]>
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 | */ | |
e688df6b | 24 | |
80c71a24 | 25 | #include "qemu/osdep.h" |
2714f13d VSO |
26 | |
27 | #define ZLIB_CONST | |
28 | #include <zlib.h> | |
29 | ||
737e150e | 30 | #include "block/block_int.h" |
609f45ea | 31 | #include "block/qdict.h" |
23588797 | 32 | #include "sysemu/block-backend.h" |
1de7afc9 | 33 | #include "qemu/module.h" |
0d8c41da | 34 | #include "qcow2.h" |
1de7afc9 | 35 | #include "qemu/error-report.h" |
e688df6b | 36 | #include "qapi/error.h" |
9af23989 | 37 | #include "qapi/qapi-events-block-core.h" |
6b673957 MA |
38 | #include "qapi/qmp/qdict.h" |
39 | #include "qapi/qmp/qstring.h" | |
3cce16f4 | 40 | #include "trace.h" |
1bd0e2d1 | 41 | #include "qemu/option_int.h" |
f348b6d1 | 42 | #include "qemu/cutils.h" |
58369e22 | 43 | #include "qemu/bswap.h" |
b76b4f60 KW |
44 | #include "qapi/qobject-input-visitor.h" |
45 | #include "qapi/qapi-visit-block-core.h" | |
0d8c41da | 46 | #include "crypto.h" |
ceb029cd | 47 | #include "block/thread-pool.h" |
585f8587 FB |
48 | |
49 | /* | |
50 | Differences with QCOW: | |
51 | ||
52 | - Support for multiple incremental snapshots. | |
53 | - Memory management by reference counts. | |
54 | - Clusters which have a reference count of one have the bit | |
55 | QCOW_OFLAG_COPIED to optimize write performance. | |
5fafdf24 | 56 | - Size of compressed clusters is stored in sectors to reduce bit usage |
585f8587 FB |
57 | in the cluster offsets. |
58 | - Support for storing additional data (such as the VM state) in the | |
3b46e624 | 59 | snapshots. |
585f8587 FB |
60 | - If a backing store is used, the cluster size is not constrained |
61 | (could be backported to QCOW). | |
62 | - L2 tables have always a size of one cluster. | |
63 | */ | |
64 | ||
9b80ddf3 AL |
65 | |
66 | typedef struct { | |
67 | uint32_t magic; | |
68 | uint32_t len; | |
c4217f64 | 69 | } QEMU_PACKED QCowExtension; |
21d82ac9 | 70 | |
7c80ab3f JS |
71 | #define QCOW2_EXT_MAGIC_END 0 |
72 | #define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA | |
cfcc4c62 | 73 | #define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 |
4652b8f3 | 74 | #define QCOW2_EXT_MAGIC_CRYPTO_HEADER 0x0537be77 |
88ddffae | 75 | #define QCOW2_EXT_MAGIC_BITMAPS 0x23852875 |
9b80ddf3 | 76 | |
c3c10f72 VSO |
77 | static int coroutine_fn |
78 | qcow2_co_preadv_compressed(BlockDriverState *bs, | |
79 | uint64_t file_cluster_offset, | |
80 | uint64_t offset, | |
81 | uint64_t bytes, | |
82 | QEMUIOVector *qiov); | |
83 | ||
7c80ab3f | 84 | static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) |
585f8587 FB |
85 | { |
86 | const QCowHeader *cow_header = (const void *)buf; | |
3b46e624 | 87 | |
585f8587 FB |
88 | if (buf_size >= sizeof(QCowHeader) && |
89 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
6744cbab | 90 | be32_to_cpu(cow_header->version) >= 2) |
585f8587 FB |
91 | return 100; |
92 | else | |
93 | return 0; | |
94 | } | |
95 | ||
9b80ddf3 | 96 | |
4652b8f3 DB |
97 | static ssize_t qcow2_crypto_hdr_read_func(QCryptoBlock *block, size_t offset, |
98 | uint8_t *buf, size_t buflen, | |
99 | void *opaque, Error **errp) | |
100 | { | |
101 | BlockDriverState *bs = opaque; | |
102 | BDRVQcow2State *s = bs->opaque; | |
103 | ssize_t ret; | |
104 | ||
105 | if ((offset + buflen) > s->crypto_header.length) { | |
106 | error_setg(errp, "Request for data outside of extension header"); | |
107 | return -1; | |
108 | } | |
109 | ||
110 | ret = bdrv_pread(bs->file, | |
111 | s->crypto_header.offset + offset, buf, buflen); | |
112 | if (ret < 0) { | |
113 | error_setg_errno(errp, -ret, "Could not read encryption header"); | |
114 | return -1; | |
115 | } | |
116 | return ret; | |
117 | } | |
118 | ||
119 | ||
120 | static ssize_t qcow2_crypto_hdr_init_func(QCryptoBlock *block, size_t headerlen, | |
121 | void *opaque, Error **errp) | |
122 | { | |
123 | BlockDriverState *bs = opaque; | |
124 | BDRVQcow2State *s = bs->opaque; | |
125 | int64_t ret; | |
126 | int64_t clusterlen; | |
127 | ||
128 | ret = qcow2_alloc_clusters(bs, headerlen); | |
129 | if (ret < 0) { | |
130 | error_setg_errno(errp, -ret, | |
131 | "Cannot allocate cluster for LUKS header size %zu", | |
132 | headerlen); | |
133 | return -1; | |
134 | } | |
135 | ||
136 | s->crypto_header.length = headerlen; | |
137 | s->crypto_header.offset = ret; | |
138 | ||
139 | /* Zero fill remaining space in cluster so it has predictable | |
140 | * content in case of future spec changes */ | |
141 | clusterlen = size_to_clusters(s, headerlen) * s->cluster_size; | |
c9b83e9c | 142 | assert(qcow2_pre_write_overlap_check(bs, 0, ret, clusterlen) == 0); |
4652b8f3 DB |
143 | ret = bdrv_pwrite_zeroes(bs->file, |
144 | ret + headerlen, | |
145 | clusterlen - headerlen, 0); | |
146 | if (ret < 0) { | |
147 | error_setg_errno(errp, -ret, "Could not zero fill encryption header"); | |
148 | return -1; | |
149 | } | |
150 | ||
151 | return ret; | |
152 | } | |
153 | ||
154 | ||
155 | static ssize_t qcow2_crypto_hdr_write_func(QCryptoBlock *block, size_t offset, | |
156 | const uint8_t *buf, size_t buflen, | |
157 | void *opaque, Error **errp) | |
158 | { | |
159 | BlockDriverState *bs = opaque; | |
160 | BDRVQcow2State *s = bs->opaque; | |
161 | ssize_t ret; | |
162 | ||
163 | if ((offset + buflen) > s->crypto_header.length) { | |
164 | error_setg(errp, "Request for data outside of extension header"); | |
165 | return -1; | |
166 | } | |
167 | ||
168 | ret = bdrv_pwrite(bs->file, | |
169 | s->crypto_header.offset + offset, buf, buflen); | |
170 | if (ret < 0) { | |
171 | error_setg_errno(errp, -ret, "Could not read encryption header"); | |
172 | return -1; | |
173 | } | |
174 | return ret; | |
175 | } | |
176 | ||
177 | ||
9b80ddf3 AL |
178 | /* |
179 | * read qcow2 extension and fill bs | |
180 | * start reading from start_offset | |
181 | * finish reading upon magic of value 0 or when end_offset reached | |
182 | * unknown magic is skipped (future extension this version knows nothing about) | |
183 | * return 0 upon success, non-0 otherwise | |
184 | */ | |
7c80ab3f | 185 | static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
3ef6c40a | 186 | uint64_t end_offset, void **p_feature_table, |
88ddffae VSO |
187 | int flags, bool *need_update_header, |
188 | Error **errp) | |
9b80ddf3 | 189 | { |
ff99129a | 190 | BDRVQcow2State *s = bs->opaque; |
9b80ddf3 AL |
191 | QCowExtension ext; |
192 | uint64_t offset; | |
75bab85c | 193 | int ret; |
88ddffae VSO |
194 | Qcow2BitmapHeaderExt bitmaps_ext; |
195 | ||
196 | if (need_update_header != NULL) { | |
197 | *need_update_header = false; | |
198 | } | |
9b80ddf3 AL |
199 | |
200 | #ifdef DEBUG_EXT | |
7c80ab3f | 201 | printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); |
9b80ddf3 AL |
202 | #endif |
203 | offset = start_offset; | |
204 | while (offset < end_offset) { | |
205 | ||
206 | #ifdef DEBUG_EXT | |
207 | /* Sanity check */ | |
208 | if (offset > s->cluster_size) | |
7c80ab3f | 209 | printf("qcow2_read_extension: suspicious offset %lu\n", offset); |
9b80ddf3 | 210 | |
9b2260cb | 211 | printf("attempting to read extended header in offset %lu\n", offset); |
9b80ddf3 AL |
212 | #endif |
213 | ||
cf2ab8fc | 214 | ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext)); |
3ef6c40a HR |
215 | if (ret < 0) { |
216 | error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: " | |
217 | "pread fail from offset %" PRIu64, offset); | |
9b80ddf3 AL |
218 | return 1; |
219 | } | |
3b698f52 PM |
220 | ext.magic = be32_to_cpu(ext.magic); |
221 | ext.len = be32_to_cpu(ext.len); | |
9b80ddf3 AL |
222 | offset += sizeof(ext); |
223 | #ifdef DEBUG_EXT | |
224 | printf("ext.magic = 0x%x\n", ext.magic); | |
225 | #endif | |
2ebafc85 | 226 | if (offset > end_offset || ext.len > end_offset - offset) { |
3ef6c40a | 227 | error_setg(errp, "Header extension too large"); |
64ca6aee KW |
228 | return -EINVAL; |
229 | } | |
230 | ||
9b80ddf3 | 231 | switch (ext.magic) { |
7c80ab3f | 232 | case QCOW2_EXT_MAGIC_END: |
9b80ddf3 | 233 | return 0; |
f965509c | 234 | |
7c80ab3f | 235 | case QCOW2_EXT_MAGIC_BACKING_FORMAT: |
f965509c | 236 | if (ext.len >= sizeof(bs->backing_format)) { |
521b2b5d HR |
237 | error_setg(errp, "ERROR: ext_backing_format: len=%" PRIu32 |
238 | " too large (>=%zu)", ext.len, | |
239 | sizeof(bs->backing_format)); | |
f965509c AL |
240 | return 2; |
241 | } | |
cf2ab8fc | 242 | ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len); |
3ef6c40a HR |
243 | if (ret < 0) { |
244 | error_setg_errno(errp, -ret, "ERROR: ext_backing_format: " | |
245 | "Could not read format name"); | |
f965509c | 246 | return 3; |
3ef6c40a | 247 | } |
f965509c | 248 | bs->backing_format[ext.len] = '\0'; |
e4603fe1 | 249 | s->image_backing_format = g_strdup(bs->backing_format); |
f965509c AL |
250 | #ifdef DEBUG_EXT |
251 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
252 | #endif | |
f965509c AL |
253 | break; |
254 | ||
cfcc4c62 KW |
255 | case QCOW2_EXT_MAGIC_FEATURE_TABLE: |
256 | if (p_feature_table != NULL) { | |
257 | void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); | |
cf2ab8fc | 258 | ret = bdrv_pread(bs->file, offset , feature_table, ext.len); |
cfcc4c62 | 259 | if (ret < 0) { |
3ef6c40a HR |
260 | error_setg_errno(errp, -ret, "ERROR: ext_feature_table: " |
261 | "Could not read table"); | |
cfcc4c62 KW |
262 | return ret; |
263 | } | |
264 | ||
265 | *p_feature_table = feature_table; | |
266 | } | |
267 | break; | |
268 | ||
4652b8f3 DB |
269 | case QCOW2_EXT_MAGIC_CRYPTO_HEADER: { |
270 | unsigned int cflags = 0; | |
271 | if (s->crypt_method_header != QCOW_CRYPT_LUKS) { | |
272 | error_setg(errp, "CRYPTO header extension only " | |
273 | "expected with LUKS encryption method"); | |
274 | return -EINVAL; | |
275 | } | |
276 | if (ext.len != sizeof(Qcow2CryptoHeaderExtension)) { | |
277 | error_setg(errp, "CRYPTO header extension size %u, " | |
278 | "but expected size %zu", ext.len, | |
279 | sizeof(Qcow2CryptoHeaderExtension)); | |
280 | return -EINVAL; | |
281 | } | |
282 | ||
283 | ret = bdrv_pread(bs->file, offset, &s->crypto_header, ext.len); | |
284 | if (ret < 0) { | |
285 | error_setg_errno(errp, -ret, | |
286 | "Unable to read CRYPTO header extension"); | |
287 | return ret; | |
288 | } | |
3b698f52 PM |
289 | s->crypto_header.offset = be64_to_cpu(s->crypto_header.offset); |
290 | s->crypto_header.length = be64_to_cpu(s->crypto_header.length); | |
4652b8f3 DB |
291 | |
292 | if ((s->crypto_header.offset % s->cluster_size) != 0) { | |
293 | error_setg(errp, "Encryption header offset '%" PRIu64 "' is " | |
294 | "not a multiple of cluster size '%u'", | |
295 | s->crypto_header.offset, s->cluster_size); | |
296 | return -EINVAL; | |
297 | } | |
298 | ||
299 | if (flags & BDRV_O_NO_IO) { | |
300 | cflags |= QCRYPTO_BLOCK_OPEN_NO_IO; | |
301 | } | |
1cd9a787 | 302 | s->crypto = qcrypto_block_open(s->crypto_opts, "encrypt.", |
4652b8f3 | 303 | qcow2_crypto_hdr_read_func, |
c972fa12 | 304 | bs, cflags, 1, errp); |
4652b8f3 DB |
305 | if (!s->crypto) { |
306 | return -EINVAL; | |
307 | } | |
308 | } break; | |
309 | ||
88ddffae VSO |
310 | case QCOW2_EXT_MAGIC_BITMAPS: |
311 | if (ext.len != sizeof(bitmaps_ext)) { | |
312 | error_setg_errno(errp, -ret, "bitmaps_ext: " | |
313 | "Invalid extension length"); | |
314 | return -EINVAL; | |
315 | } | |
316 | ||
317 | if (!(s->autoclear_features & QCOW2_AUTOCLEAR_BITMAPS)) { | |
c9ceb3ec HR |
318 | if (s->qcow_version < 3) { |
319 | /* Let's be a bit more specific */ | |
320 | warn_report("This qcow2 v2 image contains bitmaps, but " | |
321 | "they may have been modified by a program " | |
322 | "without persistent bitmap support; so now " | |
323 | "they must all be considered inconsistent"); | |
324 | } else { | |
325 | warn_report("a program lacking bitmap support " | |
326 | "modified this file, so all bitmaps are now " | |
327 | "considered inconsistent"); | |
328 | } | |
55d527a9 AF |
329 | error_printf("Some clusters may be leaked, " |
330 | "run 'qemu-img check -r' on the image " | |
88ddffae VSO |
331 | "file to fix."); |
332 | if (need_update_header != NULL) { | |
333 | /* Updating is needed to drop invalid bitmap extension. */ | |
334 | *need_update_header = true; | |
335 | } | |
336 | break; | |
337 | } | |
338 | ||
339 | ret = bdrv_pread(bs->file, offset, &bitmaps_ext, ext.len); | |
340 | if (ret < 0) { | |
341 | error_setg_errno(errp, -ret, "bitmaps_ext: " | |
342 | "Could not read ext header"); | |
343 | return ret; | |
344 | } | |
345 | ||
346 | if (bitmaps_ext.reserved32 != 0) { | |
347 | error_setg_errno(errp, -ret, "bitmaps_ext: " | |
348 | "Reserved field is not zero"); | |
349 | return -EINVAL; | |
350 | } | |
351 | ||
3b698f52 PM |
352 | bitmaps_ext.nb_bitmaps = be32_to_cpu(bitmaps_ext.nb_bitmaps); |
353 | bitmaps_ext.bitmap_directory_size = | |
354 | be64_to_cpu(bitmaps_ext.bitmap_directory_size); | |
355 | bitmaps_ext.bitmap_directory_offset = | |
356 | be64_to_cpu(bitmaps_ext.bitmap_directory_offset); | |
88ddffae VSO |
357 | |
358 | if (bitmaps_ext.nb_bitmaps > QCOW2_MAX_BITMAPS) { | |
359 | error_setg(errp, | |
360 | "bitmaps_ext: Image has %" PRIu32 " bitmaps, " | |
361 | "exceeding the QEMU supported maximum of %d", | |
362 | bitmaps_ext.nb_bitmaps, QCOW2_MAX_BITMAPS); | |
363 | return -EINVAL; | |
364 | } | |
365 | ||
366 | if (bitmaps_ext.nb_bitmaps == 0) { | |
367 | error_setg(errp, "found bitmaps extension with zero bitmaps"); | |
368 | return -EINVAL; | |
369 | } | |
370 | ||
371 | if (bitmaps_ext.bitmap_directory_offset & (s->cluster_size - 1)) { | |
372 | error_setg(errp, "bitmaps_ext: " | |
373 | "invalid bitmap directory offset"); | |
374 | return -EINVAL; | |
375 | } | |
376 | ||
377 | if (bitmaps_ext.bitmap_directory_size > | |
378 | QCOW2_MAX_BITMAP_DIRECTORY_SIZE) { | |
379 | error_setg(errp, "bitmaps_ext: " | |
380 | "bitmap directory size (%" PRIu64 ") exceeds " | |
381 | "the maximum supported size (%d)", | |
382 | bitmaps_ext.bitmap_directory_size, | |
383 | QCOW2_MAX_BITMAP_DIRECTORY_SIZE); | |
384 | return -EINVAL; | |
385 | } | |
386 | ||
387 | s->nb_bitmaps = bitmaps_ext.nb_bitmaps; | |
388 | s->bitmap_directory_offset = | |
389 | bitmaps_ext.bitmap_directory_offset; | |
390 | s->bitmap_directory_size = | |
391 | bitmaps_ext.bitmap_directory_size; | |
392 | ||
393 | #ifdef DEBUG_EXT | |
394 | printf("Qcow2: Got bitmaps extension: " | |
395 | "offset=%" PRIu64 " nb_bitmaps=%" PRIu32 "\n", | |
396 | s->bitmap_directory_offset, s->nb_bitmaps); | |
397 | #endif | |
398 | break; | |
399 | ||
9b80ddf3 | 400 | default: |
75bab85c | 401 | /* unknown magic - save it in case we need to rewrite the header */ |
4096974e EB |
402 | /* If you add a new feature, make sure to also update the fast |
403 | * path of qcow2_make_empty() to deal with it. */ | |
75bab85c KW |
404 | { |
405 | Qcow2UnknownHeaderExtension *uext; | |
406 | ||
407 | uext = g_malloc0(sizeof(*uext) + ext.len); | |
408 | uext->magic = ext.magic; | |
409 | uext->len = ext.len; | |
410 | QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); | |
411 | ||
cf2ab8fc | 412 | ret = bdrv_pread(bs->file, offset , uext->data, uext->len); |
75bab85c | 413 | if (ret < 0) { |
3ef6c40a HR |
414 | error_setg_errno(errp, -ret, "ERROR: unknown extension: " |
415 | "Could not read data"); | |
75bab85c KW |
416 | return ret; |
417 | } | |
75bab85c | 418 | } |
9b80ddf3 AL |
419 | break; |
420 | } | |
fd29b4bb KW |
421 | |
422 | offset += ((ext.len + 7) & ~7); | |
9b80ddf3 AL |
423 | } |
424 | ||
425 | return 0; | |
426 | } | |
427 | ||
75bab85c KW |
428 | static void cleanup_unknown_header_ext(BlockDriverState *bs) |
429 | { | |
ff99129a | 430 | BDRVQcow2State *s = bs->opaque; |
75bab85c KW |
431 | Qcow2UnknownHeaderExtension *uext, *next; |
432 | ||
433 | QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { | |
434 | QLIST_REMOVE(uext, next); | |
435 | g_free(uext); | |
436 | } | |
437 | } | |
9b80ddf3 | 438 | |
a55448b3 HR |
439 | static void report_unsupported_feature(Error **errp, Qcow2Feature *table, |
440 | uint64_t mask) | |
cfcc4c62 | 441 | { |
12ac6d3d KW |
442 | char *features = g_strdup(""); |
443 | char *old; | |
444 | ||
cfcc4c62 KW |
445 | while (table && table->name[0] != '\0') { |
446 | if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) { | |
12ac6d3d KW |
447 | if (mask & (1ULL << table->bit)) { |
448 | old = features; | |
449 | features = g_strdup_printf("%s%s%.46s", old, *old ? ", " : "", | |
450 | table->name); | |
451 | g_free(old); | |
452 | mask &= ~(1ULL << table->bit); | |
cfcc4c62 KW |
453 | } |
454 | } | |
455 | table++; | |
456 | } | |
457 | ||
458 | if (mask) { | |
12ac6d3d KW |
459 | old = features; |
460 | features = g_strdup_printf("%s%sUnknown incompatible feature: %" PRIx64, | |
461 | old, *old ? ", " : "", mask); | |
462 | g_free(old); | |
cfcc4c62 | 463 | } |
12ac6d3d | 464 | |
a55448b3 | 465 | error_setg(errp, "Unsupported qcow2 feature(s): %s", features); |
12ac6d3d | 466 | g_free(features); |
cfcc4c62 KW |
467 | } |
468 | ||
bfe8043e SH |
469 | /* |
470 | * Sets the dirty bit and flushes afterwards if necessary. | |
471 | * | |
472 | * The incompatible_features bit is only set if the image file header was | |
473 | * updated successfully. Therefore it is not required to check the return | |
474 | * value of this function. | |
475 | */ | |
280d3735 | 476 | int qcow2_mark_dirty(BlockDriverState *bs) |
bfe8043e | 477 | { |
ff99129a | 478 | BDRVQcow2State *s = bs->opaque; |
bfe8043e SH |
479 | uint64_t val; |
480 | int ret; | |
481 | ||
482 | assert(s->qcow_version >= 3); | |
483 | ||
484 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
485 | return 0; /* already dirty */ | |
486 | } | |
487 | ||
488 | val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); | |
d9ca2ea2 | 489 | ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features), |
bfe8043e SH |
490 | &val, sizeof(val)); |
491 | if (ret < 0) { | |
492 | return ret; | |
493 | } | |
9a4f4c31 | 494 | ret = bdrv_flush(bs->file->bs); |
bfe8043e SH |
495 | if (ret < 0) { |
496 | return ret; | |
497 | } | |
498 | ||
499 | /* Only treat image as dirty if the header was updated successfully */ | |
500 | s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; | |
501 | return 0; | |
502 | } | |
503 | ||
c61d0004 SH |
504 | /* |
505 | * Clears the dirty bit and flushes before if necessary. Only call this | |
506 | * function when there are no pending requests, it does not guard against | |
507 | * concurrent requests dirtying the image. | |
508 | */ | |
509 | static int qcow2_mark_clean(BlockDriverState *bs) | |
510 | { | |
ff99129a | 511 | BDRVQcow2State *s = bs->opaque; |
c61d0004 SH |
512 | |
513 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
4c2e5f8f KW |
514 | int ret; |
515 | ||
516 | s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; | |
517 | ||
8b220eb7 | 518 | ret = qcow2_flush_caches(bs); |
c61d0004 SH |
519 | if (ret < 0) { |
520 | return ret; | |
521 | } | |
522 | ||
c61d0004 SH |
523 | return qcow2_update_header(bs); |
524 | } | |
525 | return 0; | |
526 | } | |
527 | ||
69c98726 HR |
528 | /* |
529 | * Marks the image as corrupt. | |
530 | */ | |
531 | int qcow2_mark_corrupt(BlockDriverState *bs) | |
532 | { | |
ff99129a | 533 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
534 | |
535 | s->incompatible_features |= QCOW2_INCOMPAT_CORRUPT; | |
536 | return qcow2_update_header(bs); | |
537 | } | |
538 | ||
539 | /* | |
540 | * Marks the image as consistent, i.e., unsets the corrupt bit, and flushes | |
541 | * before if necessary. | |
542 | */ | |
543 | int qcow2_mark_consistent(BlockDriverState *bs) | |
544 | { | |
ff99129a | 545 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
546 | |
547 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { | |
8b220eb7 | 548 | int ret = qcow2_flush_caches(bs); |
69c98726 HR |
549 | if (ret < 0) { |
550 | return ret; | |
551 | } | |
552 | ||
553 | s->incompatible_features &= ~QCOW2_INCOMPAT_CORRUPT; | |
554 | return qcow2_update_header(bs); | |
555 | } | |
556 | return 0; | |
557 | } | |
558 | ||
2fd61638 PB |
559 | static int coroutine_fn qcow2_co_check_locked(BlockDriverState *bs, |
560 | BdrvCheckResult *result, | |
561 | BdrvCheckMode fix) | |
acbe5982 SH |
562 | { |
563 | int ret = qcow2_check_refcounts(bs, result, fix); | |
564 | if (ret < 0) { | |
565 | return ret; | |
566 | } | |
567 | ||
568 | if (fix && result->check_errors == 0 && result->corruptions == 0) { | |
24530f3e HR |
569 | ret = qcow2_mark_clean(bs); |
570 | if (ret < 0) { | |
571 | return ret; | |
572 | } | |
573 | return qcow2_mark_consistent(bs); | |
acbe5982 SH |
574 | } |
575 | return ret; | |
576 | } | |
577 | ||
2fd61638 PB |
578 | static int coroutine_fn qcow2_co_check(BlockDriverState *bs, |
579 | BdrvCheckResult *result, | |
580 | BdrvCheckMode fix) | |
581 | { | |
582 | BDRVQcow2State *s = bs->opaque; | |
583 | int ret; | |
584 | ||
585 | qemu_co_mutex_lock(&s->lock); | |
586 | ret = qcow2_co_check_locked(bs, result, fix); | |
587 | qemu_co_mutex_unlock(&s->lock); | |
588 | return ret; | |
589 | } | |
590 | ||
0cf0e598 AG |
591 | int qcow2_validate_table(BlockDriverState *bs, uint64_t offset, |
592 | uint64_t entries, size_t entry_len, | |
593 | int64_t max_size_bytes, const char *table_name, | |
594 | Error **errp) | |
8c7de283 | 595 | { |
ff99129a | 596 | BDRVQcow2State *s = bs->opaque; |
8c7de283 | 597 | |
0cf0e598 AG |
598 | if (entries > max_size_bytes / entry_len) { |
599 | error_setg(errp, "%s too large", table_name); | |
600 | return -EFBIG; | |
8c7de283 KW |
601 | } |
602 | ||
0cf0e598 AG |
603 | /* Use signed INT64_MAX as the maximum even for uint64_t header fields, |
604 | * because values will be passed to qemu functions taking int64_t. */ | |
605 | if ((INT64_MAX - entries * entry_len < offset) || | |
606 | (offset_into_cluster(s, offset) != 0)) { | |
607 | error_setg(errp, "%s offset invalid", table_name); | |
8c7de283 KW |
608 | return -EINVAL; |
609 | } | |
610 | ||
611 | return 0; | |
612 | } | |
613 | ||
74c4510a KW |
614 | static QemuOptsList qcow2_runtime_opts = { |
615 | .name = "qcow2", | |
616 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head), | |
617 | .desc = { | |
618 | { | |
64aa99d3 | 619 | .name = QCOW2_OPT_LAZY_REFCOUNTS, |
74c4510a KW |
620 | .type = QEMU_OPT_BOOL, |
621 | .help = "Postpone refcount updates", | |
622 | }, | |
67af674e KW |
623 | { |
624 | .name = QCOW2_OPT_DISCARD_REQUEST, | |
625 | .type = QEMU_OPT_BOOL, | |
626 | .help = "Pass guest discard requests to the layer below", | |
627 | }, | |
628 | { | |
629 | .name = QCOW2_OPT_DISCARD_SNAPSHOT, | |
630 | .type = QEMU_OPT_BOOL, | |
631 | .help = "Generate discard requests when snapshot related space " | |
632 | "is freed", | |
633 | }, | |
634 | { | |
635 | .name = QCOW2_OPT_DISCARD_OTHER, | |
636 | .type = QEMU_OPT_BOOL, | |
637 | .help = "Generate discard requests when other clusters are freed", | |
638 | }, | |
05de7e86 HR |
639 | { |
640 | .name = QCOW2_OPT_OVERLAP, | |
641 | .type = QEMU_OPT_STRING, | |
642 | .help = "Selects which overlap checks to perform from a range of " | |
643 | "templates (none, constant, cached, all)", | |
644 | }, | |
ee42b5ce HR |
645 | { |
646 | .name = QCOW2_OPT_OVERLAP_TEMPLATE, | |
647 | .type = QEMU_OPT_STRING, | |
648 | .help = "Selects which overlap checks to perform from a range of " | |
649 | "templates (none, constant, cached, all)", | |
650 | }, | |
05de7e86 HR |
651 | { |
652 | .name = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
653 | .type = QEMU_OPT_BOOL, | |
654 | .help = "Check for unintended writes into the main qcow2 header", | |
655 | }, | |
656 | { | |
657 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
658 | .type = QEMU_OPT_BOOL, | |
659 | .help = "Check for unintended writes into the active L1 table", | |
660 | }, | |
661 | { | |
662 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
663 | .type = QEMU_OPT_BOOL, | |
664 | .help = "Check for unintended writes into an active L2 table", | |
665 | }, | |
666 | { | |
667 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
668 | .type = QEMU_OPT_BOOL, | |
669 | .help = "Check for unintended writes into the refcount table", | |
670 | }, | |
671 | { | |
672 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
673 | .type = QEMU_OPT_BOOL, | |
674 | .help = "Check for unintended writes into a refcount block", | |
675 | }, | |
676 | { | |
677 | .name = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
678 | .type = QEMU_OPT_BOOL, | |
679 | .help = "Check for unintended writes into the snapshot table", | |
680 | }, | |
681 | { | |
682 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
683 | .type = QEMU_OPT_BOOL, | |
684 | .help = "Check for unintended writes into an inactive L1 table", | |
685 | }, | |
686 | { | |
687 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
688 | .type = QEMU_OPT_BOOL, | |
689 | .help = "Check for unintended writes into an inactive L2 table", | |
690 | }, | |
0e4e4318 VSO |
691 | { |
692 | .name = QCOW2_OPT_OVERLAP_BITMAP_DIRECTORY, | |
693 | .type = QEMU_OPT_BOOL, | |
694 | .help = "Check for unintended writes into the bitmap directory", | |
695 | }, | |
6c1c8d5d HR |
696 | { |
