset(BlockDriverState *bs, | |
833 | VmdkExtent *extent, | |
834 | VmdkMetaData *m_data, | |
835 | uint64_t offset, | |
836 | int allocate, | |
837 | uint64_t *cluster_offset) | |
838 | { | |
839 | unsigned int l1_index, l2_offset, l2_index; | |
840 | int min_index, i, j; | |
841 | uint32_t min_count, *l2_table, tmp = 0; | |
842 | ||
843 | if (m_data) { | |
844 | m_data->valid = 0; | |
845 | } | |
846 | if (extent->flat) { | |
847 | *cluster_offset = extent->flat_start_offset; | |
848 | return VMDK_OK; | |
849 | } | |
850 | ||
851 | offset -= (extent->end_sector - extent->sectors) * SECTOR_SIZE; | |
852 | l1_index = (offset >> 9) / extent->l1_entry_sectors; | |
853 | if (l1_index >= extent->l1_size) { | |
854 | return VMDK_ERROR; | |
855 | } | |
856 | l2_offset = extent->l1_table[l1_index]; | |
857 | if (!l2_offset) { | |
858 | return VMDK_UNALLOC; | |
859 | } | |
860 | for (i = 0; i < L2_CACHE_SIZE; i++) { | |
861 | if (l2_offset == extent->l2_cache_offsets[i]) { | |
862 | /* increment the hit count */ | |
863 | if (++extent->l2_cache_counts[i] == 0xffffffff) { | |
864 | for (j = 0; j < L2_CACHE_SIZE; j++) { | |
865 | extent->l2_cache_counts[j] >>= 1; | |
866 | } | |
867 | } | |
868 | l2_table = extent->l2_cache + (i * extent->l2_size); | |
869 | goto found; | |
870 | } | |
871 | } | |
872 | /* not found: load a new entry in the least used one */ | |
873 | min_index = 0; | |
874 | min_count = 0xffffffff; | |
875 | for (i = 0; i < L2_CACHE_SIZE; i++) { | |
876 | if (extent->l2_cache_counts[i] < min_count) { | |
877 | min_count = extent->l2_cache_counts[i]; | |
878 | min_index = i; | |
879 | } | |
880 | } | |
881 | l2_table = extent->l2_cache + (min_index * extent->l2_size); | |
882 | if (bdrv_pread( | |
883 | extent->file, | |
884 | (int64_t)l2_offset * 512, | |
885 | l2_table, | |
886 | extent->l2_size * sizeof(uint32_t) | |
887 | ) != extent->l2_size * sizeof(uint32_t)) { | |
888 | return VMDK_ERROR; | |
889 | } | |
890 | ||
891 | extent->l2_cache_offsets[min_index] = l2_offset; | |
892 | extent->l2_cache_counts[min_index] = 1; | |
893 | found: | |
894 | l2_index = ((offset >> 9) / extent->cluster_sectors) % extent->l2_size; | |
895 | *cluster_offset = le32_to_cpu(l2_table[l2_index]); | |
896 | ||
897 | if (!*cluster_offset) { | |
898 | if (!allocate) { | |
899 | return VMDK_UNALLOC; | |
900 | } | |
901 | ||
902 | /* Avoid the L2 tables update for the images that have snapshots. */ | |
903 | *cluster_offset = bdrv_getlength(extent->file); | |
904 | if (!extent->compressed) { | |
905 | bdrv_truncate( | |
906 | extent->file, | |
907 | *cluster_offset + (extent->cluster_sectors << 9) | |
908 | ); | |
909 | } | |
910 | ||
911 | *cluster_offset >>= 9; | |
912 | tmp = cpu_to_le32(*cluster_offset); | |
913 | l2_table[l2_index] = tmp; | |
914 | ||
915 | /* First of all we write grain itself, to avoid race condition | |
916 | * that may to corrupt the image. | |
917 | * This problem may occur because of insufficient space on host disk | |
918 | * or inappropriate VM shutdown. | |
919 | */ | |
920 | if (get_whole_cluster( | |
921 | bs, extent, *cluster_offset, offset, allocate) == -1) { | |
922 | return VMDK_ERROR; | |
923 | } | |
924 | ||
925 | if (m_data) { | |
926 | m_data->offset = tmp; | |
927 | m_data->l1_index = l1_index; | |
928 | m_data->l2_index = l2_index; | |
929 | m_data->l2_offset = l2_offset; | |
930 | m_data->valid = 1; | |
931 | } | |
932 | } | |
933 | *cluster_offset <<= 9; | |
934 | return VMDK_OK; | |
935 | } | |
936 | ||
937 | static VmdkExtent *find_extent(BDRVVmdkState *s, | |
938 | int64_t sector_num, VmdkExtent *start_hint) | |
939 | { | |
940 | VmdkExtent *extent = start_hint; | |
941 | ||
942 | if (!extent) { | |
943 | extent = &s->extents[0]; | |
944 | } | |
945 | while (extent < &s->extents[s->num_extents]) { | |
