]>
Commit | Line | Data |
---|---|---|
585f8587 FB |
1 | /* |
2 | * Block driver for the QCOW version 2 format | |
5fafdf24 | 3 | * |
585f8587 | 4 | * Copyright (c) 2004-2006 Fabrice Bellard |
5fafdf24 | 5 | * |
585f8587 FB |
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 | */ | |
80c71a24 | 24 | #include "qemu/osdep.h" |
737e150e | 25 | #include "block/block_int.h" |
23588797 | 26 | #include "sysemu/block-backend.h" |
1de7afc9 | 27 | #include "qemu/module.h" |
585f8587 | 28 | #include <zlib.h> |
f7d0fe02 | 29 | #include "block/qcow2.h" |
1de7afc9 | 30 | #include "qemu/error-report.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qerror.h" |
acdfb480 | 32 | #include "qapi/qmp/qbool.h" |
ffeaac9b | 33 | #include "qapi/util.h" |
85186ebd HR |
34 | #include "qapi/qmp/types.h" |
35 | #include "qapi-event.h" | |
3cce16f4 | 36 | #include "trace.h" |
1bd0e2d1 | 37 | #include "qemu/option_int.h" |
f348b6d1 | 38 | #include "qemu/cutils.h" |
58369e22 | 39 | #include "qemu/bswap.h" |
b25b387f DB |
40 | #include "qapi/opts-visitor.h" |
41 | #include "qapi-visit.h" | |
42 | #include "block/crypto.h" | |
585f8587 FB |
43 | |
44 | /* | |
45 | Differences with QCOW: | |
46 | ||
47 | - Support for multiple incremental snapshots. | |
48 | - Memory management by reference counts. | |
49 | - Clusters which have a reference count of one have the bit | |
50 | QCOW_OFLAG_COPIED to optimize write performance. | |
5fafdf24 | 51 | - Size of compressed clusters is stored in sectors to reduce bit usage |
585f8587 FB |
52 | in the cluster offsets. |
53 | - Support for storing additional data (such as the VM state) in the | |
3b46e624 | 54 | snapshots. |
585f8587 FB |
55 | - If a backing store is used, the cluster size is not constrained |
56 | (could be backported to QCOW). | |
57 | - L2 tables have always a size of one cluster. | |
58 | */ | |
59 | ||
9b80ddf3 AL |
60 | |
61 | typedef struct { | |
62 | uint32_t magic; | |
63 | uint32_t len; | |
c4217f64 | 64 | } QEMU_PACKED QCowExtension; |
21d82ac9 | 65 | |
7c80ab3f JS |
66 | #define QCOW2_EXT_MAGIC_END 0 |
67 | #define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA | |
cfcc4c62 | 68 | #define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 |
4652b8f3 | 69 | #define QCOW2_EXT_MAGIC_CRYPTO_HEADER 0x0537be77 |
9b80ddf3 | 70 | |
7c80ab3f | 71 | static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) |
585f8587 FB |
72 | { |
73 | const QCowHeader *cow_header = (const void *)buf; | |
3b46e624 | 74 | |
585f8587 FB |
75 | if (buf_size >= sizeof(QCowHeader) && |
76 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
6744cbab | 77 | be32_to_cpu(cow_header->version) >= 2) |
585f8587 FB |
78 | return 100; |
79 | else | |
80 | return 0; | |
81 | } | |
82 | ||
9b80ddf3 | 83 | |
4652b8f3 DB |
84 | static ssize_t qcow2_crypto_hdr_read_func(QCryptoBlock *block, size_t offset, |
85 | uint8_t *buf, size_t buflen, | |
86 | void *opaque, Error **errp) | |
87 | { | |
88 | BlockDriverState *bs = opaque; | |
89 | BDRVQcow2State *s = bs->opaque; | |
90 | ssize_t ret; | |
91 | ||
92 | if ((offset + buflen) > s->crypto_header.length) { | |
93 | error_setg(errp, "Request for data outside of extension header"); | |
94 | return -1; | |
95 | } | |
96 | ||
97 | ret = bdrv_pread(bs->file, | |
98 | s->crypto_header.offset + offset, buf, buflen); | |
99 | if (ret < 0) { | |
100 | error_setg_errno(errp, -ret, "Could not read encryption header"); | |
101 | return -1; | |
102 | } | |
103 | return ret; | |
104 | } | |
105 | ||
106 | ||
107 | static ssize_t qcow2_crypto_hdr_init_func(QCryptoBlock *block, size_t headerlen, | |
108 | void *opaque, Error **errp) | |
109 | { | |
110 | BlockDriverState *bs = opaque; | |
111 | BDRVQcow2State *s = bs->opaque; | |
112 | int64_t ret; | |
113 | int64_t clusterlen; | |
114 | ||
115 | ret = qcow2_alloc_clusters(bs, headerlen); | |
116 | if (ret < 0) { | |
117 | error_setg_errno(errp, -ret, | |
118 | "Cannot allocate cluster for LUKS header size %zu", | |
119 | headerlen); | |
120 | return -1; | |
121 | } | |
122 | ||
123 | s->crypto_header.length = headerlen; | |
124 | s->crypto_header.offset = ret; | |
125 | ||
126 | /* Zero fill remaining space in cluster so it has predictable | |
127 | * content in case of future spec changes */ | |
128 | clusterlen = size_to_clusters(s, headerlen) * s->cluster_size; | |
129 | ret = bdrv_pwrite_zeroes(bs->file, | |
130 | ret + headerlen, | |
131 | clusterlen - headerlen, 0); | |
132 | if (ret < 0) { | |
133 | error_setg_errno(errp, -ret, "Could not zero fill encryption header"); | |
134 | return -1; | |
135 | } | |
136 | ||
137 | return ret; | |
138 | } | |
139 | ||
140 | ||
141 | static ssize_t qcow2_crypto_hdr_write_func(QCryptoBlock *block, size_t offset, | |
142 | const uint8_t *buf, size_t buflen, | |
143 | void *opaque, Error **errp) | |
144 | { | |
145 | BlockDriverState *bs = opaque; | |
146 | BDRVQcow2State *s = bs->opaque; | |
147 | ssize_t ret; | |
148 | ||
149 | if ((offset + buflen) > s->crypto_header.length) { | |
150 | error_setg(errp, "Request for data outside of extension header"); | |
151 | return -1; | |
152 | } | |
153 | ||
154 | ret = bdrv_pwrite(bs->file, | |
155 | s->crypto_header.offset + offset, buf, buflen); | |
156 | if (ret < 0) { | |
157 | error_setg_errno(errp, -ret, "Could not read encryption header"); | |
158 | return -1; | |
159 | } | |
160 | return ret; | |
161 | } | |
162 | ||
163 | ||
9b80ddf3 AL |
164 | /* |
165 | * read qcow2 extension and fill bs | |
166 | * start reading from start_offset | |
167 | * finish reading upon magic of value 0 or when end_offset reached | |
168 | * unknown magic is skipped (future extension this version knows nothing about) | |
169 | * return 0 upon success, non-0 otherwise | |
170 | */ | |
7c80ab3f | 171 | static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
3ef6c40a | 172 | uint64_t end_offset, void **p_feature_table, |
4652b8f3 | 173 | int flags, Error **errp) |
9b80ddf3 | 174 | { |
ff99129a | 175 | BDRVQcow2State *s = bs->opaque; |
9b80ddf3 AL |
176 | QCowExtension ext; |
177 | uint64_t offset; | |
75bab85c | 178 | int ret; |
9b80ddf3 AL |
179 | |
180 | #ifdef DEBUG_EXT | |
7c80ab3f | 181 | printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); |
9b80ddf3 AL |
182 | #endif |
183 | offset = start_offset; | |
184 | while (offset < end_offset) { | |
185 | ||
186 | #ifdef DEBUG_EXT | |
187 | /* Sanity check */ | |
188 | if (offset > s->cluster_size) | |
7c80ab3f | 189 | printf("qcow2_read_extension: suspicious offset %lu\n", offset); |
9b80ddf3 | 190 | |
9b2260cb | 191 | printf("attempting to read extended header in offset %lu\n", offset); |
9b80ddf3 AL |
192 | #endif |
193 | ||
cf2ab8fc | 194 | ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext)); |
3ef6c40a HR |
195 | if (ret < 0) { |
196 | error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: " | |
197 | "pread fail from offset %" PRIu64, offset); | |
9b80ddf3 AL |
198 | return 1; |
199 | } | |
200 | be32_to_cpus(&ext.magic); | |
201 | be32_to_cpus(&ext.len); | |
202 | offset += sizeof(ext); | |
203 | #ifdef DEBUG_EXT | |
204 | printf("ext.magic = 0x%x\n", ext.magic); | |
205 | #endif | |
2ebafc85 | 206 | if (offset > end_offset || ext.len > end_offset - offset) { |
3ef6c40a | 207 | error_setg(errp, "Header extension too large"); |
64ca6aee KW |
208 | return -EINVAL; |
209 | } | |
210 | ||
9b80ddf3 | 211 | switch (ext.magic) { |
7c80ab3f | 212 | case QCOW2_EXT_MAGIC_END: |
9b80ddf3 | 213 | return 0; |
f965509c | 214 | |
7c80ab3f | 215 | case QCOW2_EXT_MAGIC_BACKING_FORMAT: |
f965509c | 216 | if (ext.len >= sizeof(bs->backing_format)) { |
521b2b5d HR |
217 | error_setg(errp, "ERROR: ext_backing_format: len=%" PRIu32 |
218 | " too large (>=%zu)", ext.len, | |
219 | sizeof(bs->backing_format)); | |
f965509c AL |
220 | return 2; |
221 | } | |
cf2ab8fc | 222 | ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len); |
3ef6c40a HR |
223 | if (ret < 0) { |
224 | error_setg_errno(errp, -ret, "ERROR: ext_backing_format: " | |
225 | "Could not read format name"); | |
f965509c | 226 | return 3; |
3ef6c40a | 227 | } |
f965509c | 228 | bs->backing_format[ext.len] = '\0'; |
e4603fe1 | 229 | s->image_backing_format = g_strdup(bs->backing_format); |
f965509c AL |
230 | #ifdef DEBUG_EXT |
231 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
232 | #endif | |
f965509c AL |
233 | break; |
234 | ||
cfcc4c62 KW |
235 | case QCOW2_EXT_MAGIC_FEATURE_TABLE: |
236 | if (p_feature_table != NULL) { | |
237 | void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); | |
cf2ab8fc | 238 | ret = bdrv_pread(bs->file, offset , feature_table, ext.len); |
cfcc4c62 | 239 | if (ret < 0) { |
3ef6c40a HR |
240 | error_setg_errno(errp, -ret, "ERROR: ext_feature_table: " |
241 | "Could not read table"); | |
cfcc4c62 KW |
242 | return ret; |
243 | } | |
244 | ||
245 | *p_feature_table = feature_table; | |
246 | } | |
247 | break; | |
248 | ||
4652b8f3 DB |
249 | case QCOW2_EXT_MAGIC_CRYPTO_HEADER: { |
250 | unsigned int cflags = 0; | |
251 | if (s->crypt_method_header != QCOW_CRYPT_LUKS) { | |
252 | error_setg(errp, "CRYPTO header extension only " | |
253 | "expected with LUKS encryption method"); | |
254 | return -EINVAL; | |
255 | } | |
256 | if (ext.len != sizeof(Qcow2CryptoHeaderExtension)) { | |
257 | error_setg(errp, "CRYPTO header extension size %u, " | |
258 | "but expected size %zu", ext.len, | |
259 | sizeof(Qcow2CryptoHeaderExtension)); | |
260 | return -EINVAL; | |
261 | } | |
262 | ||
263 | ret = bdrv_pread(bs->file, offset, &s->crypto_header, ext.len); | |
264 | if (ret < 0) { | |
265 | error_setg_errno(errp, -ret, | |
266 | "Unable to read CRYPTO header extension"); | |
267 | return ret; | |
268 | } | |
269 | be64_to_cpus(&s->crypto_header.offset); | |
270 | be64_to_cpus(&s->crypto_header.length); | |
271 | ||
272 | if ((s->crypto_header.offset % s->cluster_size) != 0) { | |
273 | error_setg(errp, "Encryption header offset '%" PRIu64 "' is " | |
274 | "not a multiple of cluster size '%u'", | |
275 | s->crypto_header.offset, s->cluster_size); | |
276 | return -EINVAL; | |
277 | } | |
278 | ||
279 | if (flags & BDRV_O_NO_IO) { | |
280 | cflags |= QCRYPTO_BLOCK_OPEN_NO_IO; | |
281 | } | |
282 | s->crypto = qcrypto_block_open(s->crypto_opts, | |
283 | qcow2_crypto_hdr_read_func, | |
284 | bs, cflags, errp); | |
285 | if (!s->crypto) { | |
286 | return -EINVAL; | |
287 | } | |
288 | } break; | |
289 | ||
9b80ddf3 | 290 | default: |
75bab85c KW |
291 | /* unknown magic - save it in case we need to rewrite the header */ |
292 | { | |
293 | Qcow2UnknownHeaderExtension *uext; | |
294 | ||
295 | uext = g_malloc0(sizeof(*uext) + ext.len); | |
296 | uext->magic = ext.magic; | |
297 | uext->len = ext.len; | |
298 | QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); | |
299 | ||
cf2ab8fc | 300 | ret = bdrv_pread(bs->file, offset , uext->data, uext->len); |
75bab85c | 301 | if (ret < 0) { |
3ef6c40a HR |
302 | error_setg_errno(errp, -ret, "ERROR: unknown extension: " |
303 | "Could not read data"); | |
75bab85c KW |
304 | return ret; |
305 | } | |
75bab85c | 306 | } |
9b80ddf3 AL |
307 | break; |
308 | } | |
fd29b4bb KW |
309 | |
310 | offset += ((ext.len + 7) & ~7); | |
9b80ddf3 AL |
311 | } |
312 | ||
313 | return 0; | |
314 | } | |
315 | ||
75bab85c KW |
316 | static void cleanup_unknown_header_ext(BlockDriverState *bs) |
317 | { | |
ff99129a | 318 | BDRVQcow2State *s = bs->opaque; |
75bab85c KW |
319 | Qcow2UnknownHeaderExtension *uext, *next; |
320 | ||
321 | QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { | |
322 | QLIST_REMOVE(uext, next); | |
323 | g_free(uext); | |
324 | } | |
325 | } | |
9b80ddf3 | 326 | |
a55448b3 HR |
327 | static void report_unsupported_feature(Error **errp, Qcow2Feature *table, |
328 | uint64_t mask) | |
cfcc4c62 | 329 | { |
12ac6d3d KW |
330 | char *features = g_strdup(""); |
331 | char *old; | |
332 | ||
cfcc4c62 KW |
333 | while (table && table->name[0] != '\0') { |
334 | if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) { | |
12ac6d3d KW |
335 | if (mask & (1ULL << table->bit)) { |
336 | old = features; | |
337 | features = g_strdup_printf("%s%s%.46s", old, *old ? ", " : "", | |
338 | table->name); | |
339 | g_free(old); | |
340 | mask &= ~(1ULL << table->bit); | |
cfcc4c62 KW |
341 | } |
342 | } | |
343 | table++; | |
344 | } | |
345 | ||
346 | if (mask) { | |
12ac6d3d KW |
347 | old = features; |
348 | features = g_strdup_printf("%s%sUnknown incompatible feature: %" PRIx64, | |
349 | old, *old ? ", " : "", mask); | |
350 | g_free(old); | |
cfcc4c62 | 351 | } |
12ac6d3d | 352 | |
a55448b3 | 353 | error_setg(errp, "Unsupported qcow2 feature(s): %s", features); |
12ac6d3d | 354 | g_free(features); |
cfcc4c62 KW |
355 | } |
356 | ||
bfe8043e SH |
357 | /* |
358 | * Sets the dirty bit and flushes afterwards if necessary. | |
359 | * | |
360 | * The incompatible_features bit is only set if the image file header was | |
361 | * updated successfully. Therefore it is not required to check the return | |
362 | * value of this function. | |
363 | */ | |
280d3735 | 364 | int qcow2_mark_dirty(BlockDriverState *bs) |
bfe8043e | 365 | { |
ff99129a | 366 | BDRVQcow2State *s = bs->opaque; |
bfe8043e SH |
367 | uint64_t val; |
368 | int ret; | |
369 | ||
370 | assert(s->qcow_version >= 3); | |
371 | ||
372 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
373 | return 0; /* already dirty */ | |
374 | } | |
375 | ||
376 | val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); | |
d9ca2ea2 | 377 | ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features), |
bfe8043e SH |
378 | &val, sizeof(val)); |
379 | if (ret < 0) { | |
380 | return ret; | |
381 | } | |
9a4f4c31 | 382 | ret = bdrv_flush(bs->file->bs); |
bfe8043e SH |
383 | if (ret < 0) { |
384 | return ret; | |
385 | } | |
386 | ||
387 | /* Only treat image as dirty if the header was updated successfully */ | |
388 | s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; | |
389 | return 0; | |
390 | } | |
391 | ||
c61d0004 SH |
392 | /* |
393 | * Clears the dirty bit and flushes before if necessary. Only call this | |
394 | * function when there are no pending requests, it does not guard against | |
395 | * concurrent requests dirtying the image. | |
396 | */ | |
397 | static int qcow2_mark_clean(BlockDriverState *bs) | |
398 | { | |
ff99129a | 399 | BDRVQcow2State *s = bs->opaque; |
c61d0004 SH |
400 | |
401 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
4c2e5f8f KW |
402 | int ret; |
403 | ||
404 | s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; | |
405 | ||
406 | ret = bdrv_flush(bs); | |
c61d0004 SH |
407 | if (ret < 0) { |
408 | return ret; | |
409 | } | |
410 | ||
c61d0004 SH |
411 | return qcow2_update_header(bs); |
412 | } | |
413 | return 0; | |
414 | } | |
415 | ||
69c98726 HR |
416 | /* |
417 | * Marks the image as corrupt. | |
418 | */ | |
419 | int qcow2_mark_corrupt(BlockDriverState *bs) | |
420 | { | |
ff99129a | 421 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
422 | |
423 | s->incompatible_features |= QCOW2_INCOMPAT_CORRUPT; | |
424 | return qcow2_update_header(bs); | |
425 | } | |
426 | ||
427 | /* | |
428 | * Marks the image as consistent, i.e., unsets the corrupt bit, and flushes | |
429 | * before if necessary. | |
430 | */ | |
431 | int qcow2_mark_consistent(BlockDriverState *bs) | |
432 | { | |
ff99129a | 433 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
434 | |
435 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { | |
436 | int ret = bdrv_flush(bs); | |
437 | if (ret < 0) { | |
438 | return ret; | |
439 | } | |
440 | ||
441 | s->incompatible_features &= ~QCOW2_INCOMPAT_CORRUPT; | |
442 | return qcow2_update_header(bs); | |
443 | } | |
444 | return 0; | |
445 | } | |
446 | ||
acbe5982 SH |
447 | static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result, |
448 | BdrvCheckMode fix) | |
449 | { | |
450 | int ret = qcow2_check_refcounts(bs, result, fix); | |
451 | if (ret < 0) { | |
452 | return ret; | |
453 | } | |
454 | ||
455 | if (fix && result->check_errors == 0 && result->corruptions == 0) { | |
24530f3e HR |
456 | ret = qcow2_mark_clean(bs); |
457 | if (ret < 0) { | |
458 | return ret; | |
459 | } | |
460 | return qcow2_mark_consistent(bs); | |
acbe5982 SH |
461 | } |
462 | return ret; | |
463 | } | |
464 | ||
8c7de283 KW |
465 | static int validate_table_offset(BlockDriverState *bs, uint64_t offset, |
466 | uint64_t entries, size_t entry_len) | |
467 | { | |
ff99129a | 468 | BDRVQcow2State *s = bs->opaque; |
8c7de283 KW |
469 | uint64_t size; |
470 | ||
471 | /* Use signed INT64_MAX as the maximum even for uint64_t header fields, | |
472 | * because values will be passed to qemu functions taking int64_t. */ | |
473 | if (entries > INT64_MAX / entry_len) { | |
474 | return -EINVAL; | |
475 | } | |
476 | ||
477 | size = entries * entry_len; | |
478 | ||
479 | if (INT64_MAX - size < offset) { | |
480 | return -EINVAL; | |
481 | } | |
482 | ||
483 | /* Tables must be cluster aligned */ | |
24990c5b | 484 | if (offset_into_cluster(s, offset) != 0) { |
8c7de283 KW |
485 | return -EINVAL; |
486 | } | |
487 | ||
488 | return 0; | |
489 | } | |
490 | ||
74c4510a KW |
491 | static QemuOptsList qcow2_runtime_opts = { |
492 | .name = "qcow2", | |
493 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head), | |
494 | .desc = { | |
495 | { | |
64aa99d3 | 496 | .name = QCOW2_OPT_LAZY_REFCOUNTS, |
74c4510a KW |
497 | .type = QEMU_OPT_BOOL, |
498 | .help = "Postpone refcount updates", | |
499 | }, | |
67af674e KW |
500 | { |
501 | .name = QCOW2_OPT_DISCARD_REQUEST, | |
502 | .type = QEMU_OPT_BOOL, | |
503 | .help = "Pass guest discard requests to the layer below", | |
504 | }, | |
505 | { | |
506 | .name = QCOW2_OPT_DISCARD_SNAPSHOT, | |
507 | .type = QEMU_OPT_BOOL, | |
508 | .help = "Generate discard requests when snapshot related space " | |
509 | "is freed", | |
510 | }, | |
511 | { | |
512 | .name = QCOW2_OPT_DISCARD_OTHER, | |
513 | .type = QEMU_OPT_BOOL, | |
514 | .help = "Generate discard requests when other clusters are freed", | |
515 | }, | |
05de7e86 HR |
516 | { |
517 | .name = QCOW2_OPT_OVERLAP, | |
518 | .type = QEMU_OPT_STRING, | |
519 | .help = "Selects which overlap checks to perform from a range of " | |
520 | "templates (none, constant, cached, all)", | |
521 | }, | |
ee42b5ce HR |
522 | { |
523 | .name = QCOW2_OPT_OVERLAP_TEMPLATE, | |
524 | .type = QEMU_OPT_STRING, | |
525 | .help = "Selects which overlap checks to perform from a range of " | |
526 | "templates (none, constant, cached, all)", | |
527 | }, | |
05de7e86 HR |
528 | { |
529 | .name = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
530 | .type = QEMU_OPT_BOOL, | |
531 | .help = "Check for unintended writes into the main qcow2 header", | |
532 | }, | |
533 | { | |
534 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
535 | .type = QEMU_OPT_BOOL, | |
536 | .help = "Check for unintended writes into the active L1 table", | |