697 | .name = QCOW2_OPT_CACHE_SIZE, | |
698 | .type = QEMU_OPT_SIZE, | |
699 | .help = "Maximum combined metadata (L2 tables and refcount blocks) " | |
700 | "cache size", | |
701 | }, | |
702 | { | |
703 | .name = QCOW2_OPT_L2_CACHE_SIZE, | |
704 | .type = QEMU_OPT_SIZE, | |
705 | .help = "Maximum L2 table cache size", | |
706 | }, | |
1221fe6f AG |
707 | { |
708 | .name = QCOW2_OPT_L2_CACHE_ENTRY_SIZE, | |
709 | .type = QEMU_OPT_SIZE, | |
710 | .help = "Size of each entry in the L2 cache", | |
711 | }, | |
6c1c8d5d HR |
712 | { |
713 | .name = QCOW2_OPT_REFCOUNT_CACHE_SIZE, | |
714 | .type = QEMU_OPT_SIZE, | |
715 | .help = "Maximum refcount block cache size", | |
716 | }, | |
279621c0 AG |
717 | { |
718 | .name = QCOW2_OPT_CACHE_CLEAN_INTERVAL, | |
719 | .type = QEMU_OPT_NUMBER, | |
720 | .help = "Clean unused cache entries after this time (in seconds)", | |
721 | }, | |
4652b8f3 DB |
722 | BLOCK_CRYPTO_OPT_DEF_KEY_SECRET("encrypt.", |
723 | "ID of secret providing qcow2 AES key or LUKS passphrase"), | |
74c4510a KW |
724 | { /* end of list */ } |
725 | }, | |
726 | }; | |
727 | ||
4092e99d | 728 | static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = { |
0e4e4318 VSO |
729 | [QCOW2_OL_MAIN_HEADER_BITNR] = QCOW2_OPT_OVERLAP_MAIN_HEADER, |
730 | [QCOW2_OL_ACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
731 | [QCOW2_OL_ACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
732 | [QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
733 | [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
734 | [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
735 | [QCOW2_OL_INACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
736 | [QCOW2_OL_INACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
737 | [QCOW2_OL_BITMAP_DIRECTORY_BITNR] = QCOW2_OPT_OVERLAP_BITMAP_DIRECTORY, | |
4092e99d HR |
738 | }; |
739 | ||
279621c0 AG |
740 | static void cache_clean_timer_cb(void *opaque) |
741 | { | |
742 | BlockDriverState *bs = opaque; | |
ff99129a | 743 | BDRVQcow2State *s = bs->opaque; |
b2f68bff AG |
744 | qcow2_cache_clean_unused(s->l2_table_cache); |
745 | qcow2_cache_clean_unused(s->refcount_block_cache); | |
279621c0 AG |
746 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + |
747 | (int64_t) s->cache_clean_interval * 1000); | |
748 | } | |
749 | ||
750 | static void cache_clean_timer_init(BlockDriverState *bs, AioContext *context) | |
751 | { | |
ff99129a | 752 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
753 | if (s->cache_clean_interval > 0) { |
754 | s->cache_clean_timer = aio_timer_new(context, QEMU_CLOCK_VIRTUAL, | |
755 | SCALE_MS, cache_clean_timer_cb, | |
756 | bs); | |
757 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
758 | (int64_t) s->cache_clean_interval * 1000); | |
759 | } | |
760 | } | |
761 | ||
762 | static void cache_clean_timer_del(BlockDriverState *bs) | |
763 | { | |
ff99129a | 764 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
765 | if (s->cache_clean_timer) { |
766 | timer_del(s->cache_clean_timer); | |
767 | timer_free(s->cache_clean_timer); | |
768 | s->cache_clean_timer = NULL; | |
769 | } | |
770 | } | |
771 | ||
772 | static void qcow2_detach_aio_context(BlockDriverState *bs) | |
773 | { | |
774 | cache_clean_timer_del(bs); | |
775 | } | |
776 | ||
777 | static void qcow2_attach_aio_context(BlockDriverState *bs, | |
778 | AioContext *new_context) | |
779 | { | |
780 | cache_clean_timer_init(bs, new_context); | |
781 | } | |
782 | ||
bc85ef26 HR |
783 | static void read_cache_sizes(BlockDriverState *bs, QemuOpts *opts, |
784 | uint64_t *l2_cache_size, | |
1221fe6f | 785 | uint64_t *l2_cache_entry_size, |
6c1c8d5d HR |
786 | uint64_t *refcount_cache_size, Error **errp) |
787 | { | |
ff99129a | 788 | BDRVQcow2State *s = bs->opaque; |
b749562d | 789 | uint64_t combined_cache_size, l2_cache_max_setting; |
6c1c8d5d | 790 | bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set; |
af39bd0d | 791 | bool l2_cache_entry_size_set; |
7af5eea9 | 792 | int min_refcount_cache = MIN_REFCOUNT_CACHE_SIZE * s->cluster_size; |
b749562d LB |
793 | uint64_t virtual_disk_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
794 | uint64_t max_l2_cache = virtual_disk_size / (s->cluster_size / 8); | |
6c1c8d5d HR |
795 | |
796 | combined_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_CACHE_SIZE); | |
797 | l2_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_SIZE); | |
798 | refcount_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
af39bd0d | 799 | l2_cache_entry_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_ENTRY_SIZE); |
6c1c8d5d HR |
800 | |
801 | combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0); | |
b749562d LB |
802 | l2_cache_max_setting = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE, |
803 | DEFAULT_L2_CACHE_MAX_SIZE); | |
6c1c8d5d HR |
804 | *refcount_cache_size = qemu_opt_get_size(opts, |
805 | QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0); | |
806 | ||
1221fe6f AG |
807 | *l2_cache_entry_size = qemu_opt_get_size( |
808 | opts, QCOW2_OPT_L2_CACHE_ENTRY_SIZE, s->cluster_size); | |
809 | ||
b749562d LB |
810 | *l2_cache_size = MIN(max_l2_cache, l2_cache_max_setting); |
811 | ||
6c1c8d5d HR |
812 | if (combined_cache_size_set) { |
813 | if (l2_cache_size_set && refcount_cache_size_set) { | |
814 | error_setg(errp, QCOW2_OPT_CACHE_SIZE ", " QCOW2_OPT_L2_CACHE_SIZE | |
815 | " and " QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not be set " | |
308999e9 | 816 | "at the same time"); |
6c1c8d5d | 817 | return; |
b749562d LB |
818 | } else if (l2_cache_size_set && |
819 | (l2_cache_max_setting > combined_cache_size)) { | |
6c1c8d5d HR |
820 | error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed " |
821 | QCOW2_OPT_CACHE_SIZE); | |
822 | return; | |
823 | } else if (*refcount_cache_size > combined_cache_size) { | |
824 | error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed " | |
825 | QCOW2_OPT_CACHE_SIZE); | |
826 | return; | |
827 | } | |
828 | ||
829 | if (l2_cache_size_set) { | |
830 | *refcount_cache_size = combined_cache_size - *l2_cache_size; | |
831 | } else if (refcount_cache_size_set) { | |
832 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
833 | } else { | |
52253998 AG |
834 | /* Assign as much memory as possible to the L2 cache, and |
835 | * use the remainder for the refcount cache */ | |
836 | if (combined_cache_size >= max_l2_cache + min_refcount_cache) { | |
837 | *l2_cache_size = max_l2_cache; | |
838 | *refcount_cache_size = combined_cache_size - *l2_cache_size; | |
839 | } else { | |
840 | *refcount_cache_size = | |
841 | MIN(combined_cache_size, min_refcount_cache); | |
842 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
843 | } | |
6c1c8d5d | 844 | } |
6c1c8d5d | 845 | } |
af39bd0d AG |
846 | |
847 | /* | |
848 | * If the L2 cache is not enough to cover the whole disk then | |
849 | * default to 4KB entries. Smaller entries reduce the cost of | |
850 | * loads and evictions and increase I/O performance. | |
851 | */ | |
852 | if (*l2_cache_size < max_l2_cache && !l2_cache_entry_size_set) { | |
853 | *l2_cache_entry_size = MIN(s->cluster_size, 4096); | |
854 | } | |
855 | ||
657ada52 LB |
856 | /* l2_cache_size and refcount_cache_size are ensured to have at least |
857 | * their minimum values in qcow2_update_options_prepare() */ | |
1221fe6f AG |
858 | |
859 | if (*l2_cache_entry_size < (1 << MIN_CLUSTER_BITS) || | |
860 | *l2_cache_entry_size > s->cluster_size || | |
861 | !is_power_of_2(*l2_cache_entry_size)) { | |
862 | error_setg(errp, "L2 cache entry size must be a power of two " | |
863 | "between %d and the cluster size (%d)", | |
864 | 1 << MIN_CLUSTER_BITS, s->cluster_size); | |
865 | return; | |
866 | } | |
6c1c8d5d HR |
867 | } |
868 | ||
ee55b173 KW |
869 | typedef struct Qcow2ReopenState { |
870 | Qcow2Cache *l2_table_cache; | |
871 | Qcow2Cache *refcount_block_cache; | |
3c2e511a | 872 | int l2_slice_size; /* Number of entries in a slice of the L2 table */ |
ee55b173 KW |
873 | bool use_lazy_refcounts; |
874 | int overlap_check; | |
875 | bool discard_passthrough[QCOW2_DISCARD_MAX]; | |
876 | uint64_t cache_clean_interval; | |
b25b387f | 877 | QCryptoBlockOpenOptions *crypto_opts; /* Disk encryption runtime options */ |
ee55b173 KW |
878 | } Qcow2ReopenState; |
879 | ||
880 | static int qcow2_update_options_prepare(BlockDriverState *bs, | |
881 | Qcow2ReopenState *r, | |
882 | QDict *options, int flags, | |
883 | Error **errp) | |
4c75d1a1 KW |
884 | { |
885 | BDRVQcow2State *s = bs->opaque; | |
94edf3fb | 886 | QemuOpts *opts = NULL; |
4c75d1a1 KW |
887 | const char *opt_overlap_check, *opt_overlap_check_template; |
888 | int overlap_check_template = 0; | |
1221fe6f | 889 | uint64_t l2_cache_size, l2_cache_entry_size, refcount_cache_size; |
4c75d1a1 | 890 | int i; |
b25b387f DB |
891 | const char *encryptfmt; |
892 | QDict *encryptopts = NULL; | |
94edf3fb | 893 | Error *local_err = NULL; |
4c75d1a1 KW |
894 | int ret; |
895 | ||
b25b387f DB |
896 | qdict_extract_subqdict(options, &encryptopts, "encrypt."); |
897 | encryptfmt = qdict_get_try_str(encryptopts, "format"); | |
898 | ||
94edf3fb KW |
899 | opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort); |
900 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
901 | if (local_err) { | |
902 | error_propagate(errp, local_err); | |
903 | ret = -EINVAL; | |
904 | goto fail; | |
905 | } | |
906 | ||
907 | /* get L2 table/refcount block cache size from command line options */ | |
1221fe6f AG |
908 | read_cache_sizes(bs, opts, &l2_cache_size, &l2_cache_entry_size, |
909 | &refcount_cache_size, &local_err); | |
94edf3fb KW |
910 | if (local_err) { |
911 | error_propagate(errp, local_err); | |
912 | ret = -EINVAL; | |
913 | goto fail; | |
914 | } | |
915 | ||
1221fe6f | 916 | l2_cache_size /= l2_cache_entry_size; |
94edf3fb KW |
917 | if (l2_cache_size < MIN_L2_CACHE_SIZE) { |
918 | l2_cache_size = MIN_L2_CACHE_SIZE; | |
919 | } | |
920 | if (l2_cache_size > INT_MAX) { | |
921 | error_setg(errp, "L2 cache size too big"); | |
922 | ret = -EINVAL; | |
923 | goto fail; | |
924 | } | |
925 | ||
926 | refcount_cache_size /= s->cluster_size; | |
927 | if (refcount_cache_size < MIN_REFCOUNT_CACHE_SIZE) { | |
928 | refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE; | |
929 | } | |
930 | if (refcount_cache_size > INT_MAX) { | |
931 | error_setg(errp, "Refcount cache size too big"); | |
932 | ret = -EINVAL; | |
933 | goto fail; | |
934 | } | |
935 | ||
5b0959a7 KW |
936 | /* alloc new L2 table/refcount block cache, flush old one */ |
937 | if (s->l2_table_cache) { | |
938 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
939 | if (ret) { | |
940 | error_setg_errno(errp, -ret, "Failed to flush the L2 table cache"); | |
941 | goto fail; | |
942 | } | |
943 | } | |
944 | ||
945 | if (s->refcount_block_cache) { | |
946 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
947 | if (ret) { | |
948 | error_setg_errno(errp, -ret, | |
949 | "Failed to flush the refcount block cache"); | |
950 | goto fail; | |
951 | } | |
952 | } | |
953 | ||
1221fe6f AG |
954 | r->l2_slice_size = l2_cache_entry_size / sizeof(uint64_t); |
955 | r->l2_table_cache = qcow2_cache_create(bs, l2_cache_size, | |
956 | l2_cache_entry_size); | |
957 | r->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size, | |
958 | s->cluster_size); | |
ee55b173 | 959 | if (r->l2_table_cache == NULL || r->refcount_block_cache == NULL) { |
94edf3fb KW |
960 | error_setg(errp, "Could not allocate metadata caches"); |
961 | ret = -ENOMEM; | |
962 | goto fail; | |
963 | } | |
964 | ||
965 | /* New interval for cache cleanup timer */ | |
ee55b173 | 966 | r->cache_clean_interval = |
5b0959a7 | 967 | qemu_opt_get_number(opts, QCOW2_OPT_CACHE_CLEAN_INTERVAL, |
e957b50b | 968 | DEFAULT_CACHE_CLEAN_INTERVAL); |
91203f08 AG |
969 | #ifndef CONFIG_LINUX |
970 | if (r->cache_clean_interval != 0) { | |
971 | error_setg(errp, QCOW2_OPT_CACHE_CLEAN_INTERVAL | |
972 | " not supported on this host"); | |
973 | ret = -EINVAL; | |
974 | goto fail; | |
975 | } | |
976 | #endif | |
ee55b173 | 977 | if (r->cache_clean_interval > UINT_MAX) { |
94edf3fb KW |
978 | error_setg(errp, "Cache clean interval too big"); |
979 | ret = -EINVAL; | |
980 | goto fail; | |
981 | } | |
94edf3fb | 982 | |
5b0959a7 | 983 | /* lazy-refcounts; flush if going from enabled to disabled */ |
ee55b173 | 984 | r->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, |
4c75d1a1 | 985 | (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); |
ee55b173 | 986 | if (r->use_lazy_refcounts && s->qcow_version < 3) { |
007dbc39 KW |
987 | error_setg(errp, "Lazy refcounts require a qcow2 image with at least " |
988 | "qemu 1.1 compatibility level"); | |
989 | ret = -EINVAL; | |
990 | goto fail; | |
991 | } | |
4c75d1a1 | 992 | |
5b0959a7 KW |
993 | if (s->use_lazy_refcounts && !r->use_lazy_refcounts) { |
994 | ret = qcow2_mark_clean(bs); | |
995 | if (ret < 0) { | |
996 | error_setg_errno(errp, -ret, "Failed to disable lazy refcounts"); | |
997 | goto fail; | |
998 | } | |
999 | } | |
1000 | ||
007dbc39 | 1001 | /* Overlap check options */ |
4c75d1a1 KW |
1002 | opt_overlap_check = qemu_opt_get(opts, QCOW2_OPT_OVERLAP); |
1003 | opt_overlap_check_template = qemu_opt_get(opts, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1004 | if (opt_overlap_check_template && opt_overlap_check && | |
1005 | strcmp(opt_overlap_check_template, opt_overlap_check)) | |
1006 | { | |
1007 | error_setg(errp, "Conflicting values for qcow2 options '" | |
1008 | QCOW2_OPT_OVERLAP "' ('%s') and '" QCOW2_OPT_OVERLAP_TEMPLATE | |
1009 | "' ('%s')", opt_overlap_check, opt_overlap_check_template); | |
1010 | ret = -EINVAL; | |
1011 | goto fail; | |
1012 | } | |
1013 | if (!opt_overlap_check) { | |
1014 | opt_overlap_check = opt_overlap_check_template ?: "cached"; | |
1015 | } | |
1016 | ||
1017 | if (!strcmp(opt_overlap_check, "none")) { | |
1018 | overlap_check_template = 0; | |
1019 | } else if (!strcmp(opt_overlap_check, "constant")) { | |
1020 | overlap_check_template = QCOW2_OL_CONSTANT; | |
1021 | } else if (!strcmp(opt_overlap_check, "cached")) { | |
1022 | overlap_check_template = QCOW2_OL_CACHED; | |
1023 | } else if (!strcmp(opt_overlap_check, "all")) { | |
1024 | overlap_check_template = QCOW2_OL_ALL; | |
1025 | } else { | |
1026 | error_setg(errp, "Unsupported value '%s' for qcow2 option " | |
1027 | "'overlap-check'. Allowed are any of the following: " | |
1028 | "none, constant, cached, all", opt_overlap_check); | |
1029 | ret = -EINVAL; | |
1030 | goto fail; | |
1031 | } | |
1032 | ||
ee55b173 | 1033 | r->overlap_check = 0; |
4c75d1a1 KW |
1034 | for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) { |
1035 | /* overlap-check defines a template bitmask, but every flag may be | |
1036 | * overwritten through the associated boolean option */ | |
ee55b173 | 1037 | r->overlap_check |= |
4c75d1a1 KW |
1038 | qemu_opt_get_bool(opts, overlap_bool_option_names[i], |
1039 | overlap_check_template & (1 << i)) << i; | |
1040 | } | |
1041 | ||
ee55b173 KW |
1042 | r->discard_passthrough[QCOW2_DISCARD_NEVER] = false; |
1043 | r->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true; | |
1044 | r->discard_passthrough[QCOW2_DISCARD_REQUEST] = | |
007dbc39 KW |
1045 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST, |
1046 | flags & BDRV_O_UNMAP); | |
ee55b173 | 1047 | r->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] = |
007dbc39 | 1048 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true); |
ee55b173 | 1049 | r->discard_passthrough[QCOW2_DISCARD_OTHER] = |
007dbc39 KW |
1050 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false); |
1051 | ||
b25b387f DB |
1052 | switch (s->crypt_method_header) { |
1053 | case QCOW_CRYPT_NONE: | |
1054 | if (encryptfmt) { | |
1055 | error_setg(errp, "No encryption in image header, but options " | |
1056 | "specified format '%s'", encryptfmt); | |
1057 | ret = -EINVAL; | |
1058 | goto fail; | |
1059 | } | |
1060 | break; | |
1061 | ||
1062 | case QCOW_CRYPT_AES: | |
1063 | if (encryptfmt && !g_str_equal(encryptfmt, "aes")) { | |
1064 | error_setg(errp, | |
1065 | "Header reported 'aes' encryption format but " | |
1066 | "options specify '%s'", encryptfmt); | |
1067 | ret = -EINVAL; | |
1068 | goto fail; | |
1069 | } | |
796d3239 MA |
1070 | qdict_put_str(encryptopts, "format", "qcow"); |
1071 | r->crypto_opts = block_crypto_open_opts_init(encryptopts, errp); | |
b25b387f DB |
1072 | break; |
1073 | ||
4652b8f3 DB |
1074 | case QCOW_CRYPT_LUKS: |
1075 | if (encryptfmt && !g_str_equal(encryptfmt, "luks")) { | |
1076 | error_setg(errp, | |
1077 | "Header reported 'luks' encryption format but " | |
1078 | "options specify '%s'", encryptfmt); | |
1079 | ret = -EINVAL; | |
1080 | goto fail; | |
1081 | } | |
796d3239 MA |
1082 | qdict_put_str(encryptopts, "format", "luks"); |
1083 | r->crypto_opts = block_crypto_open_opts_init(encryptopts, errp); | |
4652b8f3 DB |
1084 | break; |
1085 | ||
b25b387f DB |
1086 | default: |
1087 | error_setg(errp, "Unsupported encryption method %d", | |
1088 | s->crypt_method_header); | |
1089 | break; | |
1090 | } | |
1091 | if (s->crypt_method_header != QCOW_CRYPT_NONE && !r->crypto_opts) { | |
1092 | ret = -EINVAL; | |
1093 | goto fail; | |
1094 | } | |
1095 | ||
4c75d1a1 KW |
1096 | ret = 0; |
1097 | fail: | |
cb3e7f08 | 1098 | qobject_unref(encryptopts); |
94edf3fb KW |
1099 | qemu_opts_del(opts); |
1100 | opts = NULL; | |
ee55b173 KW |
1101 | return ret; |
1102 | } | |
1103 | ||
1104 | static void qcow2_update_options_commit(BlockDriverState *bs, | |
1105 | Qcow2ReopenState *r) | |
1106 | { | |
1107 | BDRVQcow2State *s = bs->opaque; | |
1108 | int i; | |
1109 | ||
5b0959a7 | 1110 | if (s->l2_table_cache) { |
e64d4072 | 1111 | qcow2_cache_destroy(s->l2_table_cache); |
5b0959a7 KW |
1112 | } |
1113 | if (s->refcount_block_cache) { | |
e64d4072 | 1114 | qcow2_cache_destroy(s->refcount_block_cache); |
5b0959a7 | 1115 | } |
ee55b173 KW |
1116 | s->l2_table_cache = r->l2_table_cache; |
1117 | s->refcount_block_cache = r->refcount_block_cache; | |
3c2e511a | 1118 | s->l2_slice_size = r->l2_slice_size; |
ee55b173 KW |
1119 | |
1120 | s->overlap_check = r->overlap_check; | |
1121 | s->use_lazy_refcounts = r->use_lazy_refcounts; | |
1122 | ||
1123 | for (i = 0; i < QCOW2_DISCARD_MAX; i++) { | |
1124 | s->discard_passthrough[i] = r->discard_passthrough[i]; | |
1125 | } | |
1126 | ||
5b0959a7 KW |
1127 | if (s->cache_clean_interval != r->cache_clean_interval) { |
1128 | cache_clean_timer_del(bs); | |
1129 | s->cache_clean_interval = r->cache_clean_interval; | |
1130 | cache_clean_timer_init(bs, bdrv_get_aio_context(bs)); | |
1131 | } | |
b25b387f DB |
1132 | |
1133 | qapi_free_QCryptoBlockOpenOptions(s->crypto_opts); | |
1134 | s->crypto_opts = r->crypto_opts; | |
ee55b173 KW |
1135 | } |
1136 | ||
1137 | static void qcow2_update_options_abort(BlockDriverState *bs, | |
1138 | Qcow2ReopenState *r) | |
1139 | { | |
1140 | if (r->l2_table_cache) { | |
e64d4072 | 1141 | qcow2_cache_destroy(r->l2_table_cache); |
ee55b173 KW |
1142 | } |
1143 | if (r->refcount_block_cache) { | |
e64d4072 | 1144 | qcow2_cache_destroy(r->refcount_block_cache); |
ee55b173 | 1145 | } |
b25b387f | 1146 | qapi_free_QCryptoBlockOpenOptions(r->crypto_opts); |
ee55b173 KW |
1147 | } |
1148 | ||
1149 | static int qcow2_update_options(BlockDriverState *bs, QDict *options, | |
1150 | int flags, Error **errp) | |
1151 | { | |
1152 | Qcow2ReopenState r = {}; | |
1153 | int ret; | |
1154 | ||
1155 | ret = qcow2_update_options_prepare(bs, &r, options, flags, errp); | |
1156 | if (ret >= 0) { | |
1157 | qcow2_update_options_commit(bs, &r); | |
1158 | } else { | |
1159 | qcow2_update_options_abort(bs, &r); | |
1160 | } | |
94edf3fb | 1161 | |
4c75d1a1 KW |
1162 | return ret; |
1163 | } | |
1164 | ||
1fafcd93 PB |
1165 | /* Called with s->lock held. */ |
1166 | static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, | |
1167 | int flags, Error **errp) | |
585f8587 | 1168 | { |
ff99129a | 1169 | BDRVQcow2State *s = bs->opaque; |
6d33e8e7 KW |
1170 | unsigned int len, i; |
1171 | int ret = 0; | |
585f8587 | 1172 | QCowHeader header; |
74c4510a | 1173 | Error *local_err = NULL; |
9b80ddf3 | 1174 | uint64_t ext_end; |
2cf7cfa1 | 1175 | uint64_t l1_vm_state_index; |
88ddffae | 1176 | bool update_header = false; |
585f8587 | 1177 | |
cf2ab8fc | 1178 | ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); |
6d85a57e | 1179 | if (ret < 0) { |
3ef6c40a | 1180 | error_setg_errno(errp, -ret, "Could not read qcow2 header"); |
585f8587 | 1181 | goto fail; |
6d85a57e | 1182 | } |
3b698f52 PM |
1183 | header.magic = be32_to_cpu(header.magic); |
1184 | header.version = be32_to_cpu(header.version); | |
1185 | header.backing_file_offset = be64_to_cpu(header.backing_file_offset); | |
1186 | header.backing_file_size = be32_to_cpu(header.backing_file_size); | |
1187 | header.size = be64_to_cpu(header.size); | |
1188 | header.cluster_bits = be32_to_cpu(header.cluster_bits); | |
1189 | header.crypt_method = be32_to_cpu(header.crypt_method); | |
1190 | header.l1_table_offset = be64_to_cpu(header.l1_table_offset); | |
1191 | header.l1_size = be32_to_cpu(header.l1_size); | |
1192 | header.refcount_table_offset = be64_to_cpu(header.refcount_table_offset); | |
1193 | header.refcount_table_clusters = | |
1194 | be32_to_cpu(header.refcount_table_clusters); | |
1195 | header.snapshots_offset = be64_to_cpu(header.snapshots_offset); | |
1196 | header.nb_snapshots = be32_to_cpu(header.nb_snapshots); | |
3b46e624 | 1197 | |
e8cdcec1 | 1198 | if (header.magic != QCOW_MAGIC) { |
3ef6c40a | 1199 | error_setg(errp, "Image is not in qcow2 format"); |
76abe407 | 1200 | ret = -EINVAL; |
585f8587 | 1201 | goto fail; |
6d85a57e | 1202 | } |
6744cbab | 1203 | if (header.version < 2 || header.version > 3) { |
a55448b3 | 1204 | error_setg(errp, "Unsupported qcow2 version %" PRIu32, header.version); |
6744cbab KW |
1205 | ret = -ENOTSUP; |
1206 | goto fail; | |
1207 | } | |
1208 | ||
1209 | s->qcow_version = header.version; | |
1210 | ||
24342f2c KW |
1211 | /* Initialise cluster size */ |
1212 | if (header.cluster_bits < MIN_CLUSTER_BITS || | |
1213 | header.cluster_bits > MAX_CLUSTER_BITS) { | |
521b2b5d HR |
1214 | error_setg(errp, "Unsupported cluster size: 2^%" PRIu32, |
1215 | header.cluster_bits); | |
24342f2c KW |
1216 | ret = -EINVAL; |
1217 | goto fail; | |
1218 | } | |
1219 | ||
1220 | s->cluster_bits = header.cluster_bits; | |
1221 | s->cluster_size = 1 << s->cluster_bits; | |
a35f87f5 | 1222 | s->cluster_sectors = 1 << (s->cluster_bits - BDRV_SECTOR_BITS); |
24342f2c | 1223 | |
6744cbab KW |
1224 | /* Initialise version 3 header fields */ |
1225 | if (header.version == 2) { | |
1226 | header.incompatible_features = 0; | |
1227 | header.compatible_features = 0; | |
1228 | header.autoclear_features = 0; | |
1229 | header.refcount_order = 4; | |
1230 | header.header_length = 72; | |
1231 | } else { | |
3b698f52 PM |
1232 | header.incompatible_features = |
1233 | be64_to_cpu(header.incompatible_features); | |
1234 | header.compatible_features = be64_to_cpu(header.compatible_features); | |
1235 | header.autoclear_features = be64_to_cpu(header.autoclear_features); | |
1236 | header.refcount_order = be32_to_cpu(header.refcount_order); | |
1237 | header.header_length = be32_to_cpu(header.header_length); | |
24342f2c KW |
1238 | |
1239 | if (header.header_length < 104) { | |
1240 | error_setg(errp, "qcow2 header too short"); | |
1241 | ret = -EINVAL; | |
1242 | goto fail; | |
1243 | } | |
1244 | } | |
1245 | ||
1246 | if (header.header_length > s->cluster_size) { | |
1247 | error_setg(errp, "qcow2 header exceeds cluster size"); | |
1248 | ret = -EINVAL; | |
1249 | goto fail; | |
6744cbab KW |
1250 | } |
1251 | ||
1252 | if (header.header_length > sizeof(header)) { | |
1253 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
1254 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
cf2ab8fc | 1255 | ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields, |
6744cbab KW |
1256 | s->unknown_header_fields_size); |
1257 | if (ret < 0) { | |
3ef6c40a HR |
1258 | error_setg_errno(errp, -ret, "Could not read unknown qcow2 header " |
1259 | "fields"); | |
6744cbab KW |
1260 | goto fail; |
1261 | } | |
1262 | } | |
1263 | ||
a1b3955c KW |
1264 | if (header.backing_file_offset > s->cluster_size) { |
1265 | error_setg(errp, "Invalid backing file offset"); | |
1266 | ret = -EINVAL; | |
1267 | goto fail; | |
1268 | } | |
1269 | ||
cfcc4c62 KW |
1270 | if (header.backing_file_offset) { |
1271 | ext_end = header.backing_file_offset; | |
1272 | } else { | |
1273 | ext_end = 1 << header.cluster_bits; | |
1274 | } | |
1275 | ||
6744cbab KW |
1276 | /* Handle feature bits */ |
1277 | s->incompatible_features = header.incompatible_features; | |
1278 | s->compatible_features = header.compatible_features; | |
1279 | s->autoclear_features = header.autoclear_features; | |
1280 | ||
c61d0004 | 1281 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { |
cfcc4c62 KW |
1282 | void *feature_table = NULL; |
1283 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
88ddffae | 1284 | &feature_table, flags, NULL, NULL); |