946 | if (sector_num < extent->end_sector) { | |
947 | return extent; | |
948 | } | |
949 | extent++; | |
950 | } | |
951 | return NULL; | |
952 | } | |
953 | ||
954 | static int coroutine_fn vmdk_co_is_allocated(BlockDriverState *bs, | |
955 | int64_t sector_num, int nb_sectors, int *pnum) | |
956 | { | |
957 | BDRVVmdkState *s = bs->opaque; | |
958 | int64_t index_in_cluster, n, ret; | |
959 | uint64_t offset; | |
960 | VmdkExtent *extent; | |
961 | ||
962 | extent = find_extent(s, sector_num, NULL); | |
963 | if (!extent) { | |
964 | return 0; | |
965 | } | |
966 | qemu_co_mutex_lock(&s->lock); | |
967 | ret = get_cluster_offset(bs, extent, NULL, | |
968 | sector_num * 512, 0, &offset); | |
969 | qemu_co_mutex_unlock(&s->lock); | |
970 | /* get_cluster_offset returning 0 means success */ | |
971 | ret = !ret; | |
972 | ||
973 | index_in_cluster = sector_num % extent->cluster_sectors; | |
974 | n = extent->cluster_sectors - index_in_cluster; | |
975 | if (n > nb_sectors) { | |
976 | n = nb_sectors; | |
977 | } | |
978 | *pnum = n; | |
979 | return ret; | |
980 | } | |
981 | ||
982 | static int vmdk_write_extent(VmdkExtent *extent, int64_t cluster_offset, | |
983 | int64_t offset_in_cluster, const uint8_t *buf, | |
984 | int nb_sectors, int64_t sector_num) | |
985 | { | |
986 | int ret; | |
987 | VmdkGrainMarker *data = NULL; | |
988 | uLongf buf_len; | |
989 | const uint8_t *write_buf = buf; | |
990 | int write_len = nb_sectors * 512; | |
991 | ||
992 | if (extent->compressed) { | |
993 | if (!extent->has_marker) { | |
994 | ret = -EINVAL; | |
995 | goto out; | |
996 | } | |
997 | buf_len = (extent->cluster_sectors << 9) * 2; | |
998 | data = g_malloc(buf_len + sizeof(VmdkGrainMarker)); | |
999 | if (compress(data->data, &buf_len, buf, nb_sectors << 9) != Z_OK || | |
1000 | buf_len == 0) { | |
1001 | ret = -EINVAL; | |
1002 | goto out; | |
1003 | } | |
1004 | data->lba = sector_num; | |
1005 | data->size = buf_len; | |
1006 | write_buf = (uint8_t *)data; | |
1007 | write_len = buf_len + sizeof(VmdkGrainMarker); | |
1008 | } | |
1009 | ret = bdrv_pwrite(extent->file, | |
1010 | cluster_offset + offset_in_cluster, | |
1011 | write_buf, | |
1012 | write_len); | |
1013 | if (ret != write_len) { | |
1014 | ret = ret < 0 ? ret : -EIO; | |
1015 | goto out; | |
1016 | } | |
1017 | ret = 0; | |
1018 | out: | |
1019 | g_free(data); | |
1020 | return ret; | |
1021 | } | |
1022 | ||
1023 | static int vmdk_read_extent(VmdkExtent *extent, int64_t cluster_offset, | |
1024 | int64_t offset_in_cluster, uint8_t *buf, | |
1025 | int nb_sectors) | |
1026 | { | |
1027 | int ret; | |
1028 | int cluster_bytes, buf_bytes; | |
1029 | uint8_t *cluster_buf, *compressed_data; | |
1030 | uint8_t *uncomp_buf; | |
1031 | uint32_t data_len; | |
1032 | VmdkGrainMarker *marker; | |
1033 | uLongf buf_len; | |
1034 | ||
1035 | ||
1036 | if (!extent->compressed) { | |
1037 | ret = bdrv_pread(extent->file, | |
1038 | cluster_offset + offset_in_cluster, | |
1039 | buf, nb_sectors * 512); | |
1040 | if (ret == nb_sectors * 512) { | |
1041 | return 0; | |
1042 | } else { | |
1043 | return -EIO; | |
1044 | } | |
1045 | } | |
1046 | cluster_bytes = extent->cluster_sectors * 512; | |
1047 | /* Read two clusters in case GrainMarker + compressed data > one cluster */ | |
1048 | buf_bytes = cluster_bytes * 2; | |
1049 | cluster_buf = g_malloc(buf_bytes); | |
1050 | uncomp_buf = g_malloc(cluster_bytes); | |
1051 | ret = bdrv_pread(extent->file, | |
1052 | cluster_offset, | |
1053 | cluster_buf, buf_bytes); | |
1054 | if (ret < 0) { | |
1055 | goto out; | |
1056 | } | |
1057 | compressed_data = cluster_buf; | |
1058 | buf_len = cluster_bytes; | |
1059 | data_len = cluster_bytes; | |
1060 | if (extent->has_marker) { | |
1061 | marker = (VmdkGrainMarker *)cluster_buf; | |
