]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Block driver for the QCOW version 2 format | |
3 | * | |
4 | * Copyright (c) 2004-2006 Fabrice Bellard | |
5 | * | |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
24 | #include "qemu-common.h" | |
25 | #include "block/block_int.h" | |
26 | #include "qemu/module.h" | |
27 | #include <zlib.h> | |
28 | #include "qemu/aes.h" | |
29 | #include "block/qcow2.h" | |
30 | #include "qemu/error-report.h" | |
31 | #include "qapi/qmp/qerror.h" | |
32 | #include "qapi/qmp/qbool.h" | |
33 | #include "trace.h" | |
34 | ||
35 | /* | |
36 | Differences with QCOW: | |
37 | ||
38 | - Support for multiple incremental snapshots. | |
39 | - Memory management by reference counts. | |
40 | - Clusters which have a reference count of one have the bit | |
41 | QCOW_OFLAG_COPIED to optimize write performance. | |
42 | - Size of compressed clusters is stored in sectors to reduce bit usage | |
43 | in the cluster offsets. | |
44 | - Support for storing additional data (such as the VM state) in the | |
45 | snapshots. | |
46 | - If a backing store is used, the cluster size is not constrained | |
47 | (could be backported to QCOW). | |
48 | - L2 tables have always a size of one cluster. | |
49 | */ | |
50 | ||
51 | ||
52 | typedef struct { | |
53 | uint32_t magic; | |
54 | uint32_t len; | |
55 | } QCowExtension; | |
56 | ||
57 | #define QCOW2_EXT_MAGIC_END 0 | |
58 | #define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA | |
59 | #define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 | |
60 | ||
61 | static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) | |
62 | { | |
63 | const QCowHeader *cow_header = (const void *)buf; | |
64 | ||
65 | if (buf_size >= sizeof(QCowHeader) && | |
66 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
67 | be32_to_cpu(cow_header->version) >= 2) | |
68 | return 100; | |
69 | else | |
70 | return 0; | |
71 | } | |
72 | ||
73 | ||
74 | /* | |
75 | * read qcow2 extension and fill bs | |
76 | * start reading from start_offset | |
77 | * finish reading upon magic of value 0 or when end_offset reached | |
78 | * unknown magic is skipped (future extension this version knows nothing about) | |
79 | * return 0 upon success, non-0 otherwise | |
80 | */ | |
81 | static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, | |
82 | uint64_t end_offset, void **p_feature_table) | |
83 | { | |
84 | BDRVQcowState *s = bs->opaque; | |
85 | QCowExtension ext; | |
86 | uint64_t offset; | |
87 | int ret; | |
88 | ||
89 | #ifdef DEBUG_EXT | |
90 | printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); | |
91 | #endif | |
92 | offset = start_offset; | |
93 | while (offset < end_offset) { | |
94 | ||
95 | #ifdef DEBUG_EXT | |
96 | /* Sanity check */ | |
97 | if (offset > s->cluster_size) | |
98 | printf("qcow2_read_extension: suspicious offset %lu\n", offset); | |
99 | ||
100 | printf("attempting to read extended header in offset %lu\n", offset); | |
101 | #endif | |
102 | ||
103 | if (bdrv_pread(bs->file, offset, &ext, sizeof(ext)) != sizeof(ext)) { | |
104 | fprintf(stderr, "qcow2_read_extension: ERROR: " | |
105 | "pread fail from offset %" PRIu64 "\n", | |
106 | offset); | |
107 | return 1; | |
108 | } | |
109 | be32_to_cpus(&ext.magic); | |
110 | be32_to_cpus(&ext.len); | |
111 | offset += sizeof(ext); | |
112 | #ifdef DEBUG_EXT | |
113 | printf("ext.magic = 0x%x\n", ext.magic); | |
114 | #endif | |
115 | if (ext.len > end_offset - offset) { | |
116 | error_report("Header extension too large"); | |
117 | return -EINVAL; | |
118 | } | |
119 | ||
120 | switch (ext.magic) { | |
121 | case QCOW2_EXT_MAGIC_END: | |
122 | return 0; | |
123 | ||
124 | case QCOW2_EXT_MAGIC_BACKING_FORMAT: | |
125 | if (ext.len >= sizeof(bs->backing_format)) { | |
126 | fprintf(stderr, "ERROR: ext_backing_format: len=%u too large" | |
127 | " (>=%zu)\n", | |
128 | ext.len, sizeof(bs->backing_format)); | |
129 | return 2; | |
130 | } | |
131 | if (bdrv_pread(bs->file, offset , bs->backing_format, | |
132 | ext.len) != ext.len) | |
133 | return 3; | |
134 | bs->backing_format[ext.len] = '\0'; | |
135 | #ifdef DEBUG_EXT | |
136 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
137 | #endif | |
138 | break; | |
139 | ||
140 | case QCOW2_EXT_MAGIC_FEATURE_TABLE: | |
141 | if (p_feature_table != NULL) { | |
142 | void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); | |
143 | ret = bdrv_pread(bs->file, offset , feature_table, ext.len); | |
144 | if (ret < 0) { | |
145 | return ret; | |
146 | } | |
147 | ||
148 | *p_feature_table = feature_table; | |
149 | } | |
150 | break; | |
151 | ||
152 | default: | |
153 | /* unknown magic - save it in case we need to rewrite the header */ | |
154 | { | |
155 | Qcow2UnknownHeaderExtension *uext; | |
156 | ||
157 | uext = g_malloc0(sizeof(*uext) + ext.len); | |
158 | uext->magic = ext.magic; | |
159 | uext->len = ext.len; | |
160 | QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); | |
161 | ||
162 | ret = bdrv_pread(bs->file, offset , uext->data, uext->len); | |
163 | if (ret < 0) { | |
164 | return ret; | |
165 | } | |
166 | } | |
167 | break; | |
168 | } | |
169 | ||
170 | offset += ((ext.len + 7) & ~7); | |
171 | } | |
172 | ||
173 | return 0; | |
174 | } | |
175 | ||
176 | static void cleanup_unknown_header_ext(BlockDriverState *bs) | |
177 | { | |
178 | BDRVQcowState *s = bs->opaque; | |
179 | Qcow2UnknownHeaderExtension *uext, *next; | |
180 | ||
181 | QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { | |
182 | QLIST_REMOVE(uext, next); | |
183 | g_free(uext); | |
184 | } | |
185 | } | |
186 | ||
187 | static void GCC_FMT_ATTR(2, 3) report_unsupported(BlockDriverState *bs, | |
188 | const char *fmt, ...) | |
189 | { | |
190 | char msg[64]; | |
191 | va_list ap; | |
192 | ||
193 | va_start(ap, fmt); | |
194 | vsnprintf(msg, sizeof(msg), fmt, ap); | |
195 | va_end(ap); | |
196 | ||
197 | qerror_report(QERR_UNKNOWN_BLOCK_FORMAT_FEATURE, | |
198 | bs->device_name, "qcow2", msg); | |
199 | } | |
200 | ||
201 | static void report_unsupported_feature(BlockDriverState *bs, | |
202 | Qcow2Feature *table, uint64_t mask) | |
203 | { | |
204 | while (table && table->name[0] != '\0') { | |
205 | if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) { | |
206 | if (mask & (1 << table->bit)) { | |
207 | report_unsupported(bs, "%.46s",table->name); | |
208 | mask &= ~(1 << table->bit); | |
209 | } | |
210 | } | |
211 | table++; | |
212 | } | |
213 | ||
214 | if (mask) { | |
215 | report_unsupported(bs, "Unknown incompatible feature: %" PRIx64, mask); | |
216 | } | |
217 | } | |
218 | ||
219 | /* | |
220 | * Sets the dirty bit and flushes afterwards if necessary. | |
221 | * | |
222 | * The incompatible_features bit is only set if the image file header was | |
223 | * updated successfully. Therefore it is not required to check the return | |
224 | * value of this function. | |
225 | */ | |
226 | int qcow2_mark_dirty(BlockDriverState *bs) | |
227 | { | |
228 | BDRVQcowState *s = bs->opaque; | |
229 | uint64_t val; | |
230 | int ret; | |
231 | ||
232 | assert(s->qcow_version >= 3); | |
233 | ||
234 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
235 | return 0; /* already dirty */ | |
236 | } | |
237 | ||
238 | val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); | |
239 | ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features), | |
240 | &val, sizeof(val)); | |
241 | if (ret < 0) { | |
242 | return ret; | |