a55448b3 | 1285 | report_unsupported_feature(errp, feature_table, |
c61d0004 SH |
1286 | s->incompatible_features & |
1287 | ~QCOW2_INCOMPAT_MASK); | |
6744cbab | 1288 | ret = -ENOTSUP; |
c5a33ee9 | 1289 | g_free(feature_table); |
6744cbab KW |
1290 | goto fail; |
1291 | } | |
1292 | ||
69c98726 HR |
1293 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { |
1294 | /* Corrupt images may not be written to unless they are being repaired | |
1295 | */ | |
1296 | if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) { | |
3ef6c40a HR |
1297 | error_setg(errp, "qcow2: Image is corrupt; cannot be opened " |
1298 | "read/write"); | |
69c98726 HR |
1299 | ret = -EACCES; |
1300 | goto fail; | |
1301 | } | |
1302 | } | |
1303 | ||
6744cbab | 1304 | /* Check support for various header values */ |
b72faf9f HR |
1305 | if (header.refcount_order > 6) { |
1306 | error_setg(errp, "Reference count entry width too large; may not " | |
1307 | "exceed 64 bits"); | |
1308 | ret = -EINVAL; | |
e8cdcec1 KW |
1309 | goto fail; |
1310 | } | |
b6481f37 | 1311 | s->refcount_order = header.refcount_order; |
346a53df HR |
1312 | s->refcount_bits = 1 << s->refcount_order; |
1313 | s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1); | |
1314 | s->refcount_max += s->refcount_max - 1; | |
6744cbab | 1315 | |
585f8587 | 1316 | s->crypt_method_header = header.crypt_method; |
6d85a57e | 1317 | if (s->crypt_method_header) { |
e6ff69bf DB |
1318 | if (bdrv_uses_whitelist() && |
1319 | s->crypt_method_header == QCOW_CRYPT_AES) { | |
8c0dcbc4 DB |
1320 | error_setg(errp, |
1321 | "Use of AES-CBC encrypted qcow2 images is no longer " | |
1322 | "supported in system emulators"); | |
1323 | error_append_hint(errp, | |
1324 | "You can use 'qemu-img convert' to convert your " | |
1325 | "image to an alternative supported format, such " | |
1326 | "as unencrypted qcow2, or raw with the LUKS " | |
1327 | "format instead.\n"); | |
1328 | ret = -ENOSYS; | |
1329 | goto fail; | |
e6ff69bf DB |
1330 | } |
1331 | ||
4652b8f3 DB |
1332 | if (s->crypt_method_header == QCOW_CRYPT_AES) { |
1333 | s->crypt_physical_offset = false; | |
1334 | } else { | |
1335 | /* Assuming LUKS and any future crypt methods we | |
1336 | * add will all use physical offsets, due to the | |
1337 | * fact that the alternative is insecure... */ | |
1338 | s->crypt_physical_offset = true; | |
1339 | } | |
1340 | ||
54115412 | 1341 | bs->encrypted = true; |
6d85a57e | 1342 | } |
24342f2c | 1343 | |
585f8587 FB |
1344 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
1345 | s->l2_size = 1 << s->l2_bits; | |
1d13d654 HR |
1346 | /* 2^(s->refcount_order - 3) is the refcount width in bytes */ |
1347 | s->refcount_block_bits = s->cluster_bits - (s->refcount_order - 3); | |
1348 | s->refcount_block_size = 1 << s->refcount_block_bits; | |
bd016b91 | 1349 | bs->total_sectors = header.size / BDRV_SECTOR_SIZE; |
585f8587 FB |
1350 | s->csize_shift = (62 - (s->cluster_bits - 8)); |
1351 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
1352 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
5dab2fad | 1353 | |
585f8587 | 1354 | s->refcount_table_offset = header.refcount_table_offset; |
5fafdf24 | 1355 | s->refcount_table_size = |
585f8587 FB |
1356 | header.refcount_table_clusters << (s->cluster_bits - 3); |
1357 | ||
951053a9 AG |
1358 | if (header.refcount_table_clusters == 0 && !(flags & BDRV_O_CHECK)) { |
1359 | error_setg(errp, "Image does not contain a reference count table"); | |
1360 | ret = -EINVAL; | |
1361 | goto fail; | |
1362 | } | |
1363 | ||
0cf0e598 AG |
1364 | ret = qcow2_validate_table(bs, s->refcount_table_offset, |
1365 | header.refcount_table_clusters, | |
1366 | s->cluster_size, QCOW_MAX_REFTABLE_SIZE, | |
1367 | "Reference count table", errp); | |
8c7de283 | 1368 | if (ret < 0) { |
8c7de283 KW |
1369 | goto fail; |
1370 | } | |
1371 | ||
0cf0e598 AG |
1372 | /* The total size in bytes of the snapshot table is checked in |
1373 | * qcow2_read_snapshots() because the size of each snapshot is | |
1374 | * variable and we don't know it yet. | |
1375 | * Here we only check the offset and number of snapshots. */ | |
1376 | ret = qcow2_validate_table(bs, header.snapshots_offset, | |
1377 | header.nb_snapshots, | |
1378 | sizeof(QCowSnapshotHeader), | |
1379 | sizeof(QCowSnapshotHeader) * QCOW_MAX_SNAPSHOTS, | |
1380 | "Snapshot table", errp); | |
ce48f2f4 | 1381 | if (ret < 0) { |
ce48f2f4 KW |
1382 | goto fail; |
1383 | } | |
1384 | ||
585f8587 | 1385 | /* read the level 1 table */ |
0cf0e598 AG |
1386 | ret = qcow2_validate_table(bs, header.l1_table_offset, |
1387 | header.l1_size, sizeof(uint64_t), | |
1388 | QCOW_MAX_L1_SIZE, "Active L1 table", errp); | |
1389 | if (ret < 0) { | |
2d51c32c KW |
1390 | goto fail; |
1391 | } | |
585f8587 | 1392 | s->l1_size = header.l1_size; |
0cf0e598 | 1393 | s->l1_table_offset = header.l1_table_offset; |
2cf7cfa1 KW |
1394 | |
1395 | l1_vm_state_index = size_to_l1(s, header.size); | |
1396 | if (l1_vm_state_index > INT_MAX) { | |
3ef6c40a | 1397 | error_setg(errp, "Image is too big"); |
2cf7cfa1 KW |
1398 | ret = -EFBIG; |
1399 | goto fail; | |
1400 | } | |
1401 | s->l1_vm_state_index = l1_vm_state_index; | |
1402 | ||
585f8587 FB |
1403 | /* the L1 table must contain at least enough entries to put |
1404 | header.size bytes */ | |
6d85a57e | 1405 | if (s->l1_size < s->l1_vm_state_index) { |
3ef6c40a | 1406 | error_setg(errp, "L1 table is too small"); |
6d85a57e | 1407 | ret = -EINVAL; |
585f8587 | 1408 | goto fail; |
6d85a57e | 1409 | } |
2d51c32c | 1410 | |
d191d12d | 1411 | if (s->l1_size > 0) { |
9a4f4c31 | 1412 | s->l1_table = qemu_try_blockalign(bs->file->bs, |
9e029689 | 1413 | ROUND_UP(s->l1_size * sizeof(uint64_t), 512)); |
de82815d KW |
1414 | if (s->l1_table == NULL) { |
1415 | error_setg(errp, "Could not allocate L1 table"); | |
1416 | ret = -ENOMEM; | |
1417 | goto fail; | |
1418 | } | |
cf2ab8fc | 1419 | ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, |
6d85a57e JS |
1420 | s->l1_size * sizeof(uint64_t)); |
1421 | if (ret < 0) { | |
3ef6c40a | 1422 | error_setg_errno(errp, -ret, "Could not read L1 table"); |
d191d12d | 1423 | goto fail; |
6d85a57e | 1424 | } |
d191d12d | 1425 | for(i = 0;i < s->l1_size; i++) { |
3b698f52 | 1426 | s->l1_table[i] = be64_to_cpu(s->l1_table[i]); |
d191d12d | 1427 | } |
585f8587 | 1428 | } |
29c1a730 | 1429 | |
94edf3fb KW |
1430 | /* Parse driver-specific options */ |
1431 | ret = qcow2_update_options(bs, options, flags, errp); | |
90efa0ea KW |
1432 | if (ret < 0) { |
1433 | goto fail; | |
1434 | } | |
1435 | ||
06d9260f | 1436 | s->flags = flags; |
3b46e624 | 1437 | |
6d85a57e JS |
1438 | ret = qcow2_refcount_init(bs); |
1439 | if (ret != 0) { | |
3ef6c40a | 1440 | error_setg_errno(errp, -ret, "Could not initialize refcount handling"); |
585f8587 | 1441 | goto fail; |
6d85a57e | 1442 | } |
585f8587 | 1443 | |
72cf2d4f | 1444 | QLIST_INIT(&s->cluster_allocs); |
0b919fae | 1445 | QTAILQ_INIT(&s->discards); |
f214978a | 1446 | |
9b80ddf3 | 1447 | /* read qcow2 extensions */ |
3ef6c40a | 1448 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL, |
88ddffae | 1449 | flags, &update_header, &local_err)) { |
3ef6c40a | 1450 | error_propagate(errp, local_err); |
6d85a57e | 1451 | ret = -EINVAL; |
9b80ddf3 | 1452 | goto fail; |
6d85a57e | 1453 | } |
9b80ddf3 | 1454 | |
4652b8f3 DB |
1455 | /* qcow2_read_extension may have set up the crypto context |
1456 | * if the crypt method needs a header region, some methods | |
1457 | * don't need header extensions, so must check here | |
1458 | */ | |
1459 | if (s->crypt_method_header && !s->crypto) { | |
1460 | if (s->crypt_method_header == QCOW_CRYPT_AES) { | |
1461 | unsigned int cflags = 0; | |
1462 | if (flags & BDRV_O_NO_IO) { | |
1463 | cflags |= QCRYPTO_BLOCK_OPEN_NO_IO; | |
1464 | } | |
1cd9a787 | 1465 | s->crypto = qcrypto_block_open(s->crypto_opts, "encrypt.", |
c972fa12 | 1466 | NULL, NULL, cflags, 1, errp); |
4652b8f3 DB |
1467 | if (!s->crypto) { |
1468 | ret = -EINVAL; | |
1469 | goto fail; | |
1470 | } | |
1471 | } else if (!(flags & BDRV_O_NO_IO)) { | |
1472 | error_setg(errp, "Missing CRYPTO header for crypt method %d", | |
1473 | s->crypt_method_header); | |
b25b387f DB |
1474 | ret = -EINVAL; |
1475 | goto fail; | |
1476 | } | |
1477 | } | |
1478 | ||
585f8587 FB |
1479 | /* read the backing file name */ |
1480 | if (header.backing_file_offset != 0) { | |
1481 | len = header.backing_file_size; | |
9a29e18f | 1482 | if (len > MIN(1023, s->cluster_size - header.backing_file_offset) || |
e729fa6a | 1483 | len >= sizeof(bs->backing_file)) { |
6d33e8e7 KW |
1484 | error_setg(errp, "Backing file name too long"); |
1485 | ret = -EINVAL; | |
1486 | goto fail; | |
6d85a57e | 1487 | } |
cf2ab8fc | 1488 | ret = bdrv_pread(bs->file, header.backing_file_offset, |
998c2019 | 1489 | bs->auto_backing_file, len); |
6d85a57e | 1490 | if (ret < 0) { |
3ef6c40a | 1491 | error_setg_errno(errp, -ret, "Could not read backing file name"); |
585f8587 | 1492 | goto fail; |
6d85a57e | 1493 | } |
998c2019 HR |
1494 | bs->auto_backing_file[len] = '\0'; |
1495 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), | |
1496 | bs->auto_backing_file); | |
1497 | s->image_backing_file = g_strdup(bs->auto_backing_file); | |
585f8587 | 1498 | } |
42deb29f | 1499 | |
11b128f4 KW |
1500 | /* Internal snapshots */ |
1501 | s->snapshots_offset = header.snapshots_offset; | |
1502 | s->nb_snapshots = header.nb_snapshots; | |
1503 | ||
42deb29f KW |
1504 | ret = qcow2_read_snapshots(bs); |
1505 | if (ret < 0) { | |
3ef6c40a | 1506 | error_setg_errno(errp, -ret, "Could not read snapshots"); |
585f8587 | 1507 | goto fail; |
6d85a57e | 1508 | } |
585f8587 | 1509 | |
af7b708d | 1510 | /* Clear unknown autoclear feature bits */ |
88ddffae | 1511 | update_header |= s->autoclear_features & ~QCOW2_AUTOCLEAR_MASK; |
d1258dd0 VSO |
1512 | update_header = |
1513 | update_header && !bs->read_only && !(flags & BDRV_O_INACTIVE); | |
1514 | if (update_header) { | |
88ddffae | 1515 | s->autoclear_features &= QCOW2_AUTOCLEAR_MASK; |
d1258dd0 VSO |
1516 | } |
1517 | ||
9c98f145 VSO |
1518 | /* == Handle persistent dirty bitmaps == |
1519 | * | |
1520 | * We want load dirty bitmaps in three cases: | |
1521 | * | |
1522 | * 1. Normal open of the disk in active mode, not related to invalidation | |
1523 | * after migration. | |
1524 | * | |
1525 | * 2. Invalidation of the target vm after pre-copy phase of migration, if | |
1526 | * bitmaps are _not_ migrating through migration channel, i.e. | |
1527 | * 'dirty-bitmaps' capability is disabled. | |
1528 | * | |
1529 | * 3. Invalidation of source vm after failed or canceled migration. | |
1530 | * This is a very interesting case. There are two possible types of | |
1531 | * bitmaps: | |
1532 | * | |
1533 | * A. Stored on inactivation and removed. They should be loaded from the | |
1534 | * image. | |
1535 | * | |
1536 | * B. Not stored: not-persistent bitmaps and bitmaps, migrated through | |
1537 | * the migration channel (with dirty-bitmaps capability). | |
1538 | * | |
1539 | * On the other hand, there are two possible sub-cases: | |
1540 | * | |
1541 | * 3.1 disk was changed by somebody else while were inactive. In this | |
1542 | * case all in-RAM dirty bitmaps (both persistent and not) are | |
1543 | * definitely invalid. And we don't have any method to determine | |
1544 | * this. | |
1545 | * | |
1546 | * Simple and safe thing is to just drop all the bitmaps of type B on | |
1547 | * inactivation. But in this case we lose bitmaps in valid 4.2 case. | |
1548 | * | |
1549 | * On the other hand, resuming source vm, if disk was already changed | |
1550 | * is a bad thing anyway: not only bitmaps, the whole vm state is | |
1551 | * out of sync with disk. | |
1552 | * | |
1553 | * This means, that user or management tool, who for some reason | |
1554 | * decided to resume source vm, after disk was already changed by | |
1555 | * target vm, should at least drop all dirty bitmaps by hand. | |
1556 | * | |
1557 | * So, we can ignore this case for now, but TODO: "generation" | |
1558 | * extension for qcow2, to determine, that image was changed after | |
1559 | * last inactivation. And if it is changed, we will drop (or at least | |
1560 | * mark as 'invalid' all the bitmaps of type B, both persistent | |
1561 | * and not). | |
1562 | * | |
1563 | * 3.2 disk was _not_ changed while were inactive. Bitmaps may be saved | |
1564 | * to disk ('dirty-bitmaps' capability disabled), or not saved | |
1565 | * ('dirty-bitmaps' capability enabled), but we don't need to care | |
1566 | * of: let's load bitmaps as always: stored bitmaps will be loaded, | |
1567 | * and not stored has flag IN_USE=1 in the image and will be skipped | |
1568 | * on loading. | |
1569 | * | |
1570 | * One remaining possible case when we don't want load bitmaps: | |
1571 | * | |
1572 | * 4. Open disk in inactive mode in target vm (bitmaps are migrating or | |
1573 | * will be loaded on invalidation, no needs try loading them before) | |
1574 | */ | |
1575 | ||
1576 | if (!(bdrv_get_flags(bs) & BDRV_O_INACTIVE)) { | |
1577 | /* It's case 1, 2 or 3.2. Or 3.1 which is BUG in management layer. */ | |
1578 | bool header_updated = qcow2_load_dirty_bitmaps(bs, &local_err); | |
1579 | ||
1580 | update_header = update_header && !header_updated; | |
d1258dd0 VSO |
1581 | } |
1582 | if (local_err != NULL) { | |
1583 | error_propagate(errp, local_err); | |
1584 | ret = -EINVAL; | |
1585 | goto fail; | |
1586 | } | |
1587 | ||
1588 | if (update_header) { | |
af7b708d SH |
1589 | ret = qcow2_update_header(bs); |
1590 | if (ret < 0) { | |
3ef6c40a | 1591 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); |
af7b708d SH |
1592 | goto fail; |
1593 | } | |
1594 | } | |
1595 | ||
e24d813b | 1596 | bs->supported_zero_flags = header.version >= 3 ? BDRV_REQ_MAY_UNMAP : 0; |
68d100e9 | 1597 | |
c61d0004 | 1598 | /* Repair image if dirty */ |
04c01a5c | 1599 | if (!(flags & (BDRV_O_CHECK | BDRV_O_INACTIVE)) && !bs->read_only && |
058f8f16 | 1600 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { |
c61d0004 SH |
1601 | BdrvCheckResult result = {0}; |
1602 | ||
2fd61638 PB |
1603 | ret = qcow2_co_check_locked(bs, &result, |
1604 | BDRV_FIX_ERRORS | BDRV_FIX_LEAKS); | |
791fff50 HR |
1605 | if (ret < 0 || result.check_errors) { |
1606 | if (ret >= 0) { | |
1607 | ret = -EIO; | |
1608 | } | |
3ef6c40a | 1609 | error_setg_errno(errp, -ret, "Could not repair dirty image"); |
c61d0004 SH |
1610 | goto fail; |
1611 | } | |
1612 | } | |
1613 | ||
585f8587 | 1614 | #ifdef DEBUG_ALLOC |
6cbc3031 PH |
1615 | { |
1616 | BdrvCheckResult result = {0}; | |
b35278f7 | 1617 | qcow2_check_refcounts(bs, &result, 0); |
6cbc3031 | 1618 | } |
585f8587 | 1619 | #endif |
ceb029cd VSO |
1620 | |
1621 | qemu_co_queue_init(&s->compress_wait_queue); | |
1622 | ||
6d85a57e | 1623 | return ret; |
585f8587 FB |
1624 | |
1625 | fail: | |
6744cbab | 1626 | g_free(s->unknown_header_fields); |
75bab85c | 1627 | cleanup_unknown_header_ext(bs); |
ed6ccf0f KW |
1628 | qcow2_free_snapshots(bs); |
1629 | qcow2_refcount_close(bs); | |
de82815d | 1630 | qemu_vfree(s->l1_table); |
cf93980e HR |
1631 | /* else pre-write overlap checks in cache_destroy may crash */ |
1632 | s->l1_table = NULL; | |
279621c0 | 1633 | cache_clean_timer_del(bs); |
29c1a730 | 1634 | if (s->l2_table_cache) { |
e64d4072 | 1635 | qcow2_cache_destroy(s->l2_table_cache); |
29c1a730 | 1636 | } |
c5a33ee9 | 1637 | if (s->refcount_block_cache) { |
e64d4072 | 1638 | qcow2_cache_destroy(s->refcount_block_cache); |
c5a33ee9 | 1639 | } |
b25b387f DB |
1640 | qcrypto_block_free(s->crypto); |
1641 | qapi_free_QCryptoBlockOpenOptions(s->crypto_opts); | |
6d85a57e | 1642 | return ret; |
585f8587 FB |
1643 | } |
1644 | ||
1fafcd93 PB |
1645 | typedef struct QCow2OpenCo { |
1646 | BlockDriverState *bs; | |
1647 | QDict *options; | |
1648 | int flags; | |
1649 | Error **errp; | |
1650 | int ret; | |
1651 | } QCow2OpenCo; | |
1652 | ||
1653 | static void coroutine_fn qcow2_open_entry(void *opaque) | |
1654 | { | |
1655 | QCow2OpenCo *qoc = opaque; | |
1656 | BDRVQcow2State *s = qoc->bs->opaque; | |
1657 | ||
1658 | qemu_co_mutex_lock(&s->lock); | |
1659 | qoc->ret = qcow2_do_open(qoc->bs, qoc->options, qoc->flags, qoc->errp); | |
1660 | qemu_co_mutex_unlock(&s->lock); | |
1661 | } | |
1662 | ||
4e4bf5c4 KW |
1663 | static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, |
1664 | Error **errp) | |
1665 | { | |
1fafcd93 PB |
1666 | BDRVQcow2State *s = bs->opaque; |
1667 | QCow2OpenCo qoc = { | |
1668 | .bs = bs, | |
1669 | .options = options, | |
1670 | .flags = flags, | |
1671 | .errp = errp, | |
1672 | .ret = -EINPROGRESS | |
1673 | }; | |
1674 | ||
4e4bf5c4 KW |
1675 | bs->file = bdrv_open_child(NULL, options, "file", bs, &child_file, |
1676 | false, errp); | |
1677 | if (!bs->file) { | |
1678 | return -EINVAL; | |
1679 | } | |
1680 | ||
1fafcd93 PB |
1681 | /* Initialise locks */ |
1682 | qemu_co_mutex_init(&s->lock); | |
1683 | ||
1684 | if (qemu_in_coroutine()) { | |
1685 | /* From bdrv_co_create. */ | |
1686 | qcow2_open_entry(&qoc); | |
1687 | } else { | |
4720cbee | 1688 | assert(qemu_get_current_aio_context() == qemu_get_aio_context()); |
1fafcd93 PB |
1689 | qemu_coroutine_enter(qemu_coroutine_create(qcow2_open_entry, &qoc)); |
1690 | BDRV_POLL_WHILE(bs, qoc.ret == -EINPROGRESS); | |
1691 | } | |
1692 | return qoc.ret; | |
4e4bf5c4 KW |
1693 | } |
1694 | ||
3baca891 | 1695 | static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd | 1696 | { |
ff99129a | 1697 | BDRVQcow2State *s = bs->opaque; |
d34682cd | 1698 | |
a84178cc EB |
1699 | if (bs->encrypted) { |
1700 | /* Encryption works on a sector granularity */ | |
6f8f015c | 1701 | bs->bl.request_alignment = qcrypto_block_get_sector_size(s->crypto); |
a84178cc | 1702 | } |
cf081fca | 1703 | bs->bl.pwrite_zeroes_alignment = s->cluster_size; |
ecdbead6 | 1704 | bs->bl.pdiscard_alignment = s->cluster_size; |
d34682cd KW |
1705 | } |
1706 | ||
21d82ac9 JC |
1707 | static int qcow2_reopen_prepare(BDRVReopenState *state, |
1708 | BlockReopenQueue *queue, Error **errp) | |
1709 | { | |
5b0959a7 | 1710 | Qcow2ReopenState *r; |
4c2e5f8f KW |
1711 | int ret; |
1712 | ||
5b0959a7 KW |
1713 | r = g_new0(Qcow2ReopenState, 1); |
1714 | state->opaque = r; | |
1715 | ||
1716 | ret = qcow2_update_options_prepare(state->bs, r, state->options, | |
1717 | state->flags, errp); | |
1718 | if (ret < 0) { | |
1719 | goto fail; | |
1720 | } | |
1721 | ||
1722 | /* We need to write out any unwritten data if we reopen read-only. */ | |
4c2e5f8f | 1723 | if ((state->flags & BDRV_O_RDWR) == 0) { |
169b8793 VSO |
1724 | ret = qcow2_reopen_bitmaps_ro(state->bs, errp); |
1725 | if (ret < 0) { | |
1726 | goto fail; | |
1727 | } | |
1728 | ||
4c2e5f8f KW |
1729 | ret = bdrv_flush(state->bs); |
1730 | if (ret < 0) { | |
5b0959a7 | 1731 | goto fail; |
4c2e5f8f KW |
1732 | } |
1733 | ||
1734 | ret = qcow2_mark_clean(state->bs); | |
1735 | if (ret < 0) { | |
5b0959a7 | 1736 | goto fail; |
4c2e5f8f KW |
1737 | } |
1738 | } | |
1739 | ||
21d82ac9 | 1740 | return 0; |
5b0959a7 KW |
1741 | |
1742 | fail: | |
1743 | qcow2_update_options_abort(state->bs, r); | |
1744 | g_free(r); | |
1745 | return ret; | |
1746 | } | |
1747 | ||
1748 | static void qcow2_reopen_commit(BDRVReopenState *state) | |
1749 | { | |
1750 | qcow2_update_options_commit(state->bs, state->opaque); | |
1751 | g_free(state->opaque); | |
1752 | } | |
1753 | ||
1754 | static void qcow2_reopen_abort(BDRVReopenState *state) | |
1755 | { | |
1756 | qcow2_update_options_abort(state->bs, state->opaque); | |
1757 | g_free(state->opaque); | |
21d82ac9 JC |
1758 | } |
1759 | ||
5365f44d KW |
1760 | static void qcow2_join_options(QDict *options, QDict *old_options) |
1761 | { | |
1762 | bool has_new_overlap_template = | |
1763 | qdict_haskey(options, QCOW2_OPT_OVERLAP) || | |
1764 | qdict_haskey(options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1765 | bool has_new_total_cache_size = | |
1766 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE); | |
1767 | bool has_all_cache_options; | |
1768 | ||
1769 | /* New overlap template overrides all old overlap options */ | |
1770 | if (has_new_overlap_template) { | |
1771 | qdict_del(old_options, QCOW2_OPT_OVERLAP); | |
1772 | qdict_del(old_options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1773 | qdict_del(old_options, QCOW2_OPT_OVERLAP_MAIN_HEADER); | |
1774 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L1); | |
1775 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L2); | |
1776 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_TABLE); | |
1777 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK); | |
1778 | qdict_del(old_options, QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE); | |
1779 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L1); | |
1780 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L2); | |
1781 | } | |
1782 | ||
1783 | /* New total cache size overrides all old options */ | |
1784 | if (qdict_haskey(options, QCOW2_OPT_CACHE_SIZE)) { | |
1785 | qdict_del(old_options, QCOW2_OPT_L2_CACHE_SIZE); | |
1786 | qdict_del(old_options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1787 | } | |
1788 | ||
1789 | qdict_join(options, old_options, false); | |
1790 | ||
1791 | /* | |
1792 | * If after merging all cache size options are set, an old total size is | |
1793 | * overwritten. Do keep all options, however, if all three are new. The | |
1794 | * resulting error message is what we want to happen. | |
1795 | */ | |
1796 | has_all_cache_options = | |
1797 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE) || | |
1798 | qdict_haskey(options, QCOW2_OPT_L2_CACHE_SIZE) || | |
1799 | qdict_haskey(options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1800 | ||
1801 | if (has_all_cache_options && !has_new_total_cache_size) { | |
1802 | qdict_del(options, QCOW2_OPT_CACHE_SIZE); | |
1803 | } | |
1804 | } | |
1805 | ||
a320fb04 EB |
1806 | static int coroutine_fn qcow2_co_block_status(BlockDriverState *bs, |
1807 | bool want_zero, | |
1808 | int64_t offset, int64_t count, | |
1809 | int64_t *pnum, int64_t *map, | |
1810 | BlockDriverState **file) | |
585f8587 | 1811 | { |
ff99129a | 1812 | BDRVQcow2State *s = bs->opaque; |
585f8587 | 1813 | uint64_t cluster_offset; |
4bc74be9 | 1814 | int index_in_cluster, ret; |
ecfe1863 | 1815 | unsigned int bytes; |
a320fb04 | 1816 | int status = 0; |
585f8587 | 1817 | |
a320fb04 | 1818 | bytes = MIN(INT_MAX, count); |
f8a2e5e3 | 1819 | qemu_co_mutex_lock(&s->lock); |
a320fb04 | 1820 | ret = qcow2_get_cluster_offset(bs, offset, &bytes, &cluster_offset); |
f8a2e5e3 | 1821 | qemu_co_mutex_unlock(&s->lock); |
1c46efaa | 1822 | if (ret < 0) { |
d663640c | 1823 | return ret; |
1c46efaa | 1824 | } |
095a9c58 | 1825 | |
a320fb04 | 1826 | *pnum = bytes; |
ecfe1863 | 1827 | |
4bc74be9 | 1828 | if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED && |
b25b387f | 1829 | !s->crypto) { |
a320fb04 EB |
1830 | index_in_cluster = offset & (s->cluster_size - 1); |
1831 | *map = cluster_offset | index_in_cluster; | |
178b4db7 | 1832 | *file = bs->file->bs; |
a320fb04 | 1833 | status |= BDRV_BLOCK_OFFSET_VALID; |
4bc74be9 | 1834 | } |
fdfab37d | 1835 | if (ret == QCOW2_CLUSTER_ZERO_PLAIN || ret == QCOW2_CLUSTER_ZERO_ALLOC) { |
4bc74be9 PB |
1836 | status |= BDRV_BLOCK_ZERO; |
1837 | } else if (ret != QCOW2_CLUSTER_UNALLOCATED) { | |
1838 | status |= BDRV_BLOCK_DATA; | |
1839 | } | |
1840 | return status; | |
585f8587 FB |
1841 | } |
1842 | ||
fd9fcd37 FZ |
1843 | static coroutine_fn int qcow2_handle_l2meta(BlockDriverState *bs, |
1844 | QCowL2Meta **pl2meta, | |
1845 | bool link_l2) | |
1846 | { | |
1847 | int ret = 0; | |
1848 | QCowL2Meta *l2meta = *pl2meta; | |
1849 | ||
1850 | while (l2meta != NULL) { | |
1851 | QCowL2Meta *next; | |
1852 | ||
354d930d | 1853 | if (link_l2) { |
fd9fcd37 FZ |
1854 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); |
1855 | if (ret) { | |
1856 | goto out; | |
1857 | } | |
8b24cd14 KW |
1858 | } else { |
1859 | qcow2_alloc_cluster_abort(bs, l2meta); | |
fd9fcd37 FZ |
1860 | } |
1861 | ||
1862 | /* Take the request off the list of running requests */ | |
1863 | if (l2meta->nb_clusters != 0) { | |
1864 | QLIST_REMOVE(l2meta, next_in_flight); | |
1865 | } | |
1866 | ||
1867 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
1868 | ||
1869 | next = l2meta->next; | |
1870 | g_free(l2meta); | |
1871 | l2meta = next; | |
1872 | } | |
1873 | out: | |
1874 | *pl2meta = l2meta; | |
1875 | return ret; | |
1876 | } | |
1877 | ||
ecfe1863 KW |
1878 | static coroutine_fn int qcow2_co_preadv(BlockDriverState *bs, uint64_t offset, |
1879 | uint64_t bytes, QEMUIOVector *qiov, | |
1880 | int flags) | |
585f8587 | 1881 | { |
ff99129a | 1882 | BDRVQcow2State *s = bs->opaque; |
546a7dc4 | 1883 | int offset_in_cluster; |
68d100e9 | 1884 | int ret; |
ecfe1863 | 1885 | unsigned int cur_bytes; /* number of bytes in current iteration */ |
c2bdd990 | 1886 | uint64_t cluster_offset = 0; |
3fc48d09 FZ |
1887 | uint64_t bytes_done = 0; |
1888 | QEMUIOVector hd_qiov; | |
1889 | uint8_t *cluster_data = NULL; | |