1062 | compressed_data = marker->data; | |
1063 | data_len = le32_to_cpu(marker->size); | |
1064 | } | |
1065 | if (!data_len || data_len > buf_bytes) { | |
1066 | ret = -EINVAL; | |
1067 | goto out; | |
1068 | } | |
1069 | ret = uncompress(uncomp_buf, &buf_len, compressed_data, data_len); | |
1070 | if (ret != Z_OK) { | |
1071 | ret = -EINVAL; | |
1072 | goto out; | |
1073 | ||
1074 | } | |
1075 | if (offset_in_cluster < 0 || | |
1076 | offset_in_cluster + nb_sectors * 512 > buf_len) { | |
1077 | ret = -EINVAL; | |
1078 | goto out; | |
1079 | } | |
1080 | memcpy(buf, uncomp_buf + offset_in_cluster, nb_sectors * 512); | |
1081 | ret = 0; | |
1082 | ||
1083 | out: | |
1084 | g_free(uncomp_buf); | |
1085 | g_free(cluster_buf); | |
1086 | return ret; | |
1087 | } | |
1088 | ||
1089 | static int vmdk_read(BlockDriverState *bs, int64_t sector_num, | |
1090 | uint8_t *buf, int nb_sectors) | |
1091 | { | |
1092 | BDRVVmdkState *s = bs->opaque; | |
1093 | int ret; | |
1094 | uint64_t n, index_in_cluster; | |
1095 | uint64_t extent_begin_sector, extent_relative_sector_num; | |
1096 | VmdkExtent *extent = NULL; | |
1097 | uint64_t cluster_offset; | |
1098 | ||
1099 | while (nb_sectors > 0) { | |
1100 | extent = find_extent(s, sector_num, extent); | |
1101 | if (!extent) { | |
1102 | return -EIO; | |
1103 | } | |
1104 | ret = get_cluster_offset( | |
1105 | bs, extent, NULL, | |
1106 | sector_num << 9, 0, &cluster_offset); | |
1107 | extent_begin_sector = extent->end_sector - extent->sectors; | |
1108 | extent_relative_sector_num = sector_num - extent_begin_sector; | |
1109 | index_in_cluster = extent_relative_sector_num % extent->cluster_sectors; | |
1110 | n = extent->cluster_sectors - index_in_cluster; | |
1111 | if (n > nb_sectors) { | |
1112 | n = nb_sectors; | |
1113 | } | |
1114 | if (ret) { | |
1115 | /* if not allocated, try to read from parent image, if exist */ | |
1116 | if (bs->backing_hd) { | |
1117 | if (!vmdk_is_cid_valid(bs)) { | |
1118 | return -EINVAL; | |
1119 | } | |
1120 | ret = bdrv_read(bs->backing_hd, sector_num, buf, n); | |
1121 | if (ret < 0) { | |
1122 | return ret; | |
1123 | } | |
1124 | } else { | |
1125 | memset(buf, 0, 512 * n); | |
1126 | } | |
1127 | } else { | |
1128 | ret = vmdk_read_extent(extent, | |
1129 | cluster_offset, index_in_cluster * 512, | |
1130 | buf, n); | |
1131 | if (ret) { | |
1132 | return ret; | |
1133 | } | |
1134 | } | |
1135 | nb_sectors -= n; | |
1136 | sector_num += n; | |
1137 | buf += n * 512; | |
1138 | } | |
1139 | return 0; | |
1140 | } | |
1141 | ||
1142 | static coroutine_fn int vmdk_co_read(BlockDriverState *bs, int64_t sector_num, | |
1143 | uint8_t *buf, int nb_sectors) | |
1144 | { | |
1145 | int ret; | |
1146 | BDRVVmdkState *s = bs->opaque; | |
1147 | qemu_co_mutex_lock(&s->lock); | |
1148 | ret = vmdk_read(bs, sector_num, buf, nb_sectors); | |
1149 | qemu_co_mutex_unlock(&s->lock); | |
1150 | return ret; | |
1151 | } | |
1152 | ||
1153 | static int vmdk_write(BlockDriverState *bs, int64_t sector_num, | |
1154 | const uint8_t *buf, int nb_sectors) | |
1155 | { | |
1156 | BDRVVmdkState *s = bs->opaque; | |
1157 | VmdkExtent *extent = NULL; | |
1158 | int n, ret; | |
1159 | int64_t index_in_cluster; | |
1160 | uint64_t extent_begin_sector, extent_relative_sector_num; | |
1161 | uint64_t cluster_offset; | |
1162 | VmdkMetaData m_data; | |
1163 | ||
1164 | if (sector_num > bs->total_sectors) { | |
1165 | fprintf(stderr, | |
1166 | "(VMDK) Wrong offset: sector_num=0x%" PRIx64 | |
1167 | " total_sectors=0x%" PRIx64 "\n", | |
1168 | sector_num, bs->total_sectors); | |
1169 | return -EIO; | |
1170 | } | |
1171 | ||
1172 | while (nb_sectors > 0) { | |
1173 | extent = find_extent(s, sector_num, extent); | |
1174 | if (!extent) { | |
1175 | return -EIO; | |
1176 | } | |