243 | } | |
244 | ret = bdrv_flush(bs->file); | |
245 | if (ret < 0) { | |
246 | return ret; | |
247 | } | |
248 | ||
249 | /* Only treat image as dirty if the header was updated successfully */ | |
250 | s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; | |
251 | return 0; | |
252 | } | |
253 | ||
254 | /* | |
255 | * Clears the dirty bit and flushes before if necessary. Only call this | |
256 | * function when there are no pending requests, it does not guard against | |
257 | * concurrent requests dirtying the image. | |
258 | */ | |
259 | static int qcow2_mark_clean(BlockDriverState *bs) | |
260 | { | |
261 | BDRVQcowState *s = bs->opaque; | |
262 | ||
263 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
264 | int ret = bdrv_flush(bs); | |
265 | if (ret < 0) { | |
266 | return ret; | |
267 | } | |
268 | ||
269 | s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; | |
270 | return qcow2_update_header(bs); | |
271 | } | |
272 | return 0; | |
273 | } | |
274 | ||
275 | static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result, | |
276 | BdrvCheckMode fix) | |
277 | { | |
278 | int ret = qcow2_check_refcounts(bs, result, fix); | |
279 | if (ret < 0) { | |
280 | return ret; | |
281 | } | |
282 | ||
283 | if (fix && result->check_errors == 0 && result->corruptions == 0) { | |
284 | return qcow2_mark_clean(bs); | |
285 | } | |
286 | return ret; | |
287 | } | |
288 | ||
289 | static QemuOptsList qcow2_runtime_opts = { | |
290 | .name = "qcow2", | |
291 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head), | |
292 | .desc = { | |
293 | { | |
294 | .name = "lazy_refcounts", | |
295 | .type = QEMU_OPT_BOOL, | |
296 | .help = "Postpone refcount updates", | |
297 | }, | |
298 | { /* end of list */ } | |
299 | }, | |
300 | }; | |
301 | ||
302 | static int qcow2_open(BlockDriverState *bs, QDict *options, int flags) | |
303 | { | |
304 | BDRVQcowState *s = bs->opaque; | |
305 | int len, i, ret = 0; | |
306 | QCowHeader header; | |
307 | QemuOpts *opts; | |
308 | Error *local_err = NULL; | |
309 | uint64_t ext_end; | |
310 | ||
311 | ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); | |
312 | if (ret < 0) { | |
313 | goto fail; | |
314 | } | |
315 | be32_to_cpus(&header.magic); | |
316 | be32_to_cpus(&header.version); | |
317 | be64_to_cpus(&header.backing_file_offset); | |
318 | be32_to_cpus(&header.backing_file_size); | |
319 | be64_to_cpus(&header.size); | |
320 | be32_to_cpus(&header.cluster_bits); | |
321 | be32_to_cpus(&header.crypt_method); | |
322 | be64_to_cpus(&header.l1_table_offset); | |
323 | be32_to_cpus(&header.l1_size); | |
324 | be64_to_cpus(&header.refcount_table_offset); | |
325 | be32_to_cpus(&header.refcount_table_clusters); | |
326 | be64_to_cpus(&header.snapshots_offset); | |
327 | be32_to_cpus(&header.nb_snapshots); | |
328 | ||
329 | if (header.magic != QCOW_MAGIC) { | |
330 | ret = -EMEDIUMTYPE; | |
331 | goto fail; | |
332 | } | |
333 | if (header.version < 2 || header.version > 3) { | |
334 | report_unsupported(bs, "QCOW version %d", header.version); | |
335 | ret = -ENOTSUP; | |
336 | goto fail; | |
337 | } | |
338 | ||
339 | s->qcow_version = header.version; | |
340 | ||
341 | /* Initialise version 3 header fields */ | |
342 | if (header.version == 2) { | |
343 | header.incompatible_features = 0; | |
344 | header.compatible_features = 0; | |
345 | header.autoclear_features = 0; | |
346 | header.refcount_order = 4; | |
347 | header.header_length = 72; | |
348 | } else { | |
349 | be64_to_cpus(&header.incompatible_features); | |
350 | be64_to_cpus(&header.compatible_features); | |
351 | be64_to_cpus(&header.autoclear_features); | |
352 | be32_to_cpus(&header.refcount_order); | |
353 | be32_to_cpus(&header.header_length); | |
354 | } | |
355 | ||
356 | if (header.header_length > sizeof(header)) { | |
357 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
358 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
359 | ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields, | |
360 | s->unknown_header_fields_size); | |
361 | if (ret < 0) { | |
362 | goto fail; | |
363 | } | |
364 | } | |
365 | ||
366 | if (header.backing_file_offset) { | |
367 | ext_end = header.backing_file_offset; | |
368 | } else { | |
369 | ext_end = 1 << header.cluster_bits; | |
370 | } | |
371 | ||
372 | /* Handle feature bits */ | |
373 | s->incompatible_features = header.incompatible_features; | |
374 | s->compatible_features = header.compatible_features; | |
375 | s->autoclear_features = header.autoclear_features; | |
376 | ||
377 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { | |
378 | void *feature_table = NULL; | |
379 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
380 | &feature_table); | |
381 | report_unsupported_feature(bs, feature_table, | |
382 | s->incompatible_features & | |
383 | ~QCOW2_INCOMPAT_MASK); | |
384 | ret = -ENOTSUP; | |
385 | goto fail; | |
386 | } | |
387 | ||
388 | /* Check support for various header values */ | |
389 | if (header.refcount_order != 4) { | |
390 | report_unsupported(bs, "%d bit reference counts", | |
391 | 1 << header.refcount_order); | |
392 | ret = -ENOTSUP; | |
393 | goto fail; | |
394 | } | |
395 | ||
396 | if (header.cluster_bits < MIN_CLUSTER_BITS || | |
397 | header.cluster_bits > MAX_CLUSTER_BITS) { | |
398 | ret = -EINVAL; | |
399 | goto fail; | |
400 | } | |
401 | if (header.crypt_method > QCOW_CRYPT_AES) { | |
402 | ret = -EINVAL; | |
403 | goto fail; | |
404 | } | |
405 | s->crypt_method_header = header.crypt_method; | |
406 | if (s->crypt_method_header) { | |
407 | bs->encrypted = 1; | |
408 | } | |
409 | s->cluster_bits = header.cluster_bits; | |
410 | s->cluster_size = 1 << s->cluster_bits; | |
411 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
412 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ | |
413 | s->l2_size = 1 << s->l2_bits; | |
414 | bs->total_sectors = header.size / 512; | |
415 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
416 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
417 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
418 | s->refcount_table_offset = header.refcount_table_offset; | |
419 | s->refcount_table_size = | |
420 | header.refcount_table_clusters << (s->cluster_bits - 3); | |
421 | ||
422 | s->snapshots_offset = header.snapshots_offset; | |
423 | s->nb_snapshots = header.nb_snapshots; | |
424 | ||
425 | /* read the level 1 table */ | |
426 | s->l1_size = header.l1_size; | |
427 | s->l1_vm_state_index = size_to_l1(s, header.size); | |
428 | /* the L1 table must contain at least enough entries to put | |
429 | header.size bytes */ | |
430 | if (s->l1_size < s->l1_vm_state_index) { | |
431 | ret = -EINVAL; | |
432 | goto fail; | |
433 | } | |
434 | s->l1_table_offset = header.l1_table_offset; | |
435 | if (s->l1_size > 0) { | |
436 | s->l1_table = g_malloc0( | |
437 | align_offset(s->l1_size * sizeof(uint64_t), 512)); | |
438 | ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, | |
439 | s->l1_size * sizeof(uint64_t)); | |
440 | if (ret < 0) { | |
441 | goto fail; | |
442 | } | |
443 | for(i = 0;i < s->l1_size; i++) { | |
444 | be64_to_cpus(&s->l1_table[i]); | |
445 | } | |
446 | } | |
447 | ||
448 | /* alloc L2 table/refcount block cache */ | |
449 | s->l2_table_cache = qcow2_cache_create(bs, L2_CACHE_SIZE); | |
450 | s->refcount_block_cache = qcow2_cache_create(bs, REFCOUNT_CACHE_SIZE); | |
451 | ||
452 | s->cluster_cache = g_malloc(s->cluster_size); | |