537 | }, | |
538 | { | |
539 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
540 | .type = QEMU_OPT_BOOL, | |
541 | .help = "Check for unintended writes into an active L2 table", | |
542 | }, | |
543 | { | |
544 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
545 | .type = QEMU_OPT_BOOL, | |
546 | .help = "Check for unintended writes into the refcount table", | |
547 | }, | |
548 | { | |
549 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
550 | .type = QEMU_OPT_BOOL, | |
551 | .help = "Check for unintended writes into a refcount block", | |
552 | }, | |
553 | { | |
554 | .name = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
555 | .type = QEMU_OPT_BOOL, | |
556 | .help = "Check for unintended writes into the snapshot table", | |
557 | }, | |
558 | { | |
559 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
560 | .type = QEMU_OPT_BOOL, | |
561 | .help = "Check for unintended writes into an inactive L1 table", | |
562 | }, | |
563 | { | |
564 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
565 | .type = QEMU_OPT_BOOL, | |
566 | .help = "Check for unintended writes into an inactive L2 table", | |
567 | }, | |
6c1c8d5d HR |
568 | { |
569 | .name = QCOW2_OPT_CACHE_SIZE, | |
570 | .type = QEMU_OPT_SIZE, | |
571 | .help = "Maximum combined metadata (L2 tables and refcount blocks) " | |
572 | "cache size", | |
573 | }, | |
574 | { | |
575 | .name = QCOW2_OPT_L2_CACHE_SIZE, | |
576 | .type = QEMU_OPT_SIZE, | |
577 | .help = "Maximum L2 table cache size", | |
578 | }, | |
579 | { | |
580 | .name = QCOW2_OPT_REFCOUNT_CACHE_SIZE, | |
581 | .type = QEMU_OPT_SIZE, | |
582 | .help = "Maximum refcount block cache size", | |
583 | }, | |
279621c0 AG |
584 | { |
585 | .name = QCOW2_OPT_CACHE_CLEAN_INTERVAL, | |
586 | .type = QEMU_OPT_NUMBER, | |
587 | .help = "Clean unused cache entries after this time (in seconds)", | |
588 | }, | |
4652b8f3 DB |
589 | BLOCK_CRYPTO_OPT_DEF_KEY_SECRET("encrypt.", |
590 | "ID of secret providing qcow2 AES key or LUKS passphrase"), | |
74c4510a KW |
591 | { /* end of list */ } |
592 | }, | |
593 | }; | |
594 | ||
4092e99d HR |
595 | static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = { |
596 | [QCOW2_OL_MAIN_HEADER_BITNR] = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
597 | [QCOW2_OL_ACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
598 | [QCOW2_OL_ACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
599 | [QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
600 | [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
601 | [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
602 | [QCOW2_OL_INACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
603 | [QCOW2_OL_INACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
604 | }; | |
605 | ||
279621c0 AG |
606 | static void cache_clean_timer_cb(void *opaque) |
607 | { | |
608 | BlockDriverState *bs = opaque; | |
ff99129a | 609 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
610 | qcow2_cache_clean_unused(bs, s->l2_table_cache); |
611 | qcow2_cache_clean_unused(bs, s->refcount_block_cache); | |
612 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
613 | (int64_t) s->cache_clean_interval * 1000); | |
614 | } | |
615 | ||
616 | static void cache_clean_timer_init(BlockDriverState *bs, AioContext *context) | |
617 | { | |
ff99129a | 618 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
619 | if (s->cache_clean_interval > 0) { |
620 | s->cache_clean_timer = aio_timer_new(context, QEMU_CLOCK_VIRTUAL, | |
621 | SCALE_MS, cache_clean_timer_cb, | |
622 | bs); | |
623 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
624 | (int64_t) s->cache_clean_interval * 1000); | |
625 | } | |
626 | } | |
627 | ||
628 | static void cache_clean_timer_del(BlockDriverState *bs) | |
629 | { | |
ff99129a | 630 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
631 | if (s->cache_clean_timer) { |
632 | timer_del(s->cache_clean_timer); | |
633 | timer_free(s->cache_clean_timer); | |
634 | s->cache_clean_timer = NULL; | |
635 | } | |
636 | } | |
637 | ||
638 | static void qcow2_detach_aio_context(BlockDriverState *bs) | |
639 | { | |
640 | cache_clean_timer_del(bs); | |
641 | } | |
642 | ||
643 | static void qcow2_attach_aio_context(BlockDriverState *bs, | |
644 | AioContext *new_context) | |
645 | { | |
646 | cache_clean_timer_init(bs, new_context); | |
647 | } | |
648 | ||
bc85ef26 HR |
649 | static void read_cache_sizes(BlockDriverState *bs, QemuOpts *opts, |
650 | uint64_t *l2_cache_size, | |
6c1c8d5d HR |
651 | uint64_t *refcount_cache_size, Error **errp) |
652 | { | |
ff99129a | 653 | BDRVQcow2State *s = bs->opaque; |
6c1c8d5d HR |
654 | uint64_t combined_cache_size; |
655 | bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set; | |
656 | ||
657 | combined_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_CACHE_SIZE); | |
658 | l2_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_SIZE); | |
659 | refcount_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
660 | ||
661 | combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0); | |
662 | *l2_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE, 0); | |
663 | *refcount_cache_size = qemu_opt_get_size(opts, | |
664 | QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0); | |
665 | ||
666 | if (combined_cache_size_set) { | |
667 | if (l2_cache_size_set && refcount_cache_size_set) { | |
668 | error_setg(errp, QCOW2_OPT_CACHE_SIZE ", " QCOW2_OPT_L2_CACHE_SIZE | |
669 | " and " QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not be set " | |
670 | "the same time"); | |
671 | return; | |
672 | } else if (*l2_cache_size > combined_cache_size) { | |
673 | error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed " | |
674 | QCOW2_OPT_CACHE_SIZE); | |
675 | return; | |
676 | } else if (*refcount_cache_size > combined_cache_size) { | |
677 | error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed " | |
678 | QCOW2_OPT_CACHE_SIZE); | |
679 | return; | |
680 | } | |
681 | ||
682 | if (l2_cache_size_set) { | |
683 | *refcount_cache_size = combined_cache_size - *l2_cache_size; | |
684 | } else if (refcount_cache_size_set) { | |
685 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
686 | } else { | |
687 | *refcount_cache_size = combined_cache_size | |
688 | / (DEFAULT_L2_REFCOUNT_SIZE_RATIO + 1); | |
689 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
690 | } | |
691 | } else { | |
692 | if (!l2_cache_size_set && !refcount_cache_size_set) { | |
bc85ef26 HR |
693 | *l2_cache_size = MAX(DEFAULT_L2_CACHE_BYTE_SIZE, |
694 | (uint64_t)DEFAULT_L2_CACHE_CLUSTERS | |
695 | * s->cluster_size); | |
6c1c8d5d HR |
696 | *refcount_cache_size = *l2_cache_size |
697 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
698 | } else if (!l2_cache_size_set) { | |
699 | *l2_cache_size = *refcount_cache_size | |
700 | * DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
701 | } else if (!refcount_cache_size_set) { | |
702 | *refcount_cache_size = *l2_cache_size | |
703 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
704 | } | |
705 | } | |
706 | } | |
707 | ||
ee55b173 KW |
708 | typedef struct Qcow2ReopenState { |
709 | Qcow2Cache *l2_table_cache; | |
710 | Qcow2Cache *refcount_block_cache; | |
711 | bool use_lazy_refcounts; | |
712 | int overlap_check; | |
713 | bool discard_passthrough[QCOW2_DISCARD_MAX]; | |
714 | uint64_t cache_clean_interval; | |
b25b387f | 715 | QCryptoBlockOpenOptions *crypto_opts; /* Disk encryption runtime options */ |
ee55b173 KW |
716 | } Qcow2ReopenState; |
717 | ||
718 | static int qcow2_update_options_prepare(BlockDriverState *bs, | |
719 | Qcow2ReopenState *r, | |
720 | QDict *options, int flags, | |
721 | Error **errp) | |
4c75d1a1 KW |
722 | { |
723 | BDRVQcow2State *s = bs->opaque; | |
94edf3fb | 724 | QemuOpts *opts = NULL; |
4c75d1a1 KW |
725 | const char *opt_overlap_check, *opt_overlap_check_template; |
726 | int overlap_check_template = 0; | |
94edf3fb | 727 | uint64_t l2_cache_size, refcount_cache_size; |
4c75d1a1 | 728 | int i; |
b25b387f DB |
729 | const char *encryptfmt; |
730 | QDict *encryptopts = NULL; | |
94edf3fb | 731 | Error *local_err = NULL; |
4c75d1a1 KW |
732 | int ret; |
733 | ||
b25b387f DB |
734 | qdict_extract_subqdict(options, &encryptopts, "encrypt."); |
735 | encryptfmt = qdict_get_try_str(encryptopts, "format"); | |
736 | ||
94edf3fb KW |
737 | opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort); |
738 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
739 | if (local_err) { | |
740 | error_propagate(errp, local_err); | |
741 | ret = -EINVAL; | |
742 | goto fail; | |
743 | } | |
744 | ||
745 | /* get L2 table/refcount block cache size from command line options */ | |
746 | read_cache_sizes(bs, opts, &l2_cache_size, &refcount_cache_size, | |
747 | &local_err); | |
748 | if (local_err) { | |
749 | error_propagate(errp, local_err); | |
750 | ret = -EINVAL; | |
751 | goto fail; | |
752 | } | |
753 | ||
754 | l2_cache_size /= s->cluster_size; | |
755 | if (l2_cache_size < MIN_L2_CACHE_SIZE) { | |
756 | l2_cache_size = MIN_L2_CACHE_SIZE; | |
757 | } | |
758 | if (l2_cache_size > INT_MAX) { | |
759 | error_setg(errp, "L2 cache size too big"); | |
760 | ret = -EINVAL; | |
761 | goto fail; | |
762 | } | |
763 | ||
764 | refcount_cache_size /= s->cluster_size; | |
765 | if (refcount_cache_size < MIN_REFCOUNT_CACHE_SIZE) { | |
766 | refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE; | |
767 | } | |
768 | if (refcount_cache_size > INT_MAX) { | |
769 | error_setg(errp, "Refcount cache size too big"); | |
770 | ret = -EINVAL; | |
771 | goto fail; | |
772 | } | |
773 | ||
5b0959a7 KW |
774 | /* alloc new L2 table/refcount block cache, flush old one */ |
775 | if (s->l2_table_cache) { | |
776 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
777 | if (ret) { | |
778 | error_setg_errno(errp, -ret, "Failed to flush the L2 table cache"); | |
779 | goto fail; | |
780 | } | |
781 | } | |
782 | ||
783 | if (s->refcount_block_cache) { | |
784 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
785 | if (ret) { | |
786 | error_setg_errno(errp, -ret, | |
787 | "Failed to flush the refcount block cache"); | |
788 | goto fail; | |
789 | } | |
790 | } | |
791 | ||
ee55b173 KW |
792 | r->l2_table_cache = qcow2_cache_create(bs, l2_cache_size); |
793 | r->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size); | |
794 | if (r->l2_table_cache == NULL || r->refcount_block_cache == NULL) { | |
94edf3fb KW |
795 | error_setg(errp, "Could not allocate metadata caches"); |
796 | ret = -ENOMEM; | |
797 | goto fail; | |
798 | } | |
799 | ||
800 | /* New interval for cache cleanup timer */ | |
ee55b173 | 801 | r->cache_clean_interval = |
5b0959a7 KW |
802 | qemu_opt_get_number(opts, QCOW2_OPT_CACHE_CLEAN_INTERVAL, |
803 | s->cache_clean_interval); | |
91203f08 AG |
804 | #ifndef CONFIG_LINUX |
805 | if (r->cache_clean_interval != 0) { | |
806 | error_setg(errp, QCOW2_OPT_CACHE_CLEAN_INTERVAL | |
807 | " not supported on this host"); | |
808 | ret = -EINVAL; | |
809 | goto fail; | |
810 | } | |
811 | #endif | |
ee55b173 | 812 | if (r->cache_clean_interval > UINT_MAX) { |
94edf3fb KW |
813 | error_setg(errp, "Cache clean interval too big"); |
814 | ret = -EINVAL; | |
815 | goto fail; | |
816 | } | |
94edf3fb | 817 | |
5b0959a7 | 818 | /* lazy-refcounts; flush if going from enabled to disabled */ |
ee55b173 | 819 | r->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, |
4c75d1a1 | 820 | (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); |
ee55b173 | 821 | if (r->use_lazy_refcounts && s->qcow_version < 3) { |
007dbc39 KW |
822 | error_setg(errp, "Lazy refcounts require a qcow2 image with at least " |
823 | "qemu 1.1 compatibility level"); | |
824 | ret = -EINVAL; | |
825 | goto fail; | |
826 | } | |
4c75d1a1 | 827 | |
5b0959a7 KW |
828 | if (s->use_lazy_refcounts && !r->use_lazy_refcounts) { |
829 | ret = qcow2_mark_clean(bs); | |
830 | if (ret < 0) { | |
831 | error_setg_errno(errp, -ret, "Failed to disable lazy refcounts"); | |
832 | goto fail; | |
833 | } | |
834 | } | |
835 | ||
007dbc39 | 836 | /* Overlap check options */ |
4c75d1a1 KW |
837 | opt_overlap_check = qemu_opt_get(opts, QCOW2_OPT_OVERLAP); |
838 | opt_overlap_check_template = qemu_opt_get(opts, QCOW2_OPT_OVERLAP_TEMPLATE); | |
839 | if (opt_overlap_check_template && opt_overlap_check && | |
840 | strcmp(opt_overlap_check_template, opt_overlap_check)) | |
841 | { | |
842 | error_setg(errp, "Conflicting values for qcow2 options '" | |
843 | QCOW2_OPT_OVERLAP "' ('%s') and '" QCOW2_OPT_OVERLAP_TEMPLATE | |
844 | "' ('%s')", opt_overlap_check, opt_overlap_check_template); | |
845 | ret = -EINVAL; | |
846 | goto fail; | |
847 | } | |
848 | if (!opt_overlap_check) { | |
849 | opt_overlap_check = opt_overlap_check_template ?: "cached"; | |
850 | } | |
851 | ||
852 | if (!strcmp(opt_overlap_check, "none")) { | |
853 | overlap_check_template = 0; | |
854 | } else if (!strcmp(opt_overlap_check, "constant")) { | |
855 | overlap_check_template = QCOW2_OL_CONSTANT; | |
856 | } else if (!strcmp(opt_overlap_check, "cached")) { | |
857 | overlap_check_template = QCOW2_OL_CACHED; | |
858 | } else if (!strcmp(opt_overlap_check, "all")) { | |
859 | overlap_check_template = QCOW2_OL_ALL; | |
860 | } else { | |
861 | error_setg(errp, "Unsupported value '%s' for qcow2 option " | |
862 | "'overlap-check'. Allowed are any of the following: " | |
863 | "none, constant, cached, all", opt_overlap_check); | |
864 | ret = -EINVAL; | |
865 | goto fail; | |
866 | } | |
867 | ||
ee55b173 | 868 | r->overlap_check = 0; |
4c75d1a1 KW |
869 | for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) { |
870 | /* overlap-check defines a template bitmask, but every flag may be | |
871 | * overwritten through the associated boolean option */ | |
ee55b173 | 872 | r->overlap_check |= |
4c75d1a1 KW |
873 | qemu_opt_get_bool(opts, overlap_bool_option_names[i], |
874 | overlap_check_template & (1 << i)) << i; | |
875 | } | |
876 | ||
ee55b173 KW |
877 | r->discard_passthrough[QCOW2_DISCARD_NEVER] = false; |
878 | r->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true; | |
879 | r->discard_passthrough[QCOW2_DISCARD_REQUEST] = | |
007dbc39 KW |
880 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST, |
881 | flags & BDRV_O_UNMAP); | |
ee55b173 | 882 | r->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] = |
007dbc39 | 883 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true); |
ee55b173 | 884 | r->discard_passthrough[QCOW2_DISCARD_OTHER] = |
007dbc39 KW |
885 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false); |
886 | ||
b25b387f DB |
887 | switch (s->crypt_method_header) { |
888 | case QCOW_CRYPT_NONE: | |
889 | if (encryptfmt) { | |
890 | error_setg(errp, "No encryption in image header, but options " | |
891 | "specified format '%s'", encryptfmt); | |
892 | ret = -EINVAL; | |
893 | goto fail; | |
894 | } | |
895 | break; | |
896 | ||
897 | case QCOW_CRYPT_AES: | |
898 | if (encryptfmt && !g_str_equal(encryptfmt, "aes")) { | |
899 | error_setg(errp, | |
900 | "Header reported 'aes' encryption format but " | |
901 | "options specify '%s'", encryptfmt); | |
902 | ret = -EINVAL; | |
903 | goto fail; | |
904 | } | |
905 | qdict_del(encryptopts, "format"); | |
906 | r->crypto_opts = block_crypto_open_opts_init( | |
907 | Q_CRYPTO_BLOCK_FORMAT_QCOW, encryptopts, errp); | |
908 | break; | |
909 | ||
4652b8f3 DB |
910 | case QCOW_CRYPT_LUKS: |
911 | if (encryptfmt && !g_str_equal(encryptfmt, "luks")) { | |
912 | error_setg(errp, | |
913 | "Header reported 'luks' encryption format but " | |
914 | "options specify '%s'", encryptfmt); | |
915 | ret = -EINVAL; | |
916 | goto fail; | |
917 | } | |
918 | qdict_del(encryptopts, "format"); | |
919 | r->crypto_opts = block_crypto_open_opts_init( | |
920 | Q_CRYPTO_BLOCK_FORMAT_LUKS, encryptopts, errp); | |
921 | break; | |
922 | ||
b25b387f DB |
923 | default: |
924 | error_setg(errp, "Unsupported encryption method %d", | |
925 | s->crypt_method_header); | |
926 | break; | |
927 | } | |
928 | if (s->crypt_method_header != QCOW_CRYPT_NONE && !r->crypto_opts) { | |
929 | ret = -EINVAL; | |
930 | goto fail; | |
931 | } | |
932 | ||
4c75d1a1 KW |
933 | ret = 0; |
934 | fail: | |
b25b387f | 935 | QDECREF(encryptopts); |
94edf3fb KW |
936 | qemu_opts_del(opts); |
937 | opts = NULL; | |
ee55b173 KW |
938 | return ret; |
939 | } | |
940 | ||
941 | static void qcow2_update_options_commit(BlockDriverState *bs, | |
942 | Qcow2ReopenState *r) | |
943 | { | |
944 | BDRVQcow2State *s = bs->opaque; | |
945 | int i; | |
946 | ||
5b0959a7 KW |
947 | if (s->l2_table_cache) { |
948 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
949 | } | |
950 | if (s->refcount_block_cache) { | |
951 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
952 | } | |
ee55b173 KW |
953 | s->l2_table_cache = r->l2_table_cache; |
954 | s->refcount_block_cache = r->refcount_block_cache; | |
955 | ||
956 | s->overlap_check = r->overlap_check; | |
957 | s->use_lazy_refcounts = r->use_lazy_refcounts; | |
958 | ||
959 | for (i = 0; i < QCOW2_DISCARD_MAX; i++) { | |
960 | s->discard_passthrough[i] = r->discard_passthrough[i]; | |
961 | } | |
962 | ||
5b0959a7 KW |
963 | if (s->cache_clean_interval != r->cache_clean_interval) { |
964 | cache_clean_timer_del(bs); | |
965 | s->cache_clean_interval = r->cache_clean_interval; | |
966 | cache_clean_timer_init(bs, bdrv_get_aio_context(bs)); | |
967 | } | |
b25b387f DB |
968 | |
969 | qapi_free_QCryptoBlockOpenOptions(s->crypto_opts); | |
970 | s->crypto_opts = r->crypto_opts; | |
ee55b173 KW |
971 | } |
972 | ||
973 | static void qcow2_update_options_abort(BlockDriverState *bs, | |
974 | Qcow2ReopenState *r) | |
975 | { | |
976 | if (r->l2_table_cache) { | |
977 | qcow2_cache_destroy(bs, r->l2_table_cache); | |
978 | } | |
979 | if (r->refcount_block_cache) { | |
980 | qcow2_cache_destroy(bs, r->refcount_block_cache); | |
981 | } | |
b25b387f | 982 | qapi_free_QCryptoBlockOpenOptions(r->crypto_opts); |
ee55b173 KW |
983 | } |
984 | ||
985 | static int qcow2_update_options(BlockDriverState *bs, QDict *options, | |
986 | int flags, Error **errp) | |
987 | { | |
988 | Qcow2ReopenState r = {}; | |
989 | int ret; | |
990 | ||
991 | ret = qcow2_update_options_prepare(bs, &r, options, flags, errp); | |
992 | if (ret >= 0) { | |
993 | qcow2_update_options_commit(bs, &r); | |
994 | } else { | |
995 | qcow2_update_options_abort(bs, &r); | |
996 | } | |
94edf3fb | 997 | |
4c75d1a1 KW |
998 | return ret; |
999 | } | |
1000 | ||
4e4bf5c4 KW |
1001 | static int qcow2_do_open(BlockDriverState *bs, QDict *options, int flags, |
1002 | Error **errp) | |
585f8587 | 1003 | { |
ff99129a | 1004 | BDRVQcow2State *s = bs->opaque; |
6d33e8e7 KW |
1005 | unsigned int len, i; |
1006 | int ret = 0; | |
585f8587 | 1007 | QCowHeader header; |
74c4510a | 1008 | Error *local_err = NULL; |
9b80ddf3 | 1009 | uint64_t ext_end; |
2cf7cfa1 | 1010 | uint64_t l1_vm_state_index; |