1177 | ret = get_cluster_offset( | |
1178 | bs, | |
1179 | extent, | |
1180 | &m_data, | |
1181 | sector_num << 9, !extent->compressed, | |
1182 | &cluster_offset); | |
1183 | if (extent->compressed) { | |
1184 | if (ret == VMDK_OK) { | |
1185 | /* Refuse write to allocated cluster for streamOptimized */ | |
1186 | fprintf(stderr, | |
1187 | "VMDK: can't write to allocated cluster" | |
1188 | " for streamOptimized\n"); | |
1189 | return -EIO; | |
1190 | } else { | |
1191 | /* allocate */ | |
1192 | ret = get_cluster_offset( | |
1193 | bs, | |
1194 | extent, | |
1195 | &m_data, | |
1196 | sector_num << 9, 1, | |
1197 | &cluster_offset); | |
1198 | } | |
1199 | } | |
1200 | if (ret) { | |
1201 | return -EINVAL; | |
1202 | } | |
1203 | extent_begin_sector = extent->end_sector - extent->sectors; | |
1204 | extent_relative_sector_num = sector_num - extent_begin_sector; | |
1205 | index_in_cluster = extent_relative_sector_num % extent->cluster_sectors; | |
1206 | n = extent->cluster_sectors - index_in_cluster; | |
1207 | if (n > nb_sectors) { | |
1208 | n = nb_sectors; | |
1209 | } | |
1210 | ||
1211 | ret = vmdk_write_extent(extent, | |
1212 | cluster_offset, index_in_cluster * 512, | |
1213 | buf, n, sector_num); | |
1214 | if (ret) { | |
1215 | return ret; | |
1216 | } | |
1217 | if (m_data.valid) { | |
1218 | /* update L2 tables */ | |
1219 | if (vmdk_L2update(extent, &m_data) == -1) { | |
1220 | return -EIO; | |
1221 | } | |
1222 | } | |
1223 | nb_sectors -= n; | |
1224 | sector_num += n; | |
1225 | buf += n * 512; | |
1226 | ||
1227 | /* update CID on the first write every time the virtual disk is | |
1228 | * opened */ | |
1229 | if (!s->cid_updated) { | |
1230 | ret = vmdk_write_cid(bs, time(NULL)); | |
1231 | if (ret < 0) { | |
1232 | return ret; | |
1233 | } | |
1234 | s->cid_updated = true; | |
1235 | } | |
1236 | } | |
1237 | return 0; | |
1238 | } | |
1239 | ||
1240 | static coroutine_fn int vmdk_co_write(BlockDriverState *bs, int64_t sector_num, | |
1241 | const uint8_t *buf, int nb_sectors) | |
1242 | { | |
1243 | int ret; | |
1244 | BDRVVmdkState *s = bs->opaque; | |
1245 | qemu_co_mutex_lock(&s->lock); | |
1246 | ret = vmdk_write(bs, sector_num, buf, nb_sectors); | |
1247 | qemu_co_mutex_unlock(&s->lock); | |
1248 | return ret; | |
1249 | } | |
1250 | ||
1251 | ||
1252 | static int vmdk_create_extent(const char *filename, int64_t filesize, | |
1253 | bool flat, bool compress) | |
1254 | { | |
1255 | int ret, i; | |
1256 | int fd = 0; | |
1257 | VMDK4Header header; | |
1258 | uint32_t tmp, magic, grains, gd_size, gt_size, gt_count; | |
1259 | ||
1260 | fd = qemu_open(filename, | |
1261 | O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, | |
1262 | 0644); | |
1263 | if (fd < 0) { | |
1264 | return -errno; | |
1265 | } | |
1266 | if (flat) { | |
1267 | ret = ftruncate(fd, filesize); | |
1268 | if (ret < 0) { | |
1269 | ret = -errno; | |
1270 | } | |
1271 | goto exit; | |
1272 | } | |
1273 | magic = cpu_to_be32(VMDK4_MAGIC); | |
1274 | memset(&header, 0, sizeof(header)); | |
1275 | header.version = 1; | |
1276 | header.flags = | |
1277 | 3 | (compress ? VMDK4_FLAG_COMPRESS | VMDK4_FLAG_MARKER : 0); | |
1278 | header.compressAlgorithm = compress ? VMDK4_COMPRESSION_DEFLATE : 0; | |
1279 | header.capacity = filesize / 512; | |
1280 | header.granularity = 128; | |
1281 | header.num_gtes_per_gte = 512; | |
1282 | ||
1283 | grains = (filesize / 512 + header.granularity - 1) / header.granularity; | |
1284 | gt_size = ((header.num_gtes_per_gte * sizeof(uint32_t)) + 511) >> 9; | |
1285 | gt_count = | |
1286 | (grains + header.num_gtes_per_gte - 1) / header.num_gtes_per_gte; | |
1287 | gd_size = (gt_count * sizeof(uint32_t) + 511) >> 9; | |
1288 | ||
1289 | header.desc_offset = 1; | |
1290 | header.desc_size = 20; | |