453 | /* one more sector for decompressed data alignment */ | |
454 | s->cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size | |
455 | + 512); | |
456 | s->cluster_cache_offset = -1; | |
457 | s->flags = flags; | |
458 | ||
459 | ret = qcow2_refcount_init(bs); | |
460 | if (ret != 0) { | |
461 | goto fail; | |
462 | } | |
463 | ||
464 | QLIST_INIT(&s->cluster_allocs); | |
465 | ||
466 | /* read qcow2 extensions */ | |
467 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL)) { | |
468 | ret = -EINVAL; | |
469 | goto fail; | |
470 | } | |
471 | ||
472 | /* read the backing file name */ | |
473 | if (header.backing_file_offset != 0) { | |
474 | len = header.backing_file_size; | |
475 | if (len > 1023) { | |
476 | len = 1023; | |
477 | } | |
478 | ret = bdrv_pread(bs->file, header.backing_file_offset, | |
479 | bs->backing_file, len); | |
480 | if (ret < 0) { | |
481 | goto fail; | |
482 | } | |
483 | bs->backing_file[len] = '\0'; | |
484 | } | |
485 | ||
486 | ret = qcow2_read_snapshots(bs); | |
487 | if (ret < 0) { | |
488 | goto fail; | |
489 | } | |
490 | ||
491 | /* Clear unknown autoclear feature bits */ | |
492 | if (!bs->read_only && s->autoclear_features != 0) { | |
493 | s->autoclear_features = 0; | |
494 | ret = qcow2_update_header(bs); | |
495 | if (ret < 0) { | |
496 | goto fail; | |
497 | } | |
498 | } | |
499 | ||
500 | /* Initialise locks */ | |
501 | qemu_co_mutex_init(&s->lock); | |
502 | ||
503 | /* Repair image if dirty */ | |
504 | if (!(flags & BDRV_O_CHECK) && !bs->read_only && | |
505 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { | |
506 | BdrvCheckResult result = {0}; | |
507 | ||
508 | ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS); | |
509 | if (ret < 0) { | |
510 | goto fail; | |
511 | } | |
512 | } | |
513 | ||
514 | /* Enable lazy_refcounts according to image and command line options */ | |
515 | opts = qemu_opts_create_nofail(&qcow2_runtime_opts); | |
516 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
517 | if (error_is_set(&local_err)) { | |
518 | qerror_report_err(local_err); | |
519 | error_free(local_err); | |
520 | ret = -EINVAL; | |
521 | goto fail; | |
522 | } | |
523 | ||
524 | s->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, | |
525 | (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); | |
526 | ||
527 | qemu_opts_del(opts); | |
528 | ||
529 | if (s->use_lazy_refcounts && s->qcow_version < 3) { | |
530 | qerror_report(ERROR_CLASS_GENERIC_ERROR, "Lazy refcounts require " | |
531 | "a qcow2 image with at least qemu 1.1 compatibility level"); | |
532 | ret = -EINVAL; | |
533 | goto fail; | |
534 | } | |
535 | ||
536 | #ifdef DEBUG_ALLOC | |
537 | { | |
538 | BdrvCheckResult result = {0}; | |
539 | qcow2_check_refcounts(bs, &result, 0); | |
540 | } | |
541 | #endif | |
542 | return ret; | |
543 | ||
544 | fail: | |
545 | g_free(s->unknown_header_fields); | |
546 | cleanup_unknown_header_ext(bs); | |
547 | qcow2_free_snapshots(bs); | |
548 | qcow2_refcount_close(bs); | |
549 | g_free(s->l1_table); | |
550 | if (s->l2_table_cache) { | |
551 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
552 | } | |
553 | g_free(s->cluster_cache); | |
554 | qemu_vfree(s->cluster_data); | |
555 | return ret; | |
556 | } | |
557 | ||
558 | static int qcow2_set_key(BlockDriverState *bs, const char *key) | |
559 | { | |
560 | BDRVQcowState *s = bs->opaque; | |
561 | uint8_t keybuf[16]; | |
562 | int len, i; | |
563 | ||
564 | memset(keybuf, 0, 16); | |
565 | len = strlen(key); | |
566 | if (len > 16) | |
567 | len = 16; | |
568 | /* XXX: we could compress the chars to 7 bits to increase | |
569 | entropy */ | |
570 | for(i = 0;i < len;i++) { | |
571 | keybuf[i] = key[i]; | |
572 | } | |
573 | s->crypt_method = s->crypt_method_header; | |
574 | ||
575 | if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0) | |
576 | return -1; | |
577 | if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0) | |
578 | return -1; | |
579 | #if 0 | |
580 | /* test */ | |
581 | { | |
582 | uint8_t in[16]; | |
583 | uint8_t out[16]; | |
584 | uint8_t tmp[16]; | |
585 | for(i=0;i<16;i++) | |
586 | in[i] = i; | |
587 | AES_encrypt(in, tmp, &s->aes_encrypt_key); | |
588 | AES_decrypt(tmp, out, &s->aes_decrypt_key); | |
589 | for(i = 0; i < 16; i++) | |
590 | printf(" %02x", tmp[i]); | |
591 | printf("\n"); | |
592 | for(i = 0; i < 16; i++) | |
593 | printf(" %02x", out[i]); | |
594 | printf("\n"); | |
595 | } | |
596 | #endif | |
597 | return 0; | |
598 | } | |
599 | ||
600 | /* We have nothing to do for QCOW2 reopen, stubs just return | |
601 | * success */ | |
602 | static int qcow2_reopen_prepare(BDRVReopenState *state, | |
603 | BlockReopenQueue *queue, Error **errp) | |
604 | { | |
605 | return 0; | |
606 | } | |
607 | ||
608 | static int coroutine_fn qcow2_co_is_allocated(BlockDriverState *bs, | |
609 | int64_t sector_num, int nb_sectors, int *pnum) | |
610 | { | |
611 | BDRVQcowState *s = bs->opaque; | |
612 | uint64_t cluster_offset; | |
613 | int ret; | |
614 | ||
615 | *pnum = nb_sectors; | |
616 | /* FIXME We can get errors here, but the bdrv_co_is_allocated interface | |
617 | * can't pass them on today */ | |
618 | qemu_co_mutex_lock(&s->lock); | |
619 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, pnum, &cluster_offset); | |
620 | qemu_co_mutex_unlock(&s->lock); | |
621 | if (ret < 0) { | |
622 | *pnum = 0; | |
623 | } | |
624 | ||
625 | return (cluster_offset != 0) || (ret == QCOW2_CLUSTER_ZERO); | |
626 | } | |
627 | ||
628 | /* handle reading after the end of the backing file */ | |
629 | int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, | |
630 | int64_t sector_num, int nb_sectors) | |
631 | { | |
632 | int n1; | |
633 | if ((sector_num + nb_sectors) <= bs->total_sectors) | |
634 | return nb_sectors; | |
635 | if (sector_num >= bs->total_sectors) | |
636 | n1 = 0; | |
637 | else | |
638 | n1 = bs->total_sectors - sector_num; | |
639 | ||
640 | qemu_iovec_memset(qiov, 512 * n1, 0, 512 * (nb_sectors - n1)); | |
641 | ||
642 | return n1; | |
643 | } | |
644 | ||
645 | static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num, | |
646 | int remaining_sectors, QEMUIOVector *qiov) | |
647 | { | |
648 | BDRVQcowState *s = bs->opaque; | |
649 | int index_in_cluster, n1; | |
650 | int ret; | |
651 | int cur_nr_sectors; /* number of sectors in current iteration */ | |
652 | uint64_t cluster_offset = 0; | |
653 | uint64_t bytes_done = 0; | |
654 | QEMUIOVector hd_qiov; | |
655 | uint8_t *cluster_data = NULL; | |
656 | ||
657 | qemu_iovec_init(&hd_qiov, qiov->niov); | |
658 | ||
659 | qemu_co_mutex_lock(&s->lock); | |
660 | ||
661 | while (remaining_sectors != 0) { | |
662 | ||
663 | /* prepare next request */ | |
664 | cur_nr_sectors = remaining_sectors; | |
665 | if (s->crypt_method) { | |
666 | cur_nr_sectors = MIN(cur_nr_sectors, | |
667 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
668 | } | |
669 | ||
670 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, | |
671 | &cur_nr_sectors, &cluster_offset); | |
672 | if (ret < 0) { | |
673 | goto fail; | |
674 | } | |
675 | ||
676 | index_in_cluster = sector_num & (s->cluster_sectors - 1); | |
677 | ||
678 | qemu_iovec_reset(&hd_qiov); | |
679 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, | |
680 | cur_nr_sectors * 512); | |
681 | ||
682 | switch (ret) { | |
683 | case QCOW2_CLUSTER_UNALLOCATED: | |
684 | ||
685 | if (bs->backing_hd) { | |
