]>
Commit | Line | Data |
---|---|---|
585f8587 FB |
1 | /* |
2 | * Block driver for the QCOW version 2 format | |
5fafdf24 | 3 | * |
585f8587 | 4 | * Copyright (c) 2004-2006 Fabrice Bellard |
5fafdf24 | 5 | * |
585f8587 FB |
6 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
7 | * of this software and associated documentation files (the "Software"), to deal | |
8 | * in the Software without restriction, including without limitation the rights | |
9 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell | |
10 | * copies of the Software, and to permit persons to whom the Software is | |
11 | * furnished to do so, subject to the following conditions: | |
12 | * | |
13 | * The above copyright notice and this permission notice shall be included in | |
14 | * all copies or substantial portions of the Software. | |
15 | * | |
16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL | |
19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, | |
21 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN | |
22 | * THE SOFTWARE. | |
23 | */ | |
80c71a24 | 24 | #include "qemu/osdep.h" |
737e150e | 25 | #include "block/block_int.h" |
23588797 | 26 | #include "sysemu/block-backend.h" |
1de7afc9 | 27 | #include "qemu/module.h" |
585f8587 | 28 | #include <zlib.h> |
f7d0fe02 | 29 | #include "block/qcow2.h" |
1de7afc9 | 30 | #include "qemu/error-report.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qerror.h" |
acdfb480 | 32 | #include "qapi/qmp/qbool.h" |
ffeaac9b | 33 | #include "qapi/util.h" |
85186ebd HR |
34 | #include "qapi/qmp/types.h" |
35 | #include "qapi-event.h" | |
3cce16f4 | 36 | #include "trace.h" |
1bd0e2d1 | 37 | #include "qemu/option_int.h" |
f348b6d1 | 38 | #include "qemu/cutils.h" |
58369e22 | 39 | #include "qemu/bswap.h" |
585f8587 FB |
40 | |
41 | /* | |
42 | Differences with QCOW: | |
43 | ||
44 | - Support for multiple incremental snapshots. | |
45 | - Memory management by reference counts. | |
46 | - Clusters which have a reference count of one have the bit | |
47 | QCOW_OFLAG_COPIED to optimize write performance. | |
5fafdf24 | 48 | - Size of compressed clusters is stored in sectors to reduce bit usage |
585f8587 FB |
49 | in the cluster offsets. |
50 | - Support for storing additional data (such as the VM state) in the | |
3b46e624 | 51 | snapshots. |
585f8587 FB |
52 | - If a backing store is used, the cluster size is not constrained |
53 | (could be backported to QCOW). | |
54 | - L2 tables have always a size of one cluster. | |
55 | */ | |
56 | ||
9b80ddf3 AL |
57 | |
58 | typedef struct { | |
59 | uint32_t magic; | |
60 | uint32_t len; | |
c4217f64 | 61 | } QEMU_PACKED QCowExtension; |
21d82ac9 | 62 | |
7c80ab3f JS |
63 | #define QCOW2_EXT_MAGIC_END 0 |
64 | #define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA | |
cfcc4c62 | 65 | #define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 |
9b80ddf3 | 66 | |
7c80ab3f | 67 | static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) |
585f8587 FB |
68 | { |
69 | const QCowHeader *cow_header = (const void *)buf; | |
3b46e624 | 70 | |
585f8587 FB |
71 | if (buf_size >= sizeof(QCowHeader) && |
72 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
6744cbab | 73 | be32_to_cpu(cow_header->version) >= 2) |
585f8587 FB |
74 | return 100; |
75 | else | |
76 | return 0; | |
77 | } | |
78 | ||
9b80ddf3 AL |
79 | |
80 | /* | |
81 | * read qcow2 extension and fill bs | |
82 | * start reading from start_offset | |
83 | * finish reading upon magic of value 0 or when end_offset reached | |
84 | * unknown magic is skipped (future extension this version knows nothing about) | |
85 | * return 0 upon success, non-0 otherwise | |
86 | */ | |
7c80ab3f | 87 | static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
3ef6c40a HR |
88 | uint64_t end_offset, void **p_feature_table, |
89 | Error **errp) | |
9b80ddf3 | 90 | { |
ff99129a | 91 | BDRVQcow2State *s = bs->opaque; |
9b80ddf3 AL |
92 | QCowExtension ext; |
93 | uint64_t offset; | |
75bab85c | 94 | int ret; |
9b80ddf3 AL |
95 | |
96 | #ifdef DEBUG_EXT | |
7c80ab3f | 97 | printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); |
9b80ddf3 AL |
98 | #endif |
99 | offset = start_offset; | |
100 | while (offset < end_offset) { | |
101 | ||
102 | #ifdef DEBUG_EXT | |
103 | /* Sanity check */ | |
104 | if (offset > s->cluster_size) | |
7c80ab3f | 105 | printf("qcow2_read_extension: suspicious offset %lu\n", offset); |
9b80ddf3 | 106 | |
9b2260cb | 107 | printf("attempting to read extended header in offset %lu\n", offset); |
9b80ddf3 AL |
108 | #endif |
109 | ||
cf2ab8fc | 110 | ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext)); |
3ef6c40a HR |
111 | if (ret < 0) { |
112 | error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: " | |
113 | "pread fail from offset %" PRIu64, offset); | |
9b80ddf3 AL |
114 | return 1; |
115 | } | |
116 | be32_to_cpus(&ext.magic); | |
117 | be32_to_cpus(&ext.len); | |
118 | offset += sizeof(ext); | |
119 | #ifdef DEBUG_EXT | |
120 | printf("ext.magic = 0x%x\n", ext.magic); | |
121 | #endif | |
2ebafc85 | 122 | if (offset > end_offset || ext.len > end_offset - offset) { |
3ef6c40a | 123 | error_setg(errp, "Header extension too large"); |
64ca6aee KW |
124 | return -EINVAL; |
125 | } | |
126 | ||
9b80ddf3 | 127 | switch (ext.magic) { |
7c80ab3f | 128 | case QCOW2_EXT_MAGIC_END: |
9b80ddf3 | 129 | return 0; |
f965509c | 130 | |
7c80ab3f | 131 | case QCOW2_EXT_MAGIC_BACKING_FORMAT: |
f965509c | 132 | if (ext.len >= sizeof(bs->backing_format)) { |
521b2b5d HR |
133 | error_setg(errp, "ERROR: ext_backing_format: len=%" PRIu32 |
134 | " too large (>=%zu)", ext.len, | |
135 | sizeof(bs->backing_format)); | |
f965509c AL |
136 | return 2; |
137 | } | |
cf2ab8fc | 138 | ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len); |
3ef6c40a HR |
139 | if (ret < 0) { |
140 | error_setg_errno(errp, -ret, "ERROR: ext_backing_format: " | |
141 | "Could not read format name"); | |
f965509c | 142 | return 3; |
3ef6c40a | 143 | } |
f965509c | 144 | bs->backing_format[ext.len] = '\0'; |
e4603fe1 | 145 | s->image_backing_format = g_strdup(bs->backing_format); |
f965509c AL |
146 | #ifdef DEBUG_EXT |
147 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
148 | #endif | |
f965509c AL |
149 | break; |
150 | ||
cfcc4c62 KW |
151 | case QCOW2_EXT_MAGIC_FEATURE_TABLE: |
152 | if (p_feature_table != NULL) { | |
153 | void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); | |
cf2ab8fc | 154 | ret = bdrv_pread(bs->file, offset , feature_table, ext.len); |
cfcc4c62 | 155 | if (ret < 0) { |
3ef6c40a HR |
156 | error_setg_errno(errp, -ret, "ERROR: ext_feature_table: " |
157 | "Could not read table"); | |
cfcc4c62 KW |
158 | return ret; |
159 | } | |
160 | ||
161 | *p_feature_table = feature_table; | |
162 | } | |
163 | break; | |
164 | ||
9b80ddf3 | 165 | default: |
75bab85c KW |
166 | /* unknown magic - save it in case we need to rewrite the header */ |
167 | { | |
168 | Qcow2UnknownHeaderExtension *uext; | |
169 | ||
170 | uext = g_malloc0(sizeof(*uext) + ext.len); | |
171 | uext->magic = ext.magic; | |
172 | uext->len = ext.len; | |
173 | QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); | |
174 | ||
cf2ab8fc | 175 | ret = bdrv_pread(bs->file, offset , uext->data, uext->len); |
75bab85c | 176 | if (ret < 0) { |
3ef6c40a HR |
177 | error_setg_errno(errp, -ret, "ERROR: unknown extension: " |
178 | "Could not read data"); | |
75bab85c KW |
179 | return ret; |
180 | } | |
75bab85c | 181 | } |
9b80ddf3 AL |
182 | break; |
183 | } | |
fd29b4bb KW |
184 | |
185 | offset += ((ext.len + 7) & ~7); | |
9b80ddf3 AL |
186 | } |
187 | ||
188 | return 0; | |
189 | } | |
190 | ||
75bab85c KW |
191 | static void cleanup_unknown_header_ext(BlockDriverState *bs) |
192 | { | |
ff99129a | 193 | BDRVQcow2State *s = bs->opaque; |
75bab85c KW |
194 | Qcow2UnknownHeaderExtension *uext, *next; |
195 | ||
196 | QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { | |
197 | QLIST_REMOVE(uext, next); | |
198 | g_free(uext); | |
199 | } | |
200 | } | |
9b80ddf3 | 201 | |
a55448b3 HR |
202 | static void report_unsupported_feature(Error **errp, Qcow2Feature *table, |
203 | uint64_t mask) | |
cfcc4c62 | 204 | { |
12ac6d3d KW |
205 | char *features = g_strdup(""); |
206 | char *old; | |
207 | ||
cfcc4c62 KW |
208 | while (table && table->name[0] != '\0') { |
209 | if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) { | |
12ac6d3d KW |
210 | if (mask & (1ULL << table->bit)) { |
211 | old = features; | |
212 | features = g_strdup_printf("%s%s%.46s", old, *old ? ", " : "", | |
213 | table->name); | |
214 | g_free(old); | |
215 | mask &= ~(1ULL << table->bit); | |
cfcc4c62 KW |
216 | } |
217 | } | |
218 | table++; | |
219 | } | |
220 | ||
221 | if (mask) { | |
12ac6d3d KW |
222 | old = features; |
223 | features = g_strdup_printf("%s%sUnknown incompatible feature: %" PRIx64, | |
224 | old, *old ? ", " : "", mask); | |
225 | g_free(old); | |
cfcc4c62 | 226 | } |
12ac6d3d | 227 | |
a55448b3 | 228 | error_setg(errp, "Unsupported qcow2 feature(s): %s", features); |
12ac6d3d | 229 | g_free(features); |
cfcc4c62 KW |
230 | } |
231 | ||
bfe8043e SH |
232 | /* |
233 | * Sets the dirty bit and flushes afterwards if necessary. | |
234 | * | |
235 | * The incompatible_features bit is only set if the image file header was | |
236 | * updated successfully. Therefore it is not required to check the return | |
237 | * value of this function. | |
238 | */ | |
280d3735 | 239 | int qcow2_mark_dirty(BlockDriverState *bs) |
bfe8043e | 240 | { |