1291 | header.rgd_offset = header.desc_offset + header.desc_size; | |
1292 | header.gd_offset = header.rgd_offset + gd_size + (gt_size * gt_count); | |
1293 | header.grain_offset = | |
1294 | ((header.gd_offset + gd_size + (gt_size * gt_count) + | |
1295 | header.granularity - 1) / header.granularity) * | |
1296 | header.granularity; | |
1297 | /* swap endianness for all header fields */ | |
1298 | header.version = cpu_to_le32(header.version); | |
1299 | header.flags = cpu_to_le32(header.flags); | |
1300 | header.capacity = cpu_to_le64(header.capacity); | |
1301 | header.granularity = cpu_to_le64(header.granularity); | |
1302 | header.num_gtes_per_gte = cpu_to_le32(header.num_gtes_per_gte); | |
1303 | header.desc_offset = cpu_to_le64(header.desc_offset); | |
1304 | header.desc_size = cpu_to_le64(header.desc_size); | |
1305 | header.rgd_offset = cpu_to_le64(header.rgd_offset); | |
1306 | header.gd_offset = cpu_to_le64(header.gd_offset); | |
1307 | header.grain_offset = cpu_to_le64(header.grain_offset); | |
1308 | header.compressAlgorithm = cpu_to_le16(header.compressAlgorithm); | |
1309 | ||
1310 | header.check_bytes[0] = 0xa; | |
1311 | header.check_bytes[1] = 0x20; | |
1312 | header.check_bytes[2] = 0xd; | |
1313 | header.check_bytes[3] = 0xa; | |
1314 | ||
1315 | /* write all the data */ | |
1316 | ret = qemu_write_full(fd, &magic, sizeof(magic)); | |
1317 | if (ret != sizeof(magic)) { | |
1318 | ret = -errno; | |
1319 | goto exit; | |
1320 | } | |
1321 | ret = qemu_write_full(fd, &header, sizeof(header)); | |
1322 | if (ret != sizeof(header)) { | |
1323 | ret = -errno; | |
1324 | goto exit; | |
1325 | } | |
1326 | ||
1327 | ret = ftruncate(fd, le64_to_cpu(header.grain_offset) << 9); | |
1328 | if (ret < 0) { | |
1329 | ret = -errno; | |
1330 | goto exit; | |
1331 | } | |
1332 | ||
1333 | /* write grain directory */ | |
1334 | lseek(fd, le64_to_cpu(header.rgd_offset) << 9, SEEK_SET); | |
1335 | for (i = 0, tmp = le64_to_cpu(header.rgd_offset) + gd_size; | |
1336 | i < gt_count; i++, tmp += gt_size) { | |
1337 | ret = qemu_write_full(fd, &tmp, sizeof(tmp)); | |
1338 | if (ret != sizeof(tmp)) { | |
1339 | ret = -errno; | |
1340 | goto exit; | |
1341 | } | |
1342 | } | |
1343 | ||
1344 | /* write backup grain directory */ | |
1345 | lseek(fd, le64_to_cpu(header.gd_offset) << 9, SEEK_SET); | |
1346 | for (i = 0, tmp = le64_to_cpu(header.gd_offset) + gd_size; | |
1347 | i < gt_count; i++, tmp += gt_size) { | |
1348 | ret = qemu_write_full(fd, &tmp, sizeof(tmp)); | |
1349 | if (ret != sizeof(tmp)) { | |
1350 | ret = -errno; | |
1351 | goto exit; | |
1352 | } | |
1353 | } | |
1354 | ||
1355 | ret = 0; | |
1356 | exit: | |
1357 | qemu_close(fd); | |
1358 | return ret; | |
1359 | } | |
1360 | ||
1361 | static int filename_decompose(const char *filename, char *path, char *prefix, | |
1362 | char *postfix, size_t buf_len) | |
1363 | { | |
1364 | const char *p, *q; | |
1365 | ||
1366 | if (filename == NULL || !strlen(filename)) { | |
1367 | fprintf(stderr, "Vmdk: no filename provided.\n"); | |
1368 | return VMDK_ERROR; | |
1369 | } | |
1370 | p = strrchr(filename, '/'); | |
1371 | if (p == NULL) { | |
1372 | p = strrchr(filename, '\\'); | |
1373 | } | |
1374 | if (p == NULL) { | |
1375 | p = strrchr(filename, ':'); | |
1376 | } | |
1377 | if (p != NULL) { | |
1378 | p++; | |
1379 | if (p - filename >= buf_len) { | |
1380 | return VMDK_ERROR; | |
1381 | } | |
1382 | pstrcpy(path, p - filename + 1, filename); | |
1383 | } else { | |
1384 | p = filename; | |
1385 | path[0] = '\0'; | |
1386 | } | |
1387 | q = strrchr(p, '.'); | |
1388 | if (q == NULL) { | |
1389 | pstrcpy(prefix, buf_len, p); | |
1390 | postfix[0] = '\0'; | |
1391 | } else { | |
1392 | if (q - p >= buf_len) { | |
1393 | return VMDK_ERROR; | |
1394 | } | |