686 | /* read from the base image */ | |
687 | n1 = qcow2_backing_read1(bs->backing_hd, &hd_qiov, | |
688 | sector_num, cur_nr_sectors); | |
689 | if (n1 > 0) { | |
690 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); | |
691 | qemu_co_mutex_unlock(&s->lock); | |
692 | ret = bdrv_co_readv(bs->backing_hd, sector_num, | |
693 | n1, &hd_qiov); | |
694 | qemu_co_mutex_lock(&s->lock); | |
695 | if (ret < 0) { | |
696 | goto fail; | |
697 | } | |
698 | } | |
699 | } else { | |
700 | /* Note: in this case, no need to wait */ | |
701 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); | |
702 | } | |
703 | break; | |
704 | ||
705 | case QCOW2_CLUSTER_ZERO: | |
706 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); | |
707 | break; | |
708 | ||
709 | case QCOW2_CLUSTER_COMPRESSED: | |
710 | /* add AIO support for compressed blocks ? */ | |
711 | ret = qcow2_decompress_cluster(bs, cluster_offset); | |
712 | if (ret < 0) { | |
713 | goto fail; | |
714 | } | |
715 | ||
716 | qemu_iovec_from_buf(&hd_qiov, 0, | |
717 | s->cluster_cache + index_in_cluster * 512, | |
718 | 512 * cur_nr_sectors); | |
719 | break; | |
720 | ||
721 | case QCOW2_CLUSTER_NORMAL: | |
722 | if ((cluster_offset & 511) != 0) { | |
723 | ret = -EIO; | |
724 | goto fail; | |
725 | } | |
726 | ||
727 | if (s->crypt_method) { | |
728 | /* | |
729 | * For encrypted images, read everything into a temporary | |
730 | * contiguous buffer on which the AES functions can work. | |
731 | */ | |
732 | if (!cluster_data) { | |
733 | cluster_data = | |
734 | qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); | |
735 | } | |
736 | ||
737 | assert(cur_nr_sectors <= | |
738 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
739 | qemu_iovec_reset(&hd_qiov); | |
740 | qemu_iovec_add(&hd_qiov, cluster_data, | |
741 | 512 * cur_nr_sectors); | |
742 | } | |
743 | ||
744 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
745 | qemu_co_mutex_unlock(&s->lock); | |
746 | ret = bdrv_co_readv(bs->file, | |
747 | (cluster_offset >> 9) + index_in_cluster, | |
748 | cur_nr_sectors, &hd_qiov); | |
749 | qemu_co_mutex_lock(&s->lock); | |
750 | if (ret < 0) { | |
751 | goto fail; | |
752 | } | |
753 | if (s->crypt_method) { | |
754 | qcow2_encrypt_sectors(s, sector_num, cluster_data, | |
755 | cluster_data, cur_nr_sectors, 0, &s->aes_decrypt_key); | |
756 | qemu_iovec_from_buf(qiov, bytes_done, | |
757 | cluster_data, 512 * cur_nr_sectors); | |
758 | } | |
759 | break; | |
760 | ||
761 | default: | |
762 | g_assert_not_reached(); | |
763 | ret = -EIO; | |
764 | goto fail; | |
765 | } | |
766 | ||
767 | remaining_sectors -= cur_nr_sectors; | |
768 | sector_num += cur_nr_sectors; | |
769 | bytes_done += cur_nr_sectors * 512; | |
770 | } | |
771 | ret = 0; | |
772 | ||
773 | fail: | |
774 | qemu_co_mutex_unlock(&s->lock); | |
775 | ||
776 | qemu_iovec_destroy(&hd_qiov); | |
777 | qemu_vfree(cluster_data); | |
778 | ||
779 | return ret; | |
780 | } | |
781 | ||
782 | static coroutine_fn int qcow2_co_writev(BlockDriverState *bs, | |
783 | int64_t sector_num, | |
784 | int remaining_sectors, | |
785 | QEMUIOVector *qiov) | |
786 | { | |
787 | BDRVQcowState *s = bs->opaque; | |
788 | int index_in_cluster; | |
789 | int n_end; | |
790 | int ret; | |
791 | int cur_nr_sectors; /* number of sectors in current iteration */ | |
792 | uint64_t cluster_offset; | |
793 | QEMUIOVector hd_qiov; | |
794 | uint64_t bytes_done = 0; | |
795 | uint8_t *cluster_data = NULL; | |
796 | QCowL2Meta *l2meta = NULL; | |
797 | ||
798 | trace_qcow2_writev_start_req(qemu_coroutine_self(), sector_num, | |
799 | remaining_sectors); | |
800 | ||
801 | qemu_iovec_init(&hd_qiov, qiov->niov); | |
802 | ||
803 | s->cluster_cache_offset = -1; /* disable compressed cache */ | |
804 | ||
805 | qemu_co_mutex_lock(&s->lock); | |
806 | ||
807 | while (remaining_sectors != 0) { | |
808 | ||
809 | l2meta = NULL; | |
810 | ||
811 | trace_qcow2_writev_start_part(qemu_coroutine_self()); | |
812 | index_in_cluster = sector_num & (s->cluster_sectors - 1); | |
813 | n_end = index_in_cluster + remaining_sectors; | |
814 | if (s->crypt_method && | |
815 | n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors) { | |
816 | n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors; | |
817 | } | |
818 | ||
819 | ret = qcow2_alloc_cluster_offset(bs, sector_num << 9, | |
820 | index_in_cluster, n_end, &cur_nr_sectors, &cluster_offset, &l2meta); | |
821 | if (ret < 0) { | |
822 | goto fail; | |
823 | } | |
824 | ||
825 | assert((cluster_offset & 511) == 0); | |
826 | ||
827 | qemu_iovec_reset(&hd_qiov); | |
828 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, | |
829 | cur_nr_sectors * 512); | |
830 | ||
831 | if (s->crypt_method) { | |
832 | if (!cluster_data) { | |
833 | cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * | |
834 | s->cluster_size); | |
835 | } | |
836 | ||
837 | assert(hd_qiov.size <= | |
838 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); | |
839 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); | |
840 | ||
841 | qcow2_encrypt_sectors(s, sector_num, cluster_data, | |
842 | cluster_data, cur_nr_sectors, 1, &s->aes_encrypt_key); | |
843 | ||
844 | qemu_iovec_reset(&hd_qiov); | |
845 | qemu_iovec_add(&hd_qiov, cluster_data, | |
846 | cur_nr_sectors * 512); | |
847 | } | |
848 | ||
849 | qemu_co_mutex_unlock(&s->lock); | |
850 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); | |
851 | trace_qcow2_writev_data(qemu_coroutine_self(), | |
852 | (cluster_offset >> 9) + index_in_cluster); | |
853 | ret = bdrv_co_writev(bs->file, | |
854 | (cluster_offset >> 9) + index_in_cluster, | |
855 | cur_nr_sectors, &hd_qiov); | |
856 | qemu_co_mutex_lock(&s->lock); | |
857 | if (ret < 0) { | |
858 | goto fail; | |
859 | } | |
860 | ||
861 | while (l2meta != NULL) { | |
862 | QCowL2Meta *next; | |
863 | ||
864 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); | |
865 | if (ret < 0) { | |
866 | goto fail; | |
867 | } | |
868 | ||
869 | /* Take the request off the list of running requests */ | |
870 | if (l2meta->nb_clusters != 0) { | |
871 | QLIST_REMOVE(l2meta, next_in_flight); | |
872 | } | |
873 | ||
874 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
875 | ||
876 | next = l2meta->next; | |
877 | g_free(l2meta); | |
878 | l2meta = next; | |
879 | } | |
880 | ||
881 | remaining_sectors -= cur_nr_sectors; | |
882 | sector_num += cur_nr_sectors; | |
883 | bytes_done += cur_nr_sectors * 512; | |
884 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_nr_sectors); | |
885 | } | |
886 | ret = 0; | |
887 | ||
888 | fail: | |
889 | qemu_co_mutex_unlock(&s->lock); | |
890 | ||
891 | while (l2meta != NULL) { | |
892 | QCowL2Meta *next; | |
893 | ||
894 | if (l2meta->nb_clusters != 0) { | |
895 | QLIST_REMOVE(l2meta, next_in_flight); | |
896 | } | |
897 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
898 | ||
899 | next = l2meta->next; | |
900 | g_free(l2meta); | |
901 | l2meta = next; | |
902 | } | |
903 | ||
904 | qemu_iovec_destroy(&hd_qiov); | |
905 | qemu_vfree(cluster_data); | |
906 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); | |
907 | ||
908 | return ret; | |
909 | } | |
910 | ||
911 | static void qcow2_close(BlockDriverState *bs) | |
912 | { | |
913 | BDRVQcowState *s = bs->opaque; | |
914 | g_free(s->l1_table); | |
915 | ||
916 | qcow2_cache_flush(bs, s->l2_table_cache); | |
917 | qcow2_cache_flush(bs, s->refcount_block_cache); | |
918 | ||
919 | qcow2_mark_clean(bs); | |