585f8587 | 1890 | |
3fc48d09 FZ |
1891 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1892 | ||
1893 | qemu_co_mutex_lock(&s->lock); | |
1894 | ||
ecfe1863 | 1895 | while (bytes != 0) { |
bd28f835 | 1896 | |
5ebaa27e | 1897 | /* prepare next request */ |
ecfe1863 | 1898 | cur_bytes = MIN(bytes, INT_MAX); |
b25b387f | 1899 | if (s->crypto) { |
ecfe1863 KW |
1900 | cur_bytes = MIN(cur_bytes, |
1901 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); | |
585f8587 | 1902 | } |
5ebaa27e | 1903 | |
ecfe1863 | 1904 | ret = qcow2_get_cluster_offset(bs, offset, &cur_bytes, &cluster_offset); |
8af36488 | 1905 | if (ret < 0) { |
3fc48d09 | 1906 | goto fail; |
8af36488 | 1907 | } |
bd28f835 | 1908 | |
ecfe1863 | 1909 | offset_in_cluster = offset_into_cluster(s, offset); |
c87c0672 | 1910 | |
3fc48d09 | 1911 | qemu_iovec_reset(&hd_qiov); |
ecfe1863 | 1912 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes); |
5ebaa27e | 1913 | |
68d000a3 KW |
1914 | switch (ret) { |
1915 | case QCOW2_CLUSTER_UNALLOCATED: | |
5ebaa27e | 1916 | |
760e0063 | 1917 | if (bs->backing) { |
546a7dc4 EK |
1918 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); |
1919 | qemu_co_mutex_unlock(&s->lock); | |
1920 | ret = bdrv_co_preadv(bs->backing, offset, cur_bytes, | |
1921 | &hd_qiov, 0); | |
1922 | qemu_co_mutex_lock(&s->lock); | |
1923 | if (ret < 0) { | |
1924 | goto fail; | |
5ebaa27e FZ |
1925 | } |
1926 | } else { | |
1927 | /* Note: in this case, no need to wait */ | |
ecfe1863 | 1928 | qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes); |
5ebaa27e | 1929 | } |
68d000a3 KW |
1930 | break; |
1931 | ||
fdfab37d EB |
1932 | case QCOW2_CLUSTER_ZERO_PLAIN: |
1933 | case QCOW2_CLUSTER_ZERO_ALLOC: | |
ecfe1863 | 1934 | qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes); |
6377af48 KW |
1935 | break; |
1936 | ||
68d000a3 | 1937 | case QCOW2_CLUSTER_COMPRESSED: |
c3c10f72 VSO |
1938 | qemu_co_mutex_unlock(&s->lock); |
1939 | ret = qcow2_co_preadv_compressed(bs, cluster_offset, | |
1940 | offset, cur_bytes, | |
1941 | &hd_qiov); | |
1942 | qemu_co_mutex_lock(&s->lock); | |
5ebaa27e | 1943 | if (ret < 0) { |
3fc48d09 | 1944 | goto fail; |
bd28f835 KW |
1945 | } |
1946 | ||
68d000a3 KW |
1947 | break; |
1948 | ||
1949 | case QCOW2_CLUSTER_NORMAL: | |
5ebaa27e | 1950 | if ((cluster_offset & 511) != 0) { |
3fc48d09 FZ |
1951 | ret = -EIO; |
1952 | goto fail; | |
5ebaa27e | 1953 | } |
bd28f835 | 1954 | |
8336aafa | 1955 | if (bs->encrypted) { |
b25b387f | 1956 | assert(s->crypto); |
8336aafa | 1957 | |
5ebaa27e FZ |
1958 | /* |
1959 | * For encrypted images, read everything into a temporary | |
1960 | * contiguous buffer on which the AES functions can work. | |
1961 | */ | |
3fc48d09 FZ |
1962 | if (!cluster_data) { |
1963 | cluster_data = | |
9a4f4c31 KW |
1964 | qemu_try_blockalign(bs->file->bs, |
1965 | QCOW_MAX_CRYPT_CLUSTERS | |
1966 | * s->cluster_size); | |
de82815d KW |
1967 | if (cluster_data == NULL) { |
1968 | ret = -ENOMEM; | |
1969 | goto fail; | |
1970 | } | |
5ebaa27e FZ |
1971 | } |
1972 | ||
ecfe1863 | 1973 | assert(cur_bytes <= QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
3fc48d09 | 1974 | qemu_iovec_reset(&hd_qiov); |
ecfe1863 | 1975 | qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes); |
5ebaa27e FZ |
1976 | } |
1977 | ||
1978 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
1979 | qemu_co_mutex_unlock(&s->lock); | |
a03ef88f | 1980 | ret = bdrv_co_preadv(bs->file, |
ecfe1863 KW |
1981 | cluster_offset + offset_in_cluster, |
1982 | cur_bytes, &hd_qiov, 0); | |
5ebaa27e FZ |
1983 | qemu_co_mutex_lock(&s->lock); |
1984 | if (ret < 0) { | |
3fc48d09 | 1985 | goto fail; |
5ebaa27e | 1986 | } |
8336aafa | 1987 | if (bs->encrypted) { |
b25b387f | 1988 | assert(s->crypto); |
ecfe1863 KW |
1989 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
1990 | assert((cur_bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
b25b387f | 1991 | if (qcrypto_block_decrypt(s->crypto, |
4652b8f3 DB |
1992 | (s->crypt_physical_offset ? |
1993 | cluster_offset + offset_in_cluster : | |
4609742a | 1994 | offset), |
446d306d | 1995 | cluster_data, |
b25b387f | 1996 | cur_bytes, |
c3a8fe33 | 1997 | NULL) < 0) { |
f6fa64f6 DB |
1998 | ret = -EIO; |
1999 | goto fail; | |
2000 | } | |
ecfe1863 | 2001 | qemu_iovec_from_buf(qiov, bytes_done, cluster_data, cur_bytes); |
5ebaa27e | 2002 | } |
68d000a3 KW |
2003 | break; |
2004 | ||
2005 | default: | |
2006 | g_assert_not_reached(); | |
2007 | ret = -EIO; | |
2008 | goto fail; | |
faf575c1 | 2009 | } |
f141eafe | 2010 | |
ecfe1863 KW |
2011 | bytes -= cur_bytes; |
2012 | offset += cur_bytes; | |
2013 | bytes_done += cur_bytes; | |
5ebaa27e | 2014 | } |
3fc48d09 | 2015 | ret = 0; |
faf575c1 | 2016 | |
3fc48d09 | 2017 | fail: |
68d100e9 | 2018 | qemu_co_mutex_unlock(&s->lock); |
42496d62 | 2019 | |
3fc48d09 | 2020 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 2021 | qemu_vfree(cluster_data); |
68d100e9 KW |
2022 | |
2023 | return ret; | |
585f8587 FB |
2024 | } |
2025 | ||
ee22a9d8 AG |
2026 | /* Check if it's possible to merge a write request with the writing of |
2027 | * the data from the COW regions */ | |
2028 | static bool merge_cow(uint64_t offset, unsigned bytes, | |
2029 | QEMUIOVector *hd_qiov, QCowL2Meta *l2meta) | |
2030 | { | |
2031 | QCowL2Meta *m; | |
2032 | ||
2033 | for (m = l2meta; m != NULL; m = m->next) { | |
2034 | /* If both COW regions are empty then there's nothing to merge */ | |
2035 | if (m->cow_start.nb_bytes == 0 && m->cow_end.nb_bytes == 0) { | |
2036 | continue; | |
2037 | } | |
2038 | ||
2039 | /* The data (middle) region must be immediately after the | |
2040 | * start region */ | |
2041 | if (l2meta_cow_start(m) + m->cow_start.nb_bytes != offset) { | |
2042 | continue; | |
2043 | } | |
2044 | ||
2045 | /* The end region must be immediately after the data (middle) | |
2046 | * region */ | |
2047 | if (m->offset + m->cow_end.offset != offset + bytes) { | |
2048 | continue; | |
2049 | } | |
2050 | ||
2051 | /* Make sure that adding both COW regions to the QEMUIOVector | |
2052 | * does not exceed IOV_MAX */ | |
2053 | if (hd_qiov->niov > IOV_MAX - 2) { | |
2054 | continue; | |
2055 | } | |
2056 | ||
2057 | m->data_qiov = hd_qiov; | |
2058 | return true; | |
2059 | } | |
2060 | ||
2061 | return false; | |
2062 | } | |
2063 | ||
d46a0bb2 KW |
2064 | static coroutine_fn int qcow2_co_pwritev(BlockDriverState *bs, uint64_t offset, |
2065 | uint64_t bytes, QEMUIOVector *qiov, | |
2066 | int flags) | |
585f8587 | 2067 | { |
ff99129a | 2068 | BDRVQcow2State *s = bs->opaque; |
d46a0bb2 | 2069 | int offset_in_cluster; |
68d100e9 | 2070 | int ret; |
d46a0bb2 | 2071 | unsigned int cur_bytes; /* number of sectors in current iteration */ |
c2bdd990 | 2072 | uint64_t cluster_offset; |
3fc48d09 FZ |
2073 | QEMUIOVector hd_qiov; |
2074 | uint64_t bytes_done = 0; | |
2075 | uint8_t *cluster_data = NULL; | |
8d2497c3 | 2076 | QCowL2Meta *l2meta = NULL; |
c2271403 | 2077 | |
d46a0bb2 | 2078 | trace_qcow2_writev_start_req(qemu_coroutine_self(), offset, bytes); |
3cce16f4 | 2079 | |
3fc48d09 FZ |
2080 | qemu_iovec_init(&hd_qiov, qiov->niov); |
2081 | ||
3fc48d09 FZ |
2082 | qemu_co_mutex_lock(&s->lock); |
2083 | ||
d46a0bb2 | 2084 | while (bytes != 0) { |
3fc48d09 | 2085 | |
f50f88b9 | 2086 | l2meta = NULL; |
cf5c1a23 | 2087 | |
3cce16f4 | 2088 | trace_qcow2_writev_start_part(qemu_coroutine_self()); |
d46a0bb2 KW |
2089 | offset_in_cluster = offset_into_cluster(s, offset); |
2090 | cur_bytes = MIN(bytes, INT_MAX); | |
2091 | if (bs->encrypted) { | |
2092 | cur_bytes = MIN(cur_bytes, | |
2093 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size | |
2094 | - offset_in_cluster); | |
5ebaa27e | 2095 | } |
095a9c58 | 2096 | |
d46a0bb2 KW |
2097 | ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes, |
2098 | &cluster_offset, &l2meta); | |
5ebaa27e | 2099 | if (ret < 0) { |
3fc48d09 | 2100 | goto fail; |
5ebaa27e | 2101 | } |
148da7ea | 2102 | |
5ebaa27e | 2103 | assert((cluster_offset & 511) == 0); |
148da7ea | 2104 | |
3fc48d09 | 2105 | qemu_iovec_reset(&hd_qiov); |
d46a0bb2 | 2106 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes); |
6f5f060b | 2107 | |
8336aafa | 2108 | if (bs->encrypted) { |
b25b387f | 2109 | assert(s->crypto); |
3fc48d09 | 2110 | if (!cluster_data) { |
9a4f4c31 | 2111 | cluster_data = qemu_try_blockalign(bs->file->bs, |
de82815d KW |
2112 | QCOW_MAX_CRYPT_CLUSTERS |
2113 | * s->cluster_size); | |
2114 | if (cluster_data == NULL) { | |
2115 | ret = -ENOMEM; | |
2116 | goto fail; | |
2117 | } | |
5ebaa27e | 2118 | } |
6f5f060b | 2119 | |
3fc48d09 | 2120 | assert(hd_qiov.size <= |
5ebaa27e | 2121 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
d5e6b161 | 2122 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); |
6f5f060b | 2123 | |
4652b8f3 DB |
2124 | if (qcrypto_block_encrypt(s->crypto, |
2125 | (s->crypt_physical_offset ? | |
2126 | cluster_offset + offset_in_cluster : | |
4609742a | 2127 | offset), |
446d306d | 2128 | cluster_data, |
c3a8fe33 | 2129 | cur_bytes, NULL) < 0) { |
f6fa64f6 DB |
2130 | ret = -EIO; |
2131 | goto fail; | |
2132 | } | |
6f5f060b | 2133 | |
3fc48d09 | 2134 | qemu_iovec_reset(&hd_qiov); |
d46a0bb2 | 2135 | qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes); |
5ebaa27e | 2136 | } |
6f5f060b | 2137 | |
231bb267 | 2138 | ret = qcow2_pre_write_overlap_check(bs, 0, |
d46a0bb2 | 2139 | cluster_offset + offset_in_cluster, cur_bytes); |
cf93980e HR |
2140 | if (ret < 0) { |
2141 | goto fail; | |
2142 | } | |
2143 | ||
ee22a9d8 AG |
2144 | /* If we need to do COW, check if it's possible to merge the |
2145 | * writing of the guest data together with that of the COW regions. | |
2146 | * If it's not possible (or not necessary) then write the | |
2147 | * guest data now. */ | |
2148 | if (!merge_cow(offset, cur_bytes, &hd_qiov, l2meta)) { | |
2149 | qemu_co_mutex_unlock(&s->lock); | |
2150 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); | |
2151 | trace_qcow2_writev_data(qemu_coroutine_self(), | |
2152 | cluster_offset + offset_in_cluster); | |
2153 | ret = bdrv_co_pwritev(bs->file, | |
2154 | cluster_offset + offset_in_cluster, | |
2155 | cur_bytes, &hd_qiov, 0); | |
2156 | qemu_co_mutex_lock(&s->lock); | |
2157 | if (ret < 0) { | |
2158 | goto fail; | |
2159 | } | |
5ebaa27e | 2160 | } |
f141eafe | 2161 | |
fd9fcd37 FZ |
2162 | ret = qcow2_handle_l2meta(bs, &l2meta, true); |
2163 | if (ret) { | |
2164 | goto fail; | |
f50f88b9 | 2165 | } |
0fa9131a | 2166 | |
d46a0bb2 KW |
2167 | bytes -= cur_bytes; |
2168 | offset += cur_bytes; | |
2169 | bytes_done += cur_bytes; | |
2170 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_bytes); | |
5ebaa27e | 2171 | } |
3fc48d09 | 2172 | ret = 0; |
faf575c1 | 2173 | |
3fc48d09 | 2174 | fail: |
fd9fcd37 | 2175 | qcow2_handle_l2meta(bs, &l2meta, false); |
0fa9131a | 2176 | |
a8c57408 PB |
2177 | qemu_co_mutex_unlock(&s->lock); |
2178 | ||
3fc48d09 | 2179 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 2180 | qemu_vfree(cluster_data); |
3cce16f4 | 2181 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
42496d62 | 2182 | |
68d100e9 | 2183 | return ret; |
585f8587 FB |
2184 | } |
2185 | ||
ec6d8912 KW |
2186 | static int qcow2_inactivate(BlockDriverState *bs) |
2187 | { | |
2188 | BDRVQcow2State *s = bs->opaque; | |
2189 | int ret, result = 0; | |
5f72826e | 2190 | Error *local_err = NULL; |
ec6d8912 | 2191 | |
83a8c775 PB |
2192 | qcow2_store_persistent_dirty_bitmaps(bs, &local_err); |
2193 | if (local_err != NULL) { | |
2194 | result = -EINVAL; | |
132adb68 VSO |
2195 | error_reportf_err(local_err, "Lost persistent bitmaps during " |
2196 | "inactivation of node '%s': ", | |
2197 | bdrv_get_device_or_node_name(bs)); | |
83a8c775 PB |
2198 | } |
2199 | ||
ec6d8912 KW |
2200 | ret = qcow2_cache_flush(bs, s->l2_table_cache); |
2201 | if (ret) { | |
2202 | result = ret; | |
2203 | error_report("Failed to flush the L2 table cache: %s", | |
2204 | strerror(-ret)); | |
2205 | } | |
2206 | ||
2207 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
2208 | if (ret) { | |
2209 | result = ret; | |
2210 | error_report("Failed to flush the refcount block cache: %s", | |
2211 | strerror(-ret)); | |
2212 | } | |
2213 | ||
2214 | if (result == 0) { | |
2215 | qcow2_mark_clean(bs); | |
2216 | } | |
2217 | ||
2218 | return result; | |
2219 | } | |
2220 | ||
7c80ab3f | 2221 | static void qcow2_close(BlockDriverState *bs) |
585f8587 | 2222 | { |
ff99129a | 2223 | BDRVQcow2State *s = bs->opaque; |
de82815d | 2224 | qemu_vfree(s->l1_table); |
cf93980e HR |
2225 | /* else pre-write overlap checks in cache_destroy may crash */ |
2226 | s->l1_table = NULL; | |
29c1a730 | 2227 | |
140fd5a6 | 2228 | if (!(s->flags & BDRV_O_INACTIVE)) { |
ec6d8912 | 2229 | qcow2_inactivate(bs); |
27eb6c09 | 2230 | } |
c61d0004 | 2231 | |
279621c0 | 2232 | cache_clean_timer_del(bs); |
e64d4072 AG |
2233 | qcow2_cache_destroy(s->l2_table_cache); |
2234 | qcow2_cache_destroy(s->refcount_block_cache); | |
29c1a730 | 2235 | |
b25b387f DB |
2236 | qcrypto_block_free(s->crypto); |
2237 | s->crypto = NULL; | |
f6fa64f6 | 2238 | |
6744cbab | 2239 | g_free(s->unknown_header_fields); |
75bab85c | 2240 | cleanup_unknown_header_ext(bs); |
6744cbab | 2241 | |
e4603fe1 KW |
2242 | g_free(s->image_backing_file); |
2243 | g_free(s->image_backing_format); | |
2244 | ||
ed6ccf0f | 2245 | qcow2_refcount_close(bs); |
28c1202b | 2246 | qcow2_free_snapshots(bs); |
585f8587 FB |
2247 | } |
2248 | ||
2b148f39 PB |
2249 | static void coroutine_fn qcow2_co_invalidate_cache(BlockDriverState *bs, |
2250 | Error **errp) | |
06d9260f | 2251 | { |
ff99129a | 2252 | BDRVQcow2State *s = bs->opaque; |
06d9260f | 2253 | int flags = s->flags; |
b25b387f | 2254 | QCryptoBlock *crypto = NULL; |
acdfb480 | 2255 | QDict *options; |
5a8a30db KW |
2256 | Error *local_err = NULL; |
2257 | int ret; | |
06d9260f AL |
2258 | |
2259 | /* | |
2260 | * Backing files are read-only which makes all of their metadata immutable, | |
2261 | * that means we don't have to worry about reopening them here. | |
2262 | */ | |
2263 | ||
b25b387f DB |
2264 | crypto = s->crypto; |
2265 | s->crypto = NULL; | |
06d9260f AL |
2266 | |
2267 | qcow2_close(bs); | |
2268 | ||
ff99129a | 2269 | memset(s, 0, sizeof(BDRVQcow2State)); |
d475e5ac | 2270 | options = qdict_clone_shallow(bs->options); |
5a8a30db | 2271 | |
140fd5a6 | 2272 | flags &= ~BDRV_O_INACTIVE; |
2b148f39 | 2273 | qemu_co_mutex_lock(&s->lock); |
4e4bf5c4 | 2274 | ret = qcow2_do_open(bs, options, flags, &local_err); |
2b148f39 | 2275 | qemu_co_mutex_unlock(&s->lock); |
cb3e7f08 | 2276 | qobject_unref(options); |
5a8a30db | 2277 | if (local_err) { |
4b576648 MA |
2278 | error_propagate_prepend(errp, local_err, |
2279 | "Could not reopen qcow2 layer: "); | |
191fb11b | 2280 | bs->drv = NULL; |
5a8a30db KW |
2281 | return; |
2282 | } else if (ret < 0) { | |
2283 | error_setg_errno(errp, -ret, "Could not reopen qcow2 layer"); | |
191fb11b | 2284 | bs->drv = NULL; |
5a8a30db KW |
2285 | return; |
2286 | } | |
acdfb480 | 2287 | |
b25b387f | 2288 | s->crypto = crypto; |
06d9260f AL |
2289 | } |
2290 | ||
e24e49e6 KW |
2291 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
2292 | size_t len, size_t buflen) | |
2293 | { | |
2294 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
2295 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
2296 | ||
2297 | if (buflen < ext_len) { | |
2298 | return -ENOSPC; | |
2299 | } | |
2300 | ||
2301 | *ext_backing_fmt = (QCowExtension) { | |
2302 | .magic = cpu_to_be32(magic), | |
2303 | .len = cpu_to_be32(len), | |
2304 | }; | |
0647d47c SH |
2305 | |
2306 | if (len) { | |
2307 | memcpy(buf + sizeof(QCowExtension), s, len); | |
2308 | } | |
e24e49e6 KW |
2309 | |
2310 | return ext_len; | |
2311 | } | |
2312 | ||
756e6736 | 2313 | /* |
e24e49e6 KW |
2314 | * Updates the qcow2 header, including the variable length parts of it, i.e. |
2315 | * the backing file name and all extensions. qcow2 was not designed to allow | |
2316 | * such changes, so if we run out of space (we can only use the first cluster) | |
2317 | * this function may fail. | |
756e6736 KW |
2318 | * |
2319 | * Returns 0 on success, -errno in error cases. | |
2320 | */ | |
e24e49e6 | 2321 | int qcow2_update_header(BlockDriverState *bs) |
756e6736 | 2322 | { |
ff99129a | 2323 | BDRVQcow2State *s = bs->opaque; |
e24e49e6 KW |
2324 | QCowHeader *header; |
2325 | char *buf; | |
2326 | size_t buflen = s->cluster_size; | |
756e6736 | 2327 | int ret; |
e24e49e6 KW |
2328 | uint64_t total_size; |
2329 | uint32_t refcount_table_clusters; | |
6744cbab | 2330 | size_t header_length; |
75bab85c | 2331 | Qcow2UnknownHeaderExtension *uext; |
756e6736 | 2332 | |
e24e49e6 | 2333 | buf = qemu_blockalign(bs, buflen); |
756e6736 | 2334 | |
e24e49e6 KW |
2335 | /* Header structure */ |
2336 | header = (QCowHeader*) buf; | |
756e6736 | 2337 | |
e24e49e6 KW |
2338 | if (buflen < sizeof(*header)) { |
2339 | ret = -ENOSPC; | |
2340 | goto fail; | |
756e6736 KW |
2341 | } |
2342 | ||
6744cbab | 2343 | header_length = sizeof(*header) + s->unknown_header_fields_size; |
e24e49e6 KW |
2344 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
2345 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
2346 | ||
2347 | *header = (QCowHeader) { | |
6744cbab | 2348 | /* Version 2 fields */ |
e24e49e6 | 2349 | .magic = cpu_to_be32(QCOW_MAGIC), |
6744cbab | 2350 | .version = cpu_to_be32(s->qcow_version), |
e24e49e6 KW |
2351 | .backing_file_offset = 0, |
2352 | .backing_file_size = 0, | |
2353 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
2354 | .size = cpu_to_be64(total_size), | |
2355 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
2356 | .l1_size = cpu_to_be32(s->l1_size), | |
2357 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
2358 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
2359 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
2360 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
2361 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
6744cbab KW |
2362 | |
2363 | /* Version 3 fields */ | |
2364 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
2365 | .compatible_features = cpu_to_be64(s->compatible_features), | |
2366 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
b6481f37 | 2367 | .refcount_order = cpu_to_be32(s->refcount_order), |
6744cbab | 2368 | .header_length = cpu_to_be32(header_length), |
e24e49e6 | 2369 | }; |
756e6736 | 2370 | |
6744cbab KW |
2371 | /* For older versions, write a shorter header */ |
2372 | switch (s->qcow_version) { | |
2373 | case 2: | |
2374 | ret = offsetof(QCowHeader, incompatible_features); | |
2375 | break; | |
2376 | case 3: | |
2377 | ret = sizeof(*header); | |
2378 | break; | |
2379 | default: | |
b6c14762 JM |
2380 | ret = -EINVAL; |
2381 | goto fail; | |
6744cbab KW |
2382 | } |
2383 | ||
2384 | buf += ret; | |
2385 | buflen -= ret; | |
2386 | memset(buf, 0, buflen); | |
2387 | ||
2388 | /* Preserve any unknown field in the header */ | |
2389 | if (s->unknown_header_fields_size) { | |
2390 | if (buflen < s->unknown_header_fields_size) { | |
2391 | ret = -ENOSPC; | |
2392 | goto fail; | |
2393 | } | |
2394 | ||
2395 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
2396 | buf += s->unknown_header_fields_size; | |
2397 | buflen -= s->unknown_header_fields_size; | |
2398 | } | |
756e6736 | 2399 | |
e24e49e6 | 2400 | /* Backing file format header extension */ |
e4603fe1 | 2401 | if (s->image_backing_format) { |
e24e49e6 | 2402 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, |
e4603fe1 KW |
2403 | s->image_backing_format, |
2404 | strlen(s->image_backing_format), | |
e24e49e6 KW |
2405 | buflen); |
2406 | if (ret < 0) { | |
2407 | goto fail; | |
756e6736 KW |
2408 | } |
2409 | ||
e24e49e6 KW |
2410 | buf += ret; |
2411 | buflen -= ret; | |
756e6736 KW |
2412 | } |
2413 | ||
4652b8f3 DB |
2414 | /* Full disk encryption header pointer extension */ |
2415 | if (s->crypto_header.offset != 0) { | |
3b698f52 PM |
2416 | s->crypto_header.offset = cpu_to_be64(s->crypto_header.offset); |
2417 | s->crypto_header.length = cpu_to_be64(s->crypto_header.length); | |
4652b8f3 DB |
2418 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_CRYPTO_HEADER, |
2419 | &s->crypto_header, sizeof(s->crypto_header), | |
2420 | buflen); | |
3b698f52 PM |
2421 | s->crypto_header.offset = be64_to_cpu(s->crypto_header.offset); |
2422 | s->crypto_header.length = be64_to_cpu(s->crypto_header.length); | |
4652b8f3 DB |
2423 | if (ret < 0) { |
2424 | goto fail; | |
2425 | } | |
2426 | buf += ret; | |
2427 | buflen -= ret; | |
2428 | } | |
2429 | ||
cfcc4c62 | 2430 | /* Feature table */ |
1a4828c7 KW |
2431 | if (s->qcow_version >= 3) { |
2432 | Qcow2Feature features[] = { | |
2433 | { | |
2434 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
2435 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
2436 | .name = "dirty bit", | |
2437 | }, | |
2438 | { | |
2439 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
2440 | .bit = QCOW2_INCOMPAT_CORRUPT_BITNR, | |
2441 | .name = "corrupt bit", | |
2442 | }, | |
2443 | { | |
2444 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
2445 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
2446 | .name = "lazy refcounts", | |
2447 | }, | |
2448 | }; | |
2449 | ||
2450 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
2451 | features, sizeof(features), buflen); | |
2452 | if (ret < 0) { | |
2453 | goto fail; | |
2454 | } | |
2455 | buf += ret; | |
2456 | buflen -= ret; | |
cfcc4c62 | 2457 | } |
cfcc4c62 | 2458 | |
88ddffae VSO |
2459 | /* Bitmap extension */ |
2460 | if (s->nb_bitmaps > 0) { | |
2461 | Qcow2BitmapHeaderExt bitmaps_header = { | |
2462 | .nb_bitmaps = cpu_to_be32(s->nb_bitmaps), | |
2463 | .bitmap_directory_size = | |
2464 | cpu_to_be64(s->bitmap_directory_size), | |
2465 | .bitmap_directory_offset = | |
2466 | cpu_to_be64(s->bitmap_directory_offset) | |
2467 | }; | |
2468 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BITMAPS, | |
2469 | &bitmaps_header, sizeof(bitmaps_header), | |
2470 | buflen); | |
2471 | if (ret < 0) { | |
2472 | goto fail; | |
2473 | } | |
2474 | buf += ret; | |
2475 | buflen -= ret; | |
2476 | } | |
2477 | ||
75bab85c KW |
2478 | /* Keep unknown header extensions */ |
2479 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
2480 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
2481 | if (ret < 0) { | |
2482 | goto fail; | |
2483 | } | |
2484 | ||
2485 | buf += ret; | |
2486 | buflen -= ret; | |
2487 | } | |
2488 | ||
e24e49e6 KW |
2489 | /* End of header extensions */ |
2490 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
756e6736 KW |
2491 | if (ret < 0) { |
2492 | goto fail; | |
2493 | } | |
2494 | ||
e24e49e6 KW |
2495 | buf += ret; |
2496 | buflen -= ret; | |
756e6736 | 2497 | |
e24e49e6 | 2498 | /* Backing file name */ |
e4603fe1 KW |
2499 | if (s->image_backing_file) { |
2500 | size_t backing_file_len = strlen(s->image_backing_file); | |
e24e49e6 KW |
2501 | |
2502 | if (buflen < backing_file_len) { | |
2503 | ret = -ENOSPC; | |
2504 | goto fail; | |
2505 | } | |
2506 | ||
00ea1881 | 2507 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ |
e4603fe1 | 2508 | strncpy(buf, s->image_backing_file, buflen); |
e24e49e6 KW |
2509 | |
2510 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
2511 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
756e6736 KW |
2512 | } |
2513 | ||
e24e49e6 | 2514 | /* Write the new header */ |
d9ca2ea2 | 2515 | ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size); |
756e6736 KW |
2516 | if (ret < 0) { |
2517 | goto fail; | |
2518 | } | |
2519 | ||
2520 | ret = 0; | |
2521 | fail: | |
e24e49e6 | 2522 | qemu_vfree(header); |
756e6736 KW |
2523 | return ret; |
2524 | } | |
2525 | ||
2526 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
2527 | const char *backing_file, const char *backing_fmt) | |
2528 | { | |
ff99129a | 2529 | BDRVQcow2State *s = bs->opaque; |
e4603fe1 | 2530 | |
4e876bcf HR |
2531 | if (backing_file && strlen(backing_file) > 1023) { |
2532 | return -EINVAL; | |
2533 | } | |
2534 | ||
998c2019 HR |
2535 | pstrcpy(bs->auto_backing_file, sizeof(bs->auto_backing_file), |
2536 | backing_file ?: ""); | |
e24e49e6 KW |
2537 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
2538 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2539 | ||
e4603fe1 KW |
2540 | g_free(s->image_backing_file); |
2541 | g_free(s->image_backing_format); | |
2542 | ||
2543 | s->image_backing_file = backing_file ? g_strdup(bs->backing_file) : NULL; | |
2544 | s->image_backing_format = backing_fmt ? g_strdup(bs->backing_format) : NULL; | |
2545 | ||
e24e49e6 | 2546 | return qcow2_update_header(bs); |
756e6736 KW |
2547 | } |
2548 | ||
4652b8f3 DB |
2549 | static int qcow2_crypt_method_from_format(const char *encryptfmt) |
2550 | { | |
2551 | if (g_str_equal(encryptfmt, "luks")) { | |
2552 | return QCOW_CRYPT_LUKS; | |
2553 | } else if (g_str_equal(encryptfmt, "aes")) { | |
2554 | return QCOW_CRYPT_AES; | |
2555 | } else { | |
2556 | return -EINVAL; | |
2557 | } | |
2558 | } | |
b25b387f | 2559 | |
60900b7b KW |
2560 | static int qcow2_set_up_encryption(BlockDriverState *bs, |