585f8587 | 1011 | |
cf2ab8fc | 1012 | ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); |
6d85a57e | 1013 | if (ret < 0) { |
3ef6c40a | 1014 | error_setg_errno(errp, -ret, "Could not read qcow2 header"); |
585f8587 | 1015 | goto fail; |
6d85a57e | 1016 | } |
585f8587 FB |
1017 | be32_to_cpus(&header.magic); |
1018 | be32_to_cpus(&header.version); | |
1019 | be64_to_cpus(&header.backing_file_offset); | |
1020 | be32_to_cpus(&header.backing_file_size); | |
1021 | be64_to_cpus(&header.size); | |
1022 | be32_to_cpus(&header.cluster_bits); | |
1023 | be32_to_cpus(&header.crypt_method); | |
1024 | be64_to_cpus(&header.l1_table_offset); | |
1025 | be32_to_cpus(&header.l1_size); | |
1026 | be64_to_cpus(&header.refcount_table_offset); | |
1027 | be32_to_cpus(&header.refcount_table_clusters); | |
1028 | be64_to_cpus(&header.snapshots_offset); | |
1029 | be32_to_cpus(&header.nb_snapshots); | |
3b46e624 | 1030 | |
e8cdcec1 | 1031 | if (header.magic != QCOW_MAGIC) { |
3ef6c40a | 1032 | error_setg(errp, "Image is not in qcow2 format"); |
76abe407 | 1033 | ret = -EINVAL; |
585f8587 | 1034 | goto fail; |
6d85a57e | 1035 | } |
6744cbab | 1036 | if (header.version < 2 || header.version > 3) { |
a55448b3 | 1037 | error_setg(errp, "Unsupported qcow2 version %" PRIu32, header.version); |
6744cbab KW |
1038 | ret = -ENOTSUP; |
1039 | goto fail; | |
1040 | } | |
1041 | ||
1042 | s->qcow_version = header.version; | |
1043 | ||
24342f2c KW |
1044 | /* Initialise cluster size */ |
1045 | if (header.cluster_bits < MIN_CLUSTER_BITS || | |
1046 | header.cluster_bits > MAX_CLUSTER_BITS) { | |
521b2b5d HR |
1047 | error_setg(errp, "Unsupported cluster size: 2^%" PRIu32, |
1048 | header.cluster_bits); | |
24342f2c KW |
1049 | ret = -EINVAL; |
1050 | goto fail; | |
1051 | } | |
1052 | ||
1053 | s->cluster_bits = header.cluster_bits; | |
1054 | s->cluster_size = 1 << s->cluster_bits; | |
1055 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
1056 | ||
6744cbab KW |
1057 | /* Initialise version 3 header fields */ |
1058 | if (header.version == 2) { | |
1059 | header.incompatible_features = 0; | |
1060 | header.compatible_features = 0; | |
1061 | header.autoclear_features = 0; | |
1062 | header.refcount_order = 4; | |
1063 | header.header_length = 72; | |
1064 | } else { | |
1065 | be64_to_cpus(&header.incompatible_features); | |
1066 | be64_to_cpus(&header.compatible_features); | |
1067 | be64_to_cpus(&header.autoclear_features); | |
1068 | be32_to_cpus(&header.refcount_order); | |
1069 | be32_to_cpus(&header.header_length); | |
24342f2c KW |
1070 | |
1071 | if (header.header_length < 104) { | |
1072 | error_setg(errp, "qcow2 header too short"); | |
1073 | ret = -EINVAL; | |
1074 | goto fail; | |
1075 | } | |
1076 | } | |
1077 | ||
1078 | if (header.header_length > s->cluster_size) { | |
1079 | error_setg(errp, "qcow2 header exceeds cluster size"); | |
1080 | ret = -EINVAL; | |
1081 | goto fail; | |
6744cbab KW |
1082 | } |
1083 | ||
1084 | if (header.header_length > sizeof(header)) { | |
1085 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
1086 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
cf2ab8fc | 1087 | ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields, |
6744cbab KW |
1088 | s->unknown_header_fields_size); |
1089 | if (ret < 0) { | |
3ef6c40a HR |
1090 | error_setg_errno(errp, -ret, "Could not read unknown qcow2 header " |
1091 | "fields"); | |
6744cbab KW |
1092 | goto fail; |
1093 | } | |
1094 | } | |
1095 | ||
a1b3955c KW |
1096 | if (header.backing_file_offset > s->cluster_size) { |
1097 | error_setg(errp, "Invalid backing file offset"); | |
1098 | ret = -EINVAL; | |
1099 | goto fail; | |
1100 | } | |
1101 | ||
cfcc4c62 KW |
1102 | if (header.backing_file_offset) { |
1103 | ext_end = header.backing_file_offset; | |
1104 | } else { | |
1105 | ext_end = 1 << header.cluster_bits; | |
1106 | } | |
1107 | ||
6744cbab KW |
1108 | /* Handle feature bits */ |
1109 | s->incompatible_features = header.incompatible_features; | |
1110 | s->compatible_features = header.compatible_features; | |
1111 | s->autoclear_features = header.autoclear_features; | |
1112 | ||
c61d0004 | 1113 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { |
cfcc4c62 KW |
1114 | void *feature_table = NULL; |
1115 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
4652b8f3 | 1116 | &feature_table, flags, NULL); |
a55448b3 | 1117 | report_unsupported_feature(errp, feature_table, |
c61d0004 SH |
1118 | s->incompatible_features & |
1119 | ~QCOW2_INCOMPAT_MASK); | |
6744cbab | 1120 | ret = -ENOTSUP; |
c5a33ee9 | 1121 | g_free(feature_table); |
6744cbab KW |
1122 | goto fail; |
1123 | } | |
1124 | ||
69c98726 HR |
1125 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { |
1126 | /* Corrupt images may not be written to unless they are being repaired | |
1127 | */ | |
1128 | if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) { | |
3ef6c40a HR |
1129 | error_setg(errp, "qcow2: Image is corrupt; cannot be opened " |
1130 | "read/write"); | |
69c98726 HR |
1131 | ret = -EACCES; |
1132 | goto fail; | |
1133 | } | |
1134 | } | |
1135 | ||
6744cbab | 1136 | /* Check support for various header values */ |
b72faf9f HR |
1137 | if (header.refcount_order > 6) { |
1138 | error_setg(errp, "Reference count entry width too large; may not " | |
1139 | "exceed 64 bits"); | |
1140 | ret = -EINVAL; | |
e8cdcec1 KW |
1141 | goto fail; |
1142 | } | |
b6481f37 | 1143 | s->refcount_order = header.refcount_order; |
346a53df HR |
1144 | s->refcount_bits = 1 << s->refcount_order; |
1145 | s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1); | |
1146 | s->refcount_max += s->refcount_max - 1; | |
6744cbab | 1147 | |
585f8587 | 1148 | s->crypt_method_header = header.crypt_method; |
6d85a57e | 1149 | if (s->crypt_method_header) { |
e6ff69bf DB |
1150 | if (bdrv_uses_whitelist() && |
1151 | s->crypt_method_header == QCOW_CRYPT_AES) { | |
8c0dcbc4 DB |
1152 | error_setg(errp, |
1153 | "Use of AES-CBC encrypted qcow2 images is no longer " | |
1154 | "supported in system emulators"); | |
1155 | error_append_hint(errp, | |
1156 | "You can use 'qemu-img convert' to convert your " | |
1157 | "image to an alternative supported format, such " | |
1158 | "as unencrypted qcow2, or raw with the LUKS " | |
1159 | "format instead.\n"); | |
1160 | ret = -ENOSYS; | |
1161 | goto fail; | |
e6ff69bf DB |
1162 | } |
1163 | ||
4652b8f3 DB |
1164 | if (s->crypt_method_header == QCOW_CRYPT_AES) { |
1165 | s->crypt_physical_offset = false; | |
1166 | } else { | |
1167 | /* Assuming LUKS and any future crypt methods we | |
1168 | * add will all use physical offsets, due to the | |
1169 | * fact that the alternative is insecure... */ | |
1170 | s->crypt_physical_offset = true; | |
1171 | } | |
1172 | ||
54115412 | 1173 | bs->encrypted = true; |
6d85a57e | 1174 | } |
24342f2c | 1175 | |
585f8587 FB |
1176 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
1177 | s->l2_size = 1 << s->l2_bits; | |
1d13d654 HR |
1178 | /* 2^(s->refcount_order - 3) is the refcount width in bytes */ |
1179 | s->refcount_block_bits = s->cluster_bits - (s->refcount_order - 3); | |
1180 | s->refcount_block_size = 1 << s->refcount_block_bits; | |
585f8587 FB |
1181 | bs->total_sectors = header.size / 512; |
1182 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
1183 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
1184 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
5dab2fad | 1185 | |
585f8587 | 1186 | s->refcount_table_offset = header.refcount_table_offset; |
5fafdf24 | 1187 | s->refcount_table_size = |
585f8587 FB |
1188 | header.refcount_table_clusters << (s->cluster_bits - 3); |
1189 | ||
2b5d5953 | 1190 | if (header.refcount_table_clusters > qcow2_max_refcount_clusters(s)) { |
5dab2fad KW |
1191 | error_setg(errp, "Reference count table too large"); |
1192 | ret = -EINVAL; | |
1193 | goto fail; | |
1194 | } | |
1195 | ||
8c7de283 KW |
1196 | ret = validate_table_offset(bs, s->refcount_table_offset, |
1197 | s->refcount_table_size, sizeof(uint64_t)); | |
1198 | if (ret < 0) { | |
1199 | error_setg(errp, "Invalid reference count table offset"); | |
1200 | goto fail; | |
1201 | } | |
1202 | ||
ce48f2f4 KW |
1203 | /* Snapshot table offset/length */ |
1204 | if (header.nb_snapshots > QCOW_MAX_SNAPSHOTS) { | |
1205 | error_setg(errp, "Too many snapshots"); | |
1206 | ret = -EINVAL; | |
1207 | goto fail; | |
1208 | } | |
1209 | ||
1210 | ret = validate_table_offset(bs, header.snapshots_offset, | |
1211 | header.nb_snapshots, | |
1212 | sizeof(QCowSnapshotHeader)); | |
1213 | if (ret < 0) { | |
1214 | error_setg(errp, "Invalid snapshot table offset"); | |
1215 | goto fail; | |
1216 | } | |
1217 | ||
585f8587 | 1218 | /* read the level 1 table */ |
87b86e7e | 1219 | if (header.l1_size > QCOW_MAX_L1_SIZE / sizeof(uint64_t)) { |
2d51c32c KW |
1220 | error_setg(errp, "Active L1 table too large"); |
1221 | ret = -EFBIG; | |
1222 | goto fail; | |
1223 | } | |
585f8587 | 1224 | s->l1_size = header.l1_size; |
2cf7cfa1 KW |
1225 | |
1226 | l1_vm_state_index = size_to_l1(s, header.size); | |
1227 | if (l1_vm_state_index > INT_MAX) { | |
3ef6c40a | 1228 | error_setg(errp, "Image is too big"); |
2cf7cfa1 KW |
1229 | ret = -EFBIG; |
1230 | goto fail; | |
1231 | } | |
1232 | s->l1_vm_state_index = l1_vm_state_index; | |
1233 | ||
585f8587 FB |
1234 | /* the L1 table must contain at least enough entries to put |
1235 | header.size bytes */ | |
6d85a57e | 1236 | if (s->l1_size < s->l1_vm_state_index) { |
3ef6c40a | 1237 | error_setg(errp, "L1 table is too small"); |
6d85a57e | 1238 | ret = -EINVAL; |
585f8587 | 1239 | goto fail; |
6d85a57e | 1240 | } |
2d51c32c KW |
1241 | |
1242 | ret = validate_table_offset(bs, header.l1_table_offset, | |
1243 | header.l1_size, sizeof(uint64_t)); | |
1244 | if (ret < 0) { | |
1245 | error_setg(errp, "Invalid L1 table offset"); | |
1246 | goto fail; | |
1247 | } | |
585f8587 | 1248 | s->l1_table_offset = header.l1_table_offset; |
2d51c32c KW |
1249 | |
1250 | ||
d191d12d | 1251 | if (s->l1_size > 0) { |
9a4f4c31 | 1252 | s->l1_table = qemu_try_blockalign(bs->file->bs, |
d191d12d | 1253 | align_offset(s->l1_size * sizeof(uint64_t), 512)); |
de82815d KW |
1254 | if (s->l1_table == NULL) { |
1255 | error_setg(errp, "Could not allocate L1 table"); | |
1256 | ret = -ENOMEM; | |
1257 | goto fail; | |
1258 | } | |
cf2ab8fc | 1259 | ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, |
6d85a57e JS |
1260 | s->l1_size * sizeof(uint64_t)); |
1261 | if (ret < 0) { | |
3ef6c40a | 1262 | error_setg_errno(errp, -ret, "Could not read L1 table"); |
d191d12d | 1263 | goto fail; |
6d85a57e | 1264 | } |
d191d12d SW |
1265 | for(i = 0;i < s->l1_size; i++) { |
1266 | be64_to_cpus(&s->l1_table[i]); | |
1267 | } | |
585f8587 | 1268 | } |
29c1a730 | 1269 | |
94edf3fb KW |
1270 | /* Parse driver-specific options */ |
1271 | ret = qcow2_update_options(bs, options, flags, errp); | |
90efa0ea KW |
1272 | if (ret < 0) { |
1273 | goto fail; | |
1274 | } | |
1275 | ||
7267c094 | 1276 | s->cluster_cache = g_malloc(s->cluster_size); |
585f8587 | 1277 | /* one more sector for decompressed data alignment */ |
9a4f4c31 | 1278 | s->cluster_data = qemu_try_blockalign(bs->file->bs, QCOW_MAX_CRYPT_CLUSTERS |
de82815d KW |
1279 | * s->cluster_size + 512); |
1280 | if (s->cluster_data == NULL) { | |
1281 | error_setg(errp, "Could not allocate temporary cluster buffer"); | |
1282 | ret = -ENOMEM; | |
1283 | goto fail; | |
1284 | } | |
1285 | ||
585f8587 | 1286 | s->cluster_cache_offset = -1; |
06d9260f | 1287 | s->flags = flags; |
3b46e624 | 1288 | |
6d85a57e JS |
1289 | ret = qcow2_refcount_init(bs); |
1290 | if (ret != 0) { | |
3ef6c40a | 1291 | error_setg_errno(errp, -ret, "Could not initialize refcount handling"); |
585f8587 | 1292 | goto fail; |
6d85a57e | 1293 | } |
585f8587 | 1294 | |
72cf2d4f | 1295 | QLIST_INIT(&s->cluster_allocs); |
0b919fae | 1296 | QTAILQ_INIT(&s->discards); |
f214978a | 1297 | |
9b80ddf3 | 1298 | /* read qcow2 extensions */ |
3ef6c40a | 1299 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL, |
4652b8f3 | 1300 | flags, &local_err)) { |
3ef6c40a | 1301 | error_propagate(errp, local_err); |
6d85a57e | 1302 | ret = -EINVAL; |
9b80ddf3 | 1303 | goto fail; |
6d85a57e | 1304 | } |
9b80ddf3 | 1305 | |
4652b8f3 DB |
1306 | /* qcow2_read_extension may have set up the crypto context |
1307 | * if the crypt method needs a header region, some methods | |
1308 | * don't need header extensions, so must check here | |
1309 | */ | |
1310 | if (s->crypt_method_header && !s->crypto) { | |
1311 | if (s->crypt_method_header == QCOW_CRYPT_AES) { | |
1312 | unsigned int cflags = 0; | |
1313 | if (flags & BDRV_O_NO_IO) { | |
1314 | cflags |= QCRYPTO_BLOCK_OPEN_NO_IO; | |
1315 | } | |
1316 | s->crypto = qcrypto_block_open(s->crypto_opts, NULL, NULL, | |
1317 | cflags, errp); | |
1318 | if (!s->crypto) { | |
1319 | ret = -EINVAL; | |
1320 | goto fail; | |
1321 | } | |
1322 | } else if (!(flags & BDRV_O_NO_IO)) { | |
1323 | error_setg(errp, "Missing CRYPTO header for crypt method %d", | |
1324 | s->crypt_method_header); | |
b25b387f DB |
1325 | ret = -EINVAL; |
1326 | goto fail; | |
1327 | } | |
1328 | } | |
1329 | ||
585f8587 FB |
1330 | /* read the backing file name */ |
1331 | if (header.backing_file_offset != 0) { | |
1332 | len = header.backing_file_size; | |
9a29e18f | 1333 | if (len > MIN(1023, s->cluster_size - header.backing_file_offset) || |
e729fa6a | 1334 | len >= sizeof(bs->backing_file)) { |
6d33e8e7 KW |
1335 | error_setg(errp, "Backing file name too long"); |
1336 | ret = -EINVAL; | |
1337 | goto fail; | |
6d85a57e | 1338 | } |
cf2ab8fc | 1339 | ret = bdrv_pread(bs->file, header.backing_file_offset, |
6d85a57e JS |
1340 | bs->backing_file, len); |
1341 | if (ret < 0) { | |
3ef6c40a | 1342 | error_setg_errno(errp, -ret, "Could not read backing file name"); |
585f8587 | 1343 | goto fail; |
6d85a57e | 1344 | } |
585f8587 | 1345 | bs->backing_file[len] = '\0'; |
e4603fe1 | 1346 | s->image_backing_file = g_strdup(bs->backing_file); |
585f8587 | 1347 | } |
42deb29f | 1348 | |
11b128f4 KW |
1349 | /* Internal snapshots */ |
1350 | s->snapshots_offset = header.snapshots_offset; | |
1351 | s->nb_snapshots = header.nb_snapshots; | |
1352 | ||
42deb29f KW |
1353 | ret = qcow2_read_snapshots(bs); |
1354 | if (ret < 0) { | |
3ef6c40a | 1355 | error_setg_errno(errp, -ret, "Could not read snapshots"); |
585f8587 | 1356 | goto fail; |
6d85a57e | 1357 | } |
585f8587 | 1358 | |
af7b708d | 1359 | /* Clear unknown autoclear feature bits */ |
04c01a5c | 1360 | if (!bs->read_only && !(flags & BDRV_O_INACTIVE) && s->autoclear_features) { |
af7b708d SH |
1361 | s->autoclear_features = 0; |
1362 | ret = qcow2_update_header(bs); | |
1363 | if (ret < 0) { | |
3ef6c40a | 1364 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); |
af7b708d SH |
1365 | goto fail; |
1366 | } | |
1367 | } | |
1368 | ||
68d100e9 KW |
1369 | /* Initialise locks */ |
1370 | qemu_co_mutex_init(&s->lock); | |
170f4b2e | 1371 | bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP; |
68d100e9 | 1372 | |
c61d0004 | 1373 | /* Repair image if dirty */ |
04c01a5c | 1374 | if (!(flags & (BDRV_O_CHECK | BDRV_O_INACTIVE)) && !bs->read_only && |
058f8f16 | 1375 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { |
c61d0004 SH |
1376 | BdrvCheckResult result = {0}; |
1377 | ||
5b84106b | 1378 | ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS); |
c61d0004 | 1379 | if (ret < 0) { |
3ef6c40a | 1380 | error_setg_errno(errp, -ret, "Could not repair dirty image"); |
c61d0004 SH |
1381 | goto fail; |
1382 | } | |
1383 | } | |
1384 | ||
585f8587 | 1385 | #ifdef DEBUG_ALLOC |
6cbc3031 PH |
1386 | { |
1387 | BdrvCheckResult result = {0}; | |
b35278f7 | 1388 | qcow2_check_refcounts(bs, &result, 0); |
6cbc3031 | 1389 | } |
585f8587 | 1390 | #endif |
6d85a57e | 1391 | return ret; |
585f8587 FB |
1392 | |
1393 | fail: | |
6744cbab | 1394 | g_free(s->unknown_header_fields); |
75bab85c | 1395 | cleanup_unknown_header_ext(bs); |
ed6ccf0f KW |
1396 | qcow2_free_snapshots(bs); |
1397 | qcow2_refcount_close(bs); | |
de82815d | 1398 | qemu_vfree(s->l1_table); |
cf93980e HR |
1399 | /* else pre-write overlap checks in cache_destroy may crash */ |
1400 | s->l1_table = NULL; | |
279621c0 | 1401 | cache_clean_timer_del(bs); |
29c1a730 KW |
1402 | if (s->l2_table_cache) { |
1403 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
1404 | } | |
c5a33ee9 PJ |
1405 | if (s->refcount_block_cache) { |
1406 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1407 | } | |
7267c094 | 1408 | g_free(s->cluster_cache); |
dea43a65 | 1409 | qemu_vfree(s->cluster_data); |
b25b387f DB |
1410 | qcrypto_block_free(s->crypto); |
1411 | qapi_free_QCryptoBlockOpenOptions(s->crypto_opts); | |
6d85a57e | 1412 | return ret; |
585f8587 FB |
1413 | } |
1414 | ||
4e4bf5c4 KW |
1415 | static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, |
1416 | Error **errp) | |
1417 | { | |
1418 | bs->file = bdrv_open_child(NULL, options, "file", bs, &child_file, | |
1419 | false, errp); | |
1420 | if (!bs->file) { | |
1421 | return -EINVAL; | |
1422 | } | |
1423 | ||
1424 | return qcow2_do_open(bs, options, flags, errp); | |
1425 | } | |
1426 | ||
3baca891 | 1427 | static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd | 1428 | { |
ff99129a | 1429 | BDRVQcow2State *s = bs->opaque; |
d34682cd | 1430 | |
a84178cc EB |
1431 | if (bs->encrypted) { |
1432 | /* Encryption works on a sector granularity */ | |
a5b8dd2c | 1433 | bs->bl.request_alignment = BDRV_SECTOR_SIZE; |
a84178cc | 1434 | } |
cf081fca | 1435 | bs->bl.pwrite_zeroes_alignment = s->cluster_size; |
ecdbead6 | 1436 | bs->bl.pdiscard_alignment = s->cluster_size; |
d34682cd KW |
1437 | } |
1438 | ||
21d82ac9 JC |
1439 | static int qcow2_reopen_prepare(BDRVReopenState *state, |
1440 | BlockReopenQueue *queue, Error **errp) | |
1441 | { | |
5b0959a7 | 1442 | Qcow2ReopenState *r; |
4c2e5f8f KW |
1443 | int ret; |
1444 | ||
5b0959a7 KW |
1445 | r = g_new0(Qcow2ReopenState, 1); |
1446 | state->opaque = r; | |
1447 | ||
1448 | ret = qcow2_update_options_prepare(state->bs, r, state->options, | |
1449 | state->flags, errp); | |
1450 | if (ret < 0) { | |
1451 | goto fail; | |
1452 | } | |
1453 | ||
1454 | /* We need to write out any unwritten data if we reopen read-only. */ | |
4c2e5f8f KW |
1455 | if ((state->flags & BDRV_O_RDWR) == 0) { |
1456 | ret = bdrv_flush(state->bs); | |
1457 | if (ret < 0) { | |
5b0959a7 | 1458 | goto fail; |
4c2e5f8f KW |
1459 | } |
1460 | ||
1461 | ret = qcow2_mark_clean(state->bs); | |
1462 | if (ret < 0) { | |
5b0959a7 | 1463 | goto fail; |
4c2e5f8f KW |
1464 | } |
1465 | } | |
1466 | ||
21d82ac9 | 1467 | return 0; |
5b0959a7 KW |
1468 | |
1469 | fail: | |
1470 | qcow2_update_options_abort(state->bs, r); | |
1471 | g_free(r); | |
1472 | return ret; | |
1473 | } | |
1474 | ||
1475 | static void qcow2_reopen_commit(BDRVReopenState *state) | |
1476 | { | |