1395 | pstrcpy(prefix, q - p + 1, p); | |
1396 | pstrcpy(postfix, buf_len, q); | |
1397 | } | |
1398 | return VMDK_OK; | |
1399 | } | |
1400 | ||
1401 | static int relative_path(char *dest, int dest_size, | |
1402 | const char *base, const char *target) | |
1403 | { | |
1404 | int i = 0; | |
1405 | int n = 0; | |
1406 | const char *p, *q; | |
1407 | #ifdef _WIN32 | |
1408 | const char *sep = "\\"; | |
1409 | #else | |
1410 | const char *sep = "/"; | |
1411 | #endif | |
1412 | ||
1413 | if (!(dest && base && target)) { | |
1414 | return VMDK_ERROR; | |
1415 | } | |
1416 | if (path_is_absolute(target)) { | |
1417 | pstrcpy(dest, dest_size, target); | |
1418 | return VMDK_OK; | |
1419 | } | |
1420 | while (base[i] == target[i]) { | |
1421 | i++; | |
1422 | } | |
1423 | p = &base[i]; | |
1424 | q = &target[i]; | |
1425 | while (*p) { | |
1426 | if (*p == *sep) { | |
1427 | n++; | |
1428 | } | |
1429 | p++; | |
1430 | } | |
1431 | dest[0] = '\0'; | |
1432 | for (; n; n--) { | |
1433 | pstrcat(dest, dest_size, ".."); | |
1434 | pstrcat(dest, dest_size, sep); | |
1435 | } | |
1436 | pstrcat(dest, dest_size, q); | |
1437 | return VMDK_OK; | |
1438 | } | |
1439 | ||
1440 | static int vmdk_create(const char *filename, QEMUOptionParameter *options) | |
1441 | { | |
1442 | int fd, idx = 0; | |
1443 | char desc[BUF_SIZE]; | |
1444 | int64_t total_size = 0, filesize; | |
1445 | const char *adapter_type = NULL; | |
1446 | const char *backing_file = NULL; | |
1447 | const char *fmt = NULL; | |
1448 | int flags = 0; | |
1449 | int ret = 0; | |
1450 | bool flat, split, compress; | |
1451 | char ext_desc_lines[BUF_SIZE] = ""; | |
1452 | char path[PATH_MAX], prefix[PATH_MAX], postfix[PATH_MAX]; | |
1453 | const int64_t split_size = 0x80000000; /* VMDK has constant split size */ | |
1454 | const char *desc_extent_line; | |
1455 | char parent_desc_line[BUF_SIZE] = ""; | |
1456 | uint32_t parent_cid = 0xffffffff; | |
1457 | uint32_t number_heads = 16; | |
1458 | const char desc_template[] = | |
1459 | "# Disk DescriptorFile\n" | |
1460 | "version=1\n" | |
1461 | "CID=%x\n" | |
1462 | "parentCID=%x\n" | |
1463 | "createType=\"%s\"\n" | |
1464 | "%s" | |
1465 | "\n" | |
1466 | "# Extent description\n" | |
1467 | "%s" | |
1468 | "\n" | |
1469 | "# The Disk Data Base\n" | |
1470 | "#DDB\n" | |
1471 | "\n" | |
1472 | "ddb.virtualHWVersion = \"%d\"\n" | |
1473 | "ddb.geometry.cylinders = \"%" PRId64 "\"\n" | |
1474 | "ddb.geometry.heads = \"%d\"\n" | |
1475 | "ddb.geometry.sectors = \"63\"\n" | |
1476 | "ddb.adapterType = \"%s\"\n"; | |
1477 | ||
1478 | if (filename_decompose(filename, path, prefix, postfix, PATH_MAX)) { | |
1479 | return -EINVAL; | |
1480 | } | |
1481 | /* Read out options */ | |
1482 | while (options && options->name) { | |
1483 | if (!strcmp(options->name, BLOCK_OPT_SIZE)) { | |
1484 | total_size = options->value.n; | |
1485 | } else if (!strcmp(options->name, BLOCK_OPT_ADAPTER_TYPE)) { | |
1486 | adapter_type = options->value.s; | |
1487 | } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) { | |
1488 | backing_file = options->value.s; | |
1489 | } else if (!strcmp(options->name, BLOCK_OPT_COMPAT6)) { | |
1490 | flags |= options->value.n ? BLOCK_FLAG_COMPAT6 : 0; | |
1491 | } else if (!strcmp(options->name, BLOCK_OPT_SUBFMT)) { | |
1492 | fmt = options->value.s; | |
1493 | } | |
1494 | options++; | |
1495 | } | |
1496 | if (!adapter_type) { | |
1497 | adapter_type = "ide"; | |
1498 | } else if (strcmp(adapter_type, "ide") && | |
1499 | strcmp(adapter_type, "buslogic") && | |
1500 | strcmp(adapter_type, "lsilogic") && | |
1501 | strcmp(adapter_type, "legacyESX")) { | |
1502 | fprintf(stderr, "VMDK: Unknown adapter type: '%s'.\n", adapter_type); | |
1503 | return -EINVAL; | |
1504 | } | |
1505 | if (strcmp(adapter_type, "ide") != 0) { | |