2561 | QCryptoBlockCreateOptions *cryptoopts, | |
2562 | Error **errp) | |
2563 | { | |
2564 | BDRVQcow2State *s = bs->opaque; | |
2565 | QCryptoBlock *crypto = NULL; | |
2566 | int fmt, ret; | |
2567 | ||
2568 | switch (cryptoopts->format) { | |
2569 | case Q_CRYPTO_BLOCK_FORMAT_LUKS: | |
2570 | fmt = QCOW_CRYPT_LUKS; | |
2571 | break; | |
2572 | case Q_CRYPTO_BLOCK_FORMAT_QCOW: | |
2573 | fmt = QCOW_CRYPT_AES; | |
2574 | break; | |
2575 | default: | |
2576 | error_setg(errp, "Crypto format not supported in qcow2"); | |
2577 | return -EINVAL; | |
b25b387f | 2578 | } |
60900b7b | 2579 | |
4652b8f3 | 2580 | s->crypt_method_header = fmt; |
b25b387f | 2581 | |
1cd9a787 | 2582 | crypto = qcrypto_block_create(cryptoopts, "encrypt.", |
4652b8f3 DB |
2583 | qcow2_crypto_hdr_init_func, |
2584 | qcow2_crypto_hdr_write_func, | |
b25b387f DB |
2585 | bs, errp); |
2586 | if (!crypto) { | |
60900b7b | 2587 | return -EINVAL; |
b25b387f DB |
2588 | } |
2589 | ||
2590 | ret = qcow2_update_header(bs); | |
2591 | if (ret < 0) { | |
2592 | error_setg_errno(errp, -ret, "Could not write encryption header"); | |
2593 | goto out; | |
2594 | } | |
2595 | ||
60900b7b | 2596 | ret = 0; |
b25b387f | 2597 | out: |
b25b387f | 2598 | qcrypto_block_free(crypto); |
b25b387f DB |
2599 | return ret; |
2600 | } | |
2601 | ||
7bc45dc1 HR |
2602 | /** |
2603 | * Preallocates metadata structures for data clusters between @offset (in the | |
2604 | * guest disk) and @new_length (which is thus generally the new guest disk | |
2605 | * size). | |
2606 | * | |
2607 | * Returns: 0 on success, -errno on failure. | |
2608 | */ | |
47e86b86 KW |
2609 | static int coroutine_fn preallocate_co(BlockDriverState *bs, uint64_t offset, |
2610 | uint64_t new_length) | |
a35e1c17 | 2611 | { |
d46a0bb2 | 2612 | uint64_t bytes; |
060bee89 | 2613 | uint64_t host_offset = 0; |
d46a0bb2 | 2614 | unsigned int cur_bytes; |
148da7ea | 2615 | int ret; |
f50f88b9 | 2616 | QCowL2Meta *meta; |
a35e1c17 | 2617 | |
7bc45dc1 HR |
2618 | assert(offset <= new_length); |
2619 | bytes = new_length - offset; | |
a35e1c17 | 2620 | |
d46a0bb2 KW |
2621 | while (bytes) { |
2622 | cur_bytes = MIN(bytes, INT_MAX); | |
2623 | ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes, | |
060bee89 | 2624 | &host_offset, &meta); |
148da7ea | 2625 | if (ret < 0) { |
47e86b86 | 2626 | return ret; |
a35e1c17 KW |
2627 | } |
2628 | ||
c792707f SH |
2629 | while (meta) { |
2630 | QCowL2Meta *next = meta->next; | |
2631 | ||
7c2bbf4a HT |
2632 | ret = qcow2_alloc_cluster_link_l2(bs, meta); |
2633 | if (ret < 0) { | |
2634 | qcow2_free_any_clusters(bs, meta->alloc_offset, | |
2635 | meta->nb_clusters, QCOW2_DISCARD_NEVER); | |
47e86b86 | 2636 | return ret; |
7c2bbf4a HT |
2637 | } |
2638 | ||
2639 | /* There are no dependent requests, but we need to remove our | |
2640 | * request from the list of in-flight requests */ | |
4e95314e | 2641 | QLIST_REMOVE(meta, next_in_flight); |
c792707f SH |
2642 | |
2643 | g_free(meta); | |
2644 | meta = next; | |
f50f88b9 | 2645 | } |
f214978a | 2646 | |
a35e1c17 KW |
2647 | /* TODO Preallocate data if requested */ |
2648 | ||
d46a0bb2 KW |
2649 | bytes -= cur_bytes; |
2650 | offset += cur_bytes; | |
a35e1c17 KW |
2651 | } |
2652 | ||
2653 | /* | |
2654 | * It is expected that the image file is large enough to actually contain | |
2655 | * all of the allocated clusters (otherwise we get failing reads after | |
2656 | * EOF). Extend the image to the last allocated sector. | |
2657 | */ | |
060bee89 | 2658 | if (host_offset != 0) { |
d46a0bb2 | 2659 | uint8_t data = 0; |
d9ca2ea2 | 2660 | ret = bdrv_pwrite(bs->file, (host_offset + cur_bytes) - 1, |
d46a0bb2 | 2661 | &data, 1); |
19dbcbf7 | 2662 | if (ret < 0) { |
47e86b86 | 2663 | return ret; |
19dbcbf7 | 2664 | } |
a35e1c17 KW |
2665 | } |
2666 | ||
47e86b86 | 2667 | return 0; |
a35e1c17 KW |
2668 | } |
2669 | ||
7c5bcc42 SH |
2670 | /* qcow2_refcount_metadata_size: |
2671 | * @clusters: number of clusters to refcount (including data and L1/L2 tables) | |
2672 | * @cluster_size: size of a cluster, in bytes | |
2673 | * @refcount_order: refcount bits power-of-2 exponent | |
12cc30a8 HR |
2674 | * @generous_increase: allow for the refcount table to be 1.5x as large as it |
2675 | * needs to be | |
7c5bcc42 SH |
2676 | * |
2677 | * Returns: Number of bytes required for refcount blocks and table metadata. | |
2678 | */ | |
12cc30a8 HR |
2679 | int64_t qcow2_refcount_metadata_size(int64_t clusters, size_t cluster_size, |
2680 | int refcount_order, bool generous_increase, | |
2681 | uint64_t *refblock_count) | |
7c5bcc42 SH |
2682 | { |
2683 | /* | |
2684 | * Every host cluster is reference-counted, including metadata (even | |
2685 | * refcount metadata is recursively included). | |
2686 | * | |
2687 | * An accurate formula for the size of refcount metadata size is difficult | |
2688 | * to derive. An easier method of calculation is finding the fixed point | |
2689 | * where no further refcount blocks or table clusters are required to | |
2690 | * reference count every cluster. | |
2691 | */ | |
2692 | int64_t blocks_per_table_cluster = cluster_size / sizeof(uint64_t); | |
2693 | int64_t refcounts_per_block = cluster_size * 8 / (1 << refcount_order); | |
2694 | int64_t table = 0; /* number of refcount table clusters */ | |
2695 | int64_t blocks = 0; /* number of refcount block clusters */ | |
2696 | int64_t last; | |
2697 | int64_t n = 0; | |
2698 | ||
2699 | do { | |
2700 | last = n; | |
2701 | blocks = DIV_ROUND_UP(clusters + table + blocks, refcounts_per_block); | |
2702 | table = DIV_ROUND_UP(blocks, blocks_per_table_cluster); | |
2703 | n = clusters + blocks + table; | |
12cc30a8 HR |
2704 | |
2705 | if (n == last && generous_increase) { | |
2706 | clusters += DIV_ROUND_UP(table, 2); | |
2707 | n = 0; /* force another loop */ | |
2708 | generous_increase = false; | |
2709 | } | |
7c5bcc42 SH |
2710 | } while (n != last); |
2711 | ||
12cc30a8 HR |
2712 | if (refblock_count) { |
2713 | *refblock_count = blocks; | |
2714 | } | |
2715 | ||
7c5bcc42 SH |
2716 | return (blocks + table) * cluster_size; |
2717 | } | |
2718 | ||
95c67e3b SH |
2719 | /** |
2720 | * qcow2_calc_prealloc_size: | |
2721 | * @total_size: virtual disk size in bytes | |
2722 | * @cluster_size: cluster size in bytes | |
2723 | * @refcount_order: refcount bits power-of-2 exponent | |
2724 | * | |
2725 | * Returns: Total number of bytes required for the fully allocated image | |
2726 | * (including metadata). | |
2727 | */ | |
2728 | static int64_t qcow2_calc_prealloc_size(int64_t total_size, | |
2729 | size_t cluster_size, | |
2730 | int refcount_order) | |
2731 | { | |
95c67e3b | 2732 | int64_t meta_size = 0; |
7c5bcc42 | 2733 | uint64_t nl1e, nl2e; |
9e029689 | 2734 | int64_t aligned_total_size = ROUND_UP(total_size, cluster_size); |
95c67e3b SH |
2735 | |
2736 | /* header: 1 cluster */ | |
2737 | meta_size += cluster_size; | |
2738 | ||
2739 | /* total size of L2 tables */ | |
2740 | nl2e = aligned_total_size / cluster_size; | |
9e029689 | 2741 | nl2e = ROUND_UP(nl2e, cluster_size / sizeof(uint64_t)); |
95c67e3b SH |
2742 | meta_size += nl2e * sizeof(uint64_t); |
2743 | ||
2744 | /* total size of L1 tables */ | |
2745 | nl1e = nl2e * sizeof(uint64_t) / cluster_size; | |
9e029689 | 2746 | nl1e = ROUND_UP(nl1e, cluster_size / sizeof(uint64_t)); |
95c67e3b SH |
2747 | meta_size += nl1e * sizeof(uint64_t); |
2748 | ||
7c5bcc42 SH |
2749 | /* total size of refcount table and blocks */ |
2750 | meta_size += qcow2_refcount_metadata_size( | |
2751 | (meta_size + aligned_total_size) / cluster_size, | |
12cc30a8 | 2752 | cluster_size, refcount_order, false, NULL); |
95c67e3b SH |
2753 | |
2754 | return meta_size + aligned_total_size; | |
2755 | } | |
2756 | ||
29ca9e45 | 2757 | static bool validate_cluster_size(size_t cluster_size, Error **errp) |
a9420734 | 2758 | { |
29ca9e45 | 2759 | int cluster_bits = ctz32(cluster_size); |
a9420734 KW |
2760 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || |
2761 | (1 << cluster_bits) != cluster_size) | |
2762 | { | |
3ef6c40a HR |
2763 | error_setg(errp, "Cluster size must be a power of two between %d and " |
2764 | "%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); | |
29ca9e45 KW |
2765 | return false; |
2766 | } | |
2767 | return true; | |
2768 | } | |
2769 | ||
2770 | static size_t qcow2_opt_get_cluster_size_del(QemuOpts *opts, Error **errp) | |
2771 | { | |
2772 | size_t cluster_size; | |
2773 | ||
2774 | cluster_size = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, | |
2775 | DEFAULT_CLUSTER_SIZE); | |
2776 | if (!validate_cluster_size(cluster_size, errp)) { | |
0eb4a8c1 SH |
2777 | return 0; |
2778 | } | |
2779 | return cluster_size; | |
2780 | } | |
2781 | ||
2782 | static int qcow2_opt_get_version_del(QemuOpts *opts, Error **errp) | |
2783 | { | |
2784 | char *buf; | |
2785 | int ret; | |
2786 | ||
2787 | buf = qemu_opt_get_del(opts, BLOCK_OPT_COMPAT_LEVEL); | |
2788 | if (!buf) { | |
2789 | ret = 3; /* default */ | |
2790 | } else if (!strcmp(buf, "0.10")) { | |
2791 | ret = 2; | |
2792 | } else if (!strcmp(buf, "1.1")) { | |
2793 | ret = 3; | |
2794 | } else { | |
2795 | error_setg(errp, "Invalid compatibility level: '%s'", buf); | |
2796 | ret = -EINVAL; | |
2797 | } | |
2798 | g_free(buf); | |
2799 | return ret; | |
2800 | } | |
2801 | ||
2802 | static uint64_t qcow2_opt_get_refcount_bits_del(QemuOpts *opts, int version, | |
2803 | Error **errp) | |
2804 | { | |
2805 | uint64_t refcount_bits; | |
2806 | ||
2807 | refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS, 16); | |
2808 | if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) { | |
2809 | error_setg(errp, "Refcount width must be a power of two and may not " | |
2810 | "exceed 64 bits"); | |
2811 | return 0; | |
2812 | } | |
2813 | ||
2814 | if (version < 3 && refcount_bits != 16) { | |
2815 | error_setg(errp, "Different refcount widths than 16 bits require " | |
2816 | "compatibility level 1.1 or above (use compat=1.1 or " | |
2817 | "greater)"); | |
2818 | return 0; | |
a9420734 KW |
2819 | } |
2820 | ||
0eb4a8c1 SH |
2821 | return refcount_bits; |
2822 | } | |
2823 | ||
c274393a | 2824 | static int coroutine_fn |
60900b7b | 2825 | qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp) |
0eb4a8c1 | 2826 | { |
29ca9e45 | 2827 | BlockdevCreateOptionsQcow2 *qcow2_opts; |
0eb4a8c1 SH |
2828 | QDict *options; |
2829 | ||
a9420734 KW |
2830 | /* |
2831 | * Open the image file and write a minimal qcow2 header. | |
2832 | * | |
2833 | * We keep things simple and start with a zero-sized image. We also | |
2834 | * do without refcount blocks or a L1 table for now. We'll fix the | |
2835 | * inconsistency later. | |
2836 | * | |
2837 | * We do need a refcount table because growing the refcount table means | |
2838 | * allocating two new refcount blocks - the seconds of which would be at | |
2839 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
2840 | * size for any qcow2 image. | |
2841 | */ | |
e1d74bc6 KW |
2842 | BlockBackend *blk = NULL; |
2843 | BlockDriverState *bs = NULL; | |
f8413b3c | 2844 | QCowHeader *header; |
29ca9e45 KW |
2845 | size_t cluster_size; |
2846 | int version; | |
2847 | int refcount_order; | |
b106ad91 | 2848 | uint64_t* refcount_table; |
3ef6c40a | 2849 | Error *local_err = NULL; |
a9420734 KW |
2850 | int ret; |
2851 | ||
29ca9e45 KW |
2852 | assert(create_options->driver == BLOCKDEV_DRIVER_QCOW2); |
2853 | qcow2_opts = &create_options->u.qcow2; | |
2854 | ||
e1d74bc6 KW |
2855 | bs = bdrv_open_blockdev_ref(qcow2_opts->file, errp); |
2856 | if (bs == NULL) { | |
2857 | return -EIO; | |
2858 | } | |
2859 | ||
2860 | /* Validate options and set default values */ | |
29ca9e45 KW |
2861 | if (!QEMU_IS_ALIGNED(qcow2_opts->size, BDRV_SECTOR_SIZE)) { |
2862 | error_setg(errp, "Image size must be a multiple of 512 bytes"); | |
2863 | ret = -EINVAL; | |
2864 | goto out; | |
2865 | } | |
2866 | ||
2867 | if (qcow2_opts->has_version) { | |
2868 | switch (qcow2_opts->version) { | |
2869 | case BLOCKDEV_QCOW2_VERSION_V2: | |
2870 | version = 2; | |
2871 | break; | |
2872 | case BLOCKDEV_QCOW2_VERSION_V3: | |
2873 | version = 3; | |
2874 | break; | |
2875 | default: | |
2876 | g_assert_not_reached(); | |
2877 | } | |
2878 | } else { | |
2879 | version = 3; | |
2880 | } | |
2881 | ||
2882 | if (qcow2_opts->has_cluster_size) { | |
2883 | cluster_size = qcow2_opts->cluster_size; | |
2884 | } else { | |
2885 | cluster_size = DEFAULT_CLUSTER_SIZE; | |
2886 | } | |
2887 | ||
2888 | if (!validate_cluster_size(cluster_size, errp)) { | |
e1d74bc6 KW |
2889 | ret = -EINVAL; |
2890 | goto out; | |
29ca9e45 KW |
2891 | } |
2892 | ||
2893 | if (!qcow2_opts->has_preallocation) { | |
2894 | qcow2_opts->preallocation = PREALLOC_MODE_OFF; | |
2895 | } | |
2896 | if (qcow2_opts->has_backing_file && | |
2897 | qcow2_opts->preallocation != PREALLOC_MODE_OFF) | |
2898 | { | |
2899 | error_setg(errp, "Backing file and preallocation cannot be used at " | |
2900 | "the same time"); | |
e1d74bc6 KW |
2901 | ret = -EINVAL; |
2902 | goto out; | |
29ca9e45 KW |
2903 | } |
2904 | if (qcow2_opts->has_backing_fmt && !qcow2_opts->has_backing_file) { | |
2905 | error_setg(errp, "Backing format cannot be used without backing file"); | |
e1d74bc6 KW |
2906 | ret = -EINVAL; |
2907 | goto out; | |
29ca9e45 KW |
2908 | } |
2909 | ||
2910 | if (!qcow2_opts->has_lazy_refcounts) { | |
2911 | qcow2_opts->lazy_refcounts = false; | |
2912 | } | |
2913 | if (version < 3 && qcow2_opts->lazy_refcounts) { | |
2914 | error_setg(errp, "Lazy refcounts only supported with compatibility " | |
b76b4f60 | 2915 | "level 1.1 and above (use version=v3 or greater)"); |
e1d74bc6 KW |
2916 | ret = -EINVAL; |
2917 | goto out; | |
29ca9e45 KW |
2918 | } |
2919 | ||
2920 | if (!qcow2_opts->has_refcount_bits) { | |
2921 | qcow2_opts->refcount_bits = 16; | |
2922 | } | |
2923 | if (qcow2_opts->refcount_bits > 64 || | |
2924 | !is_power_of_2(qcow2_opts->refcount_bits)) | |
2925 | { | |
2926 | error_setg(errp, "Refcount width must be a power of two and may not " | |
2927 | "exceed 64 bits"); | |
e1d74bc6 KW |
2928 | ret = -EINVAL; |
2929 | goto out; | |
29ca9e45 KW |
2930 | } |
2931 | if (version < 3 && qcow2_opts->refcount_bits != 16) { | |
2932 | error_setg(errp, "Different refcount widths than 16 bits require " | |
b76b4f60 | 2933 | "compatibility level 1.1 or above (use version=v3 or " |
29ca9e45 | 2934 | "greater)"); |
e1d74bc6 KW |
2935 | ret = -EINVAL; |
2936 | goto out; | |
29ca9e45 KW |
2937 | } |
2938 | refcount_order = ctz32(qcow2_opts->refcount_bits); | |
2939 | ||
2940 | ||
2941 | /* Create BlockBackend to write to the image */ | |
cbf2b7c4 KW |
2942 | blk = blk_new(BLK_PERM_WRITE | BLK_PERM_RESIZE, BLK_PERM_ALL); |
2943 | ret = blk_insert_bs(blk, bs, errp); | |
a9420734 | 2944 | if (ret < 0) { |
cbf2b7c4 | 2945 | goto out; |
a9420734 | 2946 | } |
23588797 KW |
2947 | blk_set_allow_write_beyond_eof(blk, true); |
2948 | ||
e4b5dad8 KW |
2949 | /* Clear the protocol layer and preallocate it if necessary */ |
2950 | ret = blk_truncate(blk, 0, PREALLOC_MODE_OFF, errp); | |
2951 | if (ret < 0) { | |
2952 | goto out; | |
2953 | } | |
2954 | ||
a9420734 | 2955 | /* Write the header */ |
f8413b3c KW |
2956 | QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header)); |
2957 | header = g_malloc0(cluster_size); | |
2958 | *header = (QCowHeader) { | |
2959 | .magic = cpu_to_be32(QCOW_MAGIC), | |
2960 | .version = cpu_to_be32(version), | |
0eb4a8c1 | 2961 | .cluster_bits = cpu_to_be32(ctz32(cluster_size)), |
f8413b3c KW |
2962 | .size = cpu_to_be64(0), |
2963 | .l1_table_offset = cpu_to_be64(0), | |
2964 | .l1_size = cpu_to_be32(0), | |
2965 | .refcount_table_offset = cpu_to_be64(cluster_size), | |
2966 | .refcount_table_clusters = cpu_to_be32(1), | |
bd4b167f | 2967 | .refcount_order = cpu_to_be32(refcount_order), |
f8413b3c KW |
2968 | .header_length = cpu_to_be32(sizeof(*header)), |
2969 | }; | |
a9420734 | 2970 | |
b25b387f DB |
2971 | /* We'll update this to correct value later */ |
2972 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); | |
a9420734 | 2973 | |
29ca9e45 | 2974 | if (qcow2_opts->lazy_refcounts) { |
f8413b3c | 2975 | header->compatible_features |= |
bfe8043e SH |
2976 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); |
2977 | } | |
2978 | ||
8341f00d | 2979 | ret = blk_pwrite(blk, 0, header, cluster_size, 0); |
f8413b3c | 2980 | g_free(header); |
a9420734 | 2981 | if (ret < 0) { |
3ef6c40a | 2982 | error_setg_errno(errp, -ret, "Could not write qcow2 header"); |
a9420734 KW |
2983 | goto out; |
2984 | } | |
2985 | ||
b106ad91 KW |
2986 | /* Write a refcount table with one refcount block */ |
2987 | refcount_table = g_malloc0(2 * cluster_size); | |
2988 | refcount_table[0] = cpu_to_be64(2 * cluster_size); | |
8341f00d | 2989 | ret = blk_pwrite(blk, cluster_size, refcount_table, 2 * cluster_size, 0); |
7267c094 | 2990 | g_free(refcount_table); |
a9420734 KW |
2991 | |
2992 | if (ret < 0) { | |
3ef6c40a | 2993 | error_setg_errno(errp, -ret, "Could not write refcount table"); |
a9420734 KW |
2994 | goto out; |
2995 | } | |
2996 | ||
23588797 KW |
2997 | blk_unref(blk); |
2998 | blk = NULL; | |
a9420734 KW |
2999 | |
3000 | /* | |
3001 | * And now open the image and make it consistent first (i.e. increase the | |
3002 | * refcount of the cluster that is occupied by the header and the refcount | |
3003 | * table) | |
3004 | */ | |
e6641719 | 3005 | options = qdict_new(); |
46f5ac20 | 3006 | qdict_put_str(options, "driver", "qcow2"); |
cbf2b7c4 KW |
3007 | qdict_put_str(options, "file", bs->node_name); |
3008 | blk = blk_new_open(NULL, NULL, options, | |
55880601 KW |
3009 | BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_NO_FLUSH, |
3010 | &local_err); | |
23588797 | 3011 | if (blk == NULL) { |
3ef6c40a | 3012 | error_propagate(errp, local_err); |
23588797 | 3013 | ret = -EIO; |
a9420734 KW |
3014 | goto out; |
3015 | } | |
3016 | ||
23588797 | 3017 | ret = qcow2_alloc_clusters(blk_bs(blk), 3 * cluster_size); |
a9420734 | 3018 | if (ret < 0) { |
3ef6c40a HR |
3019 | error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 " |
3020 | "header and refcount table"); | |
a9420734 KW |
3021 | goto out; |
3022 | ||
3023 | } else if (ret != 0) { | |
3024 | error_report("Huh, first cluster in empty image is already in use?"); | |
3025 | abort(); | |
3026 | } | |
3027 | ||
b527c9b3 | 3028 | /* Create a full header (including things like feature table) */ |
23588797 | 3029 | ret = qcow2_update_header(blk_bs(blk)); |
b527c9b3 KW |
3030 | if (ret < 0) { |
3031 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); | |
3032 | goto out; | |
3033 | } | |
3034 | ||
a9420734 | 3035 | /* Okay, now that we have a valid image, let's give it the right size */ |
c5e86ebc | 3036 | ret = blk_truncate(blk, qcow2_opts->size, qcow2_opts->preallocation, errp); |
a9420734 | 3037 | if (ret < 0) { |
ed3d2ec9 | 3038 | error_prepend(errp, "Could not resize image: "); |
a9420734 KW |
3039 | goto out; |
3040 | } | |
3041 | ||
3042 | /* Want a backing file? There you go.*/ | |
29ca9e45 KW |
3043 | if (qcow2_opts->has_backing_file) { |
3044 | const char *backing_format = NULL; | |
3045 | ||
3046 | if (qcow2_opts->has_backing_fmt) { | |
3047 | backing_format = BlockdevDriver_str(qcow2_opts->backing_fmt); | |
3048 | } | |
3049 | ||
3050 | ret = bdrv_change_backing_file(blk_bs(blk), qcow2_opts->backing_file, | |
3051 | backing_format); | |
a9420734 | 3052 | if (ret < 0) { |
3ef6c40a | 3053 | error_setg_errno(errp, -ret, "Could not assign backing file '%s' " |
29ca9e45 KW |
3054 | "with format '%s'", qcow2_opts->backing_file, |
3055 | backing_format); | |
a9420734 KW |
3056 | goto out; |
3057 | } | |
3058 | } | |
3059 | ||
b25b387f | 3060 | /* Want encryption? There you go. */ |
60900b7b KW |
3061 | if (qcow2_opts->has_encrypt) { |
3062 | ret = qcow2_set_up_encryption(blk_bs(blk), qcow2_opts->encrypt, errp); | |
b25b387f DB |
3063 | if (ret < 0) { |
3064 | goto out; | |
3065 | } | |
3066 | } | |
3067 | ||
23588797 KW |
3068 | blk_unref(blk); |
3069 | blk = NULL; | |
ba2ab2f2 | 3070 | |
b25b387f DB |
3071 | /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning. |
3072 | * Using BDRV_O_NO_IO, since encryption is now setup we don't want to | |
3073 | * have to setup decryption context. We're not doing any I/O on the top | |
3074 | * level BlockDriverState, only lower layers, where BDRV_O_NO_IO does | |
3075 | * not have effect. | |
3076 | */ | |
e6641719 | 3077 | options = qdict_new(); |
46f5ac20 | 3078 | qdict_put_str(options, "driver", "qcow2"); |
cbf2b7c4 KW |
3079 | qdict_put_str(options, "file", bs->node_name); |
3080 | blk = blk_new_open(NULL, NULL, options, | |
b25b387f DB |
3081 | BDRV_O_RDWR | BDRV_O_NO_BACKING | BDRV_O_NO_IO, |
3082 | &local_err); | |
23588797 | 3083 | if (blk == NULL) { |
ba2ab2f2 | 3084 | error_propagate(errp, local_err); |
23588797 | 3085 | ret = -EIO; |
ba2ab2f2 HR |
3086 | goto out; |
3087 | } | |
3088 | ||
a9420734 KW |
3089 | ret = 0; |
3090 | out: | |
e1d74bc6 KW |
3091 | blk_unref(blk); |
3092 | bdrv_unref(bs); | |
a9420734 KW |
3093 | return ret; |
3094 | } | |
de5f3f40 | 3095 | |
efc75e2a SH |
3096 | static int coroutine_fn qcow2_co_create_opts(const char *filename, QemuOpts *opts, |
3097 | Error **errp) | |
de5f3f40 | 3098 | { |
b76b4f60 | 3099 | BlockdevCreateOptions *create_options = NULL; |
92adf9db | 3100 | QDict *qdict; |
b76b4f60 | 3101 | Visitor *v; |
cbf2b7c4 | 3102 | BlockDriverState *bs = NULL; |
3ef6c40a | 3103 | Error *local_err = NULL; |
b76b4f60 | 3104 | const char *val; |
3ef6c40a | 3105 | int ret; |
de5f3f40 | 3106 | |
b76b4f60 KW |
3107 | /* Only the keyval visitor supports the dotted syntax needed for |
3108 | * encryption, so go through a QDict before getting a QAPI type. Ignore | |
3109 | * options meant for the protocol layer so that the visitor doesn't | |
3110 | * complain. */ | |
3111 | qdict = qemu_opts_to_qdict_filtered(opts, NULL, bdrv_qcow2.create_opts, | |
3112 | true); | |
3113 | ||
3114 | /* Handle encryption options */ | |
3115 | val = qdict_get_try_str(qdict, BLOCK_OPT_ENCRYPT); | |
3116 | if (val && !strcmp(val, "on")) { | |
3117 | qdict_put_str(qdict, BLOCK_OPT_ENCRYPT, "qcow"); | |
3118 | } else if (val && !strcmp(val, "off")) { | |
3119 | qdict_del(qdict, BLOCK_OPT_ENCRYPT); | |
3120 | } | |
3121 | ||
3122 | val = qdict_get_try_str(qdict, BLOCK_OPT_ENCRYPT_FORMAT); | |
3123 | if (val && !strcmp(val, "aes")) { | |
3124 | qdict_put_str(qdict, BLOCK_OPT_ENCRYPT_FORMAT, "qcow"); | |
3125 | } | |
3126 | ||
3127 | /* Convert compat=0.10/1.1 into compat=v2/v3, to be renamed into | |
3128 | * version=v2/v3 below. */ | |
3129 | val = qdict_get_try_str(qdict, BLOCK_OPT_COMPAT_LEVEL); | |
3130 | if (val && !strcmp(val, "0.10")) { | |
3131 | qdict_put_str(qdict, BLOCK_OPT_COMPAT_LEVEL, "v2"); | |
3132 | } else if (val && !strcmp(val, "1.1")) { | |
3133 | qdict_put_str(qdict, BLOCK_OPT_COMPAT_LEVEL, "v3"); | |
3134 | } | |
3135 | ||
3136 | /* Change legacy command line options into QMP ones */ | |
3137 | static const QDictRenames opt_renames[] = { | |
3138 | { BLOCK_OPT_BACKING_FILE, "backing-file" }, | |
3139 | { BLOCK_OPT_BACKING_FMT, "backing-fmt" }, | |
3140 | { BLOCK_OPT_CLUSTER_SIZE, "cluster-size" }, | |
3141 | { BLOCK_OPT_LAZY_REFCOUNTS, "lazy-refcounts" }, | |
3142 | { BLOCK_OPT_REFCOUNT_BITS, "refcount-bits" }, | |
3143 | { BLOCK_OPT_ENCRYPT, BLOCK_OPT_ENCRYPT_FORMAT }, | |
3144 | { BLOCK_OPT_COMPAT_LEVEL, "version" }, | |
3145 | { NULL, NULL }, | |
3146 | }; | |
3147 | ||
3148 | if (!qdict_rename_keys(qdict, opt_renames, errp)) { | |
29ca9e45 KW |
3149 | ret = -EINVAL; |
3150 | goto finish; | |
3151 | } | |
60900b7b | 3152 | |
b76b4f60 KW |
3153 | /* Create and open the file (protocol layer) */ |
3154 | ret = bdrv_create_file(filename, opts, errp); | |
3155 | if (ret < 0) { | |
0eb4a8c1 SH |
3156 | goto finish; |
3157 | } | |
b76b4f60 KW |
3158 | |
3159 | bs = bdrv_open(filename, NULL, NULL, | |
3160 | BDRV_O_RDWR | BDRV_O_RESIZE | BDRV_O_PROTOCOL, errp); | |
3161 | if (bs == NULL) { | |
3162 | ret = -EIO; | |
1bd0e2d1 CL |
3163 | goto finish; |
3164 | } | |
0eb4a8c1 | 3165 | |
b76b4f60 KW |
3166 | /* Set 'driver' and 'node' options */ |
3167 | qdict_put_str(qdict, "driver", "qcow2"); | |
3168 | qdict_put_str(qdict, "file", bs->node_name); | |
3169 | ||
3170 | /* Now get the QAPI type BlockdevCreateOptions */ | |
af91062e MA |
3171 | v = qobject_input_visitor_new_flat_confused(qdict, errp); |
3172 | if (!v) { | |
1bd0e2d1 CL |
3173 | ret = -EINVAL; |
3174 | goto finish; | |
3175 | } | |
3176 | ||
b76b4f60 KW |
3177 | visit_type_BlockdevCreateOptions(v, NULL, &create_options, &local_err); |
3178 | visit_free(v); | |
de5f3f40 | 3179 | |
0eb4a8c1 SH |
3180 | if (local_err) { |
3181 | error_propagate(errp, local_err); | |
bd4b167f HR |
3182 | ret = -EINVAL; |
3183 | goto finish; | |
3184 | } | |
3185 | ||
b76b4f60 KW |
3186 | /* Silently round up size */ |
3187 | create_options->u.qcow2.size = ROUND_UP(create_options->u.qcow2.size, | |
3188 | BDRV_SECTOR_SIZE); | |
cbf2b7c4 KW |
3189 | |
3190 | /* Create the qcow2 image (format layer) */ | |
b76b4f60 | 3191 | ret = qcow2_co_create(create_options, errp); |