ff99129a | 241 | BDRVQcow2State *s = bs->opaque; |
bfe8043e SH |
242 | uint64_t val; |
243 | int ret; | |
244 | ||
245 | assert(s->qcow_version >= 3); | |
246 | ||
247 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
248 | return 0; /* already dirty */ | |
249 | } | |
250 | ||
251 | val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); | |
d9ca2ea2 | 252 | ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features), |
bfe8043e SH |
253 | &val, sizeof(val)); |
254 | if (ret < 0) { | |
255 | return ret; | |
256 | } | |
9a4f4c31 | 257 | ret = bdrv_flush(bs->file->bs); |
bfe8043e SH |
258 | if (ret < 0) { |
259 | return ret; | |
260 | } | |
261 | ||
262 | /* Only treat image as dirty if the header was updated successfully */ | |
263 | s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; | |
264 | return 0; | |
265 | } | |
266 | ||
c61d0004 SH |
267 | /* |
268 | * Clears the dirty bit and flushes before if necessary. Only call this | |
269 | * function when there are no pending requests, it does not guard against | |
270 | * concurrent requests dirtying the image. | |
271 | */ | |
272 | static int qcow2_mark_clean(BlockDriverState *bs) | |
273 | { | |
ff99129a | 274 | BDRVQcow2State *s = bs->opaque; |
c61d0004 SH |
275 | |
276 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
4c2e5f8f KW |
277 | int ret; |
278 | ||
279 | s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; | |
280 | ||
281 | ret = bdrv_flush(bs); | |
c61d0004 SH |
282 | if (ret < 0) { |
283 | return ret; | |
284 | } | |
285 | ||
c61d0004 SH |
286 | return qcow2_update_header(bs); |
287 | } | |
288 | return 0; | |
289 | } | |
290 | ||
69c98726 HR |
291 | /* |
292 | * Marks the image as corrupt. | |
293 | */ | |
294 | int qcow2_mark_corrupt(BlockDriverState *bs) | |
295 | { | |
ff99129a | 296 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
297 | |
298 | s->incompatible_features |= QCOW2_INCOMPAT_CORRUPT; | |
299 | return qcow2_update_header(bs); | |
300 | } | |
301 | ||
302 | /* | |
303 | * Marks the image as consistent, i.e., unsets the corrupt bit, and flushes | |
304 | * before if necessary. | |
305 | */ | |
306 | int qcow2_mark_consistent(BlockDriverState *bs) | |
307 | { | |
ff99129a | 308 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
309 | |
310 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { | |
311 | int ret = bdrv_flush(bs); | |
312 | if (ret < 0) { | |
313 | return ret; | |
314 | } | |
315 | ||
316 | s->incompatible_features &= ~QCOW2_INCOMPAT_CORRUPT; | |
317 | return qcow2_update_header(bs); | |
318 | } | |
319 | return 0; | |
320 | } | |
321 | ||
acbe5982 SH |
322 | static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result, |
323 | BdrvCheckMode fix) | |
324 | { | |
325 | int ret = qcow2_check_refcounts(bs, result, fix); | |
326 | if (ret < 0) { | |
327 | return ret; | |
328 | } | |
329 | ||
330 | if (fix && result->check_errors == 0 && result->corruptions == 0) { | |
24530f3e HR |
331 | ret = qcow2_mark_clean(bs); |
332 | if (ret < 0) { | |
333 | return ret; | |
334 | } | |
335 | return qcow2_mark_consistent(bs); | |
acbe5982 SH |
336 | } |
337 | return ret; | |
338 | } | |
339 | ||
8c7de283 KW |
340 | static int validate_table_offset(BlockDriverState *bs, uint64_t offset, |
341 | uint64_t entries, size_t entry_len) | |
342 | { | |
ff99129a | 343 | BDRVQcow2State *s = bs->opaque; |
8c7de283 KW |
344 | uint64_t size; |
345 | ||
346 | /* Use signed INT64_MAX as the maximum even for uint64_t header fields, | |
347 | * because values will be passed to qemu functions taking int64_t. */ | |
348 | if (entries > INT64_MAX / entry_len) { | |
349 | return -EINVAL; | |
350 | } | |
351 | ||
352 | size = entries * entry_len; | |
353 | ||
354 | if (INT64_MAX - size < offset) { | |
355 | return -EINVAL; | |
356 | } | |
357 | ||
358 | /* Tables must be cluster aligned */ | |
359 | if (offset & (s->cluster_size - 1)) { | |
360 | return -EINVAL; | |
361 | } | |
362 | ||
363 | return 0; | |
364 | } | |
365 | ||
74c4510a KW |
366 | static QemuOptsList qcow2_runtime_opts = { |
367 | .name = "qcow2", | |
368 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head), | |
369 | .desc = { | |
370 | { | |
64aa99d3 | 371 | .name = QCOW2_OPT_LAZY_REFCOUNTS, |
74c4510a KW |
372 | .type = QEMU_OPT_BOOL, |
373 | .help = "Postpone refcount updates", | |
374 | }, | |
67af674e KW |
375 | { |
376 | .name = QCOW2_OPT_DISCARD_REQUEST, | |
377 | .type = QEMU_OPT_BOOL, | |
378 | .help = "Pass guest discard requests to the layer below", | |
379 | }, | |
380 | { | |
381 | .name = QCOW2_OPT_DISCARD_SNAPSHOT, | |
382 | .type = QEMU_OPT_BOOL, | |
383 | .help = "Generate discard requests when snapshot related space " | |
384 | "is freed", | |
385 | }, | |
386 | { | |
387 | .name = QCOW2_OPT_DISCARD_OTHER, | |
388 | .type = QEMU_OPT_BOOL, | |
389 | .help = "Generate discard requests when other clusters are freed", | |
390 | }, | |
05de7e86 HR |
391 | { |
392 | .name = QCOW2_OPT_OVERLAP, | |
393 | .type = QEMU_OPT_STRING, | |
394 | .help = "Selects which overlap checks to perform from a range of " | |
395 | "templates (none, constant, cached, all)", | |
396 | }, | |
ee42b5ce HR |
397 | { |
398 | .name = QCOW2_OPT_OVERLAP_TEMPLATE, | |
399 | .type = QEMU_OPT_STRING, | |
400 | .help = "Selects which overlap checks to perform from a range of " | |
401 | "templates (none, constant, cached, all)", | |
402 | }, | |
05de7e86 HR |
403 | { |
404 | .name = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
405 | .type = QEMU_OPT_BOOL, | |
406 | .help = "Check for unintended writes into the main qcow2 header", | |
407 | }, | |
408 | { | |
409 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
410 | .type = QEMU_OPT_BOOL, | |
411 | .help = "Check for unintended writes into the active L1 table", | |
412 | }, | |
413 | { | |
414 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
415 | .type = QEMU_OPT_BOOL, | |
416 | .help = "Check for unintended writes into an active L2 table", | |
417 | }, | |
418 | { | |
419 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
420 | .type = QEMU_OPT_BOOL, | |
421 | .help = "Check for unintended writes into the refcount table", | |
422 | }, | |
423 | { | |
424 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
425 | .type = QEMU_OPT_BOOL, | |
426 | .help = "Check for unintended writes into a refcount block", | |
427 | }, | |
428 | { | |
429 | .name = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
430 | .type = QEMU_OPT_BOOL, | |
431 | .help = "Check for unintended writes into the snapshot table", | |
432 | }, | |
433 | { | |
434 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
435 | .type = QEMU_OPT_BOOL, | |
436 | .help = "Check for unintended writes into an inactive L1 table", | |
437 | }, | |
438 | { | |
439 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
440 | .type = QEMU_OPT_BOOL, | |
441 | .help = "Check for unintended writes into an inactive L2 table", | |
442 | }, | |
6c1c8d5d HR |
443 | { |
444 | .name = QCOW2_OPT_CACHE_SIZE, | |
445 | .type = QEMU_OPT_SIZE, | |
446 | .help = "Maximum combined metadata (L2 tables and refcount blocks) " | |
447 | "cache size", | |
448 | }, | |
449 | { | |
450 | .name = QCOW2_OPT_L2_CACHE_SIZE, | |
451 | .type = QEMU_OPT_SIZE, | |
452 | .help = "Maximum L2 table cache size", | |
453 | }, | |
454 | { | |
455 | .name = QCOW2_OPT_REFCOUNT_CACHE_SIZE, | |
456 | .type = QEMU_OPT_SIZE, | |
457 | .help = "Maximum refcount block cache size", | |
458 | }, | |
279621c0 AG |
459 | { |
460 | .name = QCOW2_OPT_CACHE_CLEAN_INTERVAL, | |
461 | .type = QEMU_OPT_NUMBER, | |
462 | .help = "Clean unused cache entries after this time (in seconds)", | |
463 | }, | |
74c4510a KW |
464 | { /* end of list */ } |
465 | }, | |
466 | }; | |
467 | ||
4092e99d HR |
468 | static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = { |
469 | [QCOW2_OL_MAIN_HEADER_BITNR] = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
470 | [QCOW2_OL_ACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
471 | [QCOW2_OL_ACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
472 | [QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
473 | [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
474 | [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
475 | [QCOW2_OL_INACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
476 | [QCOW2_OL_INACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
477 | }; | |
478 | ||
279621c0 AG |
479 | static void cache_clean_timer_cb(void *opaque) |
480 | { | |
481 | BlockDriverState *bs = opaque; | |
ff99129a | 482 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
483 | qcow2_cache_clean_unused(bs, s->l2_table_cache); |
484 | qcow2_cache_clean_unused(bs, s->refcount_block_cache); | |
485 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
486 | (int64_t) s->cache_clean_interval * 1000); | |
487 | } | |
488 | ||
489 | static void cache_clean_timer_init(BlockDriverState *bs, AioContext *context) | |
490 | { | |
ff99129a | 491 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
492 | if (s->cache_clean_interval > 0) { |
493 | s->cache_clean_timer = aio_timer_new(context, QEMU_CLOCK_VIRTUAL, | |
494 | SCALE_MS, cache_clean_timer_cb, | |
495 | bs); | |
496 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
497 | (int64_t) s->cache_clean_interval * 1000); | |
498 | } | |
499 | } | |
500 | ||
501 | static void cache_clean_timer_del(BlockDriverState *bs) | |
502 | { | |
ff99129a | 503 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
504 | if (s->cache_clean_timer) { |
505 | timer_del(s->cache_clean_timer); | |