1477 | qcow2_update_options_commit(state->bs, state->opaque); | |
1478 | g_free(state->opaque); | |
1479 | } | |
1480 | ||
1481 | static void qcow2_reopen_abort(BDRVReopenState *state) | |
1482 | { | |
1483 | qcow2_update_options_abort(state->bs, state->opaque); | |
1484 | g_free(state->opaque); | |
21d82ac9 JC |
1485 | } |
1486 | ||
5365f44d KW |
1487 | static void qcow2_join_options(QDict *options, QDict *old_options) |
1488 | { | |
1489 | bool has_new_overlap_template = | |
1490 | qdict_haskey(options, QCOW2_OPT_OVERLAP) || | |
1491 | qdict_haskey(options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1492 | bool has_new_total_cache_size = | |
1493 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE); | |
1494 | bool has_all_cache_options; | |
1495 | ||
1496 | /* New overlap template overrides all old overlap options */ | |
1497 | if (has_new_overlap_template) { | |
1498 | qdict_del(old_options, QCOW2_OPT_OVERLAP); | |
1499 | qdict_del(old_options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1500 | qdict_del(old_options, QCOW2_OPT_OVERLAP_MAIN_HEADER); | |
1501 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L1); | |
1502 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L2); | |
1503 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_TABLE); | |
1504 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK); | |
1505 | qdict_del(old_options, QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE); | |
1506 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L1); | |
1507 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L2); | |
1508 | } | |
1509 | ||
1510 | /* New total cache size overrides all old options */ | |
1511 | if (qdict_haskey(options, QCOW2_OPT_CACHE_SIZE)) { | |
1512 | qdict_del(old_options, QCOW2_OPT_L2_CACHE_SIZE); | |
1513 | qdict_del(old_options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1514 | } | |
1515 | ||
1516 | qdict_join(options, old_options, false); | |
1517 | ||
1518 | /* | |
1519 | * If after merging all cache size options are set, an old total size is | |
1520 | * overwritten. Do keep all options, however, if all three are new. The | |
1521 | * resulting error message is what we want to happen. | |
1522 | */ | |
1523 | has_all_cache_options = | |
1524 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE) || | |
1525 | qdict_haskey(options, QCOW2_OPT_L2_CACHE_SIZE) || | |
1526 | qdict_haskey(options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1527 | ||
1528 | if (has_all_cache_options && !has_new_total_cache_size) { | |
1529 | qdict_del(options, QCOW2_OPT_CACHE_SIZE); | |
1530 | } | |
1531 | } | |
1532 | ||
b6b8a333 | 1533 | static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs, |
67a0fd2a | 1534 | int64_t sector_num, int nb_sectors, int *pnum, BlockDriverState **file) |
585f8587 | 1535 | { |
ff99129a | 1536 | BDRVQcow2State *s = bs->opaque; |
585f8587 | 1537 | uint64_t cluster_offset; |
4bc74be9 | 1538 | int index_in_cluster, ret; |
ecfe1863 | 1539 | unsigned int bytes; |
4bc74be9 | 1540 | int64_t status = 0; |
585f8587 | 1541 | |
ecfe1863 | 1542 | bytes = MIN(INT_MAX, nb_sectors * BDRV_SECTOR_SIZE); |
f8a2e5e3 | 1543 | qemu_co_mutex_lock(&s->lock); |
ecfe1863 KW |
1544 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, &bytes, |
1545 | &cluster_offset); | |
f8a2e5e3 | 1546 | qemu_co_mutex_unlock(&s->lock); |
1c46efaa | 1547 | if (ret < 0) { |
d663640c | 1548 | return ret; |
1c46efaa | 1549 | } |
095a9c58 | 1550 | |
ecfe1863 KW |
1551 | *pnum = bytes >> BDRV_SECTOR_BITS; |
1552 | ||
4bc74be9 | 1553 | if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED && |
b25b387f | 1554 | !s->crypto) { |
4bc74be9 PB |
1555 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
1556 | cluster_offset |= (index_in_cluster << BDRV_SECTOR_BITS); | |
178b4db7 | 1557 | *file = bs->file->bs; |
4bc74be9 PB |
1558 | status |= BDRV_BLOCK_OFFSET_VALID | cluster_offset; |
1559 | } | |
fdfab37d | 1560 | if (ret == QCOW2_CLUSTER_ZERO_PLAIN || ret == QCOW2_CLUSTER_ZERO_ALLOC) { |
4bc74be9 PB |
1561 | status |= BDRV_BLOCK_ZERO; |
1562 | } else if (ret != QCOW2_CLUSTER_UNALLOCATED) { | |
1563 | status |= BDRV_BLOCK_DATA; | |
1564 | } | |
1565 | return status; | |
585f8587 FB |
1566 | } |
1567 | ||
a9465922 | 1568 | /* handle reading after the end of the backing file */ |
bd28f835 | 1569 | int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, |
ecfe1863 | 1570 | int64_t offset, int bytes) |
a9465922 | 1571 | { |
ecfe1863 | 1572 | uint64_t bs_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
a9465922 | 1573 | int n1; |
ecfe1863 KW |
1574 | |
1575 | if ((offset + bytes) <= bs_size) { | |
1576 | return bytes; | |
1577 | } | |
1578 | ||
1579 | if (offset >= bs_size) { | |
a9465922 | 1580 | n1 = 0; |
ecfe1863 KW |
1581 | } else { |
1582 | n1 = bs_size - offset; | |
1583 | } | |
bd28f835 | 1584 | |
ecfe1863 | 1585 | qemu_iovec_memset(qiov, n1, 0, bytes - n1); |
bd28f835 | 1586 | |
a9465922 FB |
1587 | return n1; |
1588 | } | |
1589 | ||
ecfe1863 KW |
1590 | static coroutine_fn int qcow2_co_preadv(BlockDriverState *bs, uint64_t offset, |
1591 | uint64_t bytes, QEMUIOVector *qiov, | |
1592 | int flags) | |
585f8587 | 1593 | { |
ff99129a | 1594 | BDRVQcow2State *s = bs->opaque; |
ecfe1863 | 1595 | int offset_in_cluster, n1; |
68d100e9 | 1596 | int ret; |
ecfe1863 | 1597 | unsigned int cur_bytes; /* number of bytes in current iteration */ |
c2bdd990 | 1598 | uint64_t cluster_offset = 0; |
3fc48d09 FZ |
1599 | uint64_t bytes_done = 0; |
1600 | QEMUIOVector hd_qiov; | |
1601 | uint8_t *cluster_data = NULL; | |
585f8587 | 1602 | |
3fc48d09 FZ |
1603 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1604 | ||
1605 | qemu_co_mutex_lock(&s->lock); | |
1606 | ||
ecfe1863 | 1607 | while (bytes != 0) { |
bd28f835 | 1608 | |
5ebaa27e | 1609 | /* prepare next request */ |
ecfe1863 | 1610 | cur_bytes = MIN(bytes, INT_MAX); |
b25b387f | 1611 | if (s->crypto) { |
ecfe1863 KW |
1612 | cur_bytes = MIN(cur_bytes, |
1613 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); | |
585f8587 | 1614 | } |
5ebaa27e | 1615 | |
ecfe1863 | 1616 | ret = qcow2_get_cluster_offset(bs, offset, &cur_bytes, &cluster_offset); |
8af36488 | 1617 | if (ret < 0) { |
3fc48d09 | 1618 | goto fail; |
8af36488 | 1619 | } |
bd28f835 | 1620 | |
ecfe1863 | 1621 | offset_in_cluster = offset_into_cluster(s, offset); |
c87c0672 | 1622 | |
3fc48d09 | 1623 | qemu_iovec_reset(&hd_qiov); |
ecfe1863 | 1624 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes); |
5ebaa27e | 1625 | |
68d000a3 KW |
1626 | switch (ret) { |
1627 | case QCOW2_CLUSTER_UNALLOCATED: | |
5ebaa27e | 1628 | |
760e0063 | 1629 | if (bs->backing) { |
5ebaa27e | 1630 | /* read from the base image */ |
760e0063 | 1631 | n1 = qcow2_backing_read1(bs->backing->bs, &hd_qiov, |
ecfe1863 | 1632 | offset, cur_bytes); |
5ebaa27e | 1633 | if (n1 > 0) { |
44deba5a KW |
1634 | QEMUIOVector local_qiov; |
1635 | ||
1636 | qemu_iovec_init(&local_qiov, hd_qiov.niov); | |
ecfe1863 | 1637 | qemu_iovec_concat(&local_qiov, &hd_qiov, 0, n1); |
44deba5a | 1638 | |
5ebaa27e FZ |
1639 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); |
1640 | qemu_co_mutex_unlock(&s->lock); | |
a03ef88f | 1641 | ret = bdrv_co_preadv(bs->backing, offset, n1, |
ecfe1863 | 1642 | &local_qiov, 0); |
5ebaa27e | 1643 | qemu_co_mutex_lock(&s->lock); |
44deba5a KW |
1644 | |
1645 | qemu_iovec_destroy(&local_qiov); | |
1646 | ||
5ebaa27e | 1647 | if (ret < 0) { |
3fc48d09 | 1648 | goto fail; |
5ebaa27e FZ |
1649 | } |
1650 | } | |
1651 | } else { | |
1652 | /* Note: in this case, no need to wait */ | |
ecfe1863 | 1653 | qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes); |
5ebaa27e | 1654 | } |
68d000a3 KW |
1655 | break; |
1656 | ||
fdfab37d EB |
1657 | case QCOW2_CLUSTER_ZERO_PLAIN: |
1658 | case QCOW2_CLUSTER_ZERO_ALLOC: | |
ecfe1863 | 1659 | qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes); |
6377af48 KW |
1660 | break; |
1661 | ||
68d000a3 | 1662 | case QCOW2_CLUSTER_COMPRESSED: |
5ebaa27e FZ |
1663 | /* add AIO support for compressed blocks ? */ |
1664 | ret = qcow2_decompress_cluster(bs, cluster_offset); | |
1665 | if (ret < 0) { | |
3fc48d09 | 1666 | goto fail; |
bd28f835 KW |
1667 | } |
1668 | ||
03396148 | 1669 | qemu_iovec_from_buf(&hd_qiov, 0, |
ecfe1863 KW |
1670 | s->cluster_cache + offset_in_cluster, |
1671 | cur_bytes); | |
68d000a3 KW |
1672 | break; |
1673 | ||
1674 | case QCOW2_CLUSTER_NORMAL: | |
5ebaa27e | 1675 | if ((cluster_offset & 511) != 0) { |
3fc48d09 FZ |
1676 | ret = -EIO; |
1677 | goto fail; | |
5ebaa27e | 1678 | } |
bd28f835 | 1679 | |
8336aafa | 1680 | if (bs->encrypted) { |
b25b387f | 1681 | assert(s->crypto); |
8336aafa | 1682 | |
5ebaa27e FZ |
1683 | /* |
1684 | * For encrypted images, read everything into a temporary | |
1685 | * contiguous buffer on which the AES functions can work. | |
1686 | */ | |
3fc48d09 FZ |
1687 | if (!cluster_data) { |
1688 | cluster_data = | |
9a4f4c31 KW |
1689 | qemu_try_blockalign(bs->file->bs, |
1690 | QCOW_MAX_CRYPT_CLUSTERS | |
1691 | * s->cluster_size); | |
de82815d KW |
1692 | if (cluster_data == NULL) { |
1693 | ret = -ENOMEM; | |
1694 | goto fail; | |
1695 | } | |
5ebaa27e FZ |
1696 | } |
1697 | ||
ecfe1863 | 1698 | assert(cur_bytes <= QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
3fc48d09 | 1699 | qemu_iovec_reset(&hd_qiov); |
ecfe1863 | 1700 | qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes); |
5ebaa27e FZ |
1701 | } |
1702 | ||
1703 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
1704 | qemu_co_mutex_unlock(&s->lock); | |
a03ef88f | 1705 | ret = bdrv_co_preadv(bs->file, |
ecfe1863 KW |
1706 | cluster_offset + offset_in_cluster, |
1707 | cur_bytes, &hd_qiov, 0); | |
5ebaa27e FZ |
1708 | qemu_co_mutex_lock(&s->lock); |
1709 | if (ret < 0) { | |
3fc48d09 | 1710 | goto fail; |
5ebaa27e | 1711 | } |
8336aafa | 1712 | if (bs->encrypted) { |
b25b387f | 1713 | assert(s->crypto); |
ecfe1863 KW |
1714 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
1715 | assert((cur_bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
f6fa64f6 | 1716 | Error *err = NULL; |
b25b387f | 1717 | if (qcrypto_block_decrypt(s->crypto, |
4652b8f3 DB |
1718 | (s->crypt_physical_offset ? |
1719 | cluster_offset + offset_in_cluster : | |
1720 | offset) >> BDRV_SECTOR_BITS, | |
446d306d | 1721 | cluster_data, |
b25b387f DB |
1722 | cur_bytes, |
1723 | &err) < 0) { | |
f6fa64f6 DB |
1724 | error_free(err); |
1725 | ret = -EIO; | |
1726 | goto fail; | |
1727 | } | |
ecfe1863 | 1728 | qemu_iovec_from_buf(qiov, bytes_done, cluster_data, cur_bytes); |
5ebaa27e | 1729 | } |
68d000a3 KW |
1730 | break; |
1731 | ||
1732 | default: | |
1733 | g_assert_not_reached(); | |
1734 | ret = -EIO; | |
1735 | goto fail; | |
faf575c1 | 1736 | } |
f141eafe | 1737 | |
ecfe1863 KW |
1738 | bytes -= cur_bytes; |
1739 | offset += cur_bytes; | |
1740 | bytes_done += cur_bytes; | |
5ebaa27e | 1741 | } |
3fc48d09 | 1742 | ret = 0; |
faf575c1 | 1743 | |
3fc48d09 | 1744 | fail: |
68d100e9 | 1745 | qemu_co_mutex_unlock(&s->lock); |
42496d62 | 1746 | |
3fc48d09 | 1747 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1748 | qemu_vfree(cluster_data); |
68d100e9 KW |
1749 | |
1750 | return ret; | |
585f8587 FB |
1751 | } |
1752 | ||
ee22a9d8 AG |
1753 | /* Check if it's possible to merge a write request with the writing of |
1754 | * the data from the COW regions */ | |
1755 | static bool merge_cow(uint64_t offset, unsigned bytes, | |
1756 | QEMUIOVector *hd_qiov, QCowL2Meta *l2meta) | |
1757 | { | |
1758 | QCowL2Meta *m; | |
1759 | ||
1760 | for (m = l2meta; m != NULL; m = m->next) { | |
1761 | /* If both COW regions are empty then there's nothing to merge */ | |
1762 | if (m->cow_start.nb_bytes == 0 && m->cow_end.nb_bytes == 0) { | |
1763 | continue; | |
1764 | } | |
1765 | ||
1766 | /* The data (middle) region must be immediately after the | |
1767 | * start region */ | |
1768 | if (l2meta_cow_start(m) + m->cow_start.nb_bytes != offset) { | |
1769 | continue; | |
1770 | } | |
1771 | ||
1772 | /* The end region must be immediately after the data (middle) | |
1773 | * region */ | |
1774 | if (m->offset + m->cow_end.offset != offset + bytes) { | |
1775 | continue; | |
1776 | } | |
1777 | ||
1778 | /* Make sure that adding both COW regions to the QEMUIOVector | |
1779 | * does not exceed IOV_MAX */ | |
1780 | if (hd_qiov->niov > IOV_MAX - 2) { | |
1781 | continue; | |
1782 | } | |
1783 | ||
1784 | m->data_qiov = hd_qiov; | |
1785 | return true; | |
1786 | } | |
1787 | ||
1788 | return false; | |
1789 | } | |
1790 | ||
d46a0bb2 KW |
1791 | static coroutine_fn int qcow2_co_pwritev(BlockDriverState *bs, uint64_t offset, |
1792 | uint64_t bytes, QEMUIOVector *qiov, | |
1793 | int flags) | |
585f8587 | 1794 | { |
ff99129a | 1795 | BDRVQcow2State *s = bs->opaque; |
d46a0bb2 | 1796 | int offset_in_cluster; |
68d100e9 | 1797 | int ret; |
d46a0bb2 | 1798 | unsigned int cur_bytes; /* number of sectors in current iteration */ |
c2bdd990 | 1799 | uint64_t cluster_offset; |
3fc48d09 FZ |
1800 | QEMUIOVector hd_qiov; |
1801 | uint64_t bytes_done = 0; | |
1802 | uint8_t *cluster_data = NULL; | |
8d2497c3 | 1803 | QCowL2Meta *l2meta = NULL; |
c2271403 | 1804 | |
d46a0bb2 | 1805 | trace_qcow2_writev_start_req(qemu_coroutine_self(), offset, bytes); |
3cce16f4 | 1806 | |
3fc48d09 FZ |
1807 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1808 | ||
1809 | s->cluster_cache_offset = -1; /* disable compressed cache */ | |
3b46e624 | 1810 | |
3fc48d09 FZ |
1811 | qemu_co_mutex_lock(&s->lock); |
1812 | ||
d46a0bb2 | 1813 | while (bytes != 0) { |
3fc48d09 | 1814 | |
f50f88b9 | 1815 | l2meta = NULL; |
cf5c1a23 | 1816 | |
3cce16f4 | 1817 | trace_qcow2_writev_start_part(qemu_coroutine_self()); |
d46a0bb2 KW |
1818 | offset_in_cluster = offset_into_cluster(s, offset); |
1819 | cur_bytes = MIN(bytes, INT_MAX); | |
1820 | if (bs->encrypted) { | |
1821 | cur_bytes = MIN(cur_bytes, | |
1822 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size | |
1823 | - offset_in_cluster); | |
5ebaa27e | 1824 | } |
095a9c58 | 1825 | |
d46a0bb2 KW |
1826 | ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes, |
1827 | &cluster_offset, &l2meta); | |
5ebaa27e | 1828 | if (ret < 0) { |
3fc48d09 | 1829 | goto fail; |
5ebaa27e | 1830 | } |
148da7ea | 1831 | |
5ebaa27e | 1832 | assert((cluster_offset & 511) == 0); |
148da7ea | 1833 | |
3fc48d09 | 1834 | qemu_iovec_reset(&hd_qiov); |
d46a0bb2 | 1835 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes); |
6f5f060b | 1836 | |
8336aafa | 1837 | if (bs->encrypted) { |
f6fa64f6 | 1838 | Error *err = NULL; |
b25b387f | 1839 | assert(s->crypto); |
3fc48d09 | 1840 | if (!cluster_data) { |
9a4f4c31 | 1841 | cluster_data = qemu_try_blockalign(bs->file->bs, |
de82815d KW |
1842 | QCOW_MAX_CRYPT_CLUSTERS |
1843 | * s->cluster_size); | |
1844 | if (cluster_data == NULL) { | |
1845 | ret = -ENOMEM; | |
1846 | goto fail; | |
1847 | } | |
5ebaa27e | 1848 | } |
6f5f060b | 1849 | |
3fc48d09 | 1850 | assert(hd_qiov.size <= |
5ebaa27e | 1851 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
d5e6b161 | 1852 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); |
6f5f060b | 1853 | |
4652b8f3 DB |
1854 | if (qcrypto_block_encrypt(s->crypto, |
1855 | (s->crypt_physical_offset ? | |
1856 | cluster_offset + offset_in_cluster : | |
1857 | offset) >> BDRV_SECTOR_BITS, | |
446d306d | 1858 | cluster_data, |
b25b387f | 1859 | cur_bytes, &err) < 0) { |
f6fa64f6 DB |
1860 | error_free(err); |
1861 | ret = -EIO; | |
1862 | goto fail; | |
1863 | } | |
6f5f060b | 1864 | |
3fc48d09 | 1865 | qemu_iovec_reset(&hd_qiov); |
d46a0bb2 | 1866 | qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes); |
5ebaa27e | 1867 | } |
6f5f060b | 1868 | |
231bb267 | 1869 | ret = qcow2_pre_write_overlap_check(bs, 0, |
d46a0bb2 | 1870 | cluster_offset + offset_in_cluster, cur_bytes); |
cf93980e HR |
1871 | if (ret < 0) { |
1872 | goto fail; | |
1873 | } | |
1874 | ||
ee22a9d8 AG |
1875 | /* If we need to do COW, check if it's possible to merge the |
1876 | * writing of the guest data together with that of the COW regions. | |
1877 | * If it's not possible (or not necessary) then write the | |
1878 | * guest data now. */ | |
1879 | if (!merge_cow(offset, cur_bytes, &hd_qiov, l2meta)) { | |
1880 | qemu_co_mutex_unlock(&s->lock); | |
1881 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); | |
1882 | trace_qcow2_writev_data(qemu_coroutine_self(), | |
1883 | cluster_offset + offset_in_cluster); | |
1884 | ret = bdrv_co_pwritev(bs->file, | |
1885 | cluster_offset + offset_in_cluster, | |
1886 | cur_bytes, &hd_qiov, 0); | |
1887 | qemu_co_mutex_lock(&s->lock); | |
1888 | if (ret < 0) { | |
1889 | goto fail; | |
1890 | } | |
5ebaa27e | 1891 | } |
f141eafe | 1892 | |
88c6588c KW |
1893 | while (l2meta != NULL) { |
1894 | QCowL2Meta *next; | |
1895 | ||
f50f88b9 KW |
1896 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); |
1897 | if (ret < 0) { | |
1898 | goto fail; | |
1899 | } | |
faf575c1 | 1900 | |
4e95314e KW |
1901 | /* Take the request off the list of running requests */ |
1902 | if (l2meta->nb_clusters != 0) { | |
1903 | QLIST_REMOVE(l2meta, next_in_flight); | |
1904 | } | |
1905 | ||
4e95314e | 1906 | qemu_co_queue_restart_all(&l2meta->dependent_requests); |
4e95314e | 1907 | |
88c6588c | 1908 | next = l2meta->next; |
f50f88b9 | 1909 | g_free(l2meta); |
88c6588c | 1910 | l2meta = next; |
f50f88b9 | 1911 | } |
0fa9131a | 1912 | |
d46a0bb2 KW |
1913 | bytes -= cur_bytes; |
1914 | offset += cur_bytes; | |
1915 | bytes_done += cur_bytes; | |
1916 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_bytes); | |
5ebaa27e | 1917 | } |
3fc48d09 | 1918 | ret = 0; |
faf575c1 | 1919 | |
3fc48d09 | 1920 | fail: |
4e95314e KW |
1921 | qemu_co_mutex_unlock(&s->lock); |
1922 | ||
88c6588c KW |
1923 | while (l2meta != NULL) { |
1924 | QCowL2Meta *next; | |
1925 | ||
4e95314e KW |
1926 | if (l2meta->nb_clusters != 0) { |
1927 | QLIST_REMOVE(l2meta, next_in_flight); | |
1928 | } | |
1929 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
88c6588c KW |
1930 | |
1931 | next = l2meta->next; | |
cf5c1a23 | 1932 | g_free(l2meta); |
88c6588c | 1933 | l2meta = next; |
cf5c1a23 | 1934 | } |
0fa9131a | 1935 | |
3fc48d09 | 1936 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1937 | qemu_vfree(cluster_data); |
3cce16f4 | 1938 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
42496d62 | 1939 | |
68d100e9 | 1940 | return ret; |
585f8587 FB |
1941 | } |
1942 | ||
ec6d8912 KW |
1943 | static int qcow2_inactivate(BlockDriverState *bs) |
1944 | { | |
1945 | BDRVQcow2State *s = bs->opaque; | |
1946 | int ret, result = 0; | |
1947 | ||
1948 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
1949 | if (ret) { | |
1950 | result = ret; | |
1951 | error_report("Failed to flush the L2 table cache: %s", | |
1952 | strerror(-ret)); | |
1953 | } | |
1954 | ||
1955 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
1956 | if (ret) { | |
1957 | result = ret; | |