920 | ||
921 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
922 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
923 | ||
924 | g_free(s->unknown_header_fields); | |
925 | cleanup_unknown_header_ext(bs); | |
926 | ||
927 | g_free(s->cluster_cache); | |
928 | qemu_vfree(s->cluster_data); | |
929 | qcow2_refcount_close(bs); | |
930 | qcow2_free_snapshots(bs); | |
931 | } | |
932 | ||
933 | static void qcow2_invalidate_cache(BlockDriverState *bs) | |
934 | { | |
935 | BDRVQcowState *s = bs->opaque; | |
936 | int flags = s->flags; | |
937 | AES_KEY aes_encrypt_key; | |
938 | AES_KEY aes_decrypt_key; | |
939 | uint32_t crypt_method = 0; | |
940 | QDict *options; | |
941 | ||
942 | /* | |
943 | * Backing files are read-only which makes all of their metadata immutable, | |
944 | * that means we don't have to worry about reopening them here. | |
945 | */ | |
946 | ||
947 | if (s->crypt_method) { | |
948 | crypt_method = s->crypt_method; | |
949 | memcpy(&aes_encrypt_key, &s->aes_encrypt_key, sizeof(aes_encrypt_key)); | |
950 | memcpy(&aes_decrypt_key, &s->aes_decrypt_key, sizeof(aes_decrypt_key)); | |
951 | } | |
952 | ||
953 | qcow2_close(bs); | |
954 | ||
955 | options = qdict_new(); | |
956 | qdict_put(options, QCOW2_OPT_LAZY_REFCOUNTS, | |
957 | qbool_from_int(s->use_lazy_refcounts)); | |
958 | ||
959 | memset(s, 0, sizeof(BDRVQcowState)); | |
960 | qcow2_open(bs, options, flags); | |
961 | ||
962 | QDECREF(options); | |
963 | ||
964 | if (crypt_method) { | |
965 | s->crypt_method = crypt_method; | |
966 | memcpy(&s->aes_encrypt_key, &aes_encrypt_key, sizeof(aes_encrypt_key)); | |
967 | memcpy(&s->aes_decrypt_key, &aes_decrypt_key, sizeof(aes_decrypt_key)); | |
968 | } | |
969 | } | |
970 | ||
971 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, | |
972 | size_t len, size_t buflen) | |
973 | { | |
974 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
975 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
976 | ||
977 | if (buflen < ext_len) { | |
978 | return -ENOSPC; | |
979 | } | |
980 | ||
981 | *ext_backing_fmt = (QCowExtension) { | |
982 | .magic = cpu_to_be32(magic), | |
983 | .len = cpu_to_be32(len), | |
984 | }; | |
985 | memcpy(buf + sizeof(QCowExtension), s, len); | |
986 | ||
987 | return ext_len; | |
988 | } | |
989 | ||
990 | /* | |
991 | * Updates the qcow2 header, including the variable length parts of it, i.e. | |
992 | * the backing file name and all extensions. qcow2 was not designed to allow | |
993 | * such changes, so if we run out of space (we can only use the first cluster) | |
994 | * this function may fail. | |
995 | * | |
996 | * Returns 0 on success, -errno in error cases. | |
997 | */ | |
998 | int qcow2_update_header(BlockDriverState *bs) | |
999 | { | |
1000 | BDRVQcowState *s = bs->opaque; | |
1001 | QCowHeader *header; | |
1002 | char *buf; | |
1003 | size_t buflen = s->cluster_size; | |
1004 | int ret; | |
1005 | uint64_t total_size; | |
1006 | uint32_t refcount_table_clusters; | |
1007 | size_t header_length; | |
1008 | Qcow2UnknownHeaderExtension *uext; | |
1009 | ||
1010 | buf = qemu_blockalign(bs, buflen); | |
1011 | ||
1012 | /* Header structure */ | |
1013 | header = (QCowHeader*) buf; | |
1014 | ||
1015 | if (buflen < sizeof(*header)) { | |
1016 | ret = -ENOSPC; | |
1017 | goto fail; | |
1018 | } | |
1019 | ||
1020 | header_length = sizeof(*header) + s->unknown_header_fields_size; | |
1021 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; | |
1022 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
1023 | ||
1024 | *header = (QCowHeader) { | |
1025 | /* Version 2 fields */ | |
1026 | .magic = cpu_to_be32(QCOW_MAGIC), | |
1027 | .version = cpu_to_be32(s->qcow_version), | |
1028 | .backing_file_offset = 0, | |
1029 | .backing_file_size = 0, | |
1030 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
1031 | .size = cpu_to_be64(total_size), | |
1032 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
1033 | .l1_size = cpu_to_be32(s->l1_size), | |
1034 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
1035 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
1036 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
1037 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
1038 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
1039 | ||
1040 | /* Version 3 fields */ | |
1041 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
1042 | .compatible_features = cpu_to_be64(s->compatible_features), | |
1043 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
1044 | .refcount_order = cpu_to_be32(3 + REFCOUNT_SHIFT), | |
1045 | .header_length = cpu_to_be32(header_length), | |
1046 | }; | |
1047 | ||
1048 | /* For older versions, write a shorter header */ | |
1049 | switch (s->qcow_version) { | |
1050 | case 2: | |
1051 | ret = offsetof(QCowHeader, incompatible_features); | |
1052 | break; | |
1053 | case 3: | |
1054 | ret = sizeof(*header); | |
1055 | break; | |
1056 | default: | |
1057 | ret = -EINVAL; | |
1058 | goto fail; | |
1059 | } | |
1060 | ||
1061 | buf += ret; | |
1062 | buflen -= ret; | |
1063 | memset(buf, 0, buflen); | |
1064 | ||
1065 | /* Preserve any unknown field in the header */ | |
1066 | if (s->unknown_header_fields_size) { | |
1067 | if (buflen < s->unknown_header_fields_size) { | |
1068 | ret = -ENOSPC; | |
1069 | goto fail; | |
1070 | } | |
1071 | ||
1072 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
1073 | buf += s->unknown_header_fields_size; | |
1074 | buflen -= s->unknown_header_fields_size; | |
1075 | } | |
1076 | ||
1077 | /* Backing file format header extension */ | |
1078 | if (*bs->backing_format) { | |
1079 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, | |
1080 | bs->backing_format, strlen(bs->backing_format), | |
1081 | buflen); | |
1082 | if (ret < 0) { | |
1083 | goto fail; | |
1084 | } | |
1085 | ||
1086 | buf += ret; | |
1087 | buflen -= ret; | |
1088 | } | |
1089 | ||
1090 | /* Feature table */ | |
1091 | Qcow2Feature features[] = { | |
1092 | { | |
1093 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1094 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
1095 | .name = "dirty bit", | |
1096 | }, | |
1097 | { | |
1098 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
1099 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
1100 | .name = "lazy refcounts", | |
1101 | }, | |
1102 | }; | |
1103 | ||
1104 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
1105 | features, sizeof(features), buflen); | |
1106 | if (ret < 0) { | |
1107 | goto fail; | |
1108 | } | |
1109 | buf += ret; | |
1110 | buflen -= ret; | |
1111 | ||
1112 | /* Keep unknown header extensions */ | |
1113 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
1114 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
1115 | if (ret < 0) { | |
1116 | goto fail; | |
1117 | } | |
1118 | ||
1119 | buf += ret; | |
1120 | buflen -= ret; | |
1121 | } | |
1122 | ||
1123 | /* End of header extensions */ | |
1124 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
1125 | if (ret < 0) { | |
1126 | goto fail; | |
1127 | } | |
1128 | ||
1129 | buf += ret; | |
1130 | buflen -= ret; | |
1131 | ||
1132 | /* Backing file name */ | |
1133 | if (*bs->backing_file) { | |
1134 | size_t backing_file_len = strlen(bs->backing_file); | |
1135 | ||
1136 | if (buflen < backing_file_len) { | |
1137 | ret = -ENOSPC; | |
1138 | goto fail; | |
1139 | } | |
1140 | ||
1141 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ | |
1142 | strncpy(buf, bs->backing_file, buflen); | |
1143 | ||
1144 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
1145 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
1146 | } | |
1147 | ||
1148 | /* Write the new header */ | |
1149 | ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size); | |
1150 | if (ret < 0) { | |
1151 | goto fail; | |
1152 | } | |
1153 | ||
1154 | ret = 0; | |
1155 | fail: | |
1156 | qemu_vfree(header); | |
1157 | return ret; | |
1158 | } | |
1159 | ||
1160 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
1161 | const char *backing_file, const char *backing_fmt) | |
1162 | { | |
1163 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); | |
1164 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1165 | ||
1166 | return qcow2_update_header(bs); | |
1167 | } | |
1168 | ||
1169 | static int preallocate(BlockDriverState *bs) | |
1170 | { | |
1171 | uint64_t nb_sectors; | |
1172 | uint64_t offset; | |
1173 | uint64_t host_offset = 0; | |
1174 | int num; | |
1175 | int ret; | |
1176 | QCowL2Meta *meta; | |
1177 | ||
1178 | nb_sectors = bdrv_getlength(bs) >> 9; | |
1179 | offset = 0; | |
1180 | ||
1181 | while (nb_sectors) { | |
1182 | num = MIN(nb_sectors, INT_MAX >> 9); | |
1183 | ret = qcow2_alloc_cluster_offset(bs, offset, 0, num, &num, | |
1184 | &host_offset, &meta); | |
1185 | if (ret < 0) { | |
1186 | return ret; | |
1187 | } | |
1188 | ||
1189 | ret = qcow2_alloc_cluster_link_l2(bs, meta); | |
1190 | if (ret < 0) { | |
1191 | qcow2_free_any_clusters(bs, meta->alloc_offset, meta->nb_clusters); | |
1192 | return ret; | |
1193 | } | |
1194 | ||
1195 | /* There are no dependent requests, but we need to remove our request | |
1196 | * from the list of in-flight requests */ | |
1197 | if (meta != NULL) { | |
1198 | QLIST_REMOVE(meta, next_in_flight); | |
1199 | } | |
1200 | ||
1201 | /* TODO Preallocate data if requested */ | |
1202 | ||
1203 | nb_sectors -= num; | |
1204 | offset += num << 9; | |
1205 | } | |
1206 | ||
1207 | /* | |
1208 | * It is expected that the image file is large enough to actually contain | |
1209 | * all of the allocated clusters (otherwise we get failing reads after | |
1210 | * EOF). Extend the image to the last allocated sector. | |
1211 | */ | |
1212 | if (host_offset != 0) { | |
1213 | uint8_t buf[512]; | |
1214 | memset(buf, 0, 512); | |
1215 | ret = bdrv_write(bs->file, (host_offset >> 9) + num - 1, buf, 1); | |
1216 | if (ret < 0) { | |
1217 | return ret; | |
1218 | } | |
1219 | } | |
1220 | ||
1221 | return 0; | |
1222 | } | |
1223 | ||
1224 | static int qcow2_create2(const char *filename, int64_t total_size, | |
1225 | const char *backing_file, const char *backing_format, | |
1226 | int flags, size_t cluster_size, int prealloc, | |
1227 | QEMUOptionParameter *options, int version) | |
1228 | { | |
1229 | /* Calculate cluster_bits */ | |
1230 | int cluster_bits; | |
1231 | cluster_bits = ffs(cluster_size) - 1; | |
1232 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || | |
1233 | (1 << cluster_bits) != cluster_size) | |
1234 | { | |
1235 | error_report( | |
1236 | "Cluster size must be a power of two between %d and %dk", | |
1237 | 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); | |
1238 | return -EINVAL; | |
1239 | } | |
1240 | ||
1241 | /* | |
1242 | * Open the image file and write a minimal qcow2 header. | |
1243 | * | |
1244 | * We keep things simple and start with a zero-sized image. We also | |
1245 | * do without refcount blocks or a L1 table for now. We'll fix the | |
1246 | * inconsistency later. | |
1247 | * | |
1248 | * We do need a refcount table because growing the refcount table means | |
1249 | * allocating two new refcount blocks - the seconds of which would be at | |
1250 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
1251 | * size for any qcow2 image. | |
1252 | */ | |
1253 | BlockDriverState* bs; | |
1254 | QCowHeader header; | |
1255 | uint8_t* refcount_table; | |
1256 | int ret; | |
1257 | ||
1258 | ret = bdrv_create_file(filename, options); | |
1259 | if (ret < 0) { | |
1260 | return ret; | |
1261 | } | |
1262 | ||
1263 | ret = bdrv_file_open(&bs, filename, NULL, BDRV_O_RDWR); | |
1264 | if (ret < 0) { | |
1265 | return ret; | |
1266 | } | |
1267 | ||
1268 | /* Write the header */ | |
1269 | memset(&header, 0, sizeof(header)); | |
1270 | header.magic = cpu_to_be32(QCOW_MAGIC); | |
1271 | header.version = cpu_to_be32(version); | |
1272 | header.cluster_bits = cpu_to_be32(cluster_bits); | |
1273 | header.size = cpu_to_be64(0); | |
1274 | header.l1_table_offset = cpu_to_be64(0); | |
1275 | header.l1_size = cpu_to_be32(0); | |
1276 | header.refcount_table_offset = cpu_to_be64(cluster_size); | |
1277 | header.refcount_table_clusters = cpu_to_be32(1); | |
1278 | header.refcount_order = cpu_to_be32(3 + REFCOUNT_SHIFT); | |
1279 | header.header_length = cpu_to_be32(sizeof(header)); | |
1280 | ||
1281 | if (flags & BLOCK_FLAG_ENCRYPT) { | |
1282 | header.crypt_method = cpu_to_be32(QCOW_CRYPT_AES); | |
1283 | } else { | |
1284 | header.crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); | |
1285 | } | |
1286 | ||
1287 | if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) { | |
1288 | header.compatible_features |= | |
1289 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); | |
1290 | } | |
1291 | ||
1292 | ret = bdrv_pwrite(bs, 0, &header, sizeof(header)); | |
1293 | if (ret < 0) { | |
1294 | goto out; | |
1295 | } | |
1296 | ||
1297 | /* Write an empty refcount table */ | |
1298 | refcount_table = g_malloc0(cluster_size); | |
1299 | ret = bdrv_pwrite(bs, cluster_size, refcount_table, cluster_size); | |
1300 | g_free(refcount_table); | |
1301 | ||
1302 | if (ret < 0) { | |
1303 | goto out; | |
1304 | } | |
1305 | ||
1306 | bdrv_close(bs); | |
1307 | ||
1308 | /* | |
1309 | * And now open the image and make it consistent first (i.e. increase the | |
1310 | * refcount of the cluster that is occupied by the header and the refcount | |
1311 | * table) | |
1312 | */ | |
1313 | BlockDriver* drv = bdrv_find_format("qcow2"); | |
1314 | assert(drv != NULL); | |
1315 | ret = bdrv_open(bs, filename, NULL, | |
1316 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_FLUSH, drv); | |
1317 | if (ret < 0) { | |
1318 | goto out; | |
1319 | } | |
1320 | ||
1321 | ret = qcow2_alloc_clusters(bs, 2 * cluster_size); | |
1322 | if (ret < 0) { | |
1323 | goto out; | |
1324 | ||
1325 | } else if (ret != 0) { | |
1326 | error_report("Huh, first cluster in empty image is already in use?"); | |
1327 | abort(); | |
1328 | } | |
1329 | ||
1330 | /* Okay, now that we have a valid image, let's give it the right size */ | |
1331 | ret = bdrv_truncate(bs, total_size * BDRV_SECTOR_SIZE); | |
1332 | if (ret < 0) { | |
1333 | goto out; | |
1334 | } | |
1335 | ||
1336 | /* Want a backing file? There you go.*/ | |
1337 | if (backing_file) { | |
1338 | ret = bdrv_change_backing_file(bs, backing_file, backing_format); | |
1339 | if (ret < 0) { | |
1340 | goto out; | |
1341 | } | |
1342 | } | |
1343 | ||
1344 | /* And if we're supposed to preallocate metadata, do that now */ | |
1345 | if (prealloc) { | |
1346 | BDRVQcowState *s = bs->opaque; | |
1347 | qemu_co_mutex_lock(&s->lock); | |
1348 | ret = preallocate(bs); | |
1349 | qemu_co_mutex_unlock(&s->lock); | |
1350 | if (ret < 0) { | |
1351 | goto out; | |
1352 | } | |
1353 | } | |
1354 | ||
1355 | ret = 0; | |
1356 | out: | |
1357 | bdrv_delete(bs); | |
1358 | return ret; | |