506 | timer_free(s->cache_clean_timer); | |
507 | s->cache_clean_timer = NULL; | |
508 | } | |
509 | } | |
510 | ||
511 | static void qcow2_detach_aio_context(BlockDriverState *bs) | |
512 | { | |
513 | cache_clean_timer_del(bs); | |
514 | } | |
515 | ||
516 | static void qcow2_attach_aio_context(BlockDriverState *bs, | |
517 | AioContext *new_context) | |
518 | { | |
519 | cache_clean_timer_init(bs, new_context); | |
520 | } | |
521 | ||
bc85ef26 HR |
522 | static void read_cache_sizes(BlockDriverState *bs, QemuOpts *opts, |
523 | uint64_t *l2_cache_size, | |
6c1c8d5d HR |
524 | uint64_t *refcount_cache_size, Error **errp) |
525 | { | |
ff99129a | 526 | BDRVQcow2State *s = bs->opaque; |
6c1c8d5d HR |
527 | uint64_t combined_cache_size; |
528 | bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set; | |
529 | ||
530 | combined_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_CACHE_SIZE); | |
531 | l2_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_SIZE); | |
532 | refcount_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
533 | ||
534 | combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0); | |
535 | *l2_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE, 0); | |
536 | *refcount_cache_size = qemu_opt_get_size(opts, | |
537 | QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0); | |
538 | ||
539 | if (combined_cache_size_set) { | |
540 | if (l2_cache_size_set && refcount_cache_size_set) { | |
541 | error_setg(errp, QCOW2_OPT_CACHE_SIZE ", " QCOW2_OPT_L2_CACHE_SIZE | |
542 | " and " QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not be set " | |
543 | "the same time"); | |
544 | return; | |
545 | } else if (*l2_cache_size > combined_cache_size) { | |
546 | error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed " | |
547 | QCOW2_OPT_CACHE_SIZE); | |
548 | return; | |
549 | } else if (*refcount_cache_size > combined_cache_size) { | |
550 | error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed " | |
551 | QCOW2_OPT_CACHE_SIZE); | |
552 | return; | |
553 | } | |
554 | ||
555 | if (l2_cache_size_set) { | |
556 | *refcount_cache_size = combined_cache_size - *l2_cache_size; | |
557 | } else if (refcount_cache_size_set) { | |
558 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
559 | } else { | |
560 | *refcount_cache_size = combined_cache_size | |
561 | / (DEFAULT_L2_REFCOUNT_SIZE_RATIO + 1); | |
562 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
563 | } | |
564 | } else { | |
565 | if (!l2_cache_size_set && !refcount_cache_size_set) { | |
bc85ef26 HR |
566 | *l2_cache_size = MAX(DEFAULT_L2_CACHE_BYTE_SIZE, |
567 | (uint64_t)DEFAULT_L2_CACHE_CLUSTERS | |
568 | * s->cluster_size); | |
6c1c8d5d HR |
569 | *refcount_cache_size = *l2_cache_size |
570 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
571 | } else if (!l2_cache_size_set) { | |
572 | *l2_cache_size = *refcount_cache_size | |
573 | * DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
574 | } else if (!refcount_cache_size_set) { | |
575 | *refcount_cache_size = *l2_cache_size | |
576 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
577 | } | |
578 | } | |
579 | } | |
580 | ||
ee55b173 KW |
581 | typedef struct Qcow2ReopenState { |
582 | Qcow2Cache *l2_table_cache; | |
583 | Qcow2Cache *refcount_block_cache; | |
584 | bool use_lazy_refcounts; | |
585 | int overlap_check; | |
586 | bool discard_passthrough[QCOW2_DISCARD_MAX]; | |
587 | uint64_t cache_clean_interval; | |
588 | } Qcow2ReopenState; | |
589 | ||
590 | static int qcow2_update_options_prepare(BlockDriverState *bs, | |
591 | Qcow2ReopenState *r, | |
592 | QDict *options, int flags, | |
593 | Error **errp) | |
4c75d1a1 KW |
594 | { |
595 | BDRVQcow2State *s = bs->opaque; | |
94edf3fb | 596 | QemuOpts *opts = NULL; |
4c75d1a1 KW |
597 | const char *opt_overlap_check, *opt_overlap_check_template; |
598 | int overlap_check_template = 0; | |
94edf3fb | 599 | uint64_t l2_cache_size, refcount_cache_size; |
4c75d1a1 | 600 | int i; |
94edf3fb | 601 | Error *local_err = NULL; |
4c75d1a1 KW |
602 | int ret; |
603 | ||
94edf3fb KW |
604 | opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort); |
605 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
606 | if (local_err) { | |
607 | error_propagate(errp, local_err); | |
608 | ret = -EINVAL; | |
609 | goto fail; | |
610 | } | |
611 | ||
612 | /* get L2 table/refcount block cache size from command line options */ | |
613 | read_cache_sizes(bs, opts, &l2_cache_size, &refcount_cache_size, | |
614 | &local_err); | |
615 | if (local_err) { | |
616 | error_propagate(errp, local_err); | |
617 | ret = -EINVAL; | |
618 | goto fail; | |
619 | } | |
620 | ||
621 | l2_cache_size /= s->cluster_size; | |
622 | if (l2_cache_size < MIN_L2_CACHE_SIZE) { | |
623 | l2_cache_size = MIN_L2_CACHE_SIZE; | |
624 | } | |
625 | if (l2_cache_size > INT_MAX) { | |
626 | error_setg(errp, "L2 cache size too big"); | |
627 | ret = -EINVAL; | |
628 | goto fail; | |
629 | } | |
630 | ||
631 | refcount_cache_size /= s->cluster_size; | |
632 | if (refcount_cache_size < MIN_REFCOUNT_CACHE_SIZE) { | |
633 | refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE; | |
634 | } | |
635 | if (refcount_cache_size > INT_MAX) { | |
636 | error_setg(errp, "Refcount cache size too big"); | |
637 | ret = -EINVAL; | |
638 | goto fail; | |
639 | } | |
640 | ||
5b0959a7 KW |
641 | /* alloc new L2 table/refcount block cache, flush old one */ |
642 | if (s->l2_table_cache) { | |
643 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
644 | if (ret) { | |
645 | error_setg_errno(errp, -ret, "Failed to flush the L2 table cache"); | |
646 | goto fail; | |
647 | } | |
648 | } | |
649 | ||
650 | if (s->refcount_block_cache) { | |
651 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
652 | if (ret) { | |
653 | error_setg_errno(errp, -ret, | |
654 | "Failed to flush the refcount block cache"); | |
655 | goto fail; | |
656 | } | |
657 | } | |
658 | ||
ee55b173 KW |
659 | r->l2_table_cache = qcow2_cache_create(bs, l2_cache_size); |
660 | r->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size); | |
661 | if (r->l2_table_cache == NULL || r->refcount_block_cache == NULL) { | |
94edf3fb KW |
662 | error_setg(errp, "Could not allocate metadata caches"); |
663 | ret = -ENOMEM; | |
664 | goto fail; | |
665 | } | |
666 | ||
667 | /* New interval for cache cleanup timer */ | |
ee55b173 | 668 | r->cache_clean_interval = |
5b0959a7 KW |
669 | qemu_opt_get_number(opts, QCOW2_OPT_CACHE_CLEAN_INTERVAL, |
670 | s->cache_clean_interval); | |
ee55b173 | 671 | if (r->cache_clean_interval > UINT_MAX) { |
94edf3fb KW |
672 | error_setg(errp, "Cache clean interval too big"); |
673 | ret = -EINVAL; | |
674 | goto fail; | |
675 | } | |
94edf3fb | 676 | |
5b0959a7 | 677 | /* lazy-refcounts; flush if going from enabled to disabled */ |
ee55b173 | 678 | r->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, |
4c75d1a1 | 679 | (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); |
ee55b173 | 680 | if (r->use_lazy_refcounts && s->qcow_version < 3) { |
007dbc39 KW |
681 | error_setg(errp, "Lazy refcounts require a qcow2 image with at least " |
682 | "qemu 1.1 compatibility level"); | |
683 | ret = -EINVAL; | |
684 | goto fail; | |
685 | } | |
4c75d1a1 | 686 | |
5b0959a7 KW |
687 | if (s->use_lazy_refcounts && !r->use_lazy_refcounts) { |
688 | ret = qcow2_mark_clean(bs); | |
689 | if (ret < 0) { | |
690 | error_setg_errno(errp, -ret, "Failed to disable lazy refcounts"); | |
691 | goto fail; | |
692 | } | |
693 | } | |
694 | ||
007dbc39 | 695 | /* Overlap check options */ |
4c75d1a1 KW |
696 | opt_overlap_check = qemu_opt_get(opts, QCOW2_OPT_OVERLAP); |
697 | opt_overlap_check_template = qemu_opt_get(opts, QCOW2_OPT_OVERLAP_TEMPLATE); | |
698 | if (opt_overlap_check_template && opt_overlap_check && | |
699 | strcmp(opt_overlap_check_template, opt_overlap_check)) | |
700 | { | |
701 | error_setg(errp, "Conflicting values for qcow2 options '" | |
702 | QCOW2_OPT_OVERLAP "' ('%s') and '" QCOW2_OPT_OVERLAP_TEMPLATE | |
703 | "' ('%s')", opt_overlap_check, opt_overlap_check_template); | |
704 | ret = -EINVAL; | |
705 | goto fail; | |
706 | } | |
707 | if (!opt_overlap_check) { | |
708 | opt_overlap_check = opt_overlap_check_template ?: "cached"; | |
709 | } | |
710 | ||
711 | if (!strcmp(opt_overlap_check, "none")) { | |
712 | overlap_check_template = 0; | |
713 | } else if (!strcmp(opt_overlap_check, "constant")) { | |
714 | overlap_check_template = QCOW2_OL_CONSTANT; | |
715 | } else if (!strcmp(opt_overlap_check, "cached")) { | |
716 | overlap_check_template = QCOW2_OL_CACHED; | |
717 | } else if (!strcmp(opt_overlap_check, "all")) { | |
718 | overlap_check_template = QCOW2_OL_ALL; | |
719 | } else { | |
720 | error_setg(errp, "Unsupported value '%s' for qcow2 option " | |
721 | "'overlap-check'. Allowed are any of the following: " | |
722 | "none, constant, cached, all", opt_overlap_check); | |
723 | ret = -EINVAL; | |
724 | goto fail; | |
725 | } | |
726 | ||
ee55b173 | 727 | r->overlap_check = 0; |
4c75d1a1 KW |
728 | for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) { |
729 | /* overlap-check defines a template bitmask, but every flag may be | |
730 | * overwritten through the associated boolean option */ | |
ee55b173 | 731 | r->overlap_check |= |
4c75d1a1 KW |
732 | qemu_opt_get_bool(opts, overlap_bool_option_names[i], |
733 | overlap_check_template & (1 << i)) << i; | |
734 | } | |
735 | ||
ee55b173 KW |
736 | r->discard_passthrough[QCOW2_DISCARD_NEVER] = false; |