1958 | error_report("Failed to flush the refcount block cache: %s", | |
1959 | strerror(-ret)); | |
1960 | } | |
1961 | ||
1962 | if (result == 0) { | |
1963 | qcow2_mark_clean(bs); | |
1964 | } | |
1965 | ||
1966 | return result; | |
1967 | } | |
1968 | ||
7c80ab3f | 1969 | static void qcow2_close(BlockDriverState *bs) |
585f8587 | 1970 | { |
ff99129a | 1971 | BDRVQcow2State *s = bs->opaque; |
de82815d | 1972 | qemu_vfree(s->l1_table); |
cf93980e HR |
1973 | /* else pre-write overlap checks in cache_destroy may crash */ |
1974 | s->l1_table = NULL; | |
29c1a730 | 1975 | |
140fd5a6 | 1976 | if (!(s->flags & BDRV_O_INACTIVE)) { |
ec6d8912 | 1977 | qcow2_inactivate(bs); |
27eb6c09 | 1978 | } |
c61d0004 | 1979 | |
279621c0 | 1980 | cache_clean_timer_del(bs); |
29c1a730 KW |
1981 | qcow2_cache_destroy(bs, s->l2_table_cache); |
1982 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1983 | ||
b25b387f DB |
1984 | qcrypto_block_free(s->crypto); |
1985 | s->crypto = NULL; | |
f6fa64f6 | 1986 | |
6744cbab | 1987 | g_free(s->unknown_header_fields); |
75bab85c | 1988 | cleanup_unknown_header_ext(bs); |
6744cbab | 1989 | |
e4603fe1 KW |
1990 | g_free(s->image_backing_file); |
1991 | g_free(s->image_backing_format); | |
1992 | ||
7267c094 | 1993 | g_free(s->cluster_cache); |
dea43a65 | 1994 | qemu_vfree(s->cluster_data); |
ed6ccf0f | 1995 | qcow2_refcount_close(bs); |
28c1202b | 1996 | qcow2_free_snapshots(bs); |
585f8587 FB |
1997 | } |
1998 | ||
5a8a30db | 1999 | static void qcow2_invalidate_cache(BlockDriverState *bs, Error **errp) |
06d9260f | 2000 | { |
ff99129a | 2001 | BDRVQcow2State *s = bs->opaque; |
06d9260f | 2002 | int flags = s->flags; |
b25b387f | 2003 | QCryptoBlock *crypto = NULL; |
acdfb480 | 2004 | QDict *options; |
5a8a30db KW |
2005 | Error *local_err = NULL; |
2006 | int ret; | |
06d9260f AL |
2007 | |
2008 | /* | |
2009 | * Backing files are read-only which makes all of their metadata immutable, | |
2010 | * that means we don't have to worry about reopening them here. | |
2011 | */ | |
2012 | ||
b25b387f DB |
2013 | crypto = s->crypto; |
2014 | s->crypto = NULL; | |
06d9260f AL |
2015 | |
2016 | qcow2_close(bs); | |
2017 | ||
ff99129a | 2018 | memset(s, 0, sizeof(BDRVQcow2State)); |
d475e5ac | 2019 | options = qdict_clone_shallow(bs->options); |
5a8a30db | 2020 | |
140fd5a6 | 2021 | flags &= ~BDRV_O_INACTIVE; |
4e4bf5c4 | 2022 | ret = qcow2_do_open(bs, options, flags, &local_err); |
a1904e48 | 2023 | QDECREF(options); |
5a8a30db | 2024 | if (local_err) { |
e43bfd9c MA |
2025 | error_propagate(errp, local_err); |
2026 | error_prepend(errp, "Could not reopen qcow2 layer: "); | |
191fb11b | 2027 | bs->drv = NULL; |
5a8a30db KW |
2028 | return; |
2029 | } else if (ret < 0) { | |
2030 | error_setg_errno(errp, -ret, "Could not reopen qcow2 layer"); | |
191fb11b | 2031 | bs->drv = NULL; |
5a8a30db KW |
2032 | return; |
2033 | } | |
acdfb480 | 2034 | |
b25b387f | 2035 | s->crypto = crypto; |
06d9260f AL |
2036 | } |
2037 | ||
e24e49e6 KW |
2038 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
2039 | size_t len, size_t buflen) | |
2040 | { | |
2041 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
2042 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
2043 | ||
2044 | if (buflen < ext_len) { | |
2045 | return -ENOSPC; | |
2046 | } | |
2047 | ||
2048 | *ext_backing_fmt = (QCowExtension) { | |
2049 | .magic = cpu_to_be32(magic), | |
2050 | .len = cpu_to_be32(len), | |
2051 | }; | |
0647d47c SH |
2052 | |
2053 | if (len) { | |
2054 | memcpy(buf + sizeof(QCowExtension), s, len); | |
2055 | } | |
e24e49e6 KW |
2056 | |
2057 | return ext_len; | |
2058 | } | |
2059 | ||
756e6736 | 2060 | /* |
e24e49e6 KW |
2061 | * Updates the qcow2 header, including the variable length parts of it, i.e. |
2062 | * the backing file name and all extensions. qcow2 was not designed to allow | |
2063 | * such changes, so if we run out of space (we can only use the first cluster) | |
2064 | * this function may fail. | |
756e6736 KW |
2065 | * |
2066 | * Returns 0 on success, -errno in error cases. | |
2067 | */ | |
e24e49e6 | 2068 | int qcow2_update_header(BlockDriverState *bs) |
756e6736 | 2069 | { |
ff99129a | 2070 | BDRVQcow2State *s = bs->opaque; |
e24e49e6 KW |
2071 | QCowHeader *header; |
2072 | char *buf; | |
2073 | size_t buflen = s->cluster_size; | |
756e6736 | 2074 | int ret; |
e24e49e6 KW |
2075 | uint64_t total_size; |
2076 | uint32_t refcount_table_clusters; | |
6744cbab | 2077 | size_t header_length; |
75bab85c | 2078 | Qcow2UnknownHeaderExtension *uext; |
756e6736 | 2079 | |
e24e49e6 | 2080 | buf = qemu_blockalign(bs, buflen); |
756e6736 | 2081 | |
e24e49e6 KW |
2082 | /* Header structure */ |
2083 | header = (QCowHeader*) buf; | |
756e6736 | 2084 | |
e24e49e6 KW |
2085 | if (buflen < sizeof(*header)) { |
2086 | ret = -ENOSPC; | |
2087 | goto fail; | |
756e6736 KW |
2088 | } |
2089 | ||
6744cbab | 2090 | header_length = sizeof(*header) + s->unknown_header_fields_size; |
e24e49e6 KW |
2091 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
2092 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
2093 | ||
2094 | *header = (QCowHeader) { | |
6744cbab | 2095 | /* Version 2 fields */ |
e24e49e6 | 2096 | .magic = cpu_to_be32(QCOW_MAGIC), |
6744cbab | 2097 | .version = cpu_to_be32(s->qcow_version), |
e24e49e6 KW |
2098 | .backing_file_offset = 0, |
2099 | .backing_file_size = 0, | |
2100 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
2101 | .size = cpu_to_be64(total_size), | |
2102 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
2103 | .l1_size = cpu_to_be32(s->l1_size), | |
2104 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
2105 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
2106 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
2107 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
2108 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
6744cbab KW |
2109 | |
2110 | /* Version 3 fields */ | |
2111 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
2112 | .compatible_features = cpu_to_be64(s->compatible_features), | |
2113 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
b6481f37 | 2114 | .refcount_order = cpu_to_be32(s->refcount_order), |
6744cbab | 2115 | .header_length = cpu_to_be32(header_length), |
e24e49e6 | 2116 | }; |
756e6736 | 2117 | |
6744cbab KW |
2118 | /* For older versions, write a shorter header */ |
2119 | switch (s->qcow_version) { | |
2120 | case 2: | |
2121 | ret = offsetof(QCowHeader, incompatible_features); | |
2122 | break; | |
2123 | case 3: | |
2124 | ret = sizeof(*header); | |
2125 | break; | |
2126 | default: | |
b6c14762 JM |
2127 | ret = -EINVAL; |
2128 | goto fail; | |
6744cbab KW |
2129 | } |
2130 | ||
2131 | buf += ret; | |
2132 | buflen -= ret; | |
2133 | memset(buf, 0, buflen); | |
2134 | ||
2135 | /* Preserve any unknown field in the header */ | |
2136 | if (s->unknown_header_fields_size) { | |
2137 | if (buflen < s->unknown_header_fields_size) { | |
2138 | ret = -ENOSPC; | |
2139 | goto fail; | |
2140 | } | |
2141 | ||
2142 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
2143 | buf += s->unknown_header_fields_size; | |
2144 | buflen -= s->unknown_header_fields_size; | |
2145 | } | |
756e6736 | 2146 | |
e24e49e6 | 2147 | /* Backing file format header extension */ |
e4603fe1 | 2148 | if (s->image_backing_format) { |
e24e49e6 | 2149 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, |
e4603fe1 KW |
2150 | s->image_backing_format, |
2151 | strlen(s->image_backing_format), | |
e24e49e6 KW |
2152 | buflen); |
2153 | if (ret < 0) { | |
2154 | goto fail; | |
756e6736 KW |
2155 | } |
2156 | ||
e24e49e6 KW |
2157 | buf += ret; |
2158 | buflen -= ret; | |
756e6736 KW |
2159 | } |
2160 | ||
4652b8f3 DB |
2161 | /* Full disk encryption header pointer extension */ |
2162 | if (s->crypto_header.offset != 0) { | |
2163 | cpu_to_be64s(&s->crypto_header.offset); | |
2164 | cpu_to_be64s(&s->crypto_header.length); | |
2165 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_CRYPTO_HEADER, | |
2166 | &s->crypto_header, sizeof(s->crypto_header), | |
2167 | buflen); | |
2168 | be64_to_cpus(&s->crypto_header.offset); | |
2169 | be64_to_cpus(&s->crypto_header.length); | |
2170 | if (ret < 0) { | |
2171 | goto fail; | |
2172 | } | |
2173 | buf += ret; | |
2174 | buflen -= ret; | |
2175 | } | |
2176 | ||
cfcc4c62 | 2177 | /* Feature table */ |
1a4828c7 KW |
2178 | if (s->qcow_version >= 3) { |
2179 | Qcow2Feature features[] = { | |
2180 | { | |
2181 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
2182 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
2183 | .name = "dirty bit", | |
2184 | }, | |
2185 | { | |
2186 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
2187 | .bit = QCOW2_INCOMPAT_CORRUPT_BITNR, | |
2188 | .name = "corrupt bit", | |
2189 | }, | |
2190 | { | |
2191 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
2192 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
2193 | .name = "lazy refcounts", | |
2194 | }, | |
2195 | }; | |
2196 | ||
2197 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
2198 | features, sizeof(features), buflen); | |
2199 | if (ret < 0) { | |
2200 | goto fail; | |
2201 | } | |
2202 | buf += ret; | |
2203 | buflen -= ret; | |
cfcc4c62 | 2204 | } |
cfcc4c62 | 2205 | |
75bab85c KW |
2206 | /* Keep unknown header extensions */ |
2207 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
2208 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
2209 | if (ret < 0) { | |
2210 | goto fail; | |
2211 | } | |
2212 | ||
2213 | buf += ret; | |
2214 | buflen -= ret; | |
2215 | } | |
2216 | ||
e24e49e6 KW |
2217 | /* End of header extensions */ |
2218 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
756e6736 KW |
2219 | if (ret < 0) { |
2220 | goto fail; | |
2221 | } | |
2222 | ||
e24e49e6 KW |
2223 | buf += ret; |
2224 | buflen -= ret; | |
756e6736 | 2225 | |
e24e49e6 | 2226 | /* Backing file name */ |
e4603fe1 KW |
2227 | if (s->image_backing_file) { |
2228 | size_t backing_file_len = strlen(s->image_backing_file); | |
e24e49e6 KW |
2229 | |
2230 | if (buflen < backing_file_len) { | |
2231 | ret = -ENOSPC; | |
2232 | goto fail; | |
2233 | } | |
2234 | ||
00ea1881 | 2235 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ |
e4603fe1 | 2236 | strncpy(buf, s->image_backing_file, buflen); |
e24e49e6 KW |
2237 | |
2238 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
2239 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
756e6736 KW |
2240 | } |
2241 | ||
e24e49e6 | 2242 | /* Write the new header */ |
d9ca2ea2 | 2243 | ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size); |
756e6736 KW |
2244 | if (ret < 0) { |
2245 | goto fail; | |
2246 | } | |
2247 | ||
2248 | ret = 0; | |
2249 | fail: | |
e24e49e6 | 2250 | qemu_vfree(header); |
756e6736 KW |
2251 | return ret; |
2252 | } | |
2253 | ||
2254 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
2255 | const char *backing_file, const char *backing_fmt) | |
2256 | { | |
ff99129a | 2257 | BDRVQcow2State *s = bs->opaque; |
e4603fe1 | 2258 | |
4e876bcf HR |
2259 | if (backing_file && strlen(backing_file) > 1023) { |
2260 | return -EINVAL; | |
2261 | } | |
2262 | ||
e24e49e6 KW |
2263 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
2264 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2265 | ||
e4603fe1 KW |
2266 | g_free(s->image_backing_file); |
2267 | g_free(s->image_backing_format); | |
2268 | ||
2269 | s->image_backing_file = backing_file ? g_strdup(bs->backing_file) : NULL; | |
2270 | s->image_backing_format = backing_fmt ? g_strdup(bs->backing_format) : NULL; | |
2271 | ||
e24e49e6 | 2272 | return qcow2_update_header(bs); |
756e6736 KW |
2273 | } |
2274 | ||
4652b8f3 DB |
2275 | static int qcow2_crypt_method_from_format(const char *encryptfmt) |
2276 | { | |
2277 | if (g_str_equal(encryptfmt, "luks")) { | |
2278 | return QCOW_CRYPT_LUKS; | |
2279 | } else if (g_str_equal(encryptfmt, "aes")) { | |
2280 | return QCOW_CRYPT_AES; | |
2281 | } else { | |
2282 | return -EINVAL; | |
2283 | } | |
2284 | } | |
b25b387f DB |
2285 | |
2286 | static int qcow2_set_up_encryption(BlockDriverState *bs, const char *encryptfmt, | |
2287 | QemuOpts *opts, Error **errp) | |
2288 | { | |
2289 | BDRVQcow2State *s = bs->opaque; | |
2290 | QCryptoBlockCreateOptions *cryptoopts = NULL; | |
2291 | QCryptoBlock *crypto = NULL; | |
2292 | int ret = -EINVAL; | |
2293 | QDict *options, *encryptopts; | |
4652b8f3 | 2294 | int fmt; |
b25b387f DB |
2295 | |
2296 | options = qemu_opts_to_qdict(opts, NULL); | |
2297 | qdict_extract_subqdict(options, &encryptopts, "encrypt."); | |
2298 | QDECREF(options); | |
2299 | ||
4652b8f3 DB |
2300 | fmt = qcow2_crypt_method_from_format(encryptfmt); |
2301 | ||
2302 | switch (fmt) { | |
2303 | case QCOW_CRYPT_LUKS: | |
2304 | cryptoopts = block_crypto_create_opts_init( | |
2305 | Q_CRYPTO_BLOCK_FORMAT_LUKS, encryptopts, errp); | |
2306 | break; | |
2307 | case QCOW_CRYPT_AES: | |
2308 | cryptoopts = block_crypto_create_opts_init( | |
2309 | Q_CRYPTO_BLOCK_FORMAT_QCOW, encryptopts, errp); | |
2310 | break; | |
2311 | default: | |
2312 | error_setg(errp, "Unknown encryption format '%s'", encryptfmt); | |
2313 | break; | |
b25b387f | 2314 | } |
b25b387f DB |
2315 | if (!cryptoopts) { |
2316 | ret = -EINVAL; | |
2317 | goto out; | |
2318 | } | |
4652b8f3 | 2319 | s->crypt_method_header = fmt; |
b25b387f DB |
2320 | |
2321 | crypto = qcrypto_block_create(cryptoopts, | |
4652b8f3 DB |
2322 | qcow2_crypto_hdr_init_func, |
2323 | qcow2_crypto_hdr_write_func, | |
b25b387f DB |
2324 | bs, errp); |
2325 | if (!crypto) { | |
2326 | ret = -EINVAL; | |
2327 | goto out; | |
2328 | } | |
2329 | ||
2330 | ret = qcow2_update_header(bs); | |
2331 | if (ret < 0) { | |
2332 | error_setg_errno(errp, -ret, "Could not write encryption header"); | |
2333 | goto out; | |
2334 | } | |
2335 | ||
2336 | out: | |
2337 | QDECREF(encryptopts); | |
2338 | qcrypto_block_free(crypto); | |
2339 | qapi_free_QCryptoBlockCreateOptions(cryptoopts); | |
2340 | return ret; | |
2341 | } | |
2342 | ||
2343 | ||
a35e1c17 KW |
2344 | static int preallocate(BlockDriverState *bs) |
2345 | { | |
d46a0bb2 | 2346 | uint64_t bytes; |
a35e1c17 | 2347 | uint64_t offset; |
060bee89 | 2348 | uint64_t host_offset = 0; |
d46a0bb2 | 2349 | unsigned int cur_bytes; |
148da7ea | 2350 | int ret; |
f50f88b9 | 2351 | QCowL2Meta *meta; |
a35e1c17 | 2352 | |
d46a0bb2 | 2353 | bytes = bdrv_getlength(bs); |
a35e1c17 KW |
2354 | offset = 0; |
2355 | ||
d46a0bb2 KW |
2356 | while (bytes) { |
2357 | cur_bytes = MIN(bytes, INT_MAX); | |
2358 | ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes, | |
060bee89 | 2359 | &host_offset, &meta); |
148da7ea | 2360 | if (ret < 0) { |
19dbcbf7 | 2361 | return ret; |
a35e1c17 KW |
2362 | } |
2363 | ||
c792707f SH |
2364 | while (meta) { |
2365 | QCowL2Meta *next = meta->next; | |
2366 | ||
7c2bbf4a HT |
2367 | ret = qcow2_alloc_cluster_link_l2(bs, meta); |
2368 | if (ret < 0) { | |
2369 | qcow2_free_any_clusters(bs, meta->alloc_offset, | |
2370 | meta->nb_clusters, QCOW2_DISCARD_NEVER); | |
2371 | return ret; | |
2372 | } | |
2373 | ||
2374 | /* There are no dependent requests, but we need to remove our | |
2375 | * request from the list of in-flight requests */ | |
4e95314e | 2376 | QLIST_REMOVE(meta, next_in_flight); |
c792707f SH |
2377 | |
2378 | g_free(meta); | |
2379 | meta = next; | |
f50f88b9 | 2380 | } |
f214978a | 2381 | |
a35e1c17 KW |
2382 | /* TODO Preallocate data if requested */ |
2383 | ||
d46a0bb2 KW |
2384 | bytes -= cur_bytes; |
2385 | offset += cur_bytes; | |
a35e1c17 KW |
2386 | } |
2387 | ||
2388 | /* | |
2389 | * It is expected that the image file is large enough to actually contain | |
2390 | * all of the allocated clusters (otherwise we get failing reads after | |
2391 | * EOF). Extend the image to the last allocated sector. | |
2392 | */ | |
060bee89 | 2393 | if (host_offset != 0) { |
d46a0bb2 | 2394 | uint8_t data = 0; |
d9ca2ea2 | 2395 | ret = bdrv_pwrite(bs->file, (host_offset + cur_bytes) - 1, |
d46a0bb2 | 2396 | &data, 1); |
19dbcbf7 KW |
2397 | if (ret < 0) { |
2398 | return ret; | |
2399 | } | |
a35e1c17 KW |
2400 | } |
2401 | ||
2402 | return 0; | |
2403 | } | |
2404 | ||
7c80ab3f JS |
2405 | static int qcow2_create2(const char *filename, int64_t total_size, |
2406 | const char *backing_file, const char *backing_format, | |
ffeaac9b | 2407 | int flags, size_t cluster_size, PreallocMode prealloc, |
bd4b167f | 2408 | QemuOpts *opts, int version, int refcount_order, |
0cb8d47b | 2409 | const char *encryptfmt, Error **errp) |
a9420734 | 2410 | { |
a9420734 | 2411 | int cluster_bits; |
e6641719 HR |
2412 | QDict *options; |
2413 | ||
2414 | /* Calculate cluster_bits */ | |
786a4ea8 | 2415 | cluster_bits = ctz32(cluster_size); |
a9420734 KW |
2416 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || |
2417 | (1 << cluster_bits) != cluster_size) | |
2418 | { | |
3ef6c40a HR |
2419 | error_setg(errp, "Cluster size must be a power of two between %d and " |
2420 | "%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); | |
a9420734 KW |
2421 | return -EINVAL; |
2422 | } | |
2423 | ||
2424 | /* | |
2425 | * Open the image file and write a minimal qcow2 header. | |
2426 | * | |
2427 | * We keep things simple and start with a zero-sized image. We also | |
2428 | * do without refcount blocks or a L1 table for now. We'll fix the | |
2429 | * inconsistency later. | |
2430 | * | |
2431 | * We do need a refcount table because growing the refcount table means | |
2432 | * allocating two new refcount blocks - the seconds of which would be at | |
2433 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
2434 | * size for any qcow2 image. | |
2435 | */ | |
23588797 | 2436 | BlockBackend *blk; |
f8413b3c | 2437 | QCowHeader *header; |
b106ad91 | 2438 | uint64_t* refcount_table; |
3ef6c40a | 2439 | Error *local_err = NULL; |
a9420734 KW |
2440 | int ret; |
2441 | ||
0e4271b7 | 2442 | if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) { |
bd4b167f HR |
2443 | /* Note: The following calculation does not need to be exact; if it is a |
2444 | * bit off, either some bytes will be "leaked" (which is fine) or we | |
2445 | * will need to increase the file size by some bytes (which is fine, | |
2446 | * too, as long as the bulk is allocated here). Therefore, using | |
2447 | * floating point arithmetic is fine. */ | |
0e4271b7 | 2448 | int64_t meta_size = 0; |
92413c16 | 2449 | uint64_t nreftablee, nrefblocke, nl1e, nl2e, refblock_count; |
0e4271b7 | 2450 | int64_t aligned_total_size = align_offset(total_size, cluster_size); |
bd4b167f HR |
2451 | int refblock_bits, refblock_size; |
2452 | /* refcount entry size in bytes */ | |
2453 | double rces = (1 << refcount_order) / 8.; | |
2454 | ||
2455 | /* see qcow2_open() */ | |
2456 | refblock_bits = cluster_bits - (refcount_order - 3); | |
2457 | refblock_size = 1 << refblock_bits; | |
0e4271b7 HT |
2458 | |
2459 | /* header: 1 cluster */ | |