1359 | } | |
1360 | ||
1361 | static int qcow2_create(const char *filename, QEMUOptionParameter *options) | |
1362 | { | |
1363 | const char *backing_file = NULL; | |
1364 | const char *backing_fmt = NULL; | |
1365 | uint64_t sectors = 0; | |
1366 | int flags = 0; | |
1367 | size_t cluster_size = DEFAULT_CLUSTER_SIZE; | |
1368 | int prealloc = 0; | |
1369 | int version = 2; | |
1370 | ||
1371 | /* Read out options */ | |
1372 | while (options && options->name) { | |
1373 | if (!strcmp(options->name, BLOCK_OPT_SIZE)) { | |
1374 | sectors = options->value.n / 512; | |
1375 | } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) { | |
1376 | backing_file = options->value.s; | |
1377 | } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FMT)) { | |
1378 | backing_fmt = options->value.s; | |
1379 | } else if (!strcmp(options->name, BLOCK_OPT_ENCRYPT)) { | |
1380 | flags |= options->value.n ? BLOCK_FLAG_ENCRYPT : 0; | |
1381 | } else if (!strcmp(options->name, BLOCK_OPT_CLUSTER_SIZE)) { | |
1382 | if (options->value.n) { | |
1383 | cluster_size = options->value.n; | |
1384 | } | |
1385 | } else if (!strcmp(options->name, BLOCK_OPT_PREALLOC)) { | |
1386 | if (!options->value.s || !strcmp(options->value.s, "off")) { | |
1387 | prealloc = 0; | |
1388 | } else if (!strcmp(options->value.s, "metadata")) { | |
1389 | prealloc = 1; | |
1390 | } else { | |
1391 | fprintf(stderr, "Invalid preallocation mode: '%s'\n", | |
1392 | options->value.s); | |
1393 | return -EINVAL; | |
1394 | } | |
1395 | } else if (!strcmp(options->name, BLOCK_OPT_COMPAT_LEVEL)) { | |
1396 | if (!options->value.s || !strcmp(options->value.s, "0.10")) { | |
1397 | version = 2; | |
1398 | } else if (!strcmp(options->value.s, "1.1")) { | |
1399 | version = 3; | |
1400 | } else { | |
1401 | fprintf(stderr, "Invalid compatibility level: '%s'\n", | |
1402 | options->value.s); | |
1403 | return -EINVAL; | |
1404 | } | |
1405 | } else if (!strcmp(options->name, BLOCK_OPT_LAZY_REFCOUNTS)) { | |
1406 | flags |= options->value.n ? BLOCK_FLAG_LAZY_REFCOUNTS : 0; | |
1407 | } | |
1408 | options++; | |
1409 | } | |
1410 | ||
1411 | if (backing_file && prealloc) { | |
1412 | fprintf(stderr, "Backing file and preallocation cannot be used at " | |
1413 | "the same time\n"); | |
1414 | return -EINVAL; | |
1415 | } | |
1416 | ||
1417 | if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { | |
1418 | fprintf(stderr, "Lazy refcounts only supported with compatibility " | |
1419 | "level 1.1 and above (use compat=1.1 or greater)\n"); | |
1420 | return -EINVAL; | |
1421 | } | |
1422 | ||
1423 | return qcow2_create2(filename, sectors, backing_file, backing_fmt, flags, | |
1424 | cluster_size, prealloc, options, version); | |
1425 | } | |
1426 | ||
1427 | static int qcow2_make_empty(BlockDriverState *bs) | |
1428 | { | |
1429 | #if 0 | |
1430 | /* XXX: not correct */ | |
1431 | BDRVQcowState *s = bs->opaque; | |
1432 | uint32_t l1_length = s->l1_size * sizeof(uint64_t); | |
1433 | int ret; | |
1434 | ||
1435 | memset(s->l1_table, 0, l1_length); | |
1436 | if (bdrv_pwrite(bs->file, s->l1_table_offset, s->l1_table, l1_length) < 0) | |
1437 | return -1; | |
1438 | ret = bdrv_truncate(bs->file, s->l1_table_offset + l1_length); | |
1439 | if (ret < 0) | |
1440 | return ret; | |
1441 | ||
1442 | l2_cache_reset(bs); | |
1443 | #endif | |
1444 | return 0; | |
1445 | } | |
1446 | ||
1447 | static coroutine_fn int qcow2_co_write_zeroes(BlockDriverState *bs, | |
1448 | int64_t sector_num, int nb_sectors) | |
1449 | { | |
1450 | int ret; | |
1451 | BDRVQcowState *s = bs->opaque; | |
1452 | ||
1453 | /* Emulate misaligned zero writes */ | |
1454 | if (sector_num % s->cluster_sectors || nb_sectors % s->cluster_sectors) { | |
1455 | return -ENOTSUP; | |
1456 | } | |
1457 | ||
1458 | /* Whatever is left can use real zero clusters */ | |
1459 | qemu_co_mutex_lock(&s->lock); | |
1460 | ret = qcow2_zero_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
1461 | nb_sectors); | |
1462 | qemu_co_mutex_unlock(&s->lock); | |
1463 | ||
1464 | return ret; | |
1465 | } | |
1466 | ||
1467 | static coroutine_fn int qcow2_co_discard(BlockDriverState *bs, | |
1468 | int64_t sector_num, int nb_sectors) | |
1469 | { | |
1470 | int ret; | |
1471 | BDRVQcowState *s = bs->opaque; | |
1472 | ||
1473 | qemu_co_mutex_lock(&s->lock); | |
1474 | ret = qcow2_discard_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
1475 | nb_sectors); | |
1476 | qemu_co_mutex_unlock(&s->lock); | |
1477 | return ret; | |
1478 | } | |
1479 | ||
1480 | static int qcow2_truncate(BlockDriverState *bs, int64_t offset) | |
1481 | { | |
1482 | BDRVQcowState *s = bs->opaque; | |
1483 | int ret, new_l1_size; | |
1484 | ||
1485 | if (offset & 511) { | |
1486 | error_report("The new size must be a multiple of 512"); | |
1487 | return -EINVAL; | |
1488 | } | |
1489 | ||
1490 | /* cannot proceed if image has snapshots */ | |
1491 | if (s->nb_snapshots) { | |
1492 | error_report("Can't resize an image which has snapshots"); | |
1493 | return -ENOTSUP; | |
1494 | } | |
1495 | ||
1496 | /* shrinking is currently not supported */ | |
1497 | if (offset < bs->total_sectors * 512) { | |
1498 | error_report("qcow2 doesn't support shrinking images yet"); | |
1499 | return -ENOTSUP; | |
1500 | } | |
1501 | ||
1502 | new_l1_size = size_to_l1(s, offset); | |
1503 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); | |
1504 | if (ret < 0) { | |
1505 | return ret; | |
1506 | } | |
1507 | ||
1508 | /* write updated header.size */ | |
1509 | offset = cpu_to_be64(offset); | |
1510 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), | |
1511 | &offset, sizeof(uint64_t)); | |
1512 | if (ret < 0) { | |
1513 | return ret; | |
1514 | } | |
1515 | ||
1516 | s->l1_vm_state_index = new_l1_size; | |
1517 | return 0; | |
1518 | } | |
1519 | ||
1520 | /* XXX: put compressed sectors first, then all the cluster aligned | |
1521 | tables to avoid losing bytes in alignment */ | |
1522 | static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num, | |
1523 | const uint8_t *buf, int nb_sectors) | |
1524 | { | |
1525 | BDRVQcowState *s = bs->opaque; | |
1526 | z_stream strm; | |
1527 | int ret, out_len; | |
1528 | uint8_t *out_buf; | |
1529 | uint64_t cluster_offset; | |
1530 | ||
1531 | if (nb_sectors == 0) { | |
1532 | /* align end of file to a sector boundary to ease reading with | |
1533 | sector based I/Os */ | |
1534 | cluster_offset = bdrv_getlength(bs->file); | |
1535 | cluster_offset = (cluster_offset + 511) & ~511; | |
1536 | bdrv_truncate(bs->file, cluster_offset); | |
1537 | return 0; | |
1538 | } | |
1539 | ||
1540 | if (nb_sectors != s->cluster_sectors) { | |
1541 | ret = -EINVAL; | |
1542 | ||
1543 | /* Zero-pad last write if image size is not cluster aligned */ | |
1544 | if (sector_num + nb_sectors == bs->total_sectors && | |
1545 | nb_sectors < s->cluster_sectors) { | |
1546 | uint8_t *pad_buf = qemu_blockalign(bs, s->cluster_size); | |
1547 | memset(pad_buf, 0, s->cluster_size); | |
1548 | memcpy(pad_buf, buf, nb_sectors * BDRV_SECTOR_SIZE); | |
1549 | ret = qcow2_write_compressed(bs, sector_num, | |
1550 | pad_buf, s->cluster_sectors); | |
1551 | qemu_vfree(pad_buf); | |
1552 | } | |
1553 | return ret; | |
1554 | } | |
1555 | ||
1556 | out_buf = g_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); | |
1557 | ||
1558 | /* best compression, small window, no zlib header */ | |
1559 | memset(&strm, 0, sizeof(strm)); | |