737 | r->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true; | |
738 | r->discard_passthrough[QCOW2_DISCARD_REQUEST] = | |
007dbc39 KW |
739 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST, |
740 | flags & BDRV_O_UNMAP); | |
ee55b173 | 741 | r->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] = |
007dbc39 | 742 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true); |
ee55b173 | 743 | r->discard_passthrough[QCOW2_DISCARD_OTHER] = |
007dbc39 KW |
744 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false); |
745 | ||
4c75d1a1 KW |
746 | ret = 0; |
747 | fail: | |
94edf3fb KW |
748 | qemu_opts_del(opts); |
749 | opts = NULL; | |
ee55b173 KW |
750 | return ret; |
751 | } | |
752 | ||
753 | static void qcow2_update_options_commit(BlockDriverState *bs, | |
754 | Qcow2ReopenState *r) | |
755 | { | |
756 | BDRVQcow2State *s = bs->opaque; | |
757 | int i; | |
758 | ||
5b0959a7 KW |
759 | if (s->l2_table_cache) { |
760 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
761 | } | |
762 | if (s->refcount_block_cache) { | |
763 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
764 | } | |
ee55b173 KW |
765 | s->l2_table_cache = r->l2_table_cache; |
766 | s->refcount_block_cache = r->refcount_block_cache; | |
767 | ||
768 | s->overlap_check = r->overlap_check; | |
769 | s->use_lazy_refcounts = r->use_lazy_refcounts; | |
770 | ||
771 | for (i = 0; i < QCOW2_DISCARD_MAX; i++) { | |
772 | s->discard_passthrough[i] = r->discard_passthrough[i]; | |
773 | } | |
774 | ||
5b0959a7 KW |
775 | if (s->cache_clean_interval != r->cache_clean_interval) { |
776 | cache_clean_timer_del(bs); | |
777 | s->cache_clean_interval = r->cache_clean_interval; | |
778 | cache_clean_timer_init(bs, bdrv_get_aio_context(bs)); | |
779 | } | |
ee55b173 KW |
780 | } |
781 | ||
782 | static void qcow2_update_options_abort(BlockDriverState *bs, | |
783 | Qcow2ReopenState *r) | |
784 | { | |
785 | if (r->l2_table_cache) { | |
786 | qcow2_cache_destroy(bs, r->l2_table_cache); | |
787 | } | |
788 | if (r->refcount_block_cache) { | |
789 | qcow2_cache_destroy(bs, r->refcount_block_cache); | |
790 | } | |
791 | } | |
792 | ||
793 | static int qcow2_update_options(BlockDriverState *bs, QDict *options, | |
794 | int flags, Error **errp) | |
795 | { | |
796 | Qcow2ReopenState r = {}; | |
797 | int ret; | |
798 | ||
799 | ret = qcow2_update_options_prepare(bs, &r, options, flags, errp); | |
800 | if (ret >= 0) { | |
801 | qcow2_update_options_commit(bs, &r); | |
802 | } else { | |
803 | qcow2_update_options_abort(bs, &r); | |
804 | } | |
94edf3fb | 805 | |
4c75d1a1 KW |
806 | return ret; |
807 | } | |
808 | ||
015a1036 HR |
809 | static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, |
810 | Error **errp) | |
585f8587 | 811 | { |
ff99129a | 812 | BDRVQcow2State *s = bs->opaque; |
6d33e8e7 KW |
813 | unsigned int len, i; |
814 | int ret = 0; | |
585f8587 | 815 | QCowHeader header; |
74c4510a | 816 | Error *local_err = NULL; |
9b80ddf3 | 817 | uint64_t ext_end; |
2cf7cfa1 | 818 | uint64_t l1_vm_state_index; |
585f8587 | 819 | |
cf2ab8fc | 820 | ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); |
6d85a57e | 821 | if (ret < 0) { |
3ef6c40a | 822 | error_setg_errno(errp, -ret, "Could not read qcow2 header"); |
585f8587 | 823 | goto fail; |
6d85a57e | 824 | } |
585f8587 FB |
825 | be32_to_cpus(&header.magic); |
826 | be32_to_cpus(&header.version); | |
827 | be64_to_cpus(&header.backing_file_offset); | |
828 | be32_to_cpus(&header.backing_file_size); | |
829 | be64_to_cpus(&header.size); | |
830 | be32_to_cpus(&header.cluster_bits); | |
831 | be32_to_cpus(&header.crypt_method); | |
832 | be64_to_cpus(&header.l1_table_offset); | |
833 | be32_to_cpus(&header.l1_size); | |
834 | be64_to_cpus(&header.refcount_table_offset); | |
835 | be32_to_cpus(&header.refcount_table_clusters); | |
836 | be64_to_cpus(&header.snapshots_offset); | |
837 | be32_to_cpus(&header.nb_snapshots); | |
3b46e624 | 838 | |
e8cdcec1 | 839 | if (header.magic != QCOW_MAGIC) { |
3ef6c40a | 840 | error_setg(errp, "Image is not in qcow2 format"); |
76abe407 | 841 | ret = -EINVAL; |
585f8587 | 842 | goto fail; |
6d85a57e | 843 | } |
6744cbab | 844 | if (header.version < 2 || header.version > 3) { |
a55448b3 | 845 | error_setg(errp, "Unsupported qcow2 version %" PRIu32, header.version); |
6744cbab KW |
846 | ret = -ENOTSUP; |
847 | goto fail; | |
848 | } | |
849 | ||
850 | s->qcow_version = header.version; | |
851 | ||
24342f2c KW |
852 | /* Initialise cluster size */ |
853 | if (header.cluster_bits < MIN_CLUSTER_BITS || | |
854 | header.cluster_bits > MAX_CLUSTER_BITS) { | |
521b2b5d HR |
855 | error_setg(errp, "Unsupported cluster size: 2^%" PRIu32, |
856 | header.cluster_bits); | |
24342f2c KW |
857 | ret = -EINVAL; |
858 | goto fail; | |
859 | } | |
860 | ||
861 | s->cluster_bits = header.cluster_bits; | |
862 | s->cluster_size = 1 << s->cluster_bits; | |
863 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
864 | ||
6744cbab KW |
865 | /* Initialise version 3 header fields */ |
866 | if (header.version == 2) { | |
867 | header.incompatible_features = 0; | |
868 | header.compatible_features = 0; | |
869 | header.autoclear_features = 0; | |
870 | header.refcount_order = 4; | |
871 | header.header_length = 72; | |
872 | } else { | |
873 | be64_to_cpus(&header.incompatible_features); | |
874 | be64_to_cpus(&header.compatible_features); | |
875 | be64_to_cpus(&header.autoclear_features); | |
876 | be32_to_cpus(&header.refcount_order); | |
877 | be32_to_cpus(&header.header_length); | |
24342f2c KW |
878 | |
879 | if (header.header_length < 104) { | |
880 | error_setg(errp, "qcow2 header too short"); | |
881 | ret = -EINVAL; | |
882 | goto fail; | |
883 | } | |
884 | } | |
885 | ||
886 | if (header.header_length > s->cluster_size) { | |
887 | error_setg(errp, "qcow2 header exceeds cluster size"); | |
888 | ret = -EINVAL; | |
889 | goto fail; | |
6744cbab KW |
890 | } |
891 | ||
892 | if (header.header_length > sizeof(header)) { | |
893 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
894 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
cf2ab8fc | 895 | ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields, |
6744cbab KW |
896 | s->unknown_header_fields_size); |
897 | if (ret < 0) { | |
3ef6c40a HR |
898 | error_setg_errno(errp, -ret, "Could not read unknown qcow2 header " |
899 | "fields"); | |
6744cbab KW |
900 | goto fail; |
901 | } | |
902 | } | |
903 | ||
a1b3955c KW |
904 | if (header.backing_file_offset > s->cluster_size) { |
905 | error_setg(errp, "Invalid backing file offset"); | |
906 | ret = -EINVAL; | |
907 | goto fail; | |
908 | } | |
909 | ||
cfcc4c62 KW |
910 | if (header.backing_file_offset) { |
911 | ext_end = header.backing_file_offset; | |
912 | } else { | |
913 | ext_end = 1 << header.cluster_bits; | |
914 | } | |
915 | ||
6744cbab KW |
916 | /* Handle feature bits */ |
917 | s->incompatible_features = header.incompatible_features; | |
918 | s->compatible_features = header.compatible_features; | |
919 | s->autoclear_features = header.autoclear_features; | |
920 | ||
c61d0004 | 921 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { |
cfcc4c62 KW |
922 | void *feature_table = NULL; |
923 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
3ef6c40a | 924 | &feature_table, NULL); |
a55448b3 | 925 | report_unsupported_feature(errp, feature_table, |
c61d0004 SH |
926 | s->incompatible_features & |
927 | ~QCOW2_INCOMPAT_MASK); | |
6744cbab | 928 | ret = -ENOTSUP; |
c5a33ee9 | 929 | g_free(feature_table); |
6744cbab KW |
930 | goto fail; |
931 | } | |
932 | ||
69c98726 HR |
933 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { |
934 | /* Corrupt images may not be written to unless they are being repaired | |
935 | */ | |
936 | if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) { | |
3ef6c40a HR |
937 | error_setg(errp, "qcow2: Image is corrupt; cannot be opened " |
938 | "read/write"); | |
69c98726 HR |
939 | ret = -EACCES; |
940 | goto fail; | |
941 | } | |
942 | } | |
943 | ||
6744cbab | 944 | /* Check support for various header values */ |
b72faf9f HR |
945 | if (header.refcount_order > 6) { |
946 | error_setg(errp, "Reference count entry width too large; may not " | |
947 | "exceed 64 bits"); | |
948 | ret = -EINVAL; | |
e8cdcec1 KW |
949 | goto fail; |
950 | } | |
b6481f37 | 951 | s->refcount_order = header.refcount_order; |
346a53df HR |
952 | s->refcount_bits = 1 << s->refcount_order; |
953 | s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1); | |
954 | s->refcount_max += s->refcount_max - 1; | |
6744cbab | 955 | |
6d85a57e | 956 | if (header.crypt_method > QCOW_CRYPT_AES) { |
521b2b5d | 957 | error_setg(errp, "Unsupported encryption method: %" PRIu32, |
3ef6c40a | 958 | header.crypt_method); |
6d85a57e | 959 | ret = -EINVAL; |
585f8587 | 960 | goto fail; |
6d85a57e | 961 | } |
f844836d GA |
962 | if (!qcrypto_cipher_supports(QCRYPTO_CIPHER_ALG_AES_128, |
963 | QCRYPTO_CIPHER_MODE_CBC)) { | |
f6fa64f6 DB |
964 | error_setg(errp, "AES cipher not available"); |
965 | ret = -EINVAL; | |
966 | goto fail; | |
967 | } | |
585f8587 | 968 | s->crypt_method_header = header.crypt_method; |
6d85a57e | 969 | if (s->crypt_method_header) { |
e6ff69bf DB |
970 | if (bdrv_uses_whitelist() && |
971 | s->crypt_method_header == QCOW_CRYPT_AES) { | |
8c0dcbc4 DB |
972 | error_setg(errp, |
973 | "Use of AES-CBC encrypted qcow2 images is no longer " | |
974 | "supported in system emulators"); | |
975 | error_append_hint(errp, | |
976 | "You can use 'qemu-img convert' to convert your " | |
977 | "image to an alternative supported format, such " | |