cbf2b7c4 KW |
3192 | if (ret < 0) { |
3193 | goto finish; | |
3194 | } | |
1bd0e2d1 | 3195 | |
b76b4f60 | 3196 | ret = 0; |
1bd0e2d1 | 3197 | finish: |
cb3e7f08 | 3198 | qobject_unref(qdict); |
cbf2b7c4 | 3199 | bdrv_unref(bs); |
b76b4f60 | 3200 | qapi_free_BlockdevCreateOptions(create_options); |
3ef6c40a | 3201 | return ret; |
de5f3f40 KW |
3202 | } |
3203 | ||
2928abce | 3204 | |
f06f6b66 | 3205 | static bool is_zero(BlockDriverState *bs, int64_t offset, int64_t bytes) |
2928abce | 3206 | { |
31826642 EB |
3207 | int64_t nr; |
3208 | int res; | |
f06f6b66 EB |
3209 | |
3210 | /* Clamp to image length, before checking status of underlying sectors */ | |
8cbf74b2 EB |
3211 | if (offset + bytes > bs->total_sectors * BDRV_SECTOR_SIZE) { |
3212 | bytes = bs->total_sectors * BDRV_SECTOR_SIZE - offset; | |
fbaa6bb3 EB |
3213 | } |
3214 | ||
f06f6b66 | 3215 | if (!bytes) { |
ebb718a5 EB |
3216 | return true; |
3217 | } | |
8cbf74b2 | 3218 | res = bdrv_block_status_above(bs, NULL, offset, bytes, &nr, NULL, NULL); |
31826642 | 3219 | return res >= 0 && (res & BDRV_BLOCK_ZERO) && nr == bytes; |
2928abce DL |
3220 | } |
3221 | ||
5544b59f | 3222 | static coroutine_fn int qcow2_co_pwrite_zeroes(BlockDriverState *bs, |
f5a5ca79 | 3223 | int64_t offset, int bytes, BdrvRequestFlags flags) |
621f0589 KW |
3224 | { |
3225 | int ret; | |
ff99129a | 3226 | BDRVQcow2State *s = bs->opaque; |
621f0589 | 3227 | |
5544b59f | 3228 | uint32_t head = offset % s->cluster_size; |
f5a5ca79 | 3229 | uint32_t tail = (offset + bytes) % s->cluster_size; |
2928abce | 3230 | |
f5a5ca79 MP |
3231 | trace_qcow2_pwrite_zeroes_start_req(qemu_coroutine_self(), offset, bytes); |
3232 | if (offset + bytes == bs->total_sectors * BDRV_SECTOR_SIZE) { | |
fbaa6bb3 EB |
3233 | tail = 0; |
3234 | } | |
5a64e942 | 3235 | |
ebb718a5 | 3236 | if (head || tail) { |
ebb718a5 | 3237 | uint64_t off; |
ecfe1863 | 3238 | unsigned int nr; |
2928abce | 3239 | |
f5a5ca79 | 3240 | assert(head + bytes <= s->cluster_size); |
2928abce | 3241 | |
ebb718a5 | 3242 | /* check whether remainder of cluster already reads as zero */ |
f06f6b66 EB |
3243 | if (!(is_zero(bs, offset - head, head) && |
3244 | is_zero(bs, offset + bytes, | |
3245 | tail ? s->cluster_size - tail : 0))) { | |
2928abce DL |
3246 | return -ENOTSUP; |
3247 | } | |
3248 | ||
2928abce DL |
3249 | qemu_co_mutex_lock(&s->lock); |
3250 | /* We can have new write after previous check */ | |
f06f6b66 | 3251 | offset = QEMU_ALIGN_DOWN(offset, s->cluster_size); |
f5a5ca79 | 3252 | bytes = s->cluster_size; |
ecfe1863 | 3253 | nr = s->cluster_size; |
5544b59f | 3254 | ret = qcow2_get_cluster_offset(bs, offset, &nr, &off); |
fdfab37d EB |
3255 | if (ret != QCOW2_CLUSTER_UNALLOCATED && |
3256 | ret != QCOW2_CLUSTER_ZERO_PLAIN && | |
3257 | ret != QCOW2_CLUSTER_ZERO_ALLOC) { | |
2928abce DL |
3258 | qemu_co_mutex_unlock(&s->lock); |
3259 | return -ENOTSUP; | |
3260 | } | |
3261 | } else { | |
3262 | qemu_co_mutex_lock(&s->lock); | |
621f0589 KW |
3263 | } |
3264 | ||
f5a5ca79 | 3265 | trace_qcow2_pwrite_zeroes(qemu_coroutine_self(), offset, bytes); |
5a64e942 | 3266 | |
621f0589 | 3267 | /* Whatever is left can use real zero clusters */ |
f5a5ca79 | 3268 | ret = qcow2_cluster_zeroize(bs, offset, bytes, flags); |
621f0589 KW |
3269 | qemu_co_mutex_unlock(&s->lock); |
3270 | ||
3271 | return ret; | |
3272 | } | |
3273 | ||
82e8a788 | 3274 | static coroutine_fn int qcow2_co_pdiscard(BlockDriverState *bs, |
f5a5ca79 | 3275 | int64_t offset, int bytes) |
5ea929e3 | 3276 | { |
6db39ae2 | 3277 | int ret; |
ff99129a | 3278 | BDRVQcow2State *s = bs->opaque; |
6db39ae2 | 3279 | |
f5a5ca79 MP |
3280 | if (!QEMU_IS_ALIGNED(offset | bytes, s->cluster_size)) { |
3281 | assert(bytes < s->cluster_size); | |
048c5fd1 EB |
3282 | /* Ignore partial clusters, except for the special case of the |
3283 | * complete partial cluster at the end of an unaligned file */ | |
3284 | if (!QEMU_IS_ALIGNED(offset, s->cluster_size) || | |
f5a5ca79 | 3285 | offset + bytes != bs->total_sectors * BDRV_SECTOR_SIZE) { |
048c5fd1 EB |
3286 | return -ENOTSUP; |
3287 | } | |
49228d1e EB |
3288 | } |
3289 | ||
6db39ae2 | 3290 | qemu_co_mutex_lock(&s->lock); |
f5a5ca79 | 3291 | ret = qcow2_cluster_discard(bs, offset, bytes, QCOW2_DISCARD_REQUEST, |
d2cb36af | 3292 | false); |
6db39ae2 PB |
3293 | qemu_co_mutex_unlock(&s->lock); |
3294 | return ret; | |
5ea929e3 KW |
3295 | } |
3296 | ||
fd9fcd37 FZ |
3297 | static int coroutine_fn |
3298 | qcow2_co_copy_range_from(BlockDriverState *bs, | |
3299 | BdrvChild *src, uint64_t src_offset, | |
3300 | BdrvChild *dst, uint64_t dst_offset, | |
67b51fb9 VSO |
3301 | uint64_t bytes, BdrvRequestFlags read_flags, |
3302 | BdrvRequestFlags write_flags) | |
fd9fcd37 FZ |
3303 | { |
3304 | BDRVQcow2State *s = bs->opaque; | |
3305 | int ret; | |
3306 | unsigned int cur_bytes; /* number of bytes in current iteration */ | |
3307 | BdrvChild *child = NULL; | |
67b51fb9 | 3308 | BdrvRequestFlags cur_write_flags; |
fd9fcd37 FZ |
3309 | |
3310 | assert(!bs->encrypted); | |
3311 | qemu_co_mutex_lock(&s->lock); | |
3312 | ||
3313 | while (bytes != 0) { | |
3314 | uint64_t copy_offset = 0; | |
3315 | /* prepare next request */ | |
3316 | cur_bytes = MIN(bytes, INT_MAX); | |
67b51fb9 | 3317 | cur_write_flags = write_flags; |
fd9fcd37 FZ |
3318 | |
3319 | ret = qcow2_get_cluster_offset(bs, src_offset, &cur_bytes, ©_offset); | |
3320 | if (ret < 0) { | |
3321 | goto out; | |
3322 | } | |
3323 | ||
3324 | switch (ret) { | |
3325 | case QCOW2_CLUSTER_UNALLOCATED: | |
3326 | if (bs->backing && bs->backing->bs) { | |
3327 | int64_t backing_length = bdrv_getlength(bs->backing->bs); | |
3328 | if (src_offset >= backing_length) { | |
67b51fb9 | 3329 | cur_write_flags |= BDRV_REQ_ZERO_WRITE; |
fd9fcd37 FZ |
3330 | } else { |
3331 | child = bs->backing; | |
3332 | cur_bytes = MIN(cur_bytes, backing_length - src_offset); | |
3333 | copy_offset = src_offset; | |
3334 | } | |
3335 | } else { | |
67b51fb9 | 3336 | cur_write_flags |= BDRV_REQ_ZERO_WRITE; |
fd9fcd37 FZ |
3337 | } |
3338 | break; | |
3339 | ||
3340 | case QCOW2_CLUSTER_ZERO_PLAIN: | |
3341 | case QCOW2_CLUSTER_ZERO_ALLOC: | |
67b51fb9 | 3342 | cur_write_flags |= BDRV_REQ_ZERO_WRITE; |
fd9fcd37 FZ |
3343 | break; |
3344 | ||
3345 | case QCOW2_CLUSTER_COMPRESSED: | |
3346 | ret = -ENOTSUP; | |
3347 | goto out; | |
fd9fcd37 FZ |
3348 | |
3349 | case QCOW2_CLUSTER_NORMAL: | |
3350 | child = bs->file; | |
3351 | copy_offset += offset_into_cluster(s, src_offset); | |
3352 | if ((copy_offset & 511) != 0) { | |
3353 | ret = -EIO; | |
3354 | goto out; | |
3355 | } | |
3356 | break; | |
3357 | ||
3358 | default: | |
3359 | abort(); | |
3360 | } | |
3361 | qemu_co_mutex_unlock(&s->lock); | |
3362 | ret = bdrv_co_copy_range_from(child, | |
3363 | copy_offset, | |
3364 | dst, dst_offset, | |
67b51fb9 | 3365 | cur_bytes, read_flags, cur_write_flags); |
fd9fcd37 FZ |
3366 | qemu_co_mutex_lock(&s->lock); |
3367 | if (ret < 0) { | |
3368 | goto out; | |
3369 | } | |
3370 | ||
3371 | bytes -= cur_bytes; | |
3372 | src_offset += cur_bytes; | |
3373 | dst_offset += cur_bytes; | |
3374 | } | |
3375 | ret = 0; | |
3376 | ||
3377 | out: | |
3378 | qemu_co_mutex_unlock(&s->lock); | |
3379 | return ret; | |
3380 | } | |
3381 | ||
3382 | static int coroutine_fn | |
3383 | qcow2_co_copy_range_to(BlockDriverState *bs, | |
3384 | BdrvChild *src, uint64_t src_offset, | |
3385 | BdrvChild *dst, uint64_t dst_offset, | |
67b51fb9 VSO |
3386 | uint64_t bytes, BdrvRequestFlags read_flags, |
3387 | BdrvRequestFlags write_flags) | |
fd9fcd37 FZ |
3388 | { |
3389 | BDRVQcow2State *s = bs->opaque; | |
3390 | int offset_in_cluster; | |
3391 | int ret; | |
3392 | unsigned int cur_bytes; /* number of sectors in current iteration */ | |
3393 | uint64_t cluster_offset; | |
fd9fcd37 FZ |
3394 | QCowL2Meta *l2meta = NULL; |
3395 | ||
3396 | assert(!bs->encrypted); | |
fd9fcd37 FZ |
3397 | |
3398 | qemu_co_mutex_lock(&s->lock); | |
3399 | ||
3400 | while (bytes != 0) { | |
3401 | ||
3402 | l2meta = NULL; | |
3403 | ||
3404 | offset_in_cluster = offset_into_cluster(s, dst_offset); | |
3405 | cur_bytes = MIN(bytes, INT_MAX); | |
3406 | ||
3407 | /* TODO: | |
3408 | * If src->bs == dst->bs, we could simply copy by incrementing | |
3409 | * the refcnt, without copying user data. | |
3410 | * Or if src->bs == dst->bs->backing->bs, we could copy by discarding. */ | |
3411 | ret = qcow2_alloc_cluster_offset(bs, dst_offset, &cur_bytes, | |
3412 | &cluster_offset, &l2meta); | |
3413 | if (ret < 0) { | |
3414 | goto fail; | |
3415 | } | |
3416 | ||
3417 | assert((cluster_offset & 511) == 0); | |
3418 | ||
3419 | ret = qcow2_pre_write_overlap_check(bs, 0, | |
3420 | cluster_offset + offset_in_cluster, cur_bytes); | |
3421 | if (ret < 0) { | |
3422 | goto fail; | |
3423 | } | |
3424 | ||
3425 | qemu_co_mutex_unlock(&s->lock); | |
3426 | ret = bdrv_co_copy_range_to(src, src_offset, | |
3427 | bs->file, | |
3428 | cluster_offset + offset_in_cluster, | |
67b51fb9 | 3429 | cur_bytes, read_flags, write_flags); |
fd9fcd37 FZ |
3430 | qemu_co_mutex_lock(&s->lock); |
3431 | if (ret < 0) { | |
3432 | goto fail; | |
3433 | } | |
3434 | ||
3435 | ret = qcow2_handle_l2meta(bs, &l2meta, true); | |
3436 | if (ret) { | |
3437 | goto fail; | |
3438 | } | |
3439 | ||
3440 | bytes -= cur_bytes; | |
e06f4639 | 3441 | src_offset += cur_bytes; |
fd9fcd37 FZ |
3442 | dst_offset += cur_bytes; |
3443 | } | |
3444 | ret = 0; | |
3445 | ||
3446 | fail: | |
3447 | qcow2_handle_l2meta(bs, &l2meta, false); | |
3448 | ||
3449 | qemu_co_mutex_unlock(&s->lock); | |
3450 | ||
fd9fcd37 FZ |
3451 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
3452 | ||
3453 | return ret; | |
3454 | } | |
3455 | ||
061ca8a3 KW |
3456 | static int coroutine_fn qcow2_co_truncate(BlockDriverState *bs, int64_t offset, |
3457 | PreallocMode prealloc, Error **errp) | |
419b19d9 | 3458 | { |
ff99129a | 3459 | BDRVQcow2State *s = bs->opaque; |
95b98f34 | 3460 | uint64_t old_length; |
2cf7cfa1 KW |
3461 | int64_t new_l1_size; |
3462 | int ret; | |
45b4949c | 3463 | QDict *options; |
419b19d9 | 3464 | |
772d1f97 HR |
3465 | if (prealloc != PREALLOC_MODE_OFF && prealloc != PREALLOC_MODE_METADATA && |
3466 | prealloc != PREALLOC_MODE_FALLOC && prealloc != PREALLOC_MODE_FULL) | |
3467 | { | |
8243ccb7 | 3468 | error_setg(errp, "Unsupported preallocation mode '%s'", |
977c736f | 3469 | PreallocMode_str(prealloc)); |
8243ccb7 HR |
3470 | return -ENOTSUP; |
3471 | } | |
3472 | ||
419b19d9 | 3473 | if (offset & 511) { |
4bff28b8 | 3474 | error_setg(errp, "The new size must be a multiple of 512"); |
419b19d9 SH |
3475 | return -EINVAL; |
3476 | } | |
3477 | ||
061ca8a3 KW |
3478 | qemu_co_mutex_lock(&s->lock); |
3479 | ||
419b19d9 SH |
3480 | /* cannot proceed if image has snapshots */ |
3481 | if (s->nb_snapshots) { | |
4bff28b8 | 3482 | error_setg(errp, "Can't resize an image which has snapshots"); |
061ca8a3 KW |
3483 | ret = -ENOTSUP; |
3484 | goto fail; | |
419b19d9 SH |
3485 | } |
3486 | ||
88ddffae VSO |
3487 | /* cannot proceed if image has bitmaps */ |
3488 | if (s->nb_bitmaps) { | |
3489 | /* TODO: resize bitmaps in the image */ | |
3490 | error_setg(errp, "Can't resize an image which has bitmaps"); | |
061ca8a3 KW |
3491 | ret = -ENOTSUP; |
3492 | goto fail; | |
88ddffae VSO |
3493 | } |
3494 | ||
bd016b91 | 3495 | old_length = bs->total_sectors * BDRV_SECTOR_SIZE; |
46b732cd | 3496 | new_l1_size = size_to_l1(s, offset); |
95b98f34 | 3497 | |
95b98f34 | 3498 | if (offset < old_length) { |
163bc39d | 3499 | int64_t last_cluster, old_file_size; |
46b732cd PB |
3500 | if (prealloc != PREALLOC_MODE_OFF) { |
3501 | error_setg(errp, | |
3502 | "Preallocation can't be used for shrinking an image"); | |
061ca8a3 KW |
3503 | ret = -EINVAL; |
3504 | goto fail; | |
46b732cd | 3505 | } |
419b19d9 | 3506 | |
46b732cd PB |
3507 | ret = qcow2_cluster_discard(bs, ROUND_UP(offset, s->cluster_size), |
3508 | old_length - ROUND_UP(offset, | |
3509 | s->cluster_size), | |
3510 | QCOW2_DISCARD_ALWAYS, true); | |
3511 | if (ret < 0) { | |
3512 | error_setg_errno(errp, -ret, "Failed to discard cropped clusters"); | |
061ca8a3 | 3513 | goto fail; |
46b732cd PB |
3514 | } |
3515 | ||
3516 | ret = qcow2_shrink_l1_table(bs, new_l1_size); | |
3517 | if (ret < 0) { | |
3518 | error_setg_errno(errp, -ret, | |
3519 | "Failed to reduce the number of L2 tables"); | |
061ca8a3 | 3520 | goto fail; |
46b732cd PB |
3521 | } |
3522 | ||
3523 | ret = qcow2_shrink_reftable(bs); | |
3524 | if (ret < 0) { | |
3525 | error_setg_errno(errp, -ret, | |
3526 | "Failed to discard unused refblocks"); | |
061ca8a3 | 3527 | goto fail; |
46b732cd | 3528 | } |
163bc39d PB |
3529 | |
3530 | old_file_size = bdrv_getlength(bs->file->bs); | |
3531 | if (old_file_size < 0) { | |
3532 | error_setg_errno(errp, -old_file_size, | |
3533 | "Failed to inquire current file length"); | |
061ca8a3 KW |
3534 | ret = old_file_size; |
3535 | goto fail; | |
163bc39d PB |
3536 | } |
3537 | last_cluster = qcow2_get_last_cluster(bs, old_file_size); | |
3538 | if (last_cluster < 0) { | |
3539 | error_setg_errno(errp, -last_cluster, | |
3540 | "Failed to find the last cluster"); | |
061ca8a3 KW |
3541 | ret = last_cluster; |
3542 | goto fail; | |
163bc39d PB |
3543 | } |
3544 | if ((last_cluster + 1) * s->cluster_size < old_file_size) { | |
233521b1 HR |
3545 | Error *local_err = NULL; |
3546 | ||
061ca8a3 KW |
3547 | bdrv_co_truncate(bs->file, (last_cluster + 1) * s->cluster_size, |
3548 | PREALLOC_MODE_OFF, &local_err); | |
233521b1 HR |
3549 | if (local_err) { |
3550 | warn_reportf_err(local_err, | |
3551 | "Failed to truncate the tail of the image: "); | |
163bc39d PB |
3552 | } |
3553 | } | |
46b732cd PB |
3554 | } else { |
3555 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); | |
3556 | if (ret < 0) { | |
3557 | error_setg_errno(errp, -ret, "Failed to grow the L1 table"); | |
061ca8a3 | 3558 | goto fail; |
46b732cd | 3559 | } |
419b19d9 SH |
3560 | } |
3561 | ||
95b98f34 HR |
3562 | switch (prealloc) { |
3563 | case PREALLOC_MODE_OFF: | |
3564 | break; | |
3565 | ||
3566 | case PREALLOC_MODE_METADATA: | |
47e86b86 | 3567 | ret = preallocate_co(bs, old_length, offset); |
95b98f34 HR |
3568 | if (ret < 0) { |
3569 | error_setg_errno(errp, -ret, "Preallocation failed"); | |
061ca8a3 | 3570 | goto fail; |
95b98f34 HR |
3571 | } |
3572 | break; | |
3573 | ||
772d1f97 HR |
3574 | case PREALLOC_MODE_FALLOC: |
3575 | case PREALLOC_MODE_FULL: | |
3576 | { | |
3577 | int64_t allocation_start, host_offset, guest_offset; | |
3578 | int64_t clusters_allocated; | |
3579 | int64_t old_file_size, new_file_size; | |
3580 | uint64_t nb_new_data_clusters, nb_new_l2_tables; | |
3581 | ||
3582 | old_file_size = bdrv_getlength(bs->file->bs); | |
3583 | if (old_file_size < 0) { | |
3584 | error_setg_errno(errp, -old_file_size, | |
3585 | "Failed to inquire current file length"); | |
061ca8a3 KW |
3586 | ret = old_file_size; |
3587 | goto fail; | |
772d1f97 | 3588 | } |
e400ad1e | 3589 | old_file_size = ROUND_UP(old_file_size, s->cluster_size); |
772d1f97 HR |
3590 | |
3591 | nb_new_data_clusters = DIV_ROUND_UP(offset - old_length, | |
3592 | s->cluster_size); | |
3593 | ||
3594 | /* This is an overestimation; we will not actually allocate space for | |
3595 | * these in the file but just make sure the new refcount structures are | |
3596 | * able to cover them so we will not have to allocate new refblocks | |
3597 | * while entering the data blocks in the potentially new L2 tables. | |
3598 | * (We do not actually care where the L2 tables are placed. Maybe they | |
3599 | * are already allocated or they can be placed somewhere before | |
3600 | * @old_file_size. It does not matter because they will be fully | |
3601 | * allocated automatically, so they do not need to be covered by the | |
3602 | * preallocation. All that matters is that we will not have to allocate | |
3603 | * new refcount structures for them.) */ | |
3604 | nb_new_l2_tables = DIV_ROUND_UP(nb_new_data_clusters, | |
3605 | s->cluster_size / sizeof(uint64_t)); | |
3606 | /* The cluster range may not be aligned to L2 boundaries, so add one L2 | |
3607 | * table for a potential head/tail */ | |
3608 | nb_new_l2_tables++; | |
3609 | ||
3610 | allocation_start = qcow2_refcount_area(bs, old_file_size, | |
3611 | nb_new_data_clusters + | |
3612 | nb_new_l2_tables, | |
3613 | true, 0, 0); | |
3614 | if (allocation_start < 0) { | |
3615 | error_setg_errno(errp, -allocation_start, | |
3616 | "Failed to resize refcount structures"); | |
061ca8a3 KW |
3617 | ret = allocation_start; |
3618 | goto fail; | |
772d1f97 HR |
3619 | } |
3620 | ||
3621 | clusters_allocated = qcow2_alloc_clusters_at(bs, allocation_start, | |
3622 | nb_new_data_clusters); | |
3623 | if (clusters_allocated < 0) { | |
3624 | error_setg_errno(errp, -clusters_allocated, | |
3625 | "Failed to allocate data clusters"); | |
061ca8a3 KW |
3626 | ret = clusters_allocated; |
3627 | goto fail; | |
772d1f97 HR |
3628 | } |
3629 | ||
3630 | assert(clusters_allocated == nb_new_data_clusters); | |
3631 | ||
3632 | /* Allocate the data area */ | |
3633 | new_file_size = allocation_start + | |
3634 | nb_new_data_clusters * s->cluster_size; | |
061ca8a3 | 3635 | ret = bdrv_co_truncate(bs->file, new_file_size, prealloc, errp); |
772d1f97 HR |
3636 | if (ret < 0) { |
3637 | error_prepend(errp, "Failed to resize underlying file: "); | |
3638 | qcow2_free_clusters(bs, allocation_start, | |
3639 | nb_new_data_clusters * s->cluster_size, | |
3640 | QCOW2_DISCARD_OTHER); | |
061ca8a3 | 3641 | goto fail; |
772d1f97 HR |
3642 | } |
3643 | ||
3644 | /* Create the necessary L2 entries */ | |
3645 | host_offset = allocation_start; | |
3646 | guest_offset = old_length; | |
3647 | while (nb_new_data_clusters) { | |
13bec229 AG |
3648 | int64_t nb_clusters = MIN( |
3649 | nb_new_data_clusters, | |
3650 | s->l2_slice_size - offset_to_l2_slice_index(s, guest_offset)); | |
772d1f97 HR |
3651 | QCowL2Meta allocation = { |
3652 | .offset = guest_offset, | |
3653 | .alloc_offset = host_offset, | |
3654 | .nb_clusters = nb_clusters, | |
3655 | }; | |
3656 | qemu_co_queue_init(&allocation.dependent_requests); | |
3657 | ||
3658 | ret = qcow2_alloc_cluster_link_l2(bs, &allocation); | |
3659 | if (ret < 0) { | |
3660 | error_setg_errno(errp, -ret, "Failed to update L2 tables"); | |
3661 | qcow2_free_clusters(bs, host_offset, | |
3662 | nb_new_data_clusters * s->cluster_size, | |
3663 | QCOW2_DISCARD_OTHER); | |
061ca8a3 | 3664 | goto fail; |
772d1f97 HR |
3665 | } |
3666 | ||
3667 | guest_offset += nb_clusters * s->cluster_size; | |
3668 | host_offset += nb_clusters * s->cluster_size; | |
3669 | nb_new_data_clusters -= nb_clusters; | |
3670 | } | |
3671 | break; | |
3672 | } | |
3673 | ||
95b98f34 HR |
3674 | default: |
3675 | g_assert_not_reached(); | |
3676 | } | |
3677 | ||
3678 | if (prealloc != PREALLOC_MODE_OFF) { | |
3679 | /* Flush metadata before actually changing the image size */ | |
061ca8a3 | 3680 | ret = qcow2_write_caches(bs); |
95b98f34 HR |
3681 | if (ret < 0) { |
3682 | error_setg_errno(errp, -ret, | |
3683 | "Failed to flush the preallocated area to disk"); | |
061ca8a3 | 3684 | goto fail; |
95b98f34 HR |
3685 | } |
3686 | } | |
3687 | ||
45b4949c LB |
3688 | bs->total_sectors = offset / BDRV_SECTOR_SIZE; |
3689 | ||
419b19d9 SH |
3690 | /* write updated header.size */ |
3691 | offset = cpu_to_be64(offset); | |
d9ca2ea2 | 3692 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), |
8b3b7206 | 3693 | &offset, sizeof(uint64_t)); |
419b19d9 | 3694 | if (ret < 0) { |
f59adb32 | 3695 | error_setg_errno(errp, -ret, "Failed to update the image size"); |
061ca8a3 | 3696 | goto fail; |
419b19d9 SH |
3697 | } |
3698 | ||
3699 | s->l1_vm_state_index = new_l1_size; | |
45b4949c LB |
3700 | |
3701 | /* Update cache sizes */ | |
3702 | options = qdict_clone_shallow(bs->options); | |
3703 | ret = qcow2_update_options(bs, options, s->flags, errp); | |
3704 | qobject_unref(options); | |
3705 | if (ret < 0) { | |
3706 | goto fail; | |
3707 | } | |
061ca8a3 KW |
3708 | ret = 0; |
3709 | fail: | |
3710 | qemu_co_mutex_unlock(&s->lock); | |
3711 | return ret; | |
419b19d9 SH |
3712 | } |
3713 | ||
2714f13d VSO |
3714 | /* |
3715 | * qcow2_compress() | |
3716 | * | |
6994fd78 VSO |
3717 | * @dest - destination buffer, @dest_size bytes |
3718 | * @src - source buffer, @src_size bytes | |
2714f13d VSO |
3719 | * |
3720 | * Returns: compressed size on success | |
6994fd78 | 3721 | * -1 destination buffer is not enough to store compressed data |
2714f13d VSO |
3722 | * -2 on any other error |
3723 | */ | |
6994fd78 VSO |
3724 | static ssize_t qcow2_compress(void *dest, size_t dest_size, |
3725 | const void *src, size_t src_size) | |
2714f13d VSO |
3726 | { |
3727 | ssize_t ret; | |
3728 | z_stream strm; | |
3729 | ||
3730 | /* best compression, small window, no zlib header */ | |
3731 | memset(&strm, 0, sizeof(strm)); | |
3732 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, Z_DEFLATED, | |
3733 | -12, 9, Z_DEFAULT_STRATEGY); | |
19a44488 | 3734 | if (ret != Z_OK) { |
2714f13d VSO |
3735 | return -2; |
3736 | } | |
3737 | ||
3738 | /* strm.next_in is not const in old zlib versions, such as those used on | |
3739 | * OpenBSD/NetBSD, so cast the const away */ | |
6994fd78 | 3740 | strm.avail_in = src_size; |
2714f13d | 3741 | strm.next_in = (void *) src; |
6994fd78 | 3742 | strm.avail_out = dest_size; |
2714f13d VSO |
3743 | strm.next_out = dest; |
3744 | ||
3745 | ret = deflate(&strm, Z_FINISH); | |
3746 | if (ret == Z_STREAM_END) { | |
6994fd78 | 3747 | ret = dest_size - strm.avail_out; |
2714f13d VSO |
3748 | } else { |
3749 | ret = (ret == Z_OK ? -1 : -2); | |
3750 | } | |
3751 | ||
3752 | deflateEnd(&strm); | |
3753 | ||
3754 | return ret; | |
3755 | } | |
3756 | ||
341926ab VSO |
3757 | /* |
3758 | * qcow2_decompress() | |
3759 | * | |
3760 | * Decompress some data (not more than @src_size bytes) to produce exactly | |
3761 | * @dest_size bytes. | |
3762 | * | |
3763 | * @dest - destination buffer, @dest_size bytes | |
3764 | * @src - source buffer, @src_size bytes | |
3765 | * | |
3766 | * Returns: 0 on success | |
3767 | * -1 on fail | |
3768 | */ | |
3769 | static ssize_t qcow2_decompress(void *dest, size_t dest_size, | |
3770 | const void *src, size_t src_size) | |
f4b3e2a9 | 3771 | { |
341926ab VSO |
3772 | int ret = 0; |
3773 | z_stream strm; | |
f4b3e2a9 | 3774 | |
341926ab VSO |
3775 | memset(&strm, 0, sizeof(strm)); |
3776 | strm.avail_in = src_size; | |
3777 | strm.next_in = (void *) src; | |
3778 | strm.avail_out = dest_size; | |
3779 | strm.next_out = dest; | |
f4b3e2a9 | 3780 | |
341926ab | 3781 | ret = inflateInit2(&strm, -12); |
f4b3e2a9 VSO |
3782 | if (ret != Z_OK) { |
3783 | return -1; | |
3784 | } | |
341926ab VSO |
3785 | |
3786 | ret = inflate(&strm, Z_FINISH); | |
3787 | if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) || strm.avail_out != 0) { | |
3788 | /* We approve Z_BUF_ERROR because we need @dest buffer to be filled, but | |
3789 | * @src buffer may be processed partly (because in qcow2 we know size of | |
3790 | * compressed data with precision of one sector) */ | |
3791 | ret = -1; | |
f4b3e2a9 | 3792 | } |
341926ab VSO |
3793 | |
3794 | inflateEnd(&strm); | |
3795 | ||
3796 | return ret; | |
f4b3e2a9 VSO |
3797 | } |
3798 | ||
ceb029cd VSO |
3799 | #define MAX_COMPRESS_THREADS 4 |
3800 | ||
e23c9d7a VSO |
3801 | typedef ssize_t (*Qcow2CompressFunc)(void *dest, size_t dest_size, |
3802 | const void *src, size_t src_size); | |
ceb029cd VSO |
3803 | typedef struct Qcow2CompressData { |
3804 | void *dest; | |
6994fd78 | 3805 | size_t dest_size; |
ceb029cd | 3806 | const void *src; |
6994fd78 | 3807 | size_t src_size; |
ceb029cd | 3808 | ssize_t ret; |
e23c9d7a VSO |
3809 | |
3810 | Qcow2CompressFunc func; | |
ceb029cd VSO |
3811 | } Qcow2CompressData; |
3812 | ||
3813 | static int qcow2_compress_pool_func(void *opaque) | |
3814 | { | |
3815 | Qcow2CompressData *data = opaque; | |
3816 | ||
e23c9d7a VSO |
3817 | data->ret = data->func(data->dest, data->dest_size, |
3818 | data->src, data->src_size); | |
ceb029cd VSO |
3819 | |
3820 | return 0; | |
3821 | } | |
3822 | ||
3823 | static void qcow2_compress_complete(void *opaque, int ret) | |
3824 | { | |
3825 | qemu_coroutine_enter(opaque); | |
3826 | } | |
3827 | ||
e23c9d7a VSO |
3828 | static ssize_t coroutine_fn |
3829 | qcow2_co_do_compress(BlockDriverState *bs, void *dest, size_t dest_size, | |
3830 | const void *src, size_t src_size, Qcow2CompressFunc func) | |
ceb029cd VSO |
3831 | { |
3832 | BDRVQcow2State *s = bs->opaque; | |
3833 | BlockAIOCB *acb; | |
3834 | ThreadPool *pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); | |
3835 | Qcow2CompressData arg = { | |
3836 | .dest = dest, | |
6994fd78 | 3837 | .dest_size = dest_size, |
ceb029cd | 3838 | .src = src, |
6994fd78 | 3839 | .src_size = src_size, |
e23c9d7a | 3840 | .func = func, |
ceb029cd VSO |
3841 | }; |
3842 | ||
3843 | while (s->nb_compress_threads >= MAX_COMPRESS_THREADS) { | |
3844 | qemu_co_queue_wait(&s->compress_wait_queue, NULL); | |
3845 | } | |
3846 | ||
3847 | s->nb_compress_threads++; | |