1506 | /* that's the number of heads with which vmware operates when | |
1507 | creating, exporting, etc. vmdk files with a non-ide adapter type */ | |
1508 | number_heads = 255; | |
1509 | } | |
1510 | if (!fmt) { | |
1511 | /* Default format to monolithicSparse */ | |
1512 | fmt = "monolithicSparse"; | |
1513 | } else if (strcmp(fmt, "monolithicFlat") && | |
1514 | strcmp(fmt, "monolithicSparse") && | |
1515 | strcmp(fmt, "twoGbMaxExtentSparse") && | |
1516 | strcmp(fmt, "twoGbMaxExtentFlat") && | |
1517 | strcmp(fmt, "streamOptimized")) { | |
1518 | fprintf(stderr, "VMDK: Unknown subformat: %s\n", fmt); | |
1519 | return -EINVAL; | |
1520 | } | |
1521 | split = !(strcmp(fmt, "twoGbMaxExtentFlat") && | |
1522 | strcmp(fmt, "twoGbMaxExtentSparse")); | |
1523 | flat = !(strcmp(fmt, "monolithicFlat") && | |
1524 | strcmp(fmt, "twoGbMaxExtentFlat")); | |
1525 | compress = !strcmp(fmt, "streamOptimized"); | |
1526 | if (flat) { | |
1527 | desc_extent_line = "RW %lld FLAT \"%s\" 0\n"; | |
1528 | } else { | |
1529 | desc_extent_line = "RW %lld SPARSE \"%s\"\n"; | |
1530 | } | |
1531 | if (flat && backing_file) { | |
1532 | /* not supporting backing file for flat image */ | |
1533 | return -ENOTSUP; | |
1534 | } | |
1535 | if (backing_file) { | |
1536 | char parent_filename[PATH_MAX]; | |
1537 | BlockDriverState *bs = bdrv_new(""); | |
1538 | ret = bdrv_open(bs, backing_file, NULL, 0, NULL); | |
1539 | if (ret != 0) { | |
1540 | bdrv_delete(bs); | |
1541 | return ret; | |
1542 | } | |
1543 | if (strcmp(bs->drv->format_name, "vmdk")) { | |
1544 | bdrv_delete(bs); | |
1545 | return -EINVAL; | |
1546 | } | |
1547 | parent_cid = vmdk_read_cid(bs, 0); | |
1548 | bdrv_delete(bs); | |
1549 | relative_path(parent_filename, sizeof(parent_filename), | |
1550 | filename, backing_file); | |
1551 | snprintf(parent_desc_line, sizeof(parent_desc_line), | |
1552 | "parentFileNameHint=\"%s\"", parent_filename); | |
1553 | } | |
1554 | ||
1555 | /* Create extents */ | |
1556 | filesize = total_size; | |
1557 | while (filesize > 0) { | |
1558 | char desc_line[BUF_SIZE]; | |
1559 | char ext_filename[PATH_MAX]; | |
1560 | char desc_filename[PATH_MAX]; | |
1561 | int64_t size = filesize; | |
1562 | ||
1563 | if (split && size > split_size) { | |
1564 | size = split_size; | |
1565 | } | |
1566 | if (split) { | |
1567 | snprintf(desc_filename, sizeof(desc_filename), "%s-%c%03d%s", | |
1568 | prefix, flat ? 'f' : 's', ++idx, postfix); | |
1569 | } else if (flat) { | |
1570 | snprintf(desc_filename, sizeof(desc_filename), "%s-flat%s", | |
1571 | prefix, postfix); | |
1572 | } else { | |
1573 | snprintf(desc_filename, sizeof(desc_filename), "%s%s", | |
1574 | prefix, postfix); | |
1575 | } | |
1576 | snprintf(ext_filename, sizeof(ext_filename), "%s%s", | |
1577 | path, desc_filename); | |
1578 | ||
1579 | if (vmdk_create_extent(ext_filename, size, flat, compress)) { | |
1580 | return -EINVAL; | |
1581 | } | |
1582 | filesize -= size; | |
1583 | ||
1584 | /* Format description line */ | |
1585 | snprintf(desc_line, sizeof(desc_line), | |
1586 | desc_extent_line, size / 512, desc_filename); | |
1587 | pstrcat(ext_desc_lines, sizeof(ext_desc_lines), desc_line); | |
1588 | } | |
1589 | /* generate descriptor file */ | |
1590 | snprintf(desc, sizeof(desc), desc_template, | |
1591 | (unsigned int)time(NULL), | |
1592 | parent_cid, | |
1593 | fmt, | |
1594 | parent_desc_line, | |
1595 | ext_desc_lines, | |
1596 | (flags & BLOCK_FLAG_COMPAT6 ? 6 : 4), | |
1597 | total_size / (int64_t)(63 * number_heads * 512), number_heads, | |
1598 | adapter_type); | |
1599 | if (split || flat) { | |
1600 | fd = qemu_open(filename, | |
1601 | O_WRONLY | O_CREAT | O_TRUNC | O_BINARY | O_LARGEFILE, | |
1602 | 0644); | |
1603 | } else { | |
1604 | fd = qemu_open(filename, | |