978 | "as unencrypted qcow2, or raw with the LUKS " | |
979 | "format instead.\n"); | |
980 | ret = -ENOSYS; | |
981 | goto fail; | |
e6ff69bf DB |
982 | } |
983 | ||
54115412 | 984 | bs->encrypted = true; |
6d85a57e | 985 | } |
24342f2c | 986 | |
585f8587 FB |
987 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
988 | s->l2_size = 1 << s->l2_bits; | |
1d13d654 HR |
989 | /* 2^(s->refcount_order - 3) is the refcount width in bytes */ |
990 | s->refcount_block_bits = s->cluster_bits - (s->refcount_order - 3); | |
991 | s->refcount_block_size = 1 << s->refcount_block_bits; | |
585f8587 FB |
992 | bs->total_sectors = header.size / 512; |
993 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
994 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
995 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
5dab2fad | 996 | |
585f8587 | 997 | s->refcount_table_offset = header.refcount_table_offset; |
5fafdf24 | 998 | s->refcount_table_size = |
585f8587 FB |
999 | header.refcount_table_clusters << (s->cluster_bits - 3); |
1000 | ||
2b5d5953 | 1001 | if (header.refcount_table_clusters > qcow2_max_refcount_clusters(s)) { |
5dab2fad KW |
1002 | error_setg(errp, "Reference count table too large"); |
1003 | ret = -EINVAL; | |
1004 | goto fail; | |
1005 | } | |
1006 | ||
8c7de283 KW |
1007 | ret = validate_table_offset(bs, s->refcount_table_offset, |
1008 | s->refcount_table_size, sizeof(uint64_t)); | |
1009 | if (ret < 0) { | |
1010 | error_setg(errp, "Invalid reference count table offset"); | |
1011 | goto fail; | |
1012 | } | |
1013 | ||
ce48f2f4 KW |
1014 | /* Snapshot table offset/length */ |
1015 | if (header.nb_snapshots > QCOW_MAX_SNAPSHOTS) { | |
1016 | error_setg(errp, "Too many snapshots"); | |
1017 | ret = -EINVAL; | |
1018 | goto fail; | |
1019 | } | |
1020 | ||
1021 | ret = validate_table_offset(bs, header.snapshots_offset, | |
1022 | header.nb_snapshots, | |
1023 | sizeof(QCowSnapshotHeader)); | |
1024 | if (ret < 0) { | |
1025 | error_setg(errp, "Invalid snapshot table offset"); | |
1026 | goto fail; | |
1027 | } | |
1028 | ||
585f8587 | 1029 | /* read the level 1 table */ |
87b86e7e | 1030 | if (header.l1_size > QCOW_MAX_L1_SIZE / sizeof(uint64_t)) { |
2d51c32c KW |
1031 | error_setg(errp, "Active L1 table too large"); |
1032 | ret = -EFBIG; | |
1033 | goto fail; | |
1034 | } | |
585f8587 | 1035 | s->l1_size = header.l1_size; |
2cf7cfa1 KW |
1036 | |
1037 | l1_vm_state_index = size_to_l1(s, header.size); | |
1038 | if (l1_vm_state_index > INT_MAX) { | |
3ef6c40a | 1039 | error_setg(errp, "Image is too big"); |
2cf7cfa1 KW |
1040 | ret = -EFBIG; |
1041 | goto fail; | |
1042 | } | |
1043 | s->l1_vm_state_index = l1_vm_state_index; | |
1044 | ||
585f8587 FB |
1045 | /* the L1 table must contain at least enough entries to put |
1046 | header.size bytes */ | |
6d85a57e | 1047 | if (s->l1_size < s->l1_vm_state_index) { |
3ef6c40a | 1048 | error_setg(errp, "L1 table is too small"); |
6d85a57e | 1049 | ret = -EINVAL; |
585f8587 | 1050 | goto fail; |
6d85a57e | 1051 | } |
2d51c32c KW |
1052 | |
1053 | ret = validate_table_offset(bs, header.l1_table_offset, | |
1054 | header.l1_size, sizeof(uint64_t)); | |
1055 | if (ret < 0) { | |
1056 | error_setg(errp, "Invalid L1 table offset"); | |
1057 | goto fail; | |
1058 | } | |
585f8587 | 1059 | s->l1_table_offset = header.l1_table_offset; |
2d51c32c KW |
1060 | |
1061 | ||
d191d12d | 1062 | if (s->l1_size > 0) { |
9a4f4c31 | 1063 | s->l1_table = qemu_try_blockalign(bs->file->bs, |
d191d12d | 1064 | align_offset(s->l1_size * sizeof(uint64_t), 512)); |
de82815d KW |
1065 | if (s->l1_table == NULL) { |
1066 | error_setg(errp, "Could not allocate L1 table"); | |
1067 | ret = -ENOMEM; | |
1068 | goto fail; | |
1069 | } | |
cf2ab8fc | 1070 | ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, |
6d85a57e JS |
1071 | s->l1_size * sizeof(uint64_t)); |
1072 | if (ret < 0) { | |
3ef6c40a | 1073 | error_setg_errno(errp, -ret, "Could not read L1 table"); |
d191d12d | 1074 | goto fail; |
6d85a57e | 1075 | } |
d191d12d SW |
1076 | for(i = 0;i < s->l1_size; i++) { |
1077 | be64_to_cpus(&s->l1_table[i]); | |
1078 | } | |
585f8587 | 1079 | } |
29c1a730 | 1080 | |
94edf3fb KW |
1081 | /* Parse driver-specific options */ |
1082 | ret = qcow2_update_options(bs, options, flags, errp); | |
90efa0ea KW |
1083 | if (ret < 0) { |
1084 | goto fail; | |
1085 | } | |
1086 | ||
7267c094 | 1087 | s->cluster_cache = g_malloc(s->cluster_size); |
585f8587 | 1088 | /* one more sector for decompressed data alignment */ |
9a4f4c31 | 1089 | s->cluster_data = qemu_try_blockalign(bs->file->bs, QCOW_MAX_CRYPT_CLUSTERS |
de82815d KW |
1090 | * s->cluster_size + 512); |
1091 | if (s->cluster_data == NULL) { | |
1092 | error_setg(errp, "Could not allocate temporary cluster buffer"); | |
1093 | ret = -ENOMEM; | |
1094 | goto fail; | |
1095 | } | |
1096 | ||
585f8587 | 1097 | s->cluster_cache_offset = -1; |
06d9260f | 1098 | s->flags = flags; |
3b46e624 | 1099 | |
6d85a57e JS |
1100 | ret = qcow2_refcount_init(bs); |
1101 | if (ret != 0) { | |
3ef6c40a | 1102 | error_setg_errno(errp, -ret, "Could not initialize refcount handling"); |
585f8587 | 1103 | goto fail; |
6d85a57e | 1104 | } |
585f8587 | 1105 | |
72cf2d4f | 1106 | QLIST_INIT(&s->cluster_allocs); |
0b919fae | 1107 | QTAILQ_INIT(&s->discards); |
f214978a | 1108 | |
9b80ddf3 | 1109 | /* read qcow2 extensions */ |
3ef6c40a HR |
1110 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL, |
1111 | &local_err)) { | |
1112 | error_propagate(errp, local_err); | |
6d85a57e | 1113 | ret = -EINVAL; |
9b80ddf3 | 1114 | goto fail; |
6d85a57e | 1115 | } |
9b80ddf3 | 1116 | |
585f8587 FB |
1117 | /* read the backing file name */ |
1118 | if (header.backing_file_offset != 0) { | |
1119 | len = header.backing_file_size; | |
9a29e18f | 1120 | if (len > MIN(1023, s->cluster_size - header.backing_file_offset) || |
e729fa6a | 1121 | len >= sizeof(bs->backing_file)) { |
6d33e8e7 KW |
1122 | error_setg(errp, "Backing file name too long"); |
1123 | ret = -EINVAL; | |
1124 | goto fail; | |
6d85a57e | 1125 | } |
cf2ab8fc | 1126 | ret = bdrv_pread(bs->file, header.backing_file_offset, |
6d85a57e JS |
1127 | bs->backing_file, len); |
1128 | if (ret < 0) { | |
3ef6c40a | 1129 | error_setg_errno(errp, -ret, "Could not read backing file name"); |
585f8587 | 1130 | goto fail; |
6d85a57e | 1131 | } |
585f8587 | 1132 | bs->backing_file[len] = '\0'; |
e4603fe1 | 1133 | s->image_backing_file = g_strdup(bs->backing_file); |
585f8587 | 1134 | } |
42deb29f | 1135 | |
11b128f4 KW |
1136 | /* Internal snapshots */ |
1137 | s->snapshots_offset = header.snapshots_offset; | |
1138 | s->nb_snapshots = header.nb_snapshots; | |
1139 | ||
42deb29f KW |
1140 | ret = qcow2_read_snapshots(bs); |
1141 | if (ret < 0) { | |
3ef6c40a | 1142 | error_setg_errno(errp, -ret, "Could not read snapshots"); |
585f8587 | 1143 | goto fail; |
6d85a57e | 1144 | } |
585f8587 | 1145 | |
af7b708d | 1146 | /* Clear unknown autoclear feature bits */ |
04c01a5c | 1147 | if (!bs->read_only && !(flags & BDRV_O_INACTIVE) && s->autoclear_features) { |
af7b708d SH |
1148 | s->autoclear_features = 0; |
1149 | ret = qcow2_update_header(bs); | |
1150 | if (ret < 0) { | |
3ef6c40a | 1151 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); |
af7b708d SH |
1152 | goto fail; |
1153 | } | |
1154 | } | |
1155 | ||
68d100e9 KW |
1156 | /* Initialise locks */ |
1157 | qemu_co_mutex_init(&s->lock); | |
170f4b2e | 1158 | bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP; |
68d100e9 | 1159 | |
c61d0004 | 1160 | /* Repair image if dirty */ |
04c01a5c | 1161 | if (!(flags & (BDRV_O_CHECK | BDRV_O_INACTIVE)) && !bs->read_only && |
058f8f16 | 1162 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { |
c61d0004 SH |
1163 | BdrvCheckResult result = {0}; |
1164 | ||
5b84106b | 1165 | ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS); |
c61d0004 | 1166 | if (ret < 0) { |
3ef6c40a | 1167 | error_setg_errno(errp, -ret, "Could not repair dirty image"); |
c61d0004 SH |
1168 | goto fail; |
1169 | } | |
1170 | } | |
1171 | ||
585f8587 | 1172 | #ifdef DEBUG_ALLOC |
6cbc3031 PH |
1173 | { |
1174 | BdrvCheckResult result = {0}; | |
b35278f7 | 1175 | qcow2_check_refcounts(bs, &result, 0); |
6cbc3031 | 1176 | } |
585f8587 | 1177 | #endif |
6d85a57e | 1178 | return ret; |
585f8587 FB |
1179 | |
1180 | fail: | |
6744cbab | 1181 | g_free(s->unknown_header_fields); |
75bab85c | 1182 | cleanup_unknown_header_ext(bs); |
ed6ccf0f KW |
1183 | qcow2_free_snapshots(bs); |
1184 | qcow2_refcount_close(bs); | |
de82815d | 1185 | qemu_vfree(s->l1_table); |
cf93980e HR |
1186 | /* else pre-write overlap checks in cache_destroy may crash */ |
1187 | s->l1_table = NULL; | |
279621c0 | 1188 | cache_clean_timer_del(bs); |
29c1a730 KW |
1189 | if (s->l2_table_cache) { |
1190 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
1191 | } | |
c5a33ee9 PJ |
1192 | if (s->refcount_block_cache) { |
1193 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1194 | } | |
7267c094 | 1195 | g_free(s->cluster_cache); |
dea43a65 | 1196 | qemu_vfree(s->cluster_data); |
6d85a57e | 1197 | return ret; |
585f8587 FB |
1198 | } |
1199 | ||
3baca891 | 1200 | static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd | 1201 | { |
ff99129a | 1202 | BDRVQcow2State *s = bs->opaque; |
d34682cd | 1203 | |
a84178cc EB |
1204 | if (bs->encrypted) { |
1205 | /* Encryption works on a sector granularity */ | |
a5b8dd2c | 1206 | bs->bl.request_alignment = BDRV_SECTOR_SIZE; |
a84178cc | 1207 | } |