2460 | meta_size += cluster_size; | |
2461 | ||
2462 | /* total size of L2 tables */ | |
2463 | nl2e = aligned_total_size / cluster_size; | |
2464 | nl2e = align_offset(nl2e, cluster_size / sizeof(uint64_t)); | |
2465 | meta_size += nl2e * sizeof(uint64_t); | |
2466 | ||
2467 | /* total size of L1 tables */ | |
2468 | nl1e = nl2e * sizeof(uint64_t) / cluster_size; | |
2469 | nl1e = align_offset(nl1e, cluster_size / sizeof(uint64_t)); | |
2470 | meta_size += nl1e * sizeof(uint64_t); | |
2471 | ||
2472 | /* total size of refcount blocks | |
2473 | * | |
2474 | * note: every host cluster is reference-counted, including metadata | |
2475 | * (even refcount blocks are recursively included). | |
2476 | * Let: | |
2477 | * a = total_size (this is the guest disk size) | |
2478 | * m = meta size not including refcount blocks and refcount tables | |
2479 | * c = cluster size | |
2480 | * y1 = number of refcount blocks entries | |
2481 | * y2 = meta size including everything | |
bd4b167f | 2482 | * rces = refcount entry size in bytes |
0e4271b7 HT |
2483 | * then, |
2484 | * y1 = (y2 + a)/c | |
bd4b167f | 2485 | * y2 = y1 * rces + y1 * rces * sizeof(u64) / c + m |
0e4271b7 | 2486 | * we can get y1: |
bd4b167f | 2487 | * y1 = (a + m) / (c - rces - rces * sizeof(u64) / c) |
0e4271b7 | 2488 | */ |
bd4b167f HR |
2489 | nrefblocke = (aligned_total_size + meta_size + cluster_size) |
2490 | / (cluster_size - rces - rces * sizeof(uint64_t) | |
2491 | / cluster_size); | |
92413c16 HR |
2492 | refblock_count = DIV_ROUND_UP(nrefblocke, refblock_size); |
2493 | meta_size += refblock_count * cluster_size; | |
0e4271b7 HT |
2494 | |
2495 | /* total size of refcount tables */ | |
92413c16 HR |
2496 | nreftablee = align_offset(refblock_count, |
2497 | cluster_size / sizeof(uint64_t)); | |
0e4271b7 HT |
2498 | meta_size += nreftablee * sizeof(uint64_t); |
2499 | ||
2500 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, | |
39101f25 | 2501 | aligned_total_size + meta_size, &error_abort); |
f43e47db MA |
2502 | qemu_opt_set(opts, BLOCK_OPT_PREALLOC, PreallocMode_lookup[prealloc], |
2503 | &error_abort); | |
0e4271b7 HT |
2504 | } |
2505 | ||
c282e1fd | 2506 | ret = bdrv_create_file(filename, opts, &local_err); |
a9420734 | 2507 | if (ret < 0) { |
3ef6c40a | 2508 | error_propagate(errp, local_err); |
a9420734 KW |
2509 | return ret; |
2510 | } | |
2511 | ||
efaa7c4e | 2512 | blk = blk_new_open(filename, NULL, NULL, |
55880601 KW |
2513 | BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_PROTOCOL, |
2514 | &local_err); | |
23588797 | 2515 | if (blk == NULL) { |
3ef6c40a | 2516 | error_propagate(errp, local_err); |
23588797 | 2517 | return -EIO; |
a9420734 KW |
2518 | } |
2519 | ||
23588797 KW |
2520 | blk_set_allow_write_beyond_eof(blk, true); |
2521 | ||
a9420734 | 2522 | /* Write the header */ |
f8413b3c KW |
2523 | QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header)); |
2524 | header = g_malloc0(cluster_size); | |
2525 | *header = (QCowHeader) { | |
2526 | .magic = cpu_to_be32(QCOW_MAGIC), | |
2527 | .version = cpu_to_be32(version), | |
2528 | .cluster_bits = cpu_to_be32(cluster_bits), | |
2529 | .size = cpu_to_be64(0), | |
2530 | .l1_table_offset = cpu_to_be64(0), | |
2531 | .l1_size = cpu_to_be32(0), | |
2532 | .refcount_table_offset = cpu_to_be64(cluster_size), | |
2533 | .refcount_table_clusters = cpu_to_be32(1), | |
bd4b167f | 2534 | .refcount_order = cpu_to_be32(refcount_order), |
f8413b3c KW |
2535 | .header_length = cpu_to_be32(sizeof(*header)), |
2536 | }; | |
a9420734 | 2537 | |
b25b387f DB |
2538 | /* We'll update this to correct value later */ |
2539 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); | |
a9420734 | 2540 | |
bfe8043e | 2541 | if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) { |
f8413b3c | 2542 | header->compatible_features |= |
bfe8043e SH |
2543 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); |
2544 | } | |
2545 | ||
8341f00d | 2546 | ret = blk_pwrite(blk, 0, header, cluster_size, 0); |
f8413b3c | 2547 | g_free(header); |
a9420734 | 2548 | if (ret < 0) { |
3ef6c40a | 2549 | error_setg_errno(errp, -ret, "Could not write qcow2 header"); |
a9420734 KW |
2550 | goto out; |
2551 | } | |
2552 | ||
b106ad91 KW |
2553 | /* Write a refcount table with one refcount block */ |
2554 | refcount_table = g_malloc0(2 * cluster_size); | |
2555 | refcount_table[0] = cpu_to_be64(2 * cluster_size); | |
8341f00d | 2556 | ret = blk_pwrite(blk, cluster_size, refcount_table, 2 * cluster_size, 0); |
7267c094 | 2557 | g_free(refcount_table); |
a9420734 KW |
2558 | |
2559 | if (ret < 0) { | |
3ef6c40a | 2560 | error_setg_errno(errp, -ret, "Could not write refcount table"); |
a9420734 KW |
2561 | goto out; |
2562 | } | |
2563 | ||
23588797 KW |
2564 | blk_unref(blk); |
2565 | blk = NULL; | |
a9420734 KW |
2566 | |
2567 | /* | |
2568 | * And now open the image and make it consistent first (i.e. increase the | |
2569 | * refcount of the cluster that is occupied by the header and the refcount | |
2570 | * table) | |
2571 | */ | |
e6641719 | 2572 | options = qdict_new(); |
46f5ac20 | 2573 | qdict_put_str(options, "driver", "qcow2"); |
efaa7c4e | 2574 | blk = blk_new_open(filename, NULL, options, |
55880601 KW |
2575 | BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_NO_FLUSH, |
2576 | &local_err); | |
23588797 | 2577 | if (blk == NULL) { |
3ef6c40a | 2578 | error_propagate(errp, local_err); |
23588797 | 2579 | ret = -EIO; |
a9420734 KW |
2580 | goto out; |
2581 | } | |
2582 | ||
23588797 | 2583 | ret = qcow2_alloc_clusters(blk_bs(blk), 3 * cluster_size); |
a9420734 | 2584 | if (ret < 0) { |
3ef6c40a HR |
2585 | error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 " |
2586 | "header and refcount table"); | |
a9420734 KW |
2587 | goto out; |
2588 | ||
2589 | } else if (ret != 0) { | |
2590 | error_report("Huh, first cluster in empty image is already in use?"); | |
2591 | abort(); | |
2592 | } | |
2593 | ||
b527c9b3 | 2594 | /* Create a full header (including things like feature table) */ |
23588797 | 2595 | ret = qcow2_update_header(blk_bs(blk)); |
b527c9b3 KW |
2596 | if (ret < 0) { |
2597 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); | |
2598 | goto out; | |
2599 | } | |
2600 | ||
a9420734 | 2601 | /* Okay, now that we have a valid image, let's give it the right size */ |
ed3d2ec9 | 2602 | ret = blk_truncate(blk, total_size, errp); |
a9420734 | 2603 | if (ret < 0) { |
ed3d2ec9 | 2604 | error_prepend(errp, "Could not resize image: "); |
a9420734 KW |
2605 | goto out; |
2606 | } | |
2607 | ||
2608 | /* Want a backing file? There you go.*/ | |
2609 | if (backing_file) { | |
23588797 | 2610 | ret = bdrv_change_backing_file(blk_bs(blk), backing_file, backing_format); |
a9420734 | 2611 | if (ret < 0) { |
3ef6c40a HR |
2612 | error_setg_errno(errp, -ret, "Could not assign backing file '%s' " |
2613 | "with format '%s'", backing_file, backing_format); | |
a9420734 KW |
2614 | goto out; |
2615 | } | |
2616 | } | |
2617 | ||
b25b387f DB |
2618 | /* Want encryption? There you go. */ |
2619 | if (encryptfmt) { | |
2620 | ret = qcow2_set_up_encryption(blk_bs(blk), encryptfmt, opts, errp); | |
2621 | if (ret < 0) { | |
2622 | goto out; | |
2623 | } | |
2624 | } | |
2625 | ||
a9420734 | 2626 | /* And if we're supposed to preallocate metadata, do that now */ |
0e4271b7 | 2627 | if (prealloc != PREALLOC_MODE_OFF) { |
23588797 | 2628 | BDRVQcow2State *s = blk_bs(blk)->opaque; |
15552c4a | 2629 | qemu_co_mutex_lock(&s->lock); |
23588797 | 2630 | ret = preallocate(blk_bs(blk)); |
15552c4a | 2631 | qemu_co_mutex_unlock(&s->lock); |
a9420734 | 2632 | if (ret < 0) { |
3ef6c40a | 2633 | error_setg_errno(errp, -ret, "Could not preallocate metadata"); |
a9420734 KW |
2634 | goto out; |
2635 | } | |
2636 | } | |
2637 | ||
23588797 KW |
2638 | blk_unref(blk); |
2639 | blk = NULL; | |
ba2ab2f2 | 2640 | |
b25b387f DB |
2641 | /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning. |
2642 | * Using BDRV_O_NO_IO, since encryption is now setup we don't want to | |
2643 | * have to setup decryption context. We're not doing any I/O on the top | |
2644 | * level BlockDriverState, only lower layers, where BDRV_O_NO_IO does | |
2645 | * not have effect. | |
2646 | */ | |
e6641719 | 2647 | options = qdict_new(); |
46f5ac20 | 2648 | qdict_put_str(options, "driver", "qcow2"); |
efaa7c4e | 2649 | blk = blk_new_open(filename, NULL, options, |
b25b387f DB |
2650 | BDRV_O_RDWR | BDRV_O_NO_BACKING | BDRV_O_NO_IO, |
2651 | &local_err); | |
23588797 | 2652 | if (blk == NULL) { |
ba2ab2f2 | 2653 | error_propagate(errp, local_err); |
23588797 | 2654 | ret = -EIO; |
ba2ab2f2 HR |
2655 | goto out; |
2656 | } | |
2657 | ||
a9420734 KW |
2658 | ret = 0; |
2659 | out: | |
23588797 KW |
2660 | if (blk) { |
2661 | blk_unref(blk); | |
f67503e5 | 2662 | } |
a9420734 KW |
2663 | return ret; |
2664 | } | |
de5f3f40 | 2665 | |
1bd0e2d1 | 2666 | static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) |
de5f3f40 | 2667 | { |
1bd0e2d1 CL |
2668 | char *backing_file = NULL; |
2669 | char *backing_fmt = NULL; | |
2670 | char *buf = NULL; | |
180e9526 | 2671 | uint64_t size = 0; |
de5f3f40 | 2672 | int flags = 0; |
99cce9fa | 2673 | size_t cluster_size = DEFAULT_CLUSTER_SIZE; |
ffeaac9b | 2674 | PreallocMode prealloc; |
8ad1898c | 2675 | int version = 3; |
bd4b167f HR |
2676 | uint64_t refcount_bits = 16; |
2677 | int refcount_order; | |
0cb8d47b | 2678 | const char *encryptfmt = NULL; |
3ef6c40a HR |
2679 | Error *local_err = NULL; |
2680 | int ret; | |
de5f3f40 KW |
2681 | |
2682 | /* Read out options */ | |
180e9526 HT |
2683 | size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
2684 | BDRV_SECTOR_SIZE); | |
1bd0e2d1 CL |
2685 | backing_file = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE); |
2686 | backing_fmt = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FMT); | |
0cb8d47b DB |
2687 | encryptfmt = qemu_opt_get_del(opts, BLOCK_OPT_ENCRYPT_FORMAT); |
2688 | if (encryptfmt) { | |
2689 | if (qemu_opt_get_del(opts, BLOCK_OPT_ENCRYPT)) { | |
2690 | error_setg(errp, "Options " BLOCK_OPT_ENCRYPT " and " | |
2691 | BLOCK_OPT_ENCRYPT_FORMAT " are mutually exclusive"); | |
2692 | ret = -EINVAL; | |
2693 | goto finish; | |
2694 | } | |
2695 | } else if (qemu_opt_get_bool_del(opts, BLOCK_OPT_ENCRYPT, false)) { | |
2696 | encryptfmt = "aes"; | |
1bd0e2d1 CL |
2697 | } |
2698 | cluster_size = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, | |
2699 | DEFAULT_CLUSTER_SIZE); | |
2700 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
ffeaac9b | 2701 | prealloc = qapi_enum_parse(PreallocMode_lookup, buf, |
7fb1cf16 | 2702 | PREALLOC_MODE__MAX, PREALLOC_MODE_OFF, |
ffeaac9b HT |
2703 | &local_err); |
2704 | if (local_err) { | |
2705 | error_propagate(errp, local_err); | |
1bd0e2d1 CL |
2706 | ret = -EINVAL; |
2707 | goto finish; | |
2708 | } | |
2709 | g_free(buf); | |
2710 | buf = qemu_opt_get_del(opts, BLOCK_OPT_COMPAT_LEVEL); | |
2711 | if (!buf) { | |
2712 | /* keep the default */ | |
2713 | } else if (!strcmp(buf, "0.10")) { | |
2714 | version = 2; | |
2715 | } else if (!strcmp(buf, "1.1")) { | |
2716 | version = 3; | |
2717 | } else { | |
2718 | error_setg(errp, "Invalid compatibility level: '%s'", buf); | |
2719 | ret = -EINVAL; | |
2720 | goto finish; | |
2721 | } | |
2722 | ||
2723 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_LAZY_REFCOUNTS, false)) { | |
2724 | flags |= BLOCK_FLAG_LAZY_REFCOUNTS; | |
de5f3f40 KW |
2725 | } |
2726 | ||
ffeaac9b | 2727 | if (backing_file && prealloc != PREALLOC_MODE_OFF) { |
3ef6c40a HR |
2728 | error_setg(errp, "Backing file and preallocation cannot be used at " |
2729 | "the same time"); | |
1bd0e2d1 CL |
2730 | ret = -EINVAL; |
2731 | goto finish; | |
de5f3f40 KW |
2732 | } |
2733 | ||
bfe8043e | 2734 | if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { |
3ef6c40a HR |
2735 | error_setg(errp, "Lazy refcounts only supported with compatibility " |
2736 | "level 1.1 and above (use compat=1.1 or greater)"); | |
1bd0e2d1 CL |
2737 | ret = -EINVAL; |
2738 | goto finish; | |
bfe8043e SH |
2739 | } |
2740 | ||
06d05fa7 HR |
2741 | refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS, |
2742 | refcount_bits); | |
2743 | if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) { | |
2744 | error_setg(errp, "Refcount width must be a power of two and may not " | |
2745 | "exceed 64 bits"); | |
2746 | ret = -EINVAL; | |
2747 | goto finish; | |
2748 | } | |
2749 | ||
bd4b167f HR |
2750 | if (version < 3 && refcount_bits != 16) { |
2751 | error_setg(errp, "Different refcount widths than 16 bits require " | |
2752 | "compatibility level 1.1 or above (use compat=1.1 or " | |
2753 | "greater)"); | |
2754 | ret = -EINVAL; | |
2755 | goto finish; | |
2756 | } | |
2757 | ||
786a4ea8 | 2758 | refcount_order = ctz32(refcount_bits); |
bd4b167f | 2759 | |
180e9526 | 2760 | ret = qcow2_create2(filename, size, backing_file, backing_fmt, flags, |
bd4b167f | 2761 | cluster_size, prealloc, opts, version, refcount_order, |
0cb8d47b | 2762 | encryptfmt, &local_err); |
621ff94d | 2763 | error_propagate(errp, local_err); |
1bd0e2d1 CL |
2764 | |
2765 | finish: | |
2766 | g_free(backing_file); | |
2767 | g_free(backing_fmt); | |
2768 | g_free(buf); | |
3ef6c40a | 2769 | return ret; |
de5f3f40 KW |
2770 | } |
2771 | ||
2928abce | 2772 | |
ebb718a5 EB |
2773 | static bool is_zero_sectors(BlockDriverState *bs, int64_t start, |
2774 | uint32_t count) | |
2928abce | 2775 | { |
2928abce DL |
2776 | int nr; |
2777 | BlockDriverState *file; | |
ebb718a5 | 2778 | int64_t res; |
2928abce | 2779 | |
fbaa6bb3 EB |
2780 | if (start + count > bs->total_sectors) { |
2781 | count = bs->total_sectors - start; | |
2782 | } | |
2783 | ||
ebb718a5 EB |
2784 | if (!count) { |
2785 | return true; | |
2786 | } | |
2787 | res = bdrv_get_block_status_above(bs, NULL, start, count, | |
2788 | &nr, &file); | |
2789 | return res >= 0 && (res & BDRV_BLOCK_ZERO) && nr == count; | |
2928abce DL |
2790 | } |
2791 | ||
5544b59f | 2792 | static coroutine_fn int qcow2_co_pwrite_zeroes(BlockDriverState *bs, |
f5a5ca79 | 2793 | int64_t offset, int bytes, BdrvRequestFlags flags) |
621f0589 KW |
2794 | { |
2795 | int ret; | |
ff99129a | 2796 | BDRVQcow2State *s = bs->opaque; |
621f0589 | 2797 | |
5544b59f | 2798 | uint32_t head = offset % s->cluster_size; |
f5a5ca79 | 2799 | uint32_t tail = (offset + bytes) % s->cluster_size; |
2928abce | 2800 | |
f5a5ca79 MP |
2801 | trace_qcow2_pwrite_zeroes_start_req(qemu_coroutine_self(), offset, bytes); |
2802 | if (offset + bytes == bs->total_sectors * BDRV_SECTOR_SIZE) { | |
fbaa6bb3 EB |
2803 | tail = 0; |
2804 | } | |
5a64e942 | 2805 | |
ebb718a5 | 2806 | if (head || tail) { |
5544b59f | 2807 | int64_t cl_start = (offset - head) >> BDRV_SECTOR_BITS; |
ebb718a5 | 2808 | uint64_t off; |
ecfe1863 | 2809 | unsigned int nr; |
2928abce | 2810 | |
f5a5ca79 | 2811 | assert(head + bytes <= s->cluster_size); |
2928abce | 2812 | |
ebb718a5 | 2813 | /* check whether remainder of cluster already reads as zero */ |
5544b59f EB |
2814 | if (!(is_zero_sectors(bs, cl_start, |
2815 | DIV_ROUND_UP(head, BDRV_SECTOR_SIZE)) && | |
f5a5ca79 | 2816 | is_zero_sectors(bs, (offset + bytes) >> BDRV_SECTOR_BITS, |
5544b59f EB |
2817 | DIV_ROUND_UP(-tail & (s->cluster_size - 1), |
2818 | BDRV_SECTOR_SIZE)))) { | |
2928abce DL |
2819 | return -ENOTSUP; |
2820 | } | |
2821 | ||
2928abce DL |
2822 | qemu_co_mutex_lock(&s->lock); |
2823 | /* We can have new write after previous check */ | |
5544b59f | 2824 | offset = cl_start << BDRV_SECTOR_BITS; |
f5a5ca79 | 2825 | bytes = s->cluster_size; |
ecfe1863 | 2826 | nr = s->cluster_size; |
5544b59f | 2827 | ret = qcow2_get_cluster_offset(bs, offset, &nr, &off); |
fdfab37d EB |
2828 | if (ret != QCOW2_CLUSTER_UNALLOCATED && |
2829 | ret != QCOW2_CLUSTER_ZERO_PLAIN && | |
2830 | ret != QCOW2_CLUSTER_ZERO_ALLOC) { | |
2928abce DL |
2831 | qemu_co_mutex_unlock(&s->lock); |
2832 | return -ENOTSUP; | |
2833 | } | |
2834 | } else { | |
2835 | qemu_co_mutex_lock(&s->lock); | |
621f0589 KW |
2836 | } |
2837 | ||
f5a5ca79 | 2838 | trace_qcow2_pwrite_zeroes(qemu_coroutine_self(), offset, bytes); |
5a64e942 | 2839 | |
621f0589 | 2840 | /* Whatever is left can use real zero clusters */ |
f5a5ca79 | 2841 | ret = qcow2_cluster_zeroize(bs, offset, bytes, flags); |
621f0589 KW |
2842 | qemu_co_mutex_unlock(&s->lock); |
2843 | ||
2844 | return ret; | |
2845 | } | |
2846 | ||
82e8a788 | 2847 | static coroutine_fn int qcow2_co_pdiscard(BlockDriverState *bs, |
f5a5ca79 | 2848 | int64_t offset, int bytes) |
5ea929e3 | 2849 | { |
6db39ae2 | 2850 | int ret; |
ff99129a | 2851 | BDRVQcow2State *s = bs->opaque; |
6db39ae2 | 2852 | |
f5a5ca79 MP |
2853 | if (!QEMU_IS_ALIGNED(offset | bytes, s->cluster_size)) { |
2854 | assert(bytes < s->cluster_size); | |
048c5fd1 EB |
2855 | /* Ignore partial clusters, except for the special case of the |
2856 | * complete partial cluster at the end of an unaligned file */ | |
2857 | if (!QEMU_IS_ALIGNED(offset, s->cluster_size) || | |
f5a5ca79 | 2858 | offset + bytes != bs->total_sectors * BDRV_SECTOR_SIZE) { |
048c5fd1 EB |
2859 | return -ENOTSUP; |
2860 | } | |
49228d1e EB |
2861 | } |
2862 | ||
6db39ae2 | 2863 | qemu_co_mutex_lock(&s->lock); |
f5a5ca79 | 2864 | ret = qcow2_cluster_discard(bs, offset, bytes, QCOW2_DISCARD_REQUEST, |
d2cb36af | 2865 | false); |
6db39ae2 PB |
2866 | qemu_co_mutex_unlock(&s->lock); |
2867 | return ret; | |
5ea929e3 KW |
2868 | } |
2869 | ||
4bff28b8 | 2870 | static int qcow2_truncate(BlockDriverState *bs, int64_t offset, Error **errp) |
419b19d9 | 2871 | { |
ff99129a | 2872 | BDRVQcow2State *s = bs->opaque; |
2cf7cfa1 KW |
2873 | int64_t new_l1_size; |
2874 | int ret; | |
419b19d9 SH |
2875 | |
2876 | if (offset & 511) { | |
4bff28b8 | 2877 | error_setg(errp, "The new size must be a multiple of 512"); |
419b19d9 SH |
2878 | return -EINVAL; |
2879 | } | |
2880 | ||
2881 | /* cannot proceed if image has snapshots */ | |
2882 | if (s->nb_snapshots) { | |
4bff28b8 | 2883 | error_setg(errp, "Can't resize an image which has snapshots"); |
419b19d9 SH |
2884 | return -ENOTSUP; |
2885 | } | |
2886 | ||
2887 | /* shrinking is currently not supported */ | |
2888 | if (offset < bs->total_sectors * 512) { | |
4bff28b8 | 2889 | error_setg(errp, "qcow2 doesn't support shrinking images yet"); |
419b19d9 SH |
2890 | return -ENOTSUP; |
2891 | } | |
2892 | ||
2893 | new_l1_size = size_to_l1(s, offset); | |
72893756 | 2894 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); |
419b19d9 | 2895 | if (ret < 0) { |
f59adb32 | 2896 | error_setg_errno(errp, -ret, "Failed to grow the L1 table"); |
419b19d9 SH |
2897 | return ret; |
2898 | } | |
2899 | ||
2900 | /* write updated header.size */ | |
2901 | offset = cpu_to_be64(offset); | |
d9ca2ea2 | 2902 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), |
8b3b7206 | 2903 | &offset, sizeof(uint64_t)); |
419b19d9 | 2904 | if (ret < 0) { |