3848 | acb = thread_pool_submit_aio(pool, qcow2_compress_pool_func, &arg, | |
3849 | qcow2_compress_complete, | |
3850 | qemu_coroutine_self()); | |
3851 | ||
3852 | if (!acb) { | |
3853 | s->nb_compress_threads--; | |
3854 | return -EINVAL; | |
3855 | } | |
3856 | qemu_coroutine_yield(); | |
3857 | s->nb_compress_threads--; | |
3858 | qemu_co_queue_next(&s->compress_wait_queue); | |
3859 | ||
3860 | return arg.ret; | |
3861 | } | |
3862 | ||
e23c9d7a VSO |
3863 | static ssize_t coroutine_fn |
3864 | qcow2_co_compress(BlockDriverState *bs, void *dest, size_t dest_size, | |
3865 | const void *src, size_t src_size) | |
3866 | { | |
3867 | return qcow2_co_do_compress(bs, dest, dest_size, src, src_size, | |
3868 | qcow2_compress); | |
3869 | } | |
3870 | ||
3871 | static ssize_t coroutine_fn | |
3872 | qcow2_co_decompress(BlockDriverState *bs, void *dest, size_t dest_size, | |
3873 | const void *src, size_t src_size) | |
3874 | { | |
3875 | return qcow2_co_do_compress(bs, dest, dest_size, src, src_size, | |
3876 | qcow2_decompress); | |
3877 | } | |
3878 | ||
20d97356 BS |
3879 | /* XXX: put compressed sectors first, then all the cluster aligned |
3880 | tables to avoid losing bytes in alignment */ | |
fcccefc5 PB |
3881 | static coroutine_fn int |
3882 | qcow2_co_pwritev_compressed(BlockDriverState *bs, uint64_t offset, | |
3883 | uint64_t bytes, QEMUIOVector *qiov) | |
20d97356 | 3884 | { |
ff99129a | 3885 | BDRVQcow2State *s = bs->opaque; |
fcccefc5 | 3886 | QEMUIOVector hd_qiov; |
2714f13d VSO |
3887 | int ret; |
3888 | size_t out_len; | |
fcccefc5 | 3889 | uint8_t *buf, *out_buf; |
d0d5d0e3 | 3890 | int64_t cluster_offset; |
20d97356 | 3891 | |
fcccefc5 | 3892 | if (bytes == 0) { |
20d97356 BS |
3893 | /* align end of file to a sector boundary to ease reading with |
3894 | sector based I/Os */ | |
9a4f4c31 | 3895 | cluster_offset = bdrv_getlength(bs->file->bs); |
d0d5d0e3 EB |
3896 | if (cluster_offset < 0) { |
3897 | return cluster_offset; | |
3898 | } | |
061ca8a3 KW |
3899 | return bdrv_co_truncate(bs->file, cluster_offset, PREALLOC_MODE_OFF, |
3900 | NULL); | |
20d97356 BS |
3901 | } |
3902 | ||
3e3b838f AN |
3903 | if (offset_into_cluster(s, offset)) { |
3904 | return -EINVAL; | |
3905 | } | |
3906 | ||
a2c0ca6f | 3907 | buf = qemu_blockalign(bs, s->cluster_size); |
fcccefc5 | 3908 | if (bytes != s->cluster_size) { |
a2c0ca6f PB |
3909 | if (bytes > s->cluster_size || |
3910 | offset + bytes != bs->total_sectors << BDRV_SECTOR_BITS) | |
fcccefc5 | 3911 | { |
a2c0ca6f PB |
3912 | qemu_vfree(buf); |
3913 | return -EINVAL; | |
f4d38bef | 3914 | } |
a2c0ca6f PB |
3915 | /* Zero-pad last write if image size is not cluster aligned */ |
3916 | memset(buf + bytes, 0, s->cluster_size - bytes); | |
f4d38bef | 3917 | } |
8b2bd093 | 3918 | qemu_iovec_to_buf(qiov, 0, buf, bytes); |
20d97356 | 3919 | |
ebf7bba0 | 3920 | out_buf = g_malloc(s->cluster_size); |
20d97356 | 3921 | |
6994fd78 VSO |
3922 | out_len = qcow2_co_compress(bs, out_buf, s->cluster_size - 1, |
3923 | buf, s->cluster_size); | |
2714f13d | 3924 | if (out_len == -2) { |
8f1efd00 KW |
3925 | ret = -EINVAL; |
3926 | goto fail; | |
2714f13d | 3927 | } else if (out_len == -1) { |
20d97356 | 3928 | /* could not compress: write normal cluster */ |
fcccefc5 | 3929 | ret = qcow2_co_pwritev(bs, offset, bytes, qiov, 0); |
8f1efd00 KW |
3930 | if (ret < 0) { |
3931 | goto fail; | |
3932 | } | |
fcccefc5 PB |
3933 | goto success; |
3934 | } | |
cf93980e | 3935 | |
fcccefc5 PB |
3936 | qemu_co_mutex_lock(&s->lock); |
3937 | cluster_offset = | |
3938 | qcow2_alloc_compressed_cluster_offset(bs, offset, out_len); | |
3939 | if (!cluster_offset) { | |
3940 | qemu_co_mutex_unlock(&s->lock); | |
3941 | ret = -EIO; | |
3942 | goto fail; | |
3943 | } | |
3944 | cluster_offset &= s->cluster_offset_mask; | |
cf93980e | 3945 | |
fcccefc5 PB |
3946 | ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len); |
3947 | qemu_co_mutex_unlock(&s->lock); | |
3948 | if (ret < 0) { | |
3949 | goto fail; | |
20d97356 BS |
3950 | } |
3951 | ||
c793d4ff | 3952 | qemu_iovec_init_buf(&hd_qiov, out_buf, out_len); |
fcccefc5 PB |
3953 | |
3954 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); | |
3955 | ret = bdrv_co_pwritev(bs->file, cluster_offset, out_len, &hd_qiov, 0); | |
3956 | if (ret < 0) { | |
3957 | goto fail; | |
3958 | } | |
3959 | success: | |
8f1efd00 KW |
3960 | ret = 0; |
3961 | fail: | |
fcccefc5 | 3962 | qemu_vfree(buf); |
7267c094 | 3963 | g_free(out_buf); |
8f1efd00 | 3964 | return ret; |
20d97356 BS |
3965 | } |
3966 | ||
c3c10f72 VSO |
3967 | static int coroutine_fn |
3968 | qcow2_co_preadv_compressed(BlockDriverState *bs, | |
3969 | uint64_t file_cluster_offset, | |
3970 | uint64_t offset, | |
3971 | uint64_t bytes, | |
3972 | QEMUIOVector *qiov) | |
f4b3e2a9 VSO |
3973 | { |
3974 | BDRVQcow2State *s = bs->opaque; | |
c3c10f72 | 3975 | int ret = 0, csize, nb_csectors; |
f4b3e2a9 | 3976 | uint64_t coffset; |
c3c10f72 | 3977 | uint8_t *buf, *out_buf; |
c068a1cd | 3978 | QEMUIOVector local_qiov; |
c3c10f72 | 3979 | int offset_in_cluster = offset_into_cluster(s, offset); |
f4b3e2a9 | 3980 | |
c3c10f72 VSO |
3981 | coffset = file_cluster_offset & s->cluster_offset_mask; |
3982 | nb_csectors = ((file_cluster_offset >> s->csize_shift) & s->csize_mask) + 1; | |
3983 | csize = nb_csectors * 512 - (coffset & 511); | |
f4b3e2a9 | 3984 | |
c3c10f72 VSO |
3985 | buf = g_try_malloc(csize); |
3986 | if (!buf) { | |
3987 | return -ENOMEM; | |
3988 | } | |
c793d4ff | 3989 | qemu_iovec_init_buf(&local_qiov, buf, csize); |
f4b3e2a9 | 3990 | |
c3c10f72 | 3991 | out_buf = qemu_blockalign(bs, s->cluster_size); |
c068a1cd | 3992 | |
c3c10f72 VSO |
3993 | BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED); |
3994 | ret = bdrv_co_preadv(bs->file, coffset, csize, &local_qiov, 0); | |
3995 | if (ret < 0) { | |
3996 | goto fail; | |
f4b3e2a9 | 3997 | } |
c3c10f72 | 3998 | |
e23c9d7a | 3999 | if (qcow2_co_decompress(bs, out_buf, s->cluster_size, buf, csize) < 0) { |
c3c10f72 VSO |
4000 | ret = -EIO; |
4001 | goto fail; | |
4002 | } | |
4003 | ||
4004 | qemu_iovec_from_buf(qiov, 0, out_buf + offset_in_cluster, bytes); | |
4005 | ||
4006 | fail: | |
4007 | qemu_vfree(out_buf); | |
4008 | g_free(buf); | |
4009 | ||
4010 | return ret; | |
f4b3e2a9 VSO |
4011 | } |
4012 | ||
94054183 HR |
4013 | static int make_completely_empty(BlockDriverState *bs) |
4014 | { | |
ff99129a | 4015 | BDRVQcow2State *s = bs->opaque; |
ed3d2ec9 | 4016 | Error *local_err = NULL; |
94054183 HR |
4017 | int ret, l1_clusters; |
4018 | int64_t offset; | |
4019 | uint64_t *new_reftable = NULL; | |
4020 | uint64_t rt_entry, l1_size2; | |
4021 | struct { | |
4022 | uint64_t l1_offset; | |
4023 | uint64_t reftable_offset; | |
4024 | uint32_t reftable_clusters; | |
4025 | } QEMU_PACKED l1_ofs_rt_ofs_cls; | |
4026 | ||
4027 | ret = qcow2_cache_empty(bs, s->l2_table_cache); | |
4028 | if (ret < 0) { | |
4029 | goto fail; | |
4030 | } | |
4031 | ||
4032 | ret = qcow2_cache_empty(bs, s->refcount_block_cache); | |
4033 | if (ret < 0) { | |
4034 | goto fail; | |
4035 | } | |
4036 | ||
4037 | /* Refcounts will be broken utterly */ | |
4038 | ret = qcow2_mark_dirty(bs); | |
4039 | if (ret < 0) { | |
4040 | goto fail; | |
4041 | } | |
4042 | ||
4043 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
4044 | ||
4045 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
4046 | l1_size2 = (uint64_t)s->l1_size * sizeof(uint64_t); | |
4047 | ||
4048 | /* After this call, neither the in-memory nor the on-disk refcount | |
4049 | * information accurately describe the actual references */ | |
4050 | ||
720ff280 | 4051 | ret = bdrv_pwrite_zeroes(bs->file, s->l1_table_offset, |
74021bc4 | 4052 | l1_clusters * s->cluster_size, 0); |
94054183 HR |
4053 | if (ret < 0) { |
4054 | goto fail_broken_refcounts; | |
4055 | } | |
4056 | memset(s->l1_table, 0, l1_size2); | |
4057 | ||
4058 | BLKDBG_EVENT(bs->file, BLKDBG_EMPTY_IMAGE_PREPARE); | |
4059 | ||
4060 | /* Overwrite enough clusters at the beginning of the sectors to place | |
4061 | * the refcount table, a refcount block and the L1 table in; this may | |
4062 | * overwrite parts of the existing refcount and L1 table, which is not | |
4063 | * an issue because the dirty flag is set, complete data loss is in fact | |
4064 | * desired and partial data loss is consequently fine as well */ | |
720ff280 | 4065 | ret = bdrv_pwrite_zeroes(bs->file, s->cluster_size, |
74021bc4 | 4066 | (2 + l1_clusters) * s->cluster_size, 0); |
94054183 HR |
4067 | /* This call (even if it failed overall) may have overwritten on-disk |
4068 | * refcount structures; in that case, the in-memory refcount information | |
4069 | * will probably differ from the on-disk information which makes the BDS | |
4070 | * unusable */ | |
4071 | if (ret < 0) { | |
4072 | goto fail_broken_refcounts; | |
4073 | } | |
4074 | ||
4075 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
4076 | BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE); | |
4077 | ||
4078 | /* "Create" an empty reftable (one cluster) directly after the image | |
4079 | * header and an empty L1 table three clusters after the image header; | |
4080 | * the cluster between those two will be used as the first refblock */ | |
f1f7a1dd PM |
4081 | l1_ofs_rt_ofs_cls.l1_offset = cpu_to_be64(3 * s->cluster_size); |
4082 | l1_ofs_rt_ofs_cls.reftable_offset = cpu_to_be64(s->cluster_size); | |
4083 | l1_ofs_rt_ofs_cls.reftable_clusters = cpu_to_be32(1); | |
d9ca2ea2 | 4084 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, l1_table_offset), |
94054183 HR |
4085 | &l1_ofs_rt_ofs_cls, sizeof(l1_ofs_rt_ofs_cls)); |
4086 | if (ret < 0) { | |
4087 | goto fail_broken_refcounts; | |
4088 | } | |
4089 | ||
4090 | s->l1_table_offset = 3 * s->cluster_size; | |
4091 | ||
4092 | new_reftable = g_try_new0(uint64_t, s->cluster_size / sizeof(uint64_t)); | |
4093 | if (!new_reftable) { | |
4094 | ret = -ENOMEM; | |
4095 | goto fail_broken_refcounts; | |
4096 | } | |
4097 | ||
4098 | s->refcount_table_offset = s->cluster_size; | |
4099 | s->refcount_table_size = s->cluster_size / sizeof(uint64_t); | |
7061a078 | 4100 | s->max_refcount_table_index = 0; |
94054183 HR |
4101 | |
4102 | g_free(s->refcount_table); | |
4103 | s->refcount_table = new_reftable; | |
4104 | new_reftable = NULL; | |
4105 | ||
4106 | /* Now the in-memory refcount information again corresponds to the on-disk | |
4107 | * information (reftable is empty and no refblocks (the refblock cache is | |
4108 | * empty)); however, this means some clusters (e.g. the image header) are | |
4109 | * referenced, but not refcounted, but the normal qcow2 code assumes that | |
4110 | * the in-memory information is always correct */ | |
4111 | ||
4112 | BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC); | |
4113 | ||
4114 | /* Enter the first refblock into the reftable */ | |
4115 | rt_entry = cpu_to_be64(2 * s->cluster_size); | |
d9ca2ea2 | 4116 | ret = bdrv_pwrite_sync(bs->file, s->cluster_size, |
94054183 HR |
4117 | &rt_entry, sizeof(rt_entry)); |
4118 | if (ret < 0) { | |
4119 | goto fail_broken_refcounts; | |
4120 | } | |
4121 | s->refcount_table[0] = 2 * s->cluster_size; | |
4122 | ||
4123 | s->free_cluster_index = 0; | |
4124 | assert(3 + l1_clusters <= s->refcount_block_size); | |
4125 | offset = qcow2_alloc_clusters(bs, 3 * s->cluster_size + l1_size2); | |
4126 | if (offset < 0) { | |
4127 | ret = offset; | |
4128 | goto fail_broken_refcounts; | |
4129 | } else if (offset > 0) { | |
4130 | error_report("First cluster in emptied image is in use"); | |
4131 | abort(); | |
4132 | } | |
4133 | ||
4134 | /* Now finally the in-memory information corresponds to the on-disk | |
4135 | * structures and is correct */ | |
4136 | ret = qcow2_mark_clean(bs); | |
4137 | if (ret < 0) { | |
4138 | goto fail; | |
4139 | } | |
4140 | ||
ed3d2ec9 | 4141 | ret = bdrv_truncate(bs->file, (3 + l1_clusters) * s->cluster_size, |
7ea37c30 | 4142 | PREALLOC_MODE_OFF, &local_err); |
94054183 | 4143 | if (ret < 0) { |
ed3d2ec9 | 4144 | error_report_err(local_err); |
94054183 HR |
4145 | goto fail; |
4146 | } | |
4147 | ||
4148 | return 0; | |
4149 | ||
4150 | fail_broken_refcounts: | |
4151 | /* The BDS is unusable at this point. If we wanted to make it usable, we | |
4152 | * would have to call qcow2_refcount_close(), qcow2_refcount_init(), | |
4153 | * qcow2_check_refcounts(), qcow2_refcount_close() and qcow2_refcount_init() | |
4154 | * again. However, because the functions which could have caused this error | |
4155 | * path to be taken are used by those functions as well, it's very likely | |
4156 | * that that sequence will fail as well. Therefore, just eject the BDS. */ | |
4157 | bs->drv = NULL; | |
4158 | ||
4159 | fail: | |
4160 | g_free(new_reftable); | |
4161 | return ret; | |
4162 | } | |
4163 | ||
491d27e2 HR |
4164 | static int qcow2_make_empty(BlockDriverState *bs) |
4165 | { | |
ff99129a | 4166 | BDRVQcow2State *s = bs->opaque; |
d2cb36af EB |
4167 | uint64_t offset, end_offset; |
4168 | int step = QEMU_ALIGN_DOWN(INT_MAX, s->cluster_size); | |
94054183 HR |
4169 | int l1_clusters, ret = 0; |
4170 | ||
4171 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
4172 | ||
4096974e | 4173 | if (s->qcow_version >= 3 && !s->snapshots && !s->nb_bitmaps && |
f0603329 DB |
4174 | 3 + l1_clusters <= s->refcount_block_size && |
4175 | s->crypt_method_header != QCOW_CRYPT_LUKS) { | |
4096974e EB |
4176 | /* The following function only works for qcow2 v3 images (it |
4177 | * requires the dirty flag) and only as long as there are no | |
4178 | * features that reserve extra clusters (such as snapshots, | |
4179 | * LUKS header, or persistent bitmaps), because it completely | |
4180 | * empties the image. Furthermore, the L1 table and three | |
4181 | * additional clusters (image header, refcount table, one | |
4182 | * refcount block) have to fit inside one refcount block. */ | |
94054183 HR |
4183 | return make_completely_empty(bs); |
4184 | } | |
491d27e2 | 4185 | |
94054183 HR |
4186 | /* This fallback code simply discards every active cluster; this is slow, |
4187 | * but works in all cases */ | |
d2cb36af EB |
4188 | end_offset = bs->total_sectors * BDRV_SECTOR_SIZE; |
4189 | for (offset = 0; offset < end_offset; offset += step) { | |
491d27e2 HR |
4190 | /* As this function is generally used after committing an external |
4191 | * snapshot, QCOW2_DISCARD_SNAPSHOT seems appropriate. Also, the | |
4192 | * default action for this kind of discard is to pass the discard, | |
4193 | * which will ideally result in an actually smaller image file, as | |
4194 | * is probably desired. */ | |
d2cb36af EB |
4195 | ret = qcow2_cluster_discard(bs, offset, MIN(step, end_offset - offset), |
4196 | QCOW2_DISCARD_SNAPSHOT, true); | |
491d27e2 HR |
4197 | if (ret < 0) { |
4198 | break; | |
4199 | } | |
4200 | } | |
4201 | ||
4202 | return ret; | |
4203 | } | |
4204 | ||
a968168c | 4205 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
20d97356 | 4206 | { |
ff99129a | 4207 | BDRVQcow2State *s = bs->opaque; |
29c1a730 KW |
4208 | int ret; |
4209 | ||
8b94ff85 | 4210 | qemu_co_mutex_lock(&s->lock); |
8b220eb7 | 4211 | ret = qcow2_write_caches(bs); |
8b94ff85 | 4212 | qemu_co_mutex_unlock(&s->lock); |
29c1a730 | 4213 | |
8b220eb7 | 4214 | return ret; |
eb489bb1 KW |
4215 | } |
4216 | ||
61914f89 SH |
4217 | static ssize_t qcow2_measure_crypto_hdr_init_func(QCryptoBlock *block, |
4218 | size_t headerlen, void *opaque, Error **errp) | |
4219 | { | |
4220 | size_t *headerlenp = opaque; | |
4221 | ||
4222 | /* Stash away the payload size */ | |
4223 | *headerlenp = headerlen; | |
4224 | return 0; | |
4225 | } | |
4226 | ||
4227 | static ssize_t qcow2_measure_crypto_hdr_write_func(QCryptoBlock *block, | |
4228 | size_t offset, const uint8_t *buf, size_t buflen, | |
4229 | void *opaque, Error **errp) | |
4230 | { | |
4231 | /* Discard the bytes, we're not actually writing to an image */ | |
4232 | return buflen; | |
4233 | } | |
4234 | ||
4235 | /* Determine the number of bytes for the LUKS payload */ | |
4236 | static bool qcow2_measure_luks_headerlen(QemuOpts *opts, size_t *len, | |
4237 | Error **errp) | |
4238 | { | |
4239 | QDict *opts_qdict; | |
4240 | QDict *cryptoopts_qdict; | |
4241 | QCryptoBlockCreateOptions *cryptoopts; | |
4242 | QCryptoBlock *crypto; | |
4243 | ||
4244 | /* Extract "encrypt." options into a qdict */ | |
4245 | opts_qdict = qemu_opts_to_qdict(opts, NULL); | |
4246 | qdict_extract_subqdict(opts_qdict, &cryptoopts_qdict, "encrypt."); | |
4247 | qobject_unref(opts_qdict); | |
4248 | ||
4249 | /* Build QCryptoBlockCreateOptions object from qdict */ | |
4250 | qdict_put_str(cryptoopts_qdict, "format", "luks"); | |
4251 | cryptoopts = block_crypto_create_opts_init(cryptoopts_qdict, errp); | |
4252 | qobject_unref(cryptoopts_qdict); | |
4253 | if (!cryptoopts) { | |
4254 | return false; | |
4255 | } | |
4256 | ||
4257 | /* Fake LUKS creation in order to determine the payload size */ | |
4258 | crypto = qcrypto_block_create(cryptoopts, "encrypt.", | |
4259 | qcow2_measure_crypto_hdr_init_func, | |
4260 | qcow2_measure_crypto_hdr_write_func, | |
4261 | len, errp); | |
4262 | qapi_free_QCryptoBlockCreateOptions(cryptoopts); | |
4263 | if (!crypto) { | |
4264 | return false; | |
4265 | } | |
4266 | ||
4267 | qcrypto_block_free(crypto); | |
4268 | return true; | |
4269 | } | |
4270 | ||
c501c352 SH |
4271 | static BlockMeasureInfo *qcow2_measure(QemuOpts *opts, BlockDriverState *in_bs, |
4272 | Error **errp) | |
4273 | { | |
4274 | Error *local_err = NULL; | |
4275 | BlockMeasureInfo *info; | |
4276 | uint64_t required = 0; /* bytes that contribute to required size */ | |
4277 | uint64_t virtual_size; /* disk size as seen by guest */ | |
4278 | uint64_t refcount_bits; | |
4279 | uint64_t l2_tables; | |
61914f89 | 4280 | uint64_t luks_payload_size = 0; |
c501c352 SH |
4281 | size_t cluster_size; |
4282 | int version; | |
4283 | char *optstr; | |
4284 | PreallocMode prealloc; | |
4285 | bool has_backing_file; | |
61914f89 | 4286 | bool has_luks; |
c501c352 SH |
4287 | |
4288 | /* Parse image creation options */ | |
4289 | cluster_size = qcow2_opt_get_cluster_size_del(opts, &local_err); | |
4290 | if (local_err) { | |
4291 | goto err; | |
4292 | } | |
4293 | ||
4294 | version = qcow2_opt_get_version_del(opts, &local_err); | |
4295 | if (local_err) { | |
4296 | goto err; | |
4297 | } | |
4298 | ||
4299 | refcount_bits = qcow2_opt_get_refcount_bits_del(opts, version, &local_err); | |
4300 | if (local_err) { | |
4301 | goto err; | |
4302 | } | |
4303 | ||
4304 | optstr = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
f7abe0ec | 4305 | prealloc = qapi_enum_parse(&PreallocMode_lookup, optstr, |
06c60b6c | 4306 | PREALLOC_MODE_OFF, &local_err); |
c501c352 SH |
4307 | g_free(optstr); |
4308 | if (local_err) { | |
4309 | goto err; | |
4310 | } | |
4311 | ||
4312 | optstr = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE); | |
4313 | has_backing_file = !!optstr; | |
4314 | g_free(optstr); | |
4315 | ||
61914f89 SH |
4316 | optstr = qemu_opt_get_del(opts, BLOCK_OPT_ENCRYPT_FORMAT); |
4317 | has_luks = optstr && strcmp(optstr, "luks") == 0; | |
4318 | g_free(optstr); | |
4319 | ||
4320 | if (has_luks) { | |
4321 | size_t headerlen; | |
4322 | ||
4323 | if (!qcow2_measure_luks_headerlen(opts, &headerlen, &local_err)) { | |
4324 | goto err; | |
4325 | } | |
4326 | ||
4327 | luks_payload_size = ROUND_UP(headerlen, cluster_size); | |
4328 | } | |
4329 | ||
9e029689 AG |
4330 | virtual_size = qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0); |
4331 | virtual_size = ROUND_UP(virtual_size, cluster_size); | |
c501c352 SH |
4332 | |
4333 | /* Check that virtual disk size is valid */ | |
4334 | l2_tables = DIV_ROUND_UP(virtual_size / cluster_size, | |
4335 | cluster_size / sizeof(uint64_t)); | |
4336 | if (l2_tables * sizeof(uint64_t) > QCOW_MAX_L1_SIZE) { | |
4337 | error_setg(&local_err, "The image size is too large " | |
4338 | "(try using a larger cluster size)"); | |
4339 | goto err; | |
4340 | } | |
4341 | ||
4342 | /* Account for input image */ | |
4343 | if (in_bs) { | |
4344 | int64_t ssize = bdrv_getlength(in_bs); | |
4345 | if (ssize < 0) { | |
4346 | error_setg_errno(&local_err, -ssize, | |
4347 | "Unable to get image virtual_size"); | |
4348 | goto err; | |
4349 | } | |
4350 | ||
9e029689 | 4351 | virtual_size = ROUND_UP(ssize, cluster_size); |
c501c352 SH |
4352 | |
4353 | if (has_backing_file) { | |
4354 | /* We don't how much of the backing chain is shared by the input | |
4355 | * image and the new image file. In the worst case the new image's | |
4356 | * backing file has nothing in common with the input image. Be | |
4357 | * conservative and assume all clusters need to be written. | |
4358 | */ | |
4359 | required = virtual_size; | |
4360 | } else { | |
b85ee453 | 4361 | int64_t offset; |
31826642 | 4362 | int64_t pnum = 0; |
c501c352 | 4363 | |
31826642 EB |
4364 | for (offset = 0; offset < ssize; offset += pnum) { |
4365 | int ret; | |
c501c352 | 4366 | |
31826642 EB |
4367 | ret = bdrv_block_status_above(in_bs, NULL, offset, |
4368 | ssize - offset, &pnum, NULL, | |
4369 | NULL); | |
c501c352 SH |
4370 | if (ret < 0) { |
4371 | error_setg_errno(&local_err, -ret, | |
4372 | "Unable to get block status"); | |
4373 | goto err; | |
4374 | } | |
4375 | ||
4376 | if (ret & BDRV_BLOCK_ZERO) { | |
4377 | /* Skip zero regions (safe with no backing file) */ | |
4378 | } else if ((ret & (BDRV_BLOCK_DATA | BDRV_BLOCK_ALLOCATED)) == | |
4379 | (BDRV_BLOCK_DATA | BDRV_BLOCK_ALLOCATED)) { | |
4380 | /* Extend pnum to end of cluster for next iteration */ | |
31826642 | 4381 | pnum = ROUND_UP(offset + pnum, cluster_size) - offset; |
c501c352 SH |
4382 | |
4383 | /* Count clusters we've seen */ | |
31826642 | 4384 | required += offset % cluster_size + pnum; |
c501c352 SH |
4385 | } |
4386 | } | |
4387 | } | |
4388 | } | |
4389 | ||
4390 | /* Take into account preallocation. Nothing special is needed for | |
4391 | * PREALLOC_MODE_METADATA since metadata is always counted. | |
4392 | */ | |
4393 | if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) { | |
4394 | required = virtual_size; | |
4395 | } | |
4396 | ||
4397 | info = g_new(BlockMeasureInfo, 1); | |
4398 | info->fully_allocated = | |
4399 | qcow2_calc_prealloc_size(virtual_size, cluster_size, | |
61914f89 | 4400 | ctz32(refcount_bits)) + luks_payload_size; |
c501c352 SH |
4401 | |
4402 | /* Remove data clusters that are not required. This overestimates the | |
4403 | * required size because metadata needed for the fully allocated file is | |
4404 | * still counted. | |
4405 | */ | |
4406 | info->required = info->fully_allocated - virtual_size + required; | |
4407 | return info; | |
4408 | ||
4409 | err: | |
4410 | error_propagate(errp, local_err); | |
4411 | return NULL; | |
4412 | } | |
4413 | ||
7c80ab3f | 4414 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
20d97356 | 4415 | { |
ff99129a | 4416 | BDRVQcow2State *s = bs->opaque; |
95de6d70 | 4417 | bdi->unallocated_blocks_are_zero = true; |
20d97356 | 4418 | bdi->cluster_size = s->cluster_size; |
7c80ab3f | 4419 | bdi->vm_state_offset = qcow2_vm_state_offset(s); |
20d97356 BS |
4420 | return 0; |
4421 | } | |
4422 | ||
1bf6e9ca AS |
4423 | static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs, |
4424 | Error **errp) | |
37764dfb | 4425 | { |
ff99129a | 4426 | BDRVQcow2State *s = bs->opaque; |
0a12f6f8 DB |
4427 | ImageInfoSpecific *spec_info; |
4428 | QCryptoBlockInfo *encrypt_info = NULL; | |
1bf6e9ca | 4429 | Error *local_err = NULL; |
37764dfb | 4430 | |
0a12f6f8 | 4431 | if (s->crypto != NULL) { |
1bf6e9ca AS |
4432 | encrypt_info = qcrypto_block_get_info(s->crypto, &local_err); |
4433 | if (local_err) { | |
4434 | error_propagate(errp, local_err); | |
4435 | return NULL; | |
4436 | } | |
0a12f6f8 DB |
4437 | } |
4438 | ||
4439 | spec_info = g_new(ImageInfoSpecific, 1); | |
37764dfb | 4440 | *spec_info = (ImageInfoSpecific){ |
6a8f9661 | 4441 | .type = IMAGE_INFO_SPECIFIC_KIND_QCOW2, |
b8968c87 | 4442 | .u.qcow2.data = g_new0(ImageInfoSpecificQCow2, 1), |
37764dfb HR |
4443 | }; |
4444 | if (s->qcow_version == 2) { | |
32bafa8f | 4445 | *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){ |
0709c5a1 HR |
4446 | .compat = g_strdup("0.10"), |
4447 | .refcount_bits = s->refcount_bits, | |
37764dfb HR |
4448 | }; |
4449 | } else if (s->qcow_version == 3) { | |
b8968c87 AS |
4450 | Qcow2BitmapInfoList *bitmaps; |
4451 | bitmaps = qcow2_get_bitmap_info_list(bs, &local_err); | |
4452 | if (local_err) { | |
4453 | error_propagate(errp, local_err); | |
4454 | qapi_free_ImageInfoSpecific(spec_info); | |
4455 | return NULL; | |
4456 | } | |
32bafa8f | 4457 | *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){ |
37764dfb HR |
4458 | .compat = g_strdup("1.1"), |
4459 | .lazy_refcounts = s->compatible_features & | |
4460 | QCOW2_COMPAT_LAZY_REFCOUNTS, | |
4461 | .has_lazy_refcounts = true, | |
9009b196 HR |
4462 | .corrupt = s->incompatible_features & |
4463 | QCOW2_INCOMPAT_CORRUPT, | |
4464 | .has_corrupt = true, | |
0709c5a1 | 4465 | .refcount_bits = s->refcount_bits, |
b8968c87 AS |