cf081fca | 1208 | bs->bl.pwrite_zeroes_alignment = s->cluster_size; |
ecdbead6 | 1209 | bs->bl.pdiscard_alignment = s->cluster_size; |
d34682cd KW |
1210 | } |
1211 | ||
7c80ab3f | 1212 | static int qcow2_set_key(BlockDriverState *bs, const char *key) |
585f8587 | 1213 | { |
ff99129a | 1214 | BDRVQcow2State *s = bs->opaque; |
585f8587 FB |
1215 | uint8_t keybuf[16]; |
1216 | int len, i; | |
f6fa64f6 | 1217 | Error *err = NULL; |
3b46e624 | 1218 | |
585f8587 FB |
1219 | memset(keybuf, 0, 16); |
1220 | len = strlen(key); | |
1221 | if (len > 16) | |
1222 | len = 16; | |
1223 | /* XXX: we could compress the chars to 7 bits to increase | |
1224 | entropy */ | |
1225 | for(i = 0;i < len;i++) { | |
1226 | keybuf[i] = key[i]; | |
1227 | } | |
8336aafa | 1228 | assert(bs->encrypted); |
585f8587 | 1229 | |
f6fa64f6 DB |
1230 | qcrypto_cipher_free(s->cipher); |
1231 | s->cipher = qcrypto_cipher_new( | |
1232 | QCRYPTO_CIPHER_ALG_AES_128, | |
1233 | QCRYPTO_CIPHER_MODE_CBC, | |
1234 | keybuf, G_N_ELEMENTS(keybuf), | |
1235 | &err); | |
1236 | ||
1237 | if (!s->cipher) { | |
1238 | /* XXX would be nice if errors in this method could | |
1239 | * be properly propagate to the caller. Would need | |
1240 | * the bdrv_set_key() API signature to be fixed. */ | |
1241 | error_free(err); | |
585f8587 | 1242 | return -1; |
585f8587 | 1243 | } |
585f8587 FB |
1244 | return 0; |
1245 | } | |
1246 | ||
21d82ac9 JC |
1247 | static int qcow2_reopen_prepare(BDRVReopenState *state, |
1248 | BlockReopenQueue *queue, Error **errp) | |
1249 | { | |
5b0959a7 | 1250 | Qcow2ReopenState *r; |
4c2e5f8f KW |
1251 | int ret; |
1252 | ||
5b0959a7 KW |
1253 | r = g_new0(Qcow2ReopenState, 1); |
1254 | state->opaque = r; | |
1255 | ||
1256 | ret = qcow2_update_options_prepare(state->bs, r, state->options, | |
1257 | state->flags, errp); | |
1258 | if (ret < 0) { | |
1259 | goto fail; | |
1260 | } | |
1261 | ||
1262 | /* We need to write out any unwritten data if we reopen read-only. */ | |
4c2e5f8f KW |
1263 | if ((state->flags & BDRV_O_RDWR) == 0) { |
1264 | ret = bdrv_flush(state->bs); | |
1265 | if (ret < 0) { | |
5b0959a7 | 1266 | goto fail; |
4c2e5f8f KW |
1267 | } |
1268 | ||
1269 | ret = qcow2_mark_clean(state->bs); | |
1270 | if (ret < 0) { | |
5b0959a7 | 1271 | goto fail; |
4c2e5f8f KW |
1272 | } |
1273 | } | |
1274 | ||
21d82ac9 | 1275 | return 0; |
5b0959a7 KW |
1276 | |
1277 | fail: | |
1278 | qcow2_update_options_abort(state->bs, r); | |
1279 | g_free(r); | |
1280 | return ret; | |
1281 | } | |
1282 | ||
1283 | static void qcow2_reopen_commit(BDRVReopenState *state) | |
1284 | { | |
1285 | qcow2_update_options_commit(state->bs, state->opaque); | |
1286 | g_free(state->opaque); | |
1287 | } | |
1288 | ||
1289 | static void qcow2_reopen_abort(BDRVReopenState *state) | |
1290 | { | |
1291 | qcow2_update_options_abort(state->bs, state->opaque); | |
1292 | g_free(state->opaque); | |
21d82ac9 JC |
1293 | } |
1294 | ||
5365f44d KW |
1295 | static void qcow2_join_options(QDict *options, QDict *old_options) |
1296 | { | |
1297 | bool has_new_overlap_template = | |
1298 | qdict_haskey(options, QCOW2_OPT_OVERLAP) || | |
1299 | qdict_haskey(options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1300 | bool has_new_total_cache_size = | |
1301 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE); | |
1302 | bool has_all_cache_options; | |
1303 | ||
1304 | /* New overlap template overrides all old overlap options */ | |
1305 | if (has_new_overlap_template) { | |
1306 | qdict_del(old_options, QCOW2_OPT_OVERLAP); | |
1307 | qdict_del(old_options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1308 | qdict_del(old_options, QCOW2_OPT_OVERLAP_MAIN_HEADER); | |
1309 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L1); | |
1310 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L2); | |
1311 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_TABLE); | |
1312 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK); | |
1313 | qdict_del(old_options, QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE); | |
1314 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L1); | |
1315 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L2); | |
1316 | } | |
1317 | ||
1318 | /* New total cache size overrides all old options */ | |
1319 | if (qdict_haskey(options, QCOW2_OPT_CACHE_SIZE)) { | |
1320 | qdict_del(old_options, QCOW2_OPT_L2_CACHE_SIZE); | |
1321 | qdict_del(old_options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1322 | } | |
1323 | ||
1324 | qdict_join(options, old_options, false); | |
1325 | ||
1326 | /* | |
1327 | * If after merging all cache size options are set, an old total size is | |
1328 | * overwritten. Do keep all options, however, if all three are new. The | |
1329 | * resulting error message is what we want to happen. | |
1330 | */ | |
1331 | has_all_cache_options = | |
1332 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE) || | |
1333 | qdict_haskey(options, QCOW2_OPT_L2_CACHE_SIZE) || | |
1334 | qdict_haskey(options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1335 | ||
1336 | if (has_all_cache_options && !has_new_total_cache_size) { | |
1337 | qdict_del(options, QCOW2_OPT_CACHE_SIZE); | |
1338 | } | |
1339 | } | |
1340 | ||
b6b8a333 | 1341 | static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs, |
67a0fd2a | 1342 | int64_t sector_num, int nb_sectors, int *pnum, BlockDriverState **file) |
585f8587 | 1343 | { |
ff99129a | 1344 | BDRVQcow2State *s = bs->opaque; |
585f8587 | 1345 | uint64_t cluster_offset; |
4bc74be9 | 1346 | int index_in_cluster, ret; |
ecfe1863 | 1347 | unsigned int bytes; |
4bc74be9 | 1348 | int64_t status = 0; |
585f8587 | 1349 | |
ecfe1863 | 1350 | bytes = MIN(INT_MAX, nb_sectors * BDRV_SECTOR_SIZE); |
f8a2e5e3 | 1351 | qemu_co_mutex_lock(&s->lock); |
ecfe1863 KW |
1352 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, &bytes, |
1353 | &cluster_offset); | |
f8a2e5e3 | 1354 | qemu_co_mutex_unlock(&s->lock); |
1c46efaa | 1355 | if (ret < 0) { |
d663640c | 1356 | return ret; |
1c46efaa | 1357 | } |
095a9c58 | 1358 | |
ecfe1863 KW |
1359 | *pnum = bytes >> BDRV_SECTOR_BITS; |
1360 | ||
4bc74be9 | 1361 | if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED && |
f6fa64f6 | 1362 | !s->cipher) { |
4bc74be9 PB |
1363 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
1364 | cluster_offset |= (index_in_cluster << BDRV_SECTOR_BITS); | |
178b4db7 | 1365 | *file = bs->file->bs; |
4bc74be9 PB |
1366 | status |= BDRV_BLOCK_OFFSET_VALID | cluster_offset; |
1367 | } | |
1368 | if (ret == QCOW2_CLUSTER_ZERO) { | |
1369 | status |= BDRV_BLOCK_ZERO; | |
1370 | } else if (ret != QCOW2_CLUSTER_UNALLOCATED) { | |
1371 | status |= BDRV_BLOCK_DATA; | |
1372 | } | |
1373 | return status; | |
585f8587 FB |
1374 | } |
1375 | ||
a9465922 | 1376 | /* handle reading after the end of the backing file */ |
bd28f835 | 1377 | int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, |
ecfe1863 | 1378 | int64_t offset, int bytes) |
a9465922 | 1379 | { |
ecfe1863 | 1380 | uint64_t bs_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
a9465922 | 1381 | int n1; |
ecfe1863 KW |
1382 | |
1383 | if ((offset + bytes) <= bs_size) { | |
1384 | return bytes; | |
1385 | } | |
1386 | ||
1387 | if (offset >= bs_size) { | |
a9465922 | 1388 | n1 = 0; |
ecfe1863 KW |
1389 | } else { |
1390 | n1 = bs_size - offset; | |
1391 | } | |
bd28f835 | 1392 | |
ecfe1863 | 1393 | qemu_iovec_memset(qiov, n1, 0, bytes - n1); |
bd28f835 | 1394 | |
a9465922 FB |
1395 | return n1; |
1396 | } | |
1397 | ||
ecfe1863 KW |
1398 | static coroutine_fn int qcow2_co_preadv(BlockDriverState *bs, uint64_t offset, |
1399 | uint64_t bytes, QEMUIOVector *qiov, | |
1400 | int flags) | |
585f8587 | 1401 | { |
ff99129a | 1402 | BDRVQcow2State *s = bs->opaque; |
ecfe1863 | 1403 | int offset_in_cluster, n1; |
68d100e9 | 1404 | int ret; |
ecfe1863 | 1405 | unsigned int cur_bytes; /* number of bytes in current iteration */ |
c2bdd990 | 1406 | uint64_t cluster_offset = 0; |
3fc48d09 FZ |
1407 | uint64_t bytes_done = 0; |
1408 | QEMUIOVector hd_qiov; | |
1409 | uint8_t *cluster_data = NULL; | |
585f8587 | 1410 | |
3fc48d09 FZ |
1411 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1412 | ||
1413 | qemu_co_mutex_lock(&s->lock); | |
1414 | ||
ecfe1863 | 1415 | while (bytes != 0) { |
bd28f835 | 1416 | |
5ebaa27e | 1417 | /* prepare next request */ |
ecfe1863 | 1418 | cur_bytes = MIN(bytes, INT_MAX); |
f6fa64f6 | 1419 | if (s->cipher) { |
ecfe1863 KW |
1420 | cur_bytes = MIN(cur_bytes, |
1421 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); | |
585f8587 | 1422 | } |
5ebaa27e | 1423 | |
ecfe1863 | 1424 | ret = qcow2_get_cluster_offset(bs, offset, &cur_bytes, &cluster_offset); |
8af36488 | 1425 | if (ret < 0) { |
3fc48d09 | 1426 | goto fail; |
8af36488 | 1427 | } |
bd28f835 | 1428 | |
ecfe1863 | 1429 | offset_in_cluster = offset_into_cluster(s, offset); |
c87c0672 | 1430 | |
3fc48d09 | 1431 | qemu_iovec_reset(&hd_qiov); |
ecfe1863 | 1432 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes); |
5ebaa27e | 1433 | |
68d000a3 KW |
1434 | switch (ret) { |
1435 | case QCOW2_CLUSTER_UNALLOCATED: | |
5ebaa27e | 1436 | |
760e0063 | 1437 | if (bs->backing) { |
5ebaa27e | 1438 | /* read from the base image */ |
760e0063 | 1439 | n1 = qcow2_backing_read1(bs->backing->bs, &hd_qiov, |
ecfe1863 | 1440 | offset, cur_bytes); |
5ebaa27e | 1441 | if (n1 > 0) { |
44deba5a KW |
1442 | QEMUIOVector local_qiov; |
1443 | ||
1444 | qemu_iovec_init(&local_qiov, hd_qiov.niov); | |
ecfe1863 | 1445 | qemu_iovec_concat(&local_qiov, &hd_qiov, 0, n1); |