f59adb32 | 2905 | error_setg_errno(errp, -ret, "Failed to update the image size"); |
419b19d9 SH |
2906 | return ret; |
2907 | } | |
2908 | ||
2909 | s->l1_vm_state_index = new_l1_size; | |
2910 | return 0; | |
2911 | } | |
2912 | ||
20d97356 BS |
2913 | /* XXX: put compressed sectors first, then all the cluster aligned |
2914 | tables to avoid losing bytes in alignment */ | |
fcccefc5 PB |
2915 | static coroutine_fn int |
2916 | qcow2_co_pwritev_compressed(BlockDriverState *bs, uint64_t offset, | |
2917 | uint64_t bytes, QEMUIOVector *qiov) | |
20d97356 | 2918 | { |
ff99129a | 2919 | BDRVQcow2State *s = bs->opaque; |
fcccefc5 PB |
2920 | QEMUIOVector hd_qiov; |
2921 | struct iovec iov; | |
20d97356 BS |
2922 | z_stream strm; |
2923 | int ret, out_len; | |
fcccefc5 | 2924 | uint8_t *buf, *out_buf; |
20d97356 BS |
2925 | uint64_t cluster_offset; |
2926 | ||
fcccefc5 | 2927 | if (bytes == 0) { |
20d97356 BS |
2928 | /* align end of file to a sector boundary to ease reading with |
2929 | sector based I/Os */ | |
9a4f4c31 | 2930 | cluster_offset = bdrv_getlength(bs->file->bs); |
ed3d2ec9 | 2931 | return bdrv_truncate(bs->file, cluster_offset, NULL); |
20d97356 BS |
2932 | } |
2933 | ||
a2c0ca6f | 2934 | buf = qemu_blockalign(bs, s->cluster_size); |
fcccefc5 | 2935 | if (bytes != s->cluster_size) { |
a2c0ca6f PB |
2936 | if (bytes > s->cluster_size || |
2937 | offset + bytes != bs->total_sectors << BDRV_SECTOR_BITS) | |
fcccefc5 | 2938 | { |
a2c0ca6f PB |
2939 | qemu_vfree(buf); |
2940 | return -EINVAL; | |
f4d38bef | 2941 | } |
a2c0ca6f PB |
2942 | /* Zero-pad last write if image size is not cluster aligned */ |
2943 | memset(buf + bytes, 0, s->cluster_size - bytes); | |
f4d38bef | 2944 | } |
8b2bd093 | 2945 | qemu_iovec_to_buf(qiov, 0, buf, bytes); |
20d97356 | 2946 | |
ebf7bba0 | 2947 | out_buf = g_malloc(s->cluster_size); |
20d97356 BS |
2948 | |
2949 | /* best compression, small window, no zlib header */ | |
2950 | memset(&strm, 0, sizeof(strm)); | |
2951 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
2952 | Z_DEFLATED, -12, | |
2953 | 9, Z_DEFAULT_STRATEGY); | |
2954 | if (ret != 0) { | |
8f1efd00 KW |
2955 | ret = -EINVAL; |
2956 | goto fail; | |
20d97356 BS |
2957 | } |
2958 | ||
2959 | strm.avail_in = s->cluster_size; | |
2960 | strm.next_in = (uint8_t *)buf; | |
2961 | strm.avail_out = s->cluster_size; | |
2962 | strm.next_out = out_buf; | |
2963 | ||
2964 | ret = deflate(&strm, Z_FINISH); | |
2965 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
20d97356 | 2966 | deflateEnd(&strm); |
8f1efd00 KW |
2967 | ret = -EINVAL; |
2968 | goto fail; | |
20d97356 BS |
2969 | } |
2970 | out_len = strm.next_out - out_buf; | |
2971 | ||
2972 | deflateEnd(&strm); | |
2973 | ||
2974 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
2975 | /* could not compress: write normal cluster */ | |
fcccefc5 | 2976 | ret = qcow2_co_pwritev(bs, offset, bytes, qiov, 0); |
8f1efd00 KW |
2977 | if (ret < 0) { |
2978 | goto fail; | |
2979 | } | |
fcccefc5 PB |
2980 | goto success; |
2981 | } | |
cf93980e | 2982 | |
fcccefc5 PB |
2983 | qemu_co_mutex_lock(&s->lock); |
2984 | cluster_offset = | |
2985 | qcow2_alloc_compressed_cluster_offset(bs, offset, out_len); | |
2986 | if (!cluster_offset) { | |
2987 | qemu_co_mutex_unlock(&s->lock); | |
2988 | ret = -EIO; | |
2989 | goto fail; | |
2990 | } | |
2991 | cluster_offset &= s->cluster_offset_mask; | |
cf93980e | 2992 | |
fcccefc5 PB |
2993 | ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len); |
2994 | qemu_co_mutex_unlock(&s->lock); | |
2995 | if (ret < 0) { | |
2996 | goto fail; | |
20d97356 BS |
2997 | } |
2998 | ||
fcccefc5 PB |
2999 | iov = (struct iovec) { |
3000 | .iov_base = out_buf, | |
3001 | .iov_len = out_len, | |
3002 | }; | |
3003 | qemu_iovec_init_external(&hd_qiov, &iov, 1); | |
3004 | ||
3005 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); | |
3006 | ret = bdrv_co_pwritev(bs->file, cluster_offset, out_len, &hd_qiov, 0); | |
3007 | if (ret < 0) { | |
3008 | goto fail; | |
3009 | } | |
3010 | success: | |
8f1efd00 KW |
3011 | ret = 0; |
3012 | fail: | |
fcccefc5 | 3013 | qemu_vfree(buf); |
7267c094 | 3014 | g_free(out_buf); |
8f1efd00 | 3015 | return ret; |
20d97356 BS |
3016 | } |
3017 | ||
94054183 HR |
3018 | static int make_completely_empty(BlockDriverState *bs) |
3019 | { | |
ff99129a | 3020 | BDRVQcow2State *s = bs->opaque; |
ed3d2ec9 | 3021 | Error *local_err = NULL; |
94054183 HR |
3022 | int ret, l1_clusters; |
3023 | int64_t offset; | |
3024 | uint64_t *new_reftable = NULL; | |
3025 | uint64_t rt_entry, l1_size2; | |
3026 | struct { | |
3027 | uint64_t l1_offset; | |
3028 | uint64_t reftable_offset; | |
3029 | uint32_t reftable_clusters; | |
3030 | } QEMU_PACKED l1_ofs_rt_ofs_cls; | |
3031 | ||
3032 | ret = qcow2_cache_empty(bs, s->l2_table_cache); | |
3033 | if (ret < 0) { | |
3034 | goto fail; | |
3035 | } | |
3036 | ||
3037 | ret = qcow2_cache_empty(bs, s->refcount_block_cache); | |
3038 | if (ret < 0) { | |
3039 | goto fail; | |
3040 | } | |
3041 | ||
3042 | /* Refcounts will be broken utterly */ | |
3043 | ret = qcow2_mark_dirty(bs); | |
3044 | if (ret < 0) { | |
3045 | goto fail; | |
3046 | } | |
3047 | ||
3048 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
3049 | ||
3050 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
3051 | l1_size2 = (uint64_t)s->l1_size * sizeof(uint64_t); | |
3052 | ||
3053 | /* After this call, neither the in-memory nor the on-disk refcount | |
3054 | * information accurately describe the actual references */ | |
3055 | ||
720ff280 | 3056 | ret = bdrv_pwrite_zeroes(bs->file, s->l1_table_offset, |
74021bc4 | 3057 | l1_clusters * s->cluster_size, 0); |
94054183 HR |
3058 | if (ret < 0) { |
3059 | goto fail_broken_refcounts; | |
3060 | } | |
3061 | memset(s->l1_table, 0, l1_size2); | |
3062 | ||
3063 | BLKDBG_EVENT(bs->file, BLKDBG_EMPTY_IMAGE_PREPARE); | |
3064 | ||
3065 | /* Overwrite enough clusters at the beginning of the sectors to place | |
3066 | * the refcount table, a refcount block and the L1 table in; this may | |
3067 | * overwrite parts of the existing refcount and L1 table, which is not | |
3068 | * an issue because the dirty flag is set, complete data loss is in fact | |
3069 | * desired and partial data loss is consequently fine as well */ | |
720ff280 | 3070 | ret = bdrv_pwrite_zeroes(bs->file, s->cluster_size, |
74021bc4 | 3071 | (2 + l1_clusters) * s->cluster_size, 0); |
94054183 HR |
3072 | /* This call (even if it failed overall) may have overwritten on-disk |
3073 | * refcount structures; in that case, the in-memory refcount information | |
3074 | * will probably differ from the on-disk information which makes the BDS | |
3075 | * unusable */ | |
3076 | if (ret < 0) { | |
3077 | goto fail_broken_refcounts; | |
3078 | } | |
3079 | ||
3080 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
3081 | BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE); | |
3082 | ||
3083 | /* "Create" an empty reftable (one cluster) directly after the image | |
3084 | * header and an empty L1 table three clusters after the image header; | |
3085 | * the cluster between those two will be used as the first refblock */ | |
f1f7a1dd PM |
3086 | l1_ofs_rt_ofs_cls.l1_offset = cpu_to_be64(3 * s->cluster_size); |
3087 | l1_ofs_rt_ofs_cls.reftable_offset = cpu_to_be64(s->cluster_size); | |
3088 | l1_ofs_rt_ofs_cls.reftable_clusters = cpu_to_be32(1); | |
d9ca2ea2 | 3089 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, l1_table_offset), |
94054183 HR |
3090 | &l1_ofs_rt_ofs_cls, sizeof(l1_ofs_rt_ofs_cls)); |
3091 | if (ret < 0) { | |
3092 | goto fail_broken_refcounts; | |
3093 | } | |
3094 | ||
3095 | s->l1_table_offset = 3 * s->cluster_size; | |
3096 | ||
3097 | new_reftable = g_try_new0(uint64_t, s->cluster_size / sizeof(uint64_t)); | |
3098 | if (!new_reftable) { | |
3099 | ret = -ENOMEM; | |
3100 | goto fail_broken_refcounts; | |
3101 | } | |
3102 | ||
3103 | s->refcount_table_offset = s->cluster_size; | |
3104 | s->refcount_table_size = s->cluster_size / sizeof(uint64_t); | |
7061a078 | 3105 | s->max_refcount_table_index = 0; |
94054183 HR |
3106 | |
3107 | g_free(s->refcount_table); | |
3108 | s->refcount_table = new_reftable; | |
3109 | new_reftable = NULL; | |
3110 | ||
3111 | /* Now the in-memory refcount information again corresponds to the on-disk | |
3112 | * information (reftable is empty and no refblocks (the refblock cache is | |
3113 | * empty)); however, this means some clusters (e.g. the image header) are | |
3114 | * referenced, but not refcounted, but the normal qcow2 code assumes that | |
3115 | * the in-memory information is always correct */ | |
3116 | ||
3117 | BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC); | |
3118 | ||
3119 | /* Enter the first refblock into the reftable */ | |
3120 | rt_entry = cpu_to_be64(2 * s->cluster_size); | |
d9ca2ea2 | 3121 | ret = bdrv_pwrite_sync(bs->file, s->cluster_size, |
94054183 HR |
3122 | &rt_entry, sizeof(rt_entry)); |
3123 | if (ret < 0) { | |
3124 | goto fail_broken_refcounts; | |
3125 | } | |
3126 | s->refcount_table[0] = 2 * s->cluster_size; | |
3127 | ||
3128 | s->free_cluster_index = 0; | |
3129 | assert(3 + l1_clusters <= s->refcount_block_size); | |
3130 | offset = qcow2_alloc_clusters(bs, 3 * s->cluster_size + l1_size2); | |
3131 | if (offset < 0) { | |
3132 | ret = offset; | |
3133 | goto fail_broken_refcounts; | |
3134 | } else if (offset > 0) { | |
3135 | error_report("First cluster in emptied image is in use"); | |
3136 | abort(); | |
3137 | } | |
3138 | ||
3139 | /* Now finally the in-memory information corresponds to the on-disk | |
3140 | * structures and is correct */ | |
3141 | ret = qcow2_mark_clean(bs); | |
3142 | if (ret < 0) { | |
3143 | goto fail; | |
3144 | } | |
3145 | ||
ed3d2ec9 HR |
3146 | ret = bdrv_truncate(bs->file, (3 + l1_clusters) * s->cluster_size, |
3147 | &local_err); | |
94054183 | 3148 | if (ret < 0) { |
ed3d2ec9 | 3149 | error_report_err(local_err); |
94054183 HR |
3150 | goto fail; |
3151 | } | |
3152 | ||
3153 | return 0; | |
3154 | ||
3155 | fail_broken_refcounts: | |
3156 | /* The BDS is unusable at this point. If we wanted to make it usable, we | |
3157 | * would have to call qcow2_refcount_close(), qcow2_refcount_init(), | |
3158 | * qcow2_check_refcounts(), qcow2_refcount_close() and qcow2_refcount_init() | |
3159 | * again. However, because the functions which could have caused this error | |
3160 | * path to be taken are used by those functions as well, it's very likely | |
3161 | * that that sequence will fail as well. Therefore, just eject the BDS. */ | |
3162 | bs->drv = NULL; | |
3163 | ||
3164 | fail: | |
3165 | g_free(new_reftable); | |
3166 | return ret; | |
3167 | } | |
3168 | ||
491d27e2 HR |
3169 | static int qcow2_make_empty(BlockDriverState *bs) |
3170 | { | |
ff99129a | 3171 | BDRVQcow2State *s = bs->opaque; |
d2cb36af EB |
3172 | uint64_t offset, end_offset; |
3173 | int step = QEMU_ALIGN_DOWN(INT_MAX, s->cluster_size); | |
94054183 HR |
3174 | int l1_clusters, ret = 0; |
3175 | ||
3176 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
3177 | ||
3178 | if (s->qcow_version >= 3 && !s->snapshots && | |
3179 | 3 + l1_clusters <= s->refcount_block_size) { | |
3180 | /* The following function only works for qcow2 v3 images (it requires | |
3181 | * the dirty flag) and only as long as there are no snapshots (because | |
3182 | * it completely empties the image). Furthermore, the L1 table and three | |
3183 | * additional clusters (image header, refcount table, one refcount | |
3184 | * block) have to fit inside one refcount block. */ | |
3185 | return make_completely_empty(bs); | |
3186 | } | |
491d27e2 | 3187 | |
94054183 HR |
3188 | /* This fallback code simply discards every active cluster; this is slow, |
3189 | * but works in all cases */ | |
d2cb36af EB |
3190 | end_offset = bs->total_sectors * BDRV_SECTOR_SIZE; |
3191 | for (offset = 0; offset < end_offset; offset += step) { | |
491d27e2 HR |
3192 | /* As this function is generally used after committing an external |
3193 | * snapshot, QCOW2_DISCARD_SNAPSHOT seems appropriate. Also, the | |
3194 | * default action for this kind of discard is to pass the discard, | |
3195 | * which will ideally result in an actually smaller image file, as | |
3196 | * is probably desired. */ | |
d2cb36af EB |
3197 | ret = qcow2_cluster_discard(bs, offset, MIN(step, end_offset - offset), |
3198 | QCOW2_DISCARD_SNAPSHOT, true); | |
491d27e2 HR |
3199 | if (ret < 0) { |
3200 | break; | |
3201 | } | |
3202 | } | |
3203 | ||
3204 | return ret; | |
3205 | } | |
3206 | ||
a968168c | 3207 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
20d97356 | 3208 | { |
ff99129a | 3209 | BDRVQcow2State *s = bs->opaque; |
29c1a730 KW |
3210 | int ret; |
3211 | ||
8b94ff85 | 3212 | qemu_co_mutex_lock(&s->lock); |
f3c3b87d | 3213 | ret = qcow2_cache_write(bs, s->l2_table_cache); |
29c1a730 | 3214 | if (ret < 0) { |
c95de7e2 | 3215 | qemu_co_mutex_unlock(&s->lock); |
8b94ff85 | 3216 | return ret; |
29c1a730 KW |
3217 | } |
3218 | ||
bfe8043e | 3219 | if (qcow2_need_accurate_refcounts(s)) { |
f3c3b87d | 3220 | ret = qcow2_cache_write(bs, s->refcount_block_cache); |
bfe8043e SH |
3221 | if (ret < 0) { |
3222 | qemu_co_mutex_unlock(&s->lock); | |
3223 | return ret; | |
3224 | } | |
29c1a730 | 3225 | } |
8b94ff85 | 3226 | qemu_co_mutex_unlock(&s->lock); |
29c1a730 | 3227 | |
eb489bb1 KW |
3228 | return 0; |
3229 | } | |
3230 | ||
7c80ab3f | 3231 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
20d97356 | 3232 | { |
ff99129a | 3233 | BDRVQcow2State *s = bs->opaque; |
95de6d70 PB |
3234 | bdi->unallocated_blocks_are_zero = true; |
3235 | bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3); | |
20d97356 | 3236 | bdi->cluster_size = s->cluster_size; |
7c80ab3f | 3237 | bdi->vm_state_offset = qcow2_vm_state_offset(s); |
20d97356 BS |
3238 | return 0; |
3239 | } | |
3240 | ||
37764dfb HR |
3241 | static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs) |
3242 | { | |
ff99129a | 3243 | BDRVQcow2State *s = bs->opaque; |
37764dfb HR |
3244 | ImageInfoSpecific *spec_info = g_new(ImageInfoSpecific, 1); |
3245 | ||
3246 | *spec_info = (ImageInfoSpecific){ | |
6a8f9661 | 3247 | .type = IMAGE_INFO_SPECIFIC_KIND_QCOW2, |
32bafa8f | 3248 | .u.qcow2.data = g_new(ImageInfoSpecificQCow2, 1), |
37764dfb HR |
3249 | }; |
3250 | if (s->qcow_version == 2) { | |
32bafa8f | 3251 | *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){ |
0709c5a1 HR |
3252 | .compat = g_strdup("0.10"), |
3253 | .refcount_bits = s->refcount_bits, | |
37764dfb HR |
3254 | }; |
3255 | } else if (s->qcow_version == 3) { | |
32bafa8f | 3256 | *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){ |
37764dfb HR |
3257 | .compat = g_strdup("1.1"), |
3258 | .lazy_refcounts = s->compatible_features & | |
3259 | QCOW2_COMPAT_LAZY_REFCOUNTS, | |
3260 | .has_lazy_refcounts = true, | |
9009b196 HR |
3261 | .corrupt = s->incompatible_features & |
3262 | QCOW2_INCOMPAT_CORRUPT, | |
3263 | .has_corrupt = true, | |
0709c5a1 | 3264 | .refcount_bits = s->refcount_bits, |
37764dfb | 3265 | }; |
b1fc8f93 DL |
3266 | } else { |
3267 | /* if this assertion fails, this probably means a new version was | |
3268 | * added without having it covered here */ | |
3269 | assert(false); | |
37764dfb HR |
3270 | } |
3271 | ||
3272 | return spec_info; | |
3273 | } | |
3274 | ||
20d97356 BS |
3275 | #if 0 |
3276 | static void dump_refcounts(BlockDriverState *bs) | |
3277 | { | |
ff99129a | 3278 | BDRVQcow2State *s = bs->opaque; |
20d97356 BS |
3279 | int64_t nb_clusters, k, k1, size; |
3280 | int refcount; | |
3281 | ||
9a4f4c31 | 3282 | size = bdrv_getlength(bs->file->bs); |
20d97356 BS |
3283 | nb_clusters = size_to_clusters(s, size); |
3284 | for(k = 0; k < nb_clusters;) { | |
3285 | k1 = k; | |
3286 | refcount = get_refcount(bs, k); | |
3287 | k++; | |
3288 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
3289 | k++; | |
0bfcd599 BS |
3290 | printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount, |
3291 | k - k1); | |
20d97356 BS |
3292 | } |
3293 | } | |
3294 | #endif | |
3295 | ||
cf8074b3 KW |
3296 | static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
3297 | int64_t pos) | |
20d97356 | 3298 | { |
ff99129a | 3299 | BDRVQcow2State *s = bs->opaque; |
20d97356 | 3300 | |
66f82cee | 3301 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
734a7758 KW |
3302 | return bs->drv->bdrv_co_pwritev(bs, qcow2_vm_state_offset(s) + pos, |
3303 | qiov->size, qiov, 0); | |
20d97356 BS |
3304 | } |
3305 | ||
5ddda0b8 KW |
3306 | static int qcow2_load_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
3307 | int64_t pos) | |
20d97356 | 3308 | { |
ff99129a | 3309 | BDRVQcow2State *s = bs->opaque; |
20d97356 | 3310 | |
66f82cee | 3311 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
734a7758 KW |
3312 | return bs->drv->bdrv_co_preadv(bs, qcow2_vm_state_offset(s) + pos, |
3313 | qiov->size, qiov, 0); | |
20d97356 BS |
3314 | } |
3315 | ||
9296b3ed HR |
3316 | /* |
3317 | * Downgrades an image's version. To achieve this, any incompatible features | |
3318 | * have to be removed. | |
3319 | */ | |
4057a2b2 | 3320 | static int qcow2_downgrade(BlockDriverState *bs, int target_version, |
8b13976d | 3321 | BlockDriverAmendStatusCB *status_cb, void *cb_opaque) |
9296b3ed | 3322 | { |
ff99129a | 3323 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
3324 | int current_version = s->qcow_version; |
3325 | int ret; | |
3326 | ||
3327 | if (target_version == current_version) { | |
3328 | return 0; | |
3329 | } else if (target_version > current_version) { | |
3330 | return -EINVAL; | |
3331 | } else if (target_version != 2) { | |
3332 | return -EINVAL; | |
3333 | } | |
3334 | ||
3335 | if (s->refcount_order != 4) { | |
61ce55fc | 3336 | error_report("compat=0.10 requires refcount_bits=16"); |
9296b3ed HR |
3337 | return -ENOTSUP; |
3338 | } | |
3339 | ||
3340 | /* clear incompatible features */ | |
3341 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
3342 | ret = qcow2_mark_clean(bs); | |
3343 | if (ret < 0) { | |
3344 | return ret; | |
3345 | } | |
3346 | } | |
3347 | ||
3348 | /* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in | |
3349 | * the first place; if that happens nonetheless, returning -ENOTSUP is the | |
3350 | * best thing to do anyway */ | |
3351 | ||
3352 | if (s->incompatible_features) { | |
3353 | return -ENOTSUP; | |
3354 | } | |
3355 | ||
3356 | /* since we can ignore compatible features, we can set them to 0 as well */ | |
3357 | s->compatible_features = 0; | |
3358 | /* if lazy refcounts have been used, they have already been fixed through | |
3359 | * clearing the dirty flag */ | |
3360 | ||
3361 | /* clearing autoclear features is trivial */ | |
3362 | s->autoclear_features = 0; | |
3363 | ||
8b13976d | 3364 | ret = qcow2_expand_zero_clusters(bs, status_cb, cb_opaque); |
9296b3ed HR |
3365 | if (ret < 0) { |
3366 | return ret; | |
3367 | } | |
3368 | ||