4466 | .has_bitmaps = !!bitmaps, |
4467 | .bitmaps = bitmaps, | |
37764dfb | 4468 | }; |
b1fc8f93 DL |
4469 | } else { |
4470 | /* if this assertion fails, this probably means a new version was | |
4471 | * added without having it covered here */ | |
4472 | assert(false); | |
37764dfb HR |
4473 | } |
4474 | ||
0a12f6f8 DB |
4475 | if (encrypt_info) { |
4476 | ImageInfoSpecificQCow2Encryption *qencrypt = | |
4477 | g_new(ImageInfoSpecificQCow2Encryption, 1); | |
4478 | switch (encrypt_info->format) { | |
4479 | case Q_CRYPTO_BLOCK_FORMAT_QCOW: | |
4480 | qencrypt->format = BLOCKDEV_QCOW2_ENCRYPTION_FORMAT_AES; | |
0a12f6f8 DB |
4481 | break; |
4482 | case Q_CRYPTO_BLOCK_FORMAT_LUKS: | |
4483 | qencrypt->format = BLOCKDEV_QCOW2_ENCRYPTION_FORMAT_LUKS; | |
4484 | qencrypt->u.luks = encrypt_info->u.luks; | |
4485 | break; | |
4486 | default: | |
4487 | abort(); | |
4488 | } | |
4489 | /* Since we did shallow copy above, erase any pointers | |
4490 | * in the original info */ | |
4491 | memset(&encrypt_info->u, 0, sizeof(encrypt_info->u)); | |
4492 | qapi_free_QCryptoBlockInfo(encrypt_info); | |
4493 | ||
4494 | spec_info->u.qcow2.data->has_encrypt = true; | |
4495 | spec_info->u.qcow2.data->encrypt = qencrypt; | |
4496 | } | |
4497 | ||
37764dfb HR |
4498 | return spec_info; |
4499 | } | |
4500 | ||
cf8074b3 KW |
4501 | static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
4502 | int64_t pos) | |
20d97356 | 4503 | { |
ff99129a | 4504 | BDRVQcow2State *s = bs->opaque; |
20d97356 | 4505 | |
66f82cee | 4506 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
734a7758 KW |
4507 | return bs->drv->bdrv_co_pwritev(bs, qcow2_vm_state_offset(s) + pos, |
4508 | qiov->size, qiov, 0); | |
20d97356 BS |
4509 | } |
4510 | ||
5ddda0b8 KW |
4511 | static int qcow2_load_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
4512 | int64_t pos) | |
20d97356 | 4513 | { |
ff99129a | 4514 | BDRVQcow2State *s = bs->opaque; |
20d97356 | 4515 | |
66f82cee | 4516 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
734a7758 KW |
4517 | return bs->drv->bdrv_co_preadv(bs, qcow2_vm_state_offset(s) + pos, |
4518 | qiov->size, qiov, 0); | |
20d97356 BS |
4519 | } |
4520 | ||
9296b3ed HR |
4521 | /* |
4522 | * Downgrades an image's version. To achieve this, any incompatible features | |
4523 | * have to be removed. | |
4524 | */ | |
4057a2b2 | 4525 | static int qcow2_downgrade(BlockDriverState *bs, int target_version, |
d1402b50 HR |
4526 | BlockDriverAmendStatusCB *status_cb, void *cb_opaque, |
4527 | Error **errp) | |
9296b3ed | 4528 | { |
ff99129a | 4529 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
4530 | int current_version = s->qcow_version; |
4531 | int ret; | |
4532 | ||
d1402b50 HR |
4533 | /* This is qcow2_downgrade(), not qcow2_upgrade() */ |
4534 | assert(target_version < current_version); | |
4535 | ||
4536 | /* There are no other versions (now) that you can downgrade to */ | |
4537 | assert(target_version == 2); | |
9296b3ed HR |
4538 | |
4539 | if (s->refcount_order != 4) { | |
d1402b50 | 4540 | error_setg(errp, "compat=0.10 requires refcount_bits=16"); |
9296b3ed HR |
4541 | return -ENOTSUP; |
4542 | } | |
4543 | ||
4544 | /* clear incompatible features */ | |
4545 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
4546 | ret = qcow2_mark_clean(bs); | |
4547 | if (ret < 0) { | |
d1402b50 | 4548 | error_setg_errno(errp, -ret, "Failed to make the image clean"); |
9296b3ed HR |
4549 | return ret; |
4550 | } | |
4551 | } | |
4552 | ||
4553 | /* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in | |
4554 | * the first place; if that happens nonetheless, returning -ENOTSUP is the | |
4555 | * best thing to do anyway */ | |
4556 | ||
4557 | if (s->incompatible_features) { | |
d1402b50 HR |
4558 | error_setg(errp, "Cannot downgrade an image with incompatible features " |
4559 | "%#" PRIx64 " set", s->incompatible_features); | |
9296b3ed HR |
4560 | return -ENOTSUP; |
4561 | } | |
4562 | ||
4563 | /* since we can ignore compatible features, we can set them to 0 as well */ | |
4564 | s->compatible_features = 0; | |
4565 | /* if lazy refcounts have been used, they have already been fixed through | |
4566 | * clearing the dirty flag */ | |
4567 | ||
4568 | /* clearing autoclear features is trivial */ | |
4569 | s->autoclear_features = 0; | |
4570 | ||
8b13976d | 4571 | ret = qcow2_expand_zero_clusters(bs, status_cb, cb_opaque); |
9296b3ed | 4572 | if (ret < 0) { |
d1402b50 | 4573 | error_setg_errno(errp, -ret, "Failed to turn zero into data clusters"); |
9296b3ed HR |
4574 | return ret; |
4575 | } | |
4576 | ||
4577 | s->qcow_version = target_version; | |
4578 | ret = qcow2_update_header(bs); | |
4579 | if (ret < 0) { | |
4580 | s->qcow_version = current_version; | |
d1402b50 | 4581 | error_setg_errno(errp, -ret, "Failed to update the image header"); |
9296b3ed HR |
4582 | return ret; |
4583 | } | |
4584 | return 0; | |
4585 | } | |
4586 | ||
c293a809 HR |
4587 | typedef enum Qcow2AmendOperation { |
4588 | /* This is the value Qcow2AmendHelperCBInfo::last_operation will be | |
4589 | * statically initialized to so that the helper CB can discern the first | |
4590 | * invocation from an operation change */ | |
4591 | QCOW2_NO_OPERATION = 0, | |
4592 | ||
61ce55fc | 4593 | QCOW2_CHANGING_REFCOUNT_ORDER, |
c293a809 HR |
4594 | QCOW2_DOWNGRADING, |
4595 | } Qcow2AmendOperation; | |
4596 | ||
4597 | typedef struct Qcow2AmendHelperCBInfo { | |
4598 | /* The code coordinating the amend operations should only modify | |
4599 | * these four fields; the rest will be managed by the CB */ | |
4600 | BlockDriverAmendStatusCB *original_status_cb; | |
4601 | void *original_cb_opaque; | |
4602 | ||
4603 | Qcow2AmendOperation current_operation; | |
4604 | ||
4605 | /* Total number of operations to perform (only set once) */ | |
4606 | int total_operations; | |
4607 | ||
4608 | /* The following fields are managed by the CB */ | |
4609 | ||
4610 | /* Number of operations completed */ | |
4611 | int operations_completed; | |
4612 | ||
4613 | /* Cumulative offset of all completed operations */ | |
4614 | int64_t offset_completed; | |
4615 | ||
4616 | Qcow2AmendOperation last_operation; | |
4617 | int64_t last_work_size; | |
4618 | } Qcow2AmendHelperCBInfo; | |
4619 | ||
4620 | static void qcow2_amend_helper_cb(BlockDriverState *bs, | |
4621 | int64_t operation_offset, | |
4622 | int64_t operation_work_size, void *opaque) | |
4623 | { | |
4624 | Qcow2AmendHelperCBInfo *info = opaque; | |
4625 | int64_t current_work_size; | |
4626 | int64_t projected_work_size; | |
4627 | ||
4628 | if (info->current_operation != info->last_operation) { | |
4629 | if (info->last_operation != QCOW2_NO_OPERATION) { | |
4630 | info->offset_completed += info->last_work_size; | |
4631 | info->operations_completed++; | |
4632 | } | |
4633 | ||
4634 | info->last_operation = info->current_operation; | |
4635 | } | |
4636 | ||
4637 | assert(info->total_operations > 0); | |
4638 | assert(info->operations_completed < info->total_operations); | |
4639 | ||
4640 | info->last_work_size = operation_work_size; | |
4641 | ||
4642 | current_work_size = info->offset_completed + operation_work_size; | |
4643 | ||
4644 | /* current_work_size is the total work size for (operations_completed + 1) | |
4645 | * operations (which includes this one), so multiply it by the number of | |
4646 | * operations not covered and divide it by the number of operations | |
4647 | * covered to get a projection for the operations not covered */ | |
4648 | projected_work_size = current_work_size * (info->total_operations - | |
4649 | info->operations_completed - 1) | |
4650 | / (info->operations_completed + 1); | |
4651 | ||
4652 | info->original_status_cb(bs, info->offset_completed + operation_offset, | |
4653 | current_work_size + projected_work_size, | |
4654 | info->original_cb_opaque); | |
4655 | } | |
4656 | ||
77485434 | 4657 | static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts, |
8b13976d | 4658 | BlockDriverAmendStatusCB *status_cb, |
d1402b50 HR |
4659 | void *cb_opaque, |
4660 | Error **errp) | |
9296b3ed | 4661 | { |
ff99129a | 4662 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
4663 | int old_version = s->qcow_version, new_version = old_version; |
4664 | uint64_t new_size = 0; | |
4665 | const char *backing_file = NULL, *backing_format = NULL; | |
4666 | bool lazy_refcounts = s->use_lazy_refcounts; | |
1bd0e2d1 CL |
4667 | const char *compat = NULL; |
4668 | uint64_t cluster_size = s->cluster_size; | |
4669 | bool encrypt; | |
4652b8f3 | 4670 | int encformat; |
61ce55fc | 4671 | int refcount_bits = s->refcount_bits; |
9296b3ed | 4672 | int ret; |
1bd0e2d1 | 4673 | QemuOptDesc *desc = opts->list->desc; |
c293a809 | 4674 | Qcow2AmendHelperCBInfo helper_cb_info; |
9296b3ed | 4675 | |
1bd0e2d1 CL |
4676 | while (desc && desc->name) { |
4677 | if (!qemu_opt_find(opts, desc->name)) { | |
9296b3ed | 4678 | /* only change explicitly defined options */ |
1bd0e2d1 | 4679 | desc++; |
9296b3ed HR |
4680 | continue; |
4681 | } | |
4682 | ||
8a17b83c HR |
4683 | if (!strcmp(desc->name, BLOCK_OPT_COMPAT_LEVEL)) { |
4684 | compat = qemu_opt_get(opts, BLOCK_OPT_COMPAT_LEVEL); | |
1bd0e2d1 | 4685 | if (!compat) { |
9296b3ed | 4686 | /* preserve default */ |
1bd0e2d1 | 4687 | } else if (!strcmp(compat, "0.10")) { |
9296b3ed | 4688 | new_version = 2; |
1bd0e2d1 | 4689 | } else if (!strcmp(compat, "1.1")) { |
9296b3ed HR |
4690 | new_version = 3; |
4691 | } else { | |
d1402b50 | 4692 | error_setg(errp, "Unknown compatibility level %s", compat); |
9296b3ed HR |
4693 | return -EINVAL; |
4694 | } | |
8a17b83c | 4695 | } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) { |
d1402b50 | 4696 | error_setg(errp, "Cannot change preallocation mode"); |
9296b3ed | 4697 | return -ENOTSUP; |
8a17b83c HR |
4698 | } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) { |
4699 | new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); | |
4700 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FILE)) { | |
4701 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); | |
4702 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FMT)) { | |
4703 | backing_format = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); | |
4704 | } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT)) { | |
4705 | encrypt = qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, | |
b25b387f | 4706 | !!s->crypto); |
f6fa64f6 | 4707 | |
b25b387f | 4708 | if (encrypt != !!s->crypto) { |
d1402b50 HR |
4709 | error_setg(errp, |
4710 | "Changing the encryption flag is not supported"); | |
9296b3ed HR |
4711 | return -ENOTSUP; |
4712 | } | |
4652b8f3 DB |
4713 | } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT_FORMAT)) { |
4714 | encformat = qcow2_crypt_method_from_format( | |
4715 | qemu_opt_get(opts, BLOCK_OPT_ENCRYPT_FORMAT)); | |
4716 | ||
4717 | if (encformat != s->crypt_method_header) { | |
d1402b50 HR |
4718 | error_setg(errp, |
4719 | "Changing the encryption format is not supported"); | |
4652b8f3 DB |
4720 | return -ENOTSUP; |
4721 | } | |
f66afbe2 | 4722 | } else if (g_str_has_prefix(desc->name, "encrypt.")) { |
d1402b50 HR |
4723 | error_setg(errp, |
4724 | "Changing the encryption parameters is not supported"); | |
f66afbe2 | 4725 | return -ENOTSUP; |
8a17b83c HR |
4726 | } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) { |
4727 | cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, | |
1bd0e2d1 CL |
4728 | cluster_size); |
4729 | if (cluster_size != s->cluster_size) { | |
d1402b50 | 4730 | error_setg(errp, "Changing the cluster size is not supported"); |
9296b3ed HR |
4731 | return -ENOTSUP; |
4732 | } | |
8a17b83c HR |
4733 | } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) { |
4734 | lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS, | |
1bd0e2d1 | 4735 | lazy_refcounts); |
06d05fa7 | 4736 | } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) { |
61ce55fc HR |
4737 | refcount_bits = qemu_opt_get_number(opts, BLOCK_OPT_REFCOUNT_BITS, |
4738 | refcount_bits); | |
4739 | ||
4740 | if (refcount_bits <= 0 || refcount_bits > 64 || | |
4741 | !is_power_of_2(refcount_bits)) | |
4742 | { | |
d1402b50 HR |
4743 | error_setg(errp, "Refcount width must be a power of two and " |
4744 | "may not exceed 64 bits"); | |
61ce55fc HR |
4745 | return -EINVAL; |
4746 | } | |
9296b3ed | 4747 | } else { |
164e0f89 | 4748 | /* if this point is reached, this probably means a new option was |
9296b3ed | 4749 | * added without having it covered here */ |
164e0f89 | 4750 | abort(); |
9296b3ed | 4751 | } |
1bd0e2d1 CL |
4752 | |
4753 | desc++; | |
9296b3ed HR |
4754 | } |
4755 | ||
c293a809 HR |
4756 | helper_cb_info = (Qcow2AmendHelperCBInfo){ |
4757 | .original_status_cb = status_cb, | |
4758 | .original_cb_opaque = cb_opaque, | |
4759 | .total_operations = (new_version < old_version) | |
61ce55fc | 4760 | + (s->refcount_bits != refcount_bits) |
c293a809 HR |
4761 | }; |
4762 | ||
1038bbb8 HR |
4763 | /* Upgrade first (some features may require compat=1.1) */ |
4764 | if (new_version > old_version) { | |
4765 | s->qcow_version = new_version; | |
4766 | ret = qcow2_update_header(bs); | |
4767 | if (ret < 0) { | |
4768 | s->qcow_version = old_version; | |
d1402b50 | 4769 | error_setg_errno(errp, -ret, "Failed to update the image header"); |
1038bbb8 | 4770 | return ret; |
9296b3ed HR |
4771 | } |
4772 | } | |
4773 | ||
61ce55fc HR |
4774 | if (s->refcount_bits != refcount_bits) { |
4775 | int refcount_order = ctz32(refcount_bits); | |
61ce55fc HR |
4776 | |
4777 | if (new_version < 3 && refcount_bits != 16) { | |
d1402b50 HR |
4778 | error_setg(errp, "Refcount widths other than 16 bits require " |
4779 | "compatibility level 1.1 or above (use compat=1.1 or " | |
4780 | "greater)"); | |
61ce55fc HR |
4781 | return -EINVAL; |
4782 | } | |
4783 | ||
4784 | helper_cb_info.current_operation = QCOW2_CHANGING_REFCOUNT_ORDER; | |
4785 | ret = qcow2_change_refcount_order(bs, refcount_order, | |
4786 | &qcow2_amend_helper_cb, | |
d1402b50 | 4787 | &helper_cb_info, errp); |
61ce55fc | 4788 | if (ret < 0) { |
61ce55fc HR |
4789 | return ret; |
4790 | } | |
4791 | } | |
4792 | ||
9296b3ed | 4793 | if (backing_file || backing_format) { |
e4603fe1 KW |
4794 | ret = qcow2_change_backing_file(bs, |
4795 | backing_file ?: s->image_backing_file, | |
4796 | backing_format ?: s->image_backing_format); | |
9296b3ed | 4797 | if (ret < 0) { |
d1402b50 | 4798 | error_setg_errno(errp, -ret, "Failed to change the backing file"); |
9296b3ed HR |
4799 | return ret; |
4800 | } | |
4801 | } | |
4802 | ||
4803 | if (s->use_lazy_refcounts != lazy_refcounts) { | |
4804 | if (lazy_refcounts) { | |
1038bbb8 | 4805 | if (new_version < 3) { |
d1402b50 HR |
4806 | error_setg(errp, "Lazy refcounts only supported with " |
4807 | "compatibility level 1.1 and above (use compat=1.1 " | |
4808 | "or greater)"); | |
9296b3ed HR |
4809 | return -EINVAL; |
4810 | } | |
4811 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
4812 | ret = qcow2_update_header(bs); | |
4813 | if (ret < 0) { | |
4814 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
d1402b50 | 4815 | error_setg_errno(errp, -ret, "Failed to update the image header"); |
9296b3ed HR |
4816 | return ret; |
4817 | } | |
4818 | s->use_lazy_refcounts = true; | |
4819 | } else { | |
4820 | /* make image clean first */ | |
4821 | ret = qcow2_mark_clean(bs); | |
4822 | if (ret < 0) { | |
d1402b50 | 4823 | error_setg_errno(errp, -ret, "Failed to make the image clean"); |
9296b3ed HR |
4824 | return ret; |
4825 | } | |
4826 | /* now disallow lazy refcounts */ | |
4827 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
4828 | ret = qcow2_update_header(bs); | |
4829 | if (ret < 0) { | |
4830 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
d1402b50 | 4831 | error_setg_errno(errp, -ret, "Failed to update the image header"); |
9296b3ed HR |
4832 | return ret; |
4833 | } | |
4834 | s->use_lazy_refcounts = false; | |
4835 | } | |
4836 | } | |
4837 | ||
4838 | if (new_size) { | |
6d0eb64d | 4839 | BlockBackend *blk = blk_new(BLK_PERM_RESIZE, BLK_PERM_ALL); |
d1402b50 | 4840 | ret = blk_insert_bs(blk, bs, errp); |
d7086422 | 4841 | if (ret < 0) { |
d7086422 KW |
4842 | blk_unref(blk); |
4843 | return ret; | |
4844 | } | |
4845 | ||
d1402b50 | 4846 | ret = blk_truncate(blk, new_size, PREALLOC_MODE_OFF, errp); |
70b27f36 | 4847 | blk_unref(blk); |
9296b3ed HR |
4848 | if (ret < 0) { |
4849 | return ret; | |
4850 | } | |
4851 | } | |
4852 | ||
1038bbb8 HR |
4853 | /* Downgrade last (so unsupported features can be removed before) */ |
4854 | if (new_version < old_version) { | |
c293a809 HR |
4855 | helper_cb_info.current_operation = QCOW2_DOWNGRADING; |
4856 | ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb, | |
d1402b50 | 4857 | &helper_cb_info, errp); |
1038bbb8 HR |
4858 | if (ret < 0) { |
4859 | return ret; | |
4860 | } | |
4861 | } | |
4862 | ||
9296b3ed HR |
4863 | return 0; |
4864 | } | |
4865 | ||
85186ebd HR |
4866 | /* |
4867 | * If offset or size are negative, respectively, they will not be included in | |
4868 | * the BLOCK_IMAGE_CORRUPTED event emitted. | |
4869 | * fatal will be ignored for read-only BDS; corruptions found there will always | |
4870 | * be considered non-fatal. | |
4871 | */ | |
4872 | void qcow2_signal_corruption(BlockDriverState *bs, bool fatal, int64_t offset, | |
4873 | int64_t size, const char *message_format, ...) | |
4874 | { | |
ff99129a | 4875 | BDRVQcow2State *s = bs->opaque; |
dc881b44 | 4876 | const char *node_name; |
85186ebd HR |
4877 | char *message; |
4878 | va_list ap; | |
4879 | ||
ddf3b47e | 4880 | fatal = fatal && bdrv_is_writable(bs); |
85186ebd HR |
4881 | |
4882 | if (s->signaled_corruption && | |
4883 | (!fatal || (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT))) | |
4884 | { | |
4885 | return; | |
4886 | } | |
4887 | ||
4888 | va_start(ap, message_format); | |
4889 | message = g_strdup_vprintf(message_format, ap); | |
4890 | va_end(ap); | |
4891 | ||
4892 | if (fatal) { | |
4893 | fprintf(stderr, "qcow2: Marking image as corrupt: %s; further " | |
4894 | "corruption events will be suppressed\n", message); | |
4895 | } else { | |
4896 | fprintf(stderr, "qcow2: Image is corrupt: %s; further non-fatal " | |
4897 | "corruption events will be suppressed\n", message); | |
4898 | } | |
4899 | ||
dc881b44 AG |
4900 | node_name = bdrv_get_node_name(bs); |
4901 | qapi_event_send_block_image_corrupted(bdrv_get_device_name(bs), | |
4902 | *node_name != '\0', node_name, | |
4903 | message, offset >= 0, offset, | |
4904 | size >= 0, size, | |
3ab72385 | 4905 | fatal); |
85186ebd HR |
4906 | g_free(message); |
4907 | ||
4908 | if (fatal) { | |
4909 | qcow2_mark_corrupt(bs); | |
4910 | bs->drv = NULL; /* make BDS unusable */ | |
4911 | } | |
4912 | ||
4913 | s->signaled_corruption = true; | |
4914 | } | |
4915 | ||
1bd0e2d1 CL |
4916 | static QemuOptsList qcow2_create_opts = { |
4917 | .name = "qcow2-create-opts", | |
4918 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_create_opts.head), | |
4919 | .desc = { | |
4920 | { | |
4921 | .name = BLOCK_OPT_SIZE, | |
4922 | .type = QEMU_OPT_SIZE, | |
4923 | .help = "Virtual disk size" | |
4924 | }, | |
4925 | { | |
4926 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
4927 | .type = QEMU_OPT_STRING, | |
4928 | .help = "Compatibility level (0.10 or 1.1)" | |
4929 | }, | |
4930 | { | |
4931 | .name = BLOCK_OPT_BACKING_FILE, | |
4932 | .type = QEMU_OPT_STRING, | |
4933 | .help = "File name of a base image" | |
4934 | }, | |
4935 | { | |
4936 | .name = BLOCK_OPT_BACKING_FMT, | |
4937 | .type = QEMU_OPT_STRING, | |
4938 | .help = "Image format of the base image" | |
4939 | }, | |
4940 | { | |
4941 | .name = BLOCK_OPT_ENCRYPT, | |
4942 | .type = QEMU_OPT_BOOL, | |
0cb8d47b DB |
4943 | .help = "Encrypt the image with format 'aes'. (Deprecated " |
4944 | "in favor of " BLOCK_OPT_ENCRYPT_FORMAT "=aes)", | |
4945 | }, | |
4946 | { | |
4947 | .name = BLOCK_OPT_ENCRYPT_FORMAT, | |
4948 | .type = QEMU_OPT_STRING, | |
4652b8f3 | 4949 | .help = "Encrypt the image, format choices: 'aes', 'luks'", |
1bd0e2d1 | 4950 | }, |
4652b8f3 DB |
4951 | BLOCK_CRYPTO_OPT_DEF_KEY_SECRET("encrypt.", |
4952 | "ID of secret providing qcow AES key or LUKS passphrase"), | |
4953 | BLOCK_CRYPTO_OPT_DEF_LUKS_CIPHER_ALG("encrypt."), | |
4954 | BLOCK_CRYPTO_OPT_DEF_LUKS_CIPHER_MODE("encrypt."), | |
4955 | BLOCK_CRYPTO_OPT_DEF_LUKS_IVGEN_ALG("encrypt."), | |
4956 | BLOCK_CRYPTO_OPT_DEF_LUKS_IVGEN_HASH_ALG("encrypt."), | |
4957 | BLOCK_CRYPTO_OPT_DEF_LUKS_HASH_ALG("encrypt."), | |
4958 | BLOCK_CRYPTO_OPT_DEF_LUKS_ITER_TIME("encrypt."), | |
1bd0e2d1 CL |
4959 | { |
4960 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
4961 | .type = QEMU_OPT_SIZE, | |
4962 | .help = "qcow2 cluster size", | |
4963 | .def_value_str = stringify(DEFAULT_CLUSTER_SIZE) | |
4964 | }, | |
4965 | { | |
4966 | .name = BLOCK_OPT_PREALLOC, | |
4967 | .type = QEMU_OPT_STRING, | |
0e4271b7 HT |
4968 | .help = "Preallocation mode (allowed values: off, metadata, " |
4969 | "falloc, full)" | |
1bd0e2d1 CL |
4970 | }, |
4971 | { | |
4972 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
4973 | .type = QEMU_OPT_BOOL, | |
4974 | .help = "Postpone refcount updates", | |
4975 | .def_value_str = "off" | |
4976 | }, | |
06d05fa7 HR |
4977 | { |
4978 | .name = BLOCK_OPT_REFCOUNT_BITS, | |
4979 | .type = QEMU_OPT_NUMBER, | |
4980 | .help = "Width of a reference count entry in bits", | |
4981 | .def_value_str = "16" | |
4982 | }, | |
1bd0e2d1 CL |
4983 | { /* end of list */ } |
4984 | } | |
20d97356 BS |
4985 | }; |
4986 | ||
2654267c HR |
4987 | static const char *const qcow2_strong_runtime_opts[] = { |
4988 | "encrypt." BLOCK_CRYPTO_OPT_QCOW_KEY_SECRET, | |
4989 | ||
4990 | NULL | |
4991 | }; | |
4992 | ||
5f535a94 | 4993 | BlockDriver bdrv_qcow2 = { |
7c80ab3f | 4994 | .format_name = "qcow2", |
ff99129a | 4995 | .instance_size = sizeof(BDRVQcow2State), |
7c80ab3f JS |
4996 | .bdrv_probe = qcow2_probe, |
4997 | .bdrv_open = qcow2_open, | |
4998 | .bdrv_close = qcow2_close, | |
21d82ac9 | 4999 | .bdrv_reopen_prepare = qcow2_reopen_prepare, |
5b0959a7 KW |
5000 | .bdrv_reopen_commit = qcow2_reopen_commit, |
5001 | .bdrv_reopen_abort = qcow2_reopen_abort, | |
5365f44d | 5002 | .bdrv_join_options = qcow2_join_options, |
862f215f | 5003 | .bdrv_child_perm = bdrv_format_default_perms, |
efc75e2a | 5004 | .bdrv_co_create_opts = qcow2_co_create_opts, |
b0292b85 | 5005 | .bdrv_co_create = qcow2_co_create, |
3ac21627 | 5006 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
a320fb04 | 5007 | .bdrv_co_block_status = qcow2_co_block_status, |
7c80ab3f | 5008 | |
ecfe1863 | 5009 | .bdrv_co_preadv = qcow2_co_preadv, |
d46a0bb2 | 5010 | .bdrv_co_pwritev = qcow2_co_pwritev, |
eb489bb1 | 5011 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
419b19d9 | 5012 | |
5544b59f | 5013 | .bdrv_co_pwrite_zeroes = qcow2_co_pwrite_zeroes, |
82e8a788 | 5014 | .bdrv_co_pdiscard = qcow2_co_pdiscard, |
fd9fcd37 FZ |
5015 | .bdrv_co_copy_range_from = qcow2_co_copy_range_from, |
5016 | .bdrv_co_copy_range_to = qcow2_co_copy_range_to, | |
061ca8a3 | 5017 | .bdrv_co_truncate = qcow2_co_truncate, |
fcccefc5 | 5018 | .bdrv_co_pwritev_compressed = qcow2_co_pwritev_compressed, |
491d27e2 | 5019 | .bdrv_make_empty = qcow2_make_empty, |
20d97356 BS |
5020 | |
5021 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
5022 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
5023 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
5024 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1bd0e2d1 | 5025 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
c501c352 | 5026 | .bdrv_measure = qcow2_measure, |
1bd0e2d1 | 5027 | .bdrv_get_info = qcow2_get_info, |
37764dfb | 5028 | .bdrv_get_specific_info = qcow2_get_specific_info, |
20d97356 | 5029 | |
7c80ab3f JS |
5030 | .bdrv_save_vmstate = qcow2_save_vmstate, |
5031 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
20d97356 | 5032 | |
8ee79e70 | 5033 | .supports_backing = true, |
20d97356 BS |
5034 | .bdrv_change_backing_file = qcow2_change_backing_file, |
5035 | ||
d34682cd | 5036 | .bdrv_refresh_limits = qcow2_refresh_limits, |
2b148f39 | 5037 | .bdrv_co_invalidate_cache = qcow2_co_invalidate_cache, |
ec6d8912 | 5038 | .bdrv_inactivate = qcow2_inactivate, |
06d9260f | 5039 | |
1bd0e2d1 | 5040 | .create_opts = &qcow2_create_opts, |
2654267c | 5041 | .strong_runtime_opts = qcow2_strong_runtime_opts, |
2fd61638 | 5042 | .bdrv_co_check = qcow2_co_check, |
c282e1fd | 5043 | .bdrv_amend_options = qcow2_amend_options, |
279621c0 AG |
5044 | |
5045 | .bdrv_detach_aio_context = qcow2_detach_aio_context, | |
5046 | .bdrv_attach_aio_context = qcow2_attach_aio_context, | |
1b6b0562 VSO |
5047 | |
5048 | .bdrv_reopen_bitmaps_rw = qcow2_reopen_bitmaps_rw, | |
da0eb242 | 5049 | .bdrv_can_store_new_dirty_bitmap = qcow2_can_store_new_dirty_bitmap, |
469c71ed | 5050 | .bdrv_remove_persistent_dirty_bitmap = qcow2_remove_persistent_dirty_bitmap, |
20d97356 BS |
5051 | }; |
5052 | ||
5efa9d5a AL |
5053 | static void bdrv_qcow2_init(void) |
5054 | { | |
5055 | bdrv_register(&bdrv_qcow2); | |
5056 | } | |
5057 | ||
5058 | block_init(bdrv_qcow2_init); |