44deba5a | 1446 | |
5ebaa27e FZ |
1447 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); |
1448 | qemu_co_mutex_unlock(&s->lock); | |
a03ef88f | 1449 | ret = bdrv_co_preadv(bs->backing, offset, n1, |
ecfe1863 | 1450 | &local_qiov, 0); |
5ebaa27e | 1451 | qemu_co_mutex_lock(&s->lock); |
44deba5a KW |
1452 | |
1453 | qemu_iovec_destroy(&local_qiov); | |
1454 | ||
5ebaa27e | 1455 | if (ret < 0) { |
3fc48d09 | 1456 | goto fail; |
5ebaa27e FZ |
1457 | } |
1458 | } | |
1459 | } else { | |
1460 | /* Note: in this case, no need to wait */ | |
ecfe1863 | 1461 | qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes); |
5ebaa27e | 1462 | } |
68d000a3 KW |
1463 | break; |
1464 | ||
6377af48 | 1465 | case QCOW2_CLUSTER_ZERO: |
ecfe1863 | 1466 | qemu_iovec_memset(&hd_qiov, 0, 0, cur_bytes); |
6377af48 KW |
1467 | break; |
1468 | ||
68d000a3 | 1469 | case QCOW2_CLUSTER_COMPRESSED: |
5ebaa27e FZ |
1470 | /* add AIO support for compressed blocks ? */ |
1471 | ret = qcow2_decompress_cluster(bs, cluster_offset); | |
1472 | if (ret < 0) { | |
3fc48d09 | 1473 | goto fail; |
bd28f835 KW |
1474 | } |
1475 | ||
03396148 | 1476 | qemu_iovec_from_buf(&hd_qiov, 0, |
ecfe1863 KW |
1477 | s->cluster_cache + offset_in_cluster, |
1478 | cur_bytes); | |
68d000a3 KW |
1479 | break; |
1480 | ||
1481 | case QCOW2_CLUSTER_NORMAL: | |
5ebaa27e | 1482 | if ((cluster_offset & 511) != 0) { |
3fc48d09 FZ |
1483 | ret = -EIO; |
1484 | goto fail; | |
5ebaa27e | 1485 | } |
bd28f835 | 1486 | |
8336aafa | 1487 | if (bs->encrypted) { |
f6fa64f6 | 1488 | assert(s->cipher); |
8336aafa | 1489 | |
5ebaa27e FZ |
1490 | /* |
1491 | * For encrypted images, read everything into a temporary | |
1492 | * contiguous buffer on which the AES functions can work. | |
1493 | */ | |
3fc48d09 FZ |
1494 | if (!cluster_data) { |
1495 | cluster_data = | |
9a4f4c31 KW |
1496 | qemu_try_blockalign(bs->file->bs, |
1497 | QCOW_MAX_CRYPT_CLUSTERS | |
1498 | * s->cluster_size); | |
de82815d KW |
1499 | if (cluster_data == NULL) { |
1500 | ret = -ENOMEM; | |
1501 | goto fail; | |
1502 | } | |
5ebaa27e FZ |
1503 | } |
1504 | ||
ecfe1863 | 1505 | assert(cur_bytes <= QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
3fc48d09 | 1506 | qemu_iovec_reset(&hd_qiov); |
ecfe1863 | 1507 | qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes); |
5ebaa27e FZ |
1508 | } |
1509 | ||
1510 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
1511 | qemu_co_mutex_unlock(&s->lock); | |
a03ef88f | 1512 | ret = bdrv_co_preadv(bs->file, |
ecfe1863 KW |
1513 | cluster_offset + offset_in_cluster, |
1514 | cur_bytes, &hd_qiov, 0); | |
5ebaa27e FZ |
1515 | qemu_co_mutex_lock(&s->lock); |
1516 | if (ret < 0) { | |
3fc48d09 | 1517 | goto fail; |
5ebaa27e | 1518 | } |
8336aafa | 1519 | if (bs->encrypted) { |
f6fa64f6 | 1520 | assert(s->cipher); |
ecfe1863 KW |
1521 | assert((offset & (BDRV_SECTOR_SIZE - 1)) == 0); |
1522 | assert((cur_bytes & (BDRV_SECTOR_SIZE - 1)) == 0); | |
f6fa64f6 | 1523 | Error *err = NULL; |
ecfe1863 KW |
1524 | if (qcow2_encrypt_sectors(s, offset >> BDRV_SECTOR_BITS, |
1525 | cluster_data, cluster_data, | |
1526 | cur_bytes >> BDRV_SECTOR_BITS, | |
1527 | false, &err) < 0) { | |
f6fa64f6 DB |
1528 | error_free(err); |
1529 | ret = -EIO; | |
1530 | goto fail; | |
1531 | } | |
ecfe1863 | 1532 | qemu_iovec_from_buf(qiov, bytes_done, cluster_data, cur_bytes); |
5ebaa27e | 1533 | } |
68d000a3 KW |
1534 | break; |
1535 | ||
1536 | default: | |
1537 | g_assert_not_reached(); | |
1538 | ret = -EIO; | |
1539 | goto fail; | |
faf575c1 | 1540 | } |
f141eafe | 1541 | |
ecfe1863 KW |
1542 | bytes -= cur_bytes; |
1543 | offset += cur_bytes; | |
1544 | bytes_done += cur_bytes; | |
5ebaa27e | 1545 | } |
3fc48d09 | 1546 | ret = 0; |
faf575c1 | 1547 | |
3fc48d09 | 1548 | fail: |
68d100e9 | 1549 | qemu_co_mutex_unlock(&s->lock); |
42496d62 | 1550 | |
3fc48d09 | 1551 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1552 | qemu_vfree(cluster_data); |
68d100e9 KW |
1553 | |
1554 | return ret; | |
585f8587 FB |
1555 | } |
1556 | ||
d46a0bb2 KW |
1557 | static coroutine_fn int qcow2_co_pwritev(BlockDriverState *bs, uint64_t offset, |
1558 | uint64_t bytes, QEMUIOVector *qiov, | |
1559 | int flags) | |
585f8587 | 1560 | { |
ff99129a | 1561 | BDRVQcow2State *s = bs->opaque; |
d46a0bb2 | 1562 | int offset_in_cluster; |
68d100e9 | 1563 | int ret; |
d46a0bb2 | 1564 | unsigned int cur_bytes; /* number of sectors in current iteration */ |
c2bdd990 | 1565 | uint64_t cluster_offset; |
3fc48d09 FZ |
1566 | QEMUIOVector hd_qiov; |
1567 | uint64_t bytes_done = 0; | |
1568 | uint8_t *cluster_data = NULL; | |
8d2497c3 | 1569 | QCowL2Meta *l2meta = NULL; |
c2271403 | 1570 | |
d46a0bb2 | 1571 | trace_qcow2_writev_start_req(qemu_coroutine_self(), offset, bytes); |
3cce16f4 | 1572 | |
3fc48d09 FZ |
1573 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1574 | ||
1575 | s->cluster_cache_offset = -1; /* disable compressed cache */ | |
3b46e624 | 1576 | |
3fc48d09 FZ |
1577 | qemu_co_mutex_lock(&s->lock); |
1578 | ||
d46a0bb2 | 1579 | while (bytes != 0) { |
3fc48d09 | 1580 | |
f50f88b9 | 1581 | l2meta = NULL; |
cf5c1a23 | 1582 | |
3cce16f4 | 1583 | trace_qcow2_writev_start_part(qemu_coroutine_self()); |
d46a0bb2 KW |
1584 | offset_in_cluster = offset_into_cluster(s, offset); |
1585 | cur_bytes = MIN(bytes, INT_MAX); | |
1586 | if (bs->encrypted) { | |
1587 | cur_bytes = MIN(cur_bytes, | |
1588 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size | |
1589 | - offset_in_cluster); | |
5ebaa27e | 1590 | } |
095a9c58 | 1591 | |
d46a0bb2 KW |
1592 | ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes, |
1593 | &cluster_offset, &l2meta); | |
5ebaa27e | 1594 | if (ret < 0) { |
3fc48d09 | 1595 | goto fail; |
5ebaa27e | 1596 | } |
148da7ea | 1597 | |
5ebaa27e | 1598 | assert((cluster_offset & 511) == 0); |
148da7ea | 1599 | |
3fc48d09 | 1600 | qemu_iovec_reset(&hd_qiov); |
d46a0bb2 | 1601 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, cur_bytes); |
6f5f060b | 1602 | |
8336aafa | 1603 | if (bs->encrypted) { |
f6fa64f6 DB |
1604 | Error *err = NULL; |
1605 | assert(s->cipher); | |
3fc48d09 | 1606 | if (!cluster_data) { |
9a4f4c31 | 1607 | cluster_data = qemu_try_blockalign(bs->file->bs, |
de82815d KW |
1608 | QCOW_MAX_CRYPT_CLUSTERS |
1609 | * s->cluster_size); | |
1610 | if (cluster_data == NULL) { | |
1611 | ret = -ENOMEM; | |
1612 | goto fail; | |
1613 | } | |
5ebaa27e | 1614 | } |
6f5f060b | 1615 | |
3fc48d09 | 1616 | assert(hd_qiov.size <= |
5ebaa27e | 1617 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
d5e6b161 | 1618 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); |
6f5f060b | 1619 | |
d46a0bb2 KW |
1620 | if (qcow2_encrypt_sectors(s, offset >> BDRV_SECTOR_BITS, |
1621 | cluster_data, cluster_data, | |
1622 | cur_bytes >>BDRV_SECTOR_BITS, | |
f6fa64f6 DB |
1623 | true, &err) < 0) { |
1624 | error_free(err); | |
1625 | ret = -EIO; | |
1626 | goto fail; | |
1627 | } | |
6f5f060b | 1628 | |
3fc48d09 | 1629 | qemu_iovec_reset(&hd_qiov); |
d46a0bb2 | 1630 | qemu_iovec_add(&hd_qiov, cluster_data, cur_bytes); |
5ebaa27e | 1631 | } |
6f5f060b | 1632 | |
231bb267 | 1633 | ret = qcow2_pre_write_overlap_check(bs, 0, |
d46a0bb2 | 1634 | cluster_offset + offset_in_cluster, cur_bytes); |
cf93980e HR |
1635 | if (ret < 0) { |
1636 | goto fail; | |
1637 | } | |
1638 | ||
5ebaa27e | 1639 | qemu_co_mutex_unlock(&s->lock); |
67a7a0eb | 1640 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); |
3cce16f4 | 1641 | trace_qcow2_writev_data(qemu_coroutine_self(), |
d46a0bb2 | 1642 | cluster_offset + offset_in_cluster); |
a03ef88f | 1643 | ret = bdrv_co_pwritev(bs->file, |
d46a0bb2 KW |
1644 | cluster_offset + offset_in_cluster, |
1645 | cur_bytes, &hd_qiov, 0); | |
5ebaa27e FZ |
1646 | qemu_co_mutex_lock(&s->lock); |
1647 | if (ret < 0) { | |
3fc48d09 | 1648 | goto fail; |
5ebaa27e | 1649 | } |
f141eafe | 1650 | |
88c6588c KW |
1651 | while (l2meta != NULL) { |
1652 | QCowL2Meta *next; | |
1653 | ||
f50f88b9 KW |
1654 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); |
1655 | if (ret < 0) { | |
1656 | goto fail; | |
1657 | } | |
faf575c1 | 1658 | |
4e95314e KW |
1659 | /* Take the request off the list of running requests */ |
1660 | if (l2meta->nb_clusters != 0) { | |
1661 | QLIST_REMOVE(l2meta, next_in_flight); | |
1662 | } | |
1663 | ||
4e95314e | 1664 | qemu_co_queue_restart_all(&l2meta->dependent_requests); |
4e95314e | 1665 | |
88c6588c | 1666 | next = l2meta->next; |
f50f88b9 | 1667 | g_free(l2meta); |
88c6588c | 1668 | l2meta = next; |
f50f88b9 | 1669 | } |
0fa9131a | 1670 | |
d46a0bb2 KW |
1671 | bytes -= cur_bytes; |
1672 | offset += cur_bytes; | |
1673 | bytes_done += cur_bytes; | |
1674 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_bytes); | |
5ebaa27e | 1675 | } |
3fc48d09 | 1676 | ret = 0; |
faf575c1 | 1677 | |
3fc48d09 | 1678 | fail: |
4e95314e KW |
1679 | qemu_co_mutex_unlock(&s->lock); |
1680 | ||
88c6588c KW |
1681 | while (l2meta != NULL) { |
1682 | QCowL2Meta *next; | |
1683 | ||
4e95314e KW |
1684 | if (l2meta->nb_clusters != 0) { |
1685 | QLIST_REMOVE(l2meta, next_in_flight); | |
1686 | } | |
1687 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
88c6588c KW |
1688 | |
1689 | next = l2meta->next; | |
cf5c1a23 | 1690 | g_free(l2meta); |
88c6588c | 1691 | l2meta = next; |
cf5c1a23 | 1692 | } |
0fa9131a | 1693 | |
3fc48d09 | 1694 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1695 | qemu_vfree(cluster_data); |
3cce16f4 | 1696 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
42496d62 | 1697 | |
68d100e9 | 1698 | return ret; |
585f8587 FB |
1699 | } |
1700 | ||
ec6d8912 KW |
1701 | static int qcow2_inactivate(BlockDriverState *bs) |
1702 | { | |
1703 | BDRVQcow2State *s = bs->opaque; | |
1704 | int ret, result = 0; | |
1705 | ||
1706 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
1707 | if (ret) { | |
1708 | result = ret; | |
1709 | error_report("Failed to flush the L2 table cache: %s", | |
1710 | strerror(-ret)); | |
1711 | } | |
1712 | ||
1713 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
1714 | if (ret) { | |
1715 | result = ret; | |
1716 | error_report("Failed to flush the refcount block cache: %s", | |
1717 | strerror(-ret)); | |
1718 | } | |
1719 | ||
1720 | if (result == 0) { | |
1721 | qcow2_mark_clean(bs); | |
1722 | } | |
1723 | ||
1724 | return result; | |
1725 | } | |
1726 | ||
7c80ab3f | 1727 | static void qcow2_close(BlockDriverState *bs) |
585f8587 | 1728 | { |
ff99129a | 1729 | BDRVQcow2State *s = bs->opaque; |
de82815d | 1730 | qemu_vfree(s->l1_table); |
cf93980e HR |
1731 | /* else pre-write overlap checks in cache_destroy may crash */ |
1732 | s->l1_table = NULL; | |
29c1a730 | 1733 | |
140fd5a6 | 1734 | if (!(s->flags & BDRV_O_INACTIVE)) { |
ec6d8912 | 1735 | qcow2_inactivate(bs); |
27eb6c09 | 1736 | } |
c61d0004 | 1737 | |
279621c0 | 1738 | cache_clean_timer_del(bs); |
29c1a730 KW |
1739 | qcow2_cache_destroy(bs, s->l2_table_cache); |
1740 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1741 | ||
f6fa64f6 DB |
1742 | qcrypto_cipher_free(s->cipher); |
1743 | s->cipher = NULL; | |
1744 | ||
6744cbab | 1745 | g_free(s->unknown_header_fields); |
75bab85c | 1746 | cleanup_unknown_header_ext(bs); |
6744cbab | 1747 | |
e4603fe1 KW |
1748 | g_free(s->image_backing_file); |
1749 | g_free(s->image_backing_format); | |
1750 | ||
7267c094 | 1751 | g_free(s->cluster_cache); |
dea43a65 | 1752 | qemu_vfree(s->cluster_data); |
ed6ccf0f | 1753 | qcow2_refcount_close(bs); |
28c1202b | 1754 | qcow2_free_snapshots(bs); |
585f8587 FB |
1755 | } |
1756 | ||
5a8a30db | 1757 | static void qcow2_invalidate_cache(BlockDriverState *bs, Error **errp) |
06d9260f | 1758 | { |
ff99129a | 1759 | BDRVQcow2State *s = bs->opaque; |
06d9260f | 1760 | int flags = s->flags; |
f6fa64f6 | 1761 | QCryptoCipher *cipher = NULL; |
acdfb480 | 1762 | QDict *options; |
5a8a30db KW |
1763 | Error *local_err = NULL; |
1764 | int ret; | |
06d9260f AL |
1765 | |
1766 | /* | |
1767 | * Backing files are read-only which makes all of their metadata immutable, | |
1768 | * that means we don't have to worry about reopening them here. | |
1769 | */ | |
1770 | ||
f6fa64f6 DB |
1771 | cipher = s->cipher; |
1772 | s->cipher = NULL; | |
06d9260f AL |
1773 | |
1774 | qcow2_close(bs); | |
1775 | ||
ff99129a | 1776 | memset(s, 0, sizeof(BDRVQcow2State)); |
d475e5ac | 1777 | options = qdict_clone_shallow(bs->options); |
5a8a30db | 1778 | |
140fd5a6 | 1779 | flags &= ~BDRV_O_INACTIVE; |
5a8a30db | 1780 | ret = qcow2_open(bs, options, flags, &local_err); |
a1904e48 | 1781 | QDECREF(options); |
5a8a30db | 1782 | if (local_err) { |
e43bfd9c MA |
1783 | error_propagate(errp, local_err); |
1784 | error_prepend(errp, "Could not reopen qcow2 layer: "); | |
191fb11b | 1785 | bs->drv = NULL; |
5a8a30db KW |
1786 | return; |
1787 | } else if (ret < 0) { | |
1788 | error_setg_errno(errp, -ret, "Could not reopen qcow2 layer"); | |
191fb11b | 1789 | bs->drv = NULL; |
5a8a30db KW |
1790 | return; |
1791 | } | |
acdfb480 | 1792 | |
f6fa64f6 | 1793 | s->cipher = cipher; |
06d9260f AL |
1794 | } |
1795 | ||
e24e49e6 KW |
1796 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
1797 | size_t len, size_t buflen) | |
1798 | { | |
1799 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
1800 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
1801 | ||
1802 | if (buflen < ext_len) { | |
1803 | return -ENOSPC; | |
1804 | } | |
1805 | ||
1806 | *ext_backing_fmt = (QCowExtension) { | |
1807 | .magic = cpu_to_be32(magic), | |
1808 | .len = cpu_to_be32(len), | |
1809 | }; | |
0647d47c SH |
1810 | |
1811 | if (len) { | |
1812 | memcpy(buf + sizeof(QCowExtension), s, len); | |
1813 | } | |
e24e49e6 KW |
1814 | |
1815 | return ext_len; | |
1816 | } | |
1817 | ||
756e6736 | 1818 | /* |
e24e49e6 KW |
1819 | * Updates the qcow2 header, including the variable length parts of it, i.e. |
1820 | * the backing file name and all extensions. qcow2 was not designed to allow | |
1821 | * such changes, so if we run out of space (we can only use the first cluster) | |
1822 | * this function may fail. | |
756e6736 KW |
1823 | * |
1824 | * Returns 0 on success, -errno in error cases. | |
1825 | */ | |
e24e49e6 | 1826 | int qcow2_update_header(BlockDriverState *bs) |
756e6736 | 1827 | { |
ff99129a | 1828 | BDRVQcow2State *s = bs->opaque; |
e24e49e6 KW |
1829 | QCowHeader *header; |
1830 | char *buf; | |
1831 | size_t buflen = s->cluster_size; | |
756e6736 | 1832 | int ret; |
e24e49e6 KW |
1833 | uint64_t total_size; |
1834 | uint32_t refcount_table_clusters; | |
6744cbab | 1835 | size_t header_length; |
75bab85c | 1836 | Qcow2UnknownHeaderExtension *uext; |
756e6736 | 1837 | |
e24e49e6 | 1838 | buf = qemu_blockalign(bs, buflen); |
756e6736 | 1839 | |
e24e49e6 KW |
1840 | /* Header structure */ |
1841 | header = (QCowHeader*) buf; | |
756e6736 | 1842 | |
e24e49e6 KW |
1843 | if (buflen < sizeof(*header)) { |
1844 | ret = -ENOSPC; | |
1845 | goto fail; | |
756e6736 KW |
1846 | } |
1847 | ||
6744cbab | 1848 | header_length = sizeof(*header) + s->unknown_header_fields_size; |
e24e49e6 KW |
1849 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
1850 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
1851 | ||
1852 | *header = (QCowHeader) { | |
6744cbab | 1853 | /* Version 2 fields */ |
e24e49e6 | 1854 | .magic = cpu_to_be32(QCOW_MAGIC), |
6744cbab | 1855 | .version = cpu_to_be32(s->qcow_version), |
e24e49e6 KW |
1856 | .backing_file_offset = 0, |
1857 | .backing_file_size = 0, | |
1858 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
1859 | .size = cpu_to_be64(total_size), | |
1860 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
1861 | .l1_size = cpu_to_be32(s->l1_size), | |
1862 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
1863 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
1864 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
1865 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
1866 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
6744cbab KW |
1867 | |
1868 | /* Version 3 fields */ | |
1869 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
1870 | .compatible_features = cpu_to_be64(s->compatible_features), | |
1871 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
b6481f37 | 1872 | .refcount_order = cpu_to_be32(s->refcount_order), |
6744cbab | 1873 | .header_length = cpu_to_be32(header_length), |
e24e49e6 | 1874 | }; |
756e6736 | 1875 | |
6744cbab KW |
1876 | /* For older versions, write a shorter header */ |
1877 | switch (s->qcow_version) { | |
1878 | case 2: | |
1879 | ret = offsetof(QCowHeader, incompatible_features); | |
1880 | break; | |
1881 | case 3: | |
1882 | ret = sizeof(*header); | |
1883 | break; | |
1884 | default: | |
b6c14762 JM |
1885 | ret = -EINVAL; |
1886 | goto fail; | |
6744cbab KW |
1887 | } |
1888 | ||
1889 | buf += ret; | |
1890 | buflen -= ret; | |
1891 | memset(buf, 0, buflen); | |
1892 | ||
1893 | /* Preserve any unknown field in the header */ | |
1894 | if (s->unknown_header_fields_size) { | |
1895 | if (buflen < s->unknown_header_fields_size) { | |
1896 | ret = -ENOSPC; | |
1897 | goto fail; | |
1898 | } | |
1899 | ||
1900 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
1901 | buf += s->unknown_header_fields_size; | |
1902 | buflen -= s->unknown_header_fields_size; | |
1903 | } | |
756e6736 | 1904 | |
e24e49e6 | 1905 | /* Backing file format header extension */ |
e4603fe1 | 1906 | if (s->image_backing_format) { |
e24e49e6 | 1907 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, |
e4603fe1 KW |
1908 | s->image_backing_format, |
1909 | strlen(s->image_backing_format), | |
e24e49e6 KW |
1910 | buflen); |
1911 | if (ret < 0) { | |
1912 | goto fail; | |
756e6736 KW |
1913 | } |
1914 | ||
e24e49e6 KW |
1915 | buf += ret; |
1916 | buflen -= ret; | |
756e6736 KW |
1917 | } |
1918 | ||
cfcc4c62 | 1919 | /* Feature table */ |
1a4828c7 KW |
1920 | if (s->qcow_version >= 3) { |
1921 | Qcow2Feature features[] = { | |
1922 | { | |
1923 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1924 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
1925 | .name = "dirty bit", | |
1926 | }, | |
1927 | { | |
1928 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1929 | .bit = QCOW2_INCOMPAT_CORRUPT_BITNR, | |
1930 | .name = "corrupt bit", | |
1931 | }, | |
1932 | { | |
1933 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
1934 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
1935 | .name = "lazy refcounts", | |
1936 | }, | |
1937 | }; | |
1938 | ||
1939 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
1940 | features, sizeof(features), buflen); | |
1941 | if (ret < 0) { | |
1942 | goto fail; | |
1943 | } | |
1944 | buf += ret; | |
1945 | buflen -= ret; | |
cfcc4c62 | 1946 | } |
cfcc4c62 | 1947 | |
75bab85c KW |
1948 | /* Keep unknown header extensions */ |
1949 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
1950 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
1951 | if (ret < 0) { | |
1952 | goto fail; | |
1953 | } | |
1954 | ||
1955 | buf += ret; | |
1956 | buflen -= ret; | |
1957 | } | |
1958 | ||
e24e49e6 KW |
1959 | /* End of