3369 | s->qcow_version = target_version; | |
3370 | ret = qcow2_update_header(bs); | |
3371 | if (ret < 0) { | |
3372 | s->qcow_version = current_version; | |
3373 | return ret; | |
3374 | } | |
3375 | return 0; | |
3376 | } | |
3377 | ||
c293a809 HR |
3378 | typedef enum Qcow2AmendOperation { |
3379 | /* This is the value Qcow2AmendHelperCBInfo::last_operation will be | |
3380 | * statically initialized to so that the helper CB can discern the first | |
3381 | * invocation from an operation change */ | |
3382 | QCOW2_NO_OPERATION = 0, | |
3383 | ||
61ce55fc | 3384 | QCOW2_CHANGING_REFCOUNT_ORDER, |
c293a809 HR |
3385 | QCOW2_DOWNGRADING, |
3386 | } Qcow2AmendOperation; | |
3387 | ||
3388 | typedef struct Qcow2AmendHelperCBInfo { | |
3389 | /* The code coordinating the amend operations should only modify | |
3390 | * these four fields; the rest will be managed by the CB */ | |
3391 | BlockDriverAmendStatusCB *original_status_cb; | |
3392 | void *original_cb_opaque; | |
3393 | ||
3394 | Qcow2AmendOperation current_operation; | |
3395 | ||
3396 | /* Total number of operations to perform (only set once) */ | |
3397 | int total_operations; | |
3398 | ||
3399 | /* The following fields are managed by the CB */ | |
3400 | ||
3401 | /* Number of operations completed */ | |
3402 | int operations_completed; | |
3403 | ||
3404 | /* Cumulative offset of all completed operations */ | |
3405 | int64_t offset_completed; | |
3406 | ||
3407 | Qcow2AmendOperation last_operation; | |
3408 | int64_t last_work_size; | |
3409 | } Qcow2AmendHelperCBInfo; | |
3410 | ||
3411 | static void qcow2_amend_helper_cb(BlockDriverState *bs, | |
3412 | int64_t operation_offset, | |
3413 | int64_t operation_work_size, void *opaque) | |
3414 | { | |
3415 | Qcow2AmendHelperCBInfo *info = opaque; | |
3416 | int64_t current_work_size; | |
3417 | int64_t projected_work_size; | |
3418 | ||
3419 | if (info->current_operation != info->last_operation) { | |
3420 | if (info->last_operation != QCOW2_NO_OPERATION) { | |
3421 | info->offset_completed += info->last_work_size; | |
3422 | info->operations_completed++; | |
3423 | } | |
3424 | ||
3425 | info->last_operation = info->current_operation; | |
3426 | } | |
3427 | ||
3428 | assert(info->total_operations > 0); | |
3429 | assert(info->operations_completed < info->total_operations); | |
3430 | ||
3431 | info->last_work_size = operation_work_size; | |
3432 | ||
3433 | current_work_size = info->offset_completed + operation_work_size; | |
3434 | ||
3435 | /* current_work_size is the total work size for (operations_completed + 1) | |
3436 | * operations (which includes this one), so multiply it by the number of | |
3437 | * operations not covered and divide it by the number of operations | |
3438 | * covered to get a projection for the operations not covered */ | |
3439 | projected_work_size = current_work_size * (info->total_operations - | |
3440 | info->operations_completed - 1) | |
3441 | / (info->operations_completed + 1); | |
3442 | ||
3443 | info->original_status_cb(bs, info->offset_completed + operation_offset, | |
3444 | current_work_size + projected_work_size, | |
3445 | info->original_cb_opaque); | |
3446 | } | |
3447 | ||
77485434 | 3448 | static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts, |
8b13976d HR |
3449 | BlockDriverAmendStatusCB *status_cb, |
3450 | void *cb_opaque) | |
9296b3ed | 3451 | { |
ff99129a | 3452 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
3453 | int old_version = s->qcow_version, new_version = old_version; |
3454 | uint64_t new_size = 0; | |
3455 | const char *backing_file = NULL, *backing_format = NULL; | |
3456 | bool lazy_refcounts = s->use_lazy_refcounts; | |
1bd0e2d1 CL |
3457 | const char *compat = NULL; |
3458 | uint64_t cluster_size = s->cluster_size; | |
3459 | bool encrypt; | |
4652b8f3 | 3460 | int encformat; |
61ce55fc | 3461 | int refcount_bits = s->refcount_bits; |
d7086422 | 3462 | Error *local_err = NULL; |
9296b3ed | 3463 | int ret; |
1bd0e2d1 | 3464 | QemuOptDesc *desc = opts->list->desc; |
c293a809 | 3465 | Qcow2AmendHelperCBInfo helper_cb_info; |
9296b3ed | 3466 | |
1bd0e2d1 CL |
3467 | while (desc && desc->name) { |
3468 | if (!qemu_opt_find(opts, desc->name)) { | |
9296b3ed | 3469 | /* only change explicitly defined options */ |
1bd0e2d1 | 3470 | desc++; |
9296b3ed HR |
3471 | continue; |
3472 | } | |
3473 | ||
8a17b83c HR |
3474 | if (!strcmp(desc->name, BLOCK_OPT_COMPAT_LEVEL)) { |
3475 | compat = qemu_opt_get(opts, BLOCK_OPT_COMPAT_LEVEL); | |
1bd0e2d1 | 3476 | if (!compat) { |
9296b3ed | 3477 | /* preserve default */ |
1bd0e2d1 | 3478 | } else if (!strcmp(compat, "0.10")) { |
9296b3ed | 3479 | new_version = 2; |
1bd0e2d1 | 3480 | } else if (!strcmp(compat, "1.1")) { |
9296b3ed HR |
3481 | new_version = 3; |
3482 | } else { | |
29d72431 | 3483 | error_report("Unknown compatibility level %s", compat); |
9296b3ed HR |
3484 | return -EINVAL; |
3485 | } | |
8a17b83c | 3486 | } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) { |
29d72431 | 3487 | error_report("Cannot change preallocation mode"); |
9296b3ed | 3488 | return -ENOTSUP; |
8a17b83c HR |
3489 | } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) { |
3490 | new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); | |
3491 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FILE)) { | |
3492 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); | |
3493 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FMT)) { | |
3494 | backing_format = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); | |
3495 | } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT)) { | |
3496 | encrypt = qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, | |
b25b387f | 3497 | !!s->crypto); |
f6fa64f6 | 3498 | |
b25b387f | 3499 | if (encrypt != !!s->crypto) { |
29d72431 | 3500 | error_report("Changing the encryption flag is not supported"); |
9296b3ed HR |
3501 | return -ENOTSUP; |
3502 | } | |
4652b8f3 DB |
3503 | } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT_FORMAT)) { |
3504 | encformat = qcow2_crypt_method_from_format( | |
3505 | qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT)); | |
3506 | ||
3507 | if (encformat != s->crypt_method_header) { | |
3508 | error_report("Changing the encryption format is not supported"); | |
3509 | return -ENOTSUP; | |
3510 | } | |
8a17b83c HR |
3511 | } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) { |
3512 | cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, | |
1bd0e2d1 CL |
3513 | cluster_size); |
3514 | if (cluster_size != s->cluster_size) { | |
29d72431 | 3515 | error_report("Changing the cluster size is not supported"); |
9296b3ed HR |
3516 | return -ENOTSUP; |
3517 | } | |
8a17b83c HR |
3518 | } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) { |
3519 | lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS, | |
1bd0e2d1 | 3520 | lazy_refcounts); |
06d05fa7 | 3521 | } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) { |
61ce55fc HR |
3522 | refcount_bits = qemu_opt_get_number(opts, BLOCK_OPT_REFCOUNT_BITS, |
3523 | refcount_bits); | |
3524 | ||
3525 | if (refcount_bits <= 0 || refcount_bits > 64 || | |
3526 | !is_power_of_2(refcount_bits)) | |
3527 | { | |
3528 | error_report("Refcount width must be a power of two and may " | |
3529 | "not exceed 64 bits"); | |
3530 | return -EINVAL; | |
3531 | } | |
9296b3ed | 3532 | } else { |
164e0f89 | 3533 | /* if this point is reached, this probably means a new option was |
9296b3ed | 3534 | * added without having it covered here */ |
164e0f89 | 3535 | abort(); |
9296b3ed | 3536 | } |
1bd0e2d1 CL |
3537 | |
3538 | desc++; | |
9296b3ed HR |
3539 | } |
3540 | ||
c293a809 HR |
3541 | helper_cb_info = (Qcow2AmendHelperCBInfo){ |
3542 | .original_status_cb = status_cb, | |
3543 | .original_cb_opaque = cb_opaque, | |
3544 | .total_operations = (new_version < old_version) | |
61ce55fc | 3545 | + (s->refcount_bits != refcount_bits) |
c293a809 HR |
3546 | }; |
3547 | ||
1038bbb8 HR |
3548 | /* Upgrade first (some features may require compat=1.1) */ |
3549 | if (new_version > old_version) { | |
3550 | s->qcow_version = new_version; | |
3551 | ret = qcow2_update_header(bs); | |
3552 | if (ret < 0) { | |
3553 | s->qcow_version = old_version; | |
3554 | return ret; | |
9296b3ed HR |
3555 | } |
3556 | } | |
3557 | ||
61ce55fc HR |
3558 | if (s->refcount_bits != refcount_bits) { |
3559 | int refcount_order = ctz32(refcount_bits); | |
61ce55fc HR |
3560 | |
3561 | if (new_version < 3 && refcount_bits != 16) { | |
3562 | error_report("Different refcount widths than 16 bits require " | |
3563 | "compatibility level 1.1 or above (use compat=1.1 or " | |
3564 | "greater)"); | |
3565 | return -EINVAL; | |
3566 | } | |
3567 | ||
3568 | helper_cb_info.current_operation = QCOW2_CHANGING_REFCOUNT_ORDER; | |
3569 | ret = qcow2_change_refcount_order(bs, refcount_order, | |
3570 | &qcow2_amend_helper_cb, | |
a7a6a2bf | 3571 | &helper_cb_info, &local_err); |
61ce55fc | 3572 | if (ret < 0) { |
a7a6a2bf | 3573 | error_report_err(local_err); |
61ce55fc HR |
3574 | return ret; |
3575 | } | |
3576 | } | |
3577 | ||
9296b3ed | 3578 | if (backing_file || backing_format) { |
e4603fe1 KW |
3579 | ret = qcow2_change_backing_file(bs, |
3580 | backing_file ?: s->image_backing_file, | |
3581 | backing_format ?: s->image_backing_format); | |
9296b3ed HR |
3582 | if (ret < 0) { |
3583 | return ret; | |
3584 | } | |
3585 | } | |
3586 | ||
3587 | if (s->use_lazy_refcounts != lazy_refcounts) { | |
3588 | if (lazy_refcounts) { | |
1038bbb8 | 3589 | if (new_version < 3) { |
29d72431 HR |
3590 | error_report("Lazy refcounts only supported with compatibility " |
3591 | "level 1.1 and above (use compat=1.1 or greater)"); | |
9296b3ed HR |
3592 | return -EINVAL; |
3593 | } | |
3594 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3595 | ret = qcow2_update_header(bs); | |
3596 | if (ret < 0) { | |
3597 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3598 | return ret; | |
3599 | } | |
3600 | s->use_lazy_refcounts = true; | |
3601 | } else { | |
3602 | /* make image clean first */ | |
3603 | ret = qcow2_mark_clean(bs); | |
3604 | if (ret < 0) { | |
3605 | return ret; | |
3606 | } | |
3607 | /* now disallow lazy refcounts */ | |
3608 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3609 | ret = qcow2_update_header(bs); | |
3610 | if (ret < 0) { | |
3611 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3612 | return ret; | |
3613 | } | |
3614 | s->use_lazy_refcounts = false; | |
3615 | } | |
3616 | } | |
3617 | ||
3618 | if (new_size) { | |
6d0eb64d | 3619 | BlockBackend *blk = blk_new(BLK_PERM_RESIZE, BLK_PERM_ALL); |
d7086422 KW |
3620 | ret = blk_insert_bs(blk, bs, &local_err); |
3621 | if (ret < 0) { | |
3622 | error_report_err(local_err); | |
3623 | blk_unref(blk); | |
3624 | return ret; | |
3625 | } | |
3626 | ||
ed3d2ec9 | 3627 | ret = blk_truncate(blk, new_size, &local_err); |
70b27f36 | 3628 | blk_unref(blk); |
9296b3ed | 3629 | if (ret < 0) { |
ed3d2ec9 | 3630 | error_report_err(local_err); |
9296b3ed HR |
3631 | return ret; |
3632 | } | |
3633 | } | |
3634 | ||
1038bbb8 HR |
3635 | /* Downgrade last (so unsupported features can be removed before) */ |
3636 | if (new_version < old_version) { | |
c293a809 HR |
3637 | helper_cb_info.current_operation = QCOW2_DOWNGRADING; |
3638 | ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb, | |
3639 | &helper_cb_info); | |
1038bbb8 HR |
3640 | if (ret < 0) { |
3641 | return ret; | |
3642 | } | |
3643 | } | |
3644 | ||
9296b3ed HR |
3645 | return 0; |
3646 | } | |
3647 | ||
85186ebd HR |
3648 | /* |
3649 | * If offset or size are negative, respectively, they will not be included in | |
3650 | * the BLOCK_IMAGE_CORRUPTED event emitted. | |
3651 | * fatal will be ignored for read-only BDS; corruptions found there will always | |
3652 | * be considered non-fatal. | |
3653 | */ | |
3654 | void qcow2_signal_corruption(BlockDriverState *bs, bool fatal, int64_t offset, | |
3655 | int64_t size, const char *message_format, ...) | |
3656 | { | |
ff99129a | 3657 | BDRVQcow2State *s = bs->opaque; |
dc881b44 | 3658 | const char *node_name; |
85186ebd HR |
3659 | char *message; |
3660 | va_list ap; | |
3661 | ||
3662 | fatal = fatal && !bs->read_only; | |
3663 | ||
3664 | if (s->signaled_corruption && | |
3665 | (!fatal || (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT))) | |
3666 | { | |
3667 | return; | |
3668 | } | |
3669 | ||
3670 | va_start(ap, message_format); | |
3671 | message = g_strdup_vprintf(message_format, ap); | |
3672 | va_end(ap); | |
3673 | ||
3674 | if (fatal) { | |
3675 | fprintf(stderr, "qcow2: Marking image as corrupt: %s; further " | |
3676 | "corruption events will be suppressed\n", message); | |
3677 | } else { | |
3678 | fprintf(stderr, "qcow2: Image is corrupt: %s; further non-fatal " | |
3679 | "corruption events will be suppressed\n", message); | |
3680 | } | |
3681 | ||
dc881b44 AG |
3682 | node_name = bdrv_get_node_name(bs); |
3683 | qapi_event_send_block_image_corrupted(bdrv_get_device_name(bs), | |
3684 | *node_name != '\0', node_name, | |
3685 | message, offset >= 0, offset, | |
3686 | size >= 0, size, | |
85186ebd HR |
3687 | fatal, &error_abort); |
3688 | g_free(message); | |
3689 | ||
3690 | if (fatal) { | |
3691 | qcow2_mark_corrupt(bs); | |
3692 | bs->drv = NULL; /* make BDS unusable */ | |
3693 | } | |
3694 | ||
3695 | s->signaled_corruption = true; | |
3696 | } | |
3697 | ||
1bd0e2d1 CL |
3698 | static QemuOptsList qcow2_create_opts = { |
3699 | .name = "qcow2-create-opts", | |
3700 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_create_opts.head), | |
3701 | .desc = { | |
3702 | { | |
3703 | .name = BLOCK_OPT_SIZE, | |
3704 | .type = QEMU_OPT_SIZE, | |
3705 | .help = "Virtual disk size" | |
3706 | }, | |
3707 | { | |
3708 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
3709 | .type = QEMU_OPT_STRING, | |
3710 | .help = "Compatibility level (0.10 or 1.1)" | |
3711 | }, | |
3712 | { | |
3713 | .name = BLOCK_OPT_BACKING_FILE, | |
3714 | .type = QEMU_OPT_STRING, | |
3715 | .help = "File name of a base image" | |
3716 | }, | |
3717 | { | |
3718 | .name = BLOCK_OPT_BACKING_FMT, | |
3719 | .type = QEMU_OPT_STRING, | |
3720 | .help = "Image format of the base image" | |
3721 | }, | |
3722 | { | |
3723 | .name = BLOCK_OPT_ENCRYPT, | |
3724 | .type = QEMU_OPT_BOOL, | |
0cb8d47b DB |
3725 | .help = "Encrypt the image with format 'aes'. (Deprecated " |
3726 | "in favor of " BLOCK_OPT_ENCRYPT_FORMAT "=aes)", | |
3727 | }, | |
3728 | { | |
3729 | .name = BLOCK_OPT_ENCRYPT_FORMAT, | |
3730 | .type = QEMU_OPT_STRING, | |
4652b8f3 | 3731 | .help = "Encrypt the image, format choices: 'aes', 'luks'", |
1bd0e2d1 | 3732 | }, |
4652b8f3 DB |
3733 | BLOCK_CRYPTO_OPT_DEF_KEY_SECRET("encrypt.", |
3734 | "ID of secret providing qcow AES key or LUKS passphrase"), | |
3735 | BLOCK_CRYPTO_OPT_DEF_LUKS_CIPHER_ALG("encrypt."), | |
3736 | BLOCK_CRYPTO_OPT_DEF_LUKS_CIPHER_MODE("encrypt."), | |
3737 | BLOCK_CRYPTO_OPT_DEF_LUKS_IVGEN_ALG("encrypt."), | |
3738 | BLOCK_CRYPTO_OPT_DEF_LUKS_IVGEN_HASH_ALG("encrypt."), | |
3739 | BLOCK_CRYPTO_OPT_DEF_LUKS_HASH_ALG("encrypt."), | |
3740 | BLOCK_CRYPTO_OPT_DEF_LUKS_ITER_TIME("encrypt."), | |
1bd0e2d1 CL |
3741 | { |
3742 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
3743 | .type = QEMU_OPT_SIZE, | |
3744 | .help = "qcow2 cluster size", | |
3745 | .def_value_str = stringify(DEFAULT_CLUSTER_SIZE) | |
3746 | }, | |
3747 | { | |
3748 | .name = BLOCK_OPT_PREALLOC, | |
3749 | .type = QEMU_OPT_STRING, | |
0e4271b7 HT |
3750 | .help = "Preallocation mode (allowed values: off, metadata, " |
3751 | "falloc, full)" | |
1bd0e2d1 CL |
3752 | }, |
3753 | { | |
3754 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
3755 | .type = QEMU_OPT_BOOL, | |
3756 | .help = "Postpone refcount updates", | |
3757 | .def_value_str = "off" | |
3758 | }, | |
06d05fa7 HR |
3759 | { |
3760 | .name = BLOCK_OPT_REFCOUNT_BITS, | |
3761 | .type = QEMU_OPT_NUMBER, | |
3762 | .help = "Width of a reference count entry in bits", | |
3763 | .def_value_str = "16" | |
3764 | }, | |
1bd0e2d1 CL |
3765 | { /* end of list */ } |
3766 | } | |
20d97356 BS |
3767 | }; |
3768 | ||
5f535a94 | 3769 | BlockDriver bdrv_qcow2 = { |
7c80ab3f | 3770 | .format_name = "qcow2", |
ff99129a | 3771 | .instance_size = sizeof(BDRVQcow2State), |
7c80ab3f JS |
3772 | .bdrv_probe = qcow2_probe, |
3773 | .bdrv_open = qcow2_open, | |
3774 | .bdrv_close = qcow2_close, | |
21d82ac9 | 3775 | .bdrv_reopen_prepare = qcow2_reopen_prepare, |
5b0959a7 KW |
3776 | .bdrv_reopen_commit = qcow2_reopen_commit, |
3777 | .bdrv_reopen_abort = qcow2_reopen_abort, | |
5365f44d | 3778 | .bdrv_join_options = qcow2_join_options, |
862f215f | 3779 | .bdrv_child_perm = bdrv_format_default_perms, |
c282e1fd | 3780 | .bdrv_create = qcow2_create, |
3ac21627 | 3781 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 3782 | .bdrv_co_get_block_status = qcow2_co_get_block_status, |
7c80ab3f | 3783 | |
ecfe1863 | 3784 | .bdrv_co_preadv = qcow2_co_preadv, |
d46a0bb2 | 3785 | .bdrv_co_pwritev = qcow2_co_pwritev, |
eb489bb1 | 3786 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
419b19d9 | 3787 | |
5544b59f | 3788 | .bdrv_co_pwrite_zeroes = qcow2_co_pwrite_zeroes, |
82e8a788 | 3789 | .bdrv_co_pdiscard = qcow2_co_pdiscard, |
419b19d9 | 3790 | .bdrv_truncate = qcow2_truncate, |
fcccefc5 | 3791 | .bdrv_co_pwritev_compressed = qcow2_co_pwritev_compressed, |
491d27e2 | 3792 | .bdrv_make_empty = qcow2_make_empty, |
20d97356 BS |
3793 | |
3794 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
3795 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
3796 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
3797 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1bd0e2d1 CL |
3798 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
3799 | .bdrv_get_info = qcow2_get_info, | |
37764dfb | 3800 | .bdrv_get_specific_info = qcow2_get_specific_info, |
20d97356 | 3801 | |
7c80ab3f JS |
3802 | .bdrv_save_vmstate = qcow2_save_vmstate, |
3803 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
20d97356 | 3804 | |
8ee79e70 | 3805 | .supports_backing = true, |
20d97356 BS |
3806 | .bdrv_change_backing_file = qcow2_change_backing_file, |
3807 | ||
d34682cd | 3808 | .bdrv_refresh_limits = qcow2_refresh_limits, |
06d9260f | 3809 | .bdrv_invalidate_cache = qcow2_invalidate_cache, |
ec6d8912 | 3810 | .bdrv_inactivate = qcow2_inactivate, |
06d9260f | 3811 | |
1bd0e2d1 CL |
3812 | .create_opts = &qcow2_create_opts, |
3813 | .bdrv_check = qcow2_check, | |
c282e1fd | 3814 | .bdrv_amend_options = qcow2_amend_options, |
279621c0 AG |
3815 | |
3816 | .bdrv_detach_aio_context = qcow2_detach_aio_context, | |
3817 | .bdrv_attach_aio_context = qcow2_attach_aio_context, | |
20d97356 BS |
3818 | }; |
3819 | ||
5efa9d5a AL |
3820 | static void bdrv_qcow2_init(void) |
3821 | { | |
3822 | bdrv_register(&bdrv_qcow2); | |
3823 | } | |
3824 | ||
3825 | block_init(bdrv_qcow2_init); |