1560 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
1561 | Z_DEFLATED, -12, | |
1562 | 9, Z_DEFAULT_STRATEGY); | |
1563 | if (ret != 0) { | |
1564 | ret = -EINVAL; | |
1565 | goto fail; | |
1566 | } | |
1567 | ||
1568 | strm.avail_in = s->cluster_size; | |
1569 | strm.next_in = (uint8_t *)buf; | |
1570 | strm.avail_out = s->cluster_size; | |
1571 | strm.next_out = out_buf; | |
1572 | ||
1573 | ret = deflate(&strm, Z_FINISH); | |
1574 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
1575 | deflateEnd(&strm); | |
1576 | ret = -EINVAL; | |
1577 | goto fail; | |
1578 | } | |
1579 | out_len = strm.next_out - out_buf; | |
1580 | ||
1581 | deflateEnd(&strm); | |
1582 | ||
1583 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
1584 | /* could not compress: write normal cluster */ | |
1585 | ret = bdrv_write(bs, sector_num, buf, s->cluster_sectors); | |
1586 | if (ret < 0) { | |
1587 | goto fail; | |
1588 | } | |
1589 | } else { | |
1590 | cluster_offset = qcow2_alloc_compressed_cluster_offset(bs, | |
1591 | sector_num << 9, out_len); | |
1592 | if (!cluster_offset) { | |
1593 | ret = -EIO; | |
1594 | goto fail; | |
1595 | } | |
1596 | cluster_offset &= s->cluster_offset_mask; | |
1597 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); | |
1598 | ret = bdrv_pwrite(bs->file, cluster_offset, out_buf, out_len); | |
1599 | if (ret < 0) { | |
1600 | goto fail; | |
1601 | } | |
1602 | } | |
1603 | ||
1604 | ret = 0; | |
1605 | fail: | |
1606 | g_free(out_buf); | |
1607 | return ret; | |
1608 | } | |
1609 | ||
1610 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) | |
1611 | { | |
1612 | BDRVQcowState *s = bs->opaque; | |
1613 | int ret; | |
1614 | ||
1615 | qemu_co_mutex_lock(&s->lock); | |
1616 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
1617 | if (ret < 0) { | |
1618 | qemu_co_mutex_unlock(&s->lock); | |
1619 | return ret; | |
1620 | } | |
1621 | ||
1622 | if (qcow2_need_accurate_refcounts(s)) { | |
1623 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
1624 | if (ret < 0) { | |
1625 | qemu_co_mutex_unlock(&s->lock); | |
1626 | return ret; | |
1627 | } | |
1628 | } | |
1629 | qemu_co_mutex_unlock(&s->lock); | |
1630 | ||
1631 | return 0; | |
1632 | } | |
1633 | ||
1634 | static int64_t qcow2_vm_state_offset(BDRVQcowState *s) | |
1635 | { | |
1636 | return (int64_t)s->l1_vm_state_index << (s->cluster_bits + s->l2_bits); | |
1637 | } | |
1638 | ||
1639 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) | |
1640 | { | |
1641 | BDRVQcowState *s = bs->opaque; | |
1642 | bdi->cluster_size = s->cluster_size; | |
1643 | bdi->vm_state_offset = qcow2_vm_state_offset(s); | |
1644 | return 0; | |
1645 | } | |
1646 | ||
1647 | #if 0 | |
1648 | static void dump_refcounts(BlockDriverState *bs) | |
1649 | { | |
1650 | BDRVQcowState *s = bs->opaque; | |
1651 | int64_t nb_clusters, k, k1, size; | |
1652 | int refcount; | |
1653 | ||
1654 | size = bdrv_getlength(bs->file); | |
1655 | nb_clusters = size_to_clusters(s, size); | |
1656 | for(k = 0; k < nb_clusters;) { | |
1657 | k1 = k; | |
1658 | refcount = get_refcount(bs, k); | |
1659 | k++; | |
1660 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
1661 | k++; | |
1662 | printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount, | |
1663 | k - k1); | |
1664 | } | |
1665 | } | |
1666 | #endif | |
1667 | ||
1668 | static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, | |
1669 | int64_t pos) | |
1670 | { | |
1671 | BDRVQcowState *s = bs->opaque; | |
1672 | int growable = bs->growable; | |
1673 | int ret; | |
1674 | ||
1675 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); | |
1676 | bs->growable = 1; | |
1677 | ret = bdrv_pwritev(bs, qcow2_vm_state_offset(s) + pos, qiov); | |
1678 | bs->growable = growable; | |
1679 | ||
1680 | return ret; | |
1681 | } | |
1682 | ||
1683 | static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf, | |
1684 | int64_t pos, int size) | |
1685 | { | |
1686 | BDRVQcowState *s = bs->opaque; | |
1687 | int growable = bs->growable; | |
1688 | int ret; | |
1689 | ||
1690 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); | |
1691 | bs->growable = 1; | |
1692 | ret = bdrv_pread(bs, qcow2_vm_state_offset(s) + pos, buf, size); | |
1693 | bs->growable = growable; | |
1694 | ||
1695 | return ret; | |
1696 | } | |
1697 | ||
1698 | static QEMUOptionParameter qcow2_create_options[] = { | |
1699 | { | |
1700 | .name = BLOCK_OPT_SIZE, | |
1701 | .type = OPT_SIZE, | |
1702 | .help = "Virtual disk size" | |
1703 | }, | |
1704 | { | |
1705 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
1706 | .type = OPT_STRING, | |
1707 | .help = "Compatibility level (0.10 or 1.1)" | |
1708 | }, | |
1709 | { | |
1710 | .name = BLOCK_OPT_BACKING_FILE, | |
1711 | .type = OPT_STRING, | |
1712 | .help = "File name of a base image" | |
1713 | }, | |
1714 | { | |
1715 | .name = BLOCK_OPT_BACKING_FMT, | |
1716 | .type = OPT_STRING, | |
1717 | .help = "Image format of the base image" | |
1718 | }, | |
1719 | { | |
1720 | .name = BLOCK_OPT_ENCRYPT, | |
1721 | .type = OPT_FLAG, | |
1722 | .help = "Encrypt the image" | |
1723 | }, | |
1724 | { | |
1725 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
1726 | .type = OPT_SIZE, | |
1727 | .help = "qcow2 cluster size", | |
1728 | .value = { .n = DEFAULT_CLUSTER_SIZE }, | |
1729 | }, | |
1730 | { | |
1731 | .name = BLOCK_OPT_PREALLOC, | |
1732 | .type = OPT_STRING, | |
1733 | .help = "Preallocation mode (allowed values: off, metadata)" | |
1734 | }, | |
1735 | { | |
1736 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
1737 | .type = OPT_FLAG, | |
1738 | .help = "Postpone refcount updates", | |
1739 | }, | |
1740 | { NULL } | |
1741 | }; | |
1742 | ||
1743 | static BlockDriver bdrv_qcow2 = { | |
1744 | .format_name = "qcow2", | |
1745 | .instance_size = sizeof(BDRVQcowState), | |
1746 | .bdrv_probe = qcow2_probe, | |
1747 | .bdrv_open = qcow2_open, | |
1748 | .bdrv_close = qcow2_close, | |
1749 | .bdrv_reopen_prepare = qcow2_reopen_prepare, | |
1750 | .bdrv_create = qcow2_create, | |
1751 | .bdrv_co_is_allocated = qcow2_co_is_allocated, | |
1752 | .bdrv_set_key = qcow2_set_key, | |
1753 | .bdrv_make_empty = qcow2_make_empty, | |
1754 | ||
1755 | .bdrv_co_readv = qcow2_co_readv, | |
1756 | .bdrv_co_writev = qcow2_co_writev, | |
1757 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, | |
1758 | ||
1759 | .bdrv_co_write_zeroes = qcow2_co_write_zeroes, | |
1760 | .bdrv_co_discard = qcow2_co_discard, | |
1761 | .bdrv_truncate = qcow2_truncate, | |
1762 | .bdrv_write_compressed = qcow2_write_compressed, | |
1763 | ||
1764 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
1765 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
1766 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
1767 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1768 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, | |
1769 | .bdrv_get_info = qcow2_get_info, | |
1770 | ||
1771 | .bdrv_save_vmstate = qcow2_save_vmstate, | |
1772 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
1773 | ||
1774 | .bdrv_change_backing_file = qcow2_change_backing_file, | |
1775 | ||
1776 | .bdrv_invalidate_cache = qcow2_invalidate_cache, | |
1777 | ||
1778 | .create_options = qcow2_create_options, | |
1779 | .bdrv_check = qcow2_check, | |
1780 | }; | |
1781 | ||
1782 | static void bdrv_qcow2_init(void) | |
1783 | { | |
1784 | bdrv_register(&bdrv_qcow2); | |
1785 | } | |
1786 | ||
1787 | block_init(bdrv_qcow2_init); |