1605 | O_WRONLY | O_BINARY | O_LARGEFILE, | |
1606 | 0644); | |
1607 | } | |
1608 | if (fd < 0) { | |
1609 | return -errno; | |
1610 | } | |
1611 | /* the descriptor offset = 0x200 */ | |
1612 | if (!split && !flat && 0x200 != lseek(fd, 0x200, SEEK_SET)) { | |
1613 | ret = -errno; | |
1614 | goto exit; | |
1615 | } | |
1616 | ret = qemu_write_full(fd, desc, strlen(desc)); | |
1617 | if (ret != strlen(desc)) { | |
1618 | ret = -errno; | |
1619 | goto exit; | |
1620 | } | |
1621 | ret = 0; | |
1622 | exit: | |
1623 | qemu_close(fd); | |
1624 | return ret; | |
1625 | } | |
1626 | ||
1627 | static void vmdk_close(BlockDriverState *bs) | |
1628 | { | |
1629 | BDRVVmdkState *s = bs->opaque; | |
1630 | ||
1631 | vmdk_free_extents(bs); | |
1632 | ||
1633 | migrate_del_blocker(s->migration_blocker); | |
1634 | error_free(s->migration_blocker); | |
1635 | } | |
1636 | ||
1637 | static coroutine_fn int vmdk_co_flush(BlockDriverState *bs) | |
1638 | { | |
1639 | BDRVVmdkState *s = bs->opaque; | |
1640 | int i, err; | |
1641 | int ret = 0; | |
1642 | ||
1643 | for (i = 0; i < s->num_extents; i++) { | |
1644 | err = bdrv_co_flush(s->extents[i].file); | |
1645 | if (err < 0) { | |
1646 | ret = err; | |
1647 | } | |
1648 | } | |
1649 | return ret; | |
1650 | } | |
1651 | ||
1652 | static int64_t vmdk_get_allocated_file_size(BlockDriverState *bs) | |
1653 | { | |
1654 | int i; | |
1655 | int64_t ret = 0; | |
1656 | int64_t r; | |
1657 | BDRVVmdkState *s = bs->opaque; | |
1658 | ||
1659 | ret = bdrv_get_allocated_file_size(bs->file); | |
1660 | if (ret < 0) { | |
1661 | return ret; | |
1662 | } | |
1663 | for (i = 0; i < s->num_extents; i++) { | |
1664 | if (s->extents[i].file == bs->file) { | |
1665 | continue; | |
1666 | } | |
1667 | r = bdrv_get_allocated_file_size(s->extents[i].file); | |
1668 | if (r < 0) { | |
1669 | return r; | |
1670 | } | |
1671 | ret += r; | |
1672 | } | |
1673 | return ret; | |
1674 | } | |
1675 | ||
1676 | static QEMUOptionParameter vmdk_create_options[] = { | |
1677 | { | |
1678 | .name = BLOCK_OPT_SIZE, | |
1679 | .type = OPT_SIZE, | |
1680 | .help = "Virtual disk size" | |
1681 | }, | |
1682 | { | |
1683 | .name = BLOCK_OPT_ADAPTER_TYPE, | |
1684 | .type = OPT_STRING, | |
1685 | .help = "Virtual adapter type, can be one of " | |
1686 | "ide (default), lsilogic, buslogic or legacyESX" | |
1687 | }, | |
1688 | { | |
1689 | .name = BLOCK_OPT_BACKING_FILE, | |
1690 | .type = OPT_STRING, | |
1691 | .help = "File name of a base image" | |
1692 | }, | |
1693 | { | |
1694 | .name = BLOCK_OPT_COMPAT6, | |
1695 | .type = OPT_FLAG, | |
1696 | .help = "VMDK version 6 image" | |
1697 | }, | |
1698 | { | |
1699 | .name = BLOCK_OPT_SUBFMT, | |
1700 | .type = OPT_STRING, | |
1701 | .help = | |
1702 | "VMDK flat extent format, can be one of " | |
1703 | "{monolithicSparse (default) | monolithicFlat | twoGbMaxExtentSparse | twoGbMaxExtentFlat | streamOptimized} " | |
1704 | }, | |
1705 | { NULL } | |
1706 | }; | |
1707 | ||
1708 | static BlockDriver bdrv_vmdk = { | |
1709 | .format_name = "vmdk", | |
1710 | .instance_size = sizeof(BDRVVmdkState), | |
1711 | .bdrv_probe = vmdk_probe, | |
1712 | .bdrv_open = vmdk_open, | |
1713 | .bdrv_reopen_prepare = vmdk_reopen_prepare, | |
1714 | .bdrv_read = vmdk_co_read, | |
1715 | .bdrv_write = vmdk_co_write, | |
1716 | .bdrv_close = vmdk_close, | |
1717 | .bdrv_create = vmdk_create, | |
1718 | .bdrv_co_flush_to_disk = vmdk_co_flush, | |
1719 | .bdrv_co_is_allocated = vmdk_co_is_allocated, | |
1720 | .bdrv_get_allocated_file_size = vmdk_get_allocated_file_size, | |
1721 | ||
1722 | .create_options = vmdk_create_options, | |
1723 | }; | |
1724 | ||
1725 | static void bdrv_vmdk_init(void) | |
1726 | { | |
1727 | bdrv_register(&bdrv_vmdk); | |
1728 | } | |
1729 | ||
1730 | block_init(bdrv_vmdk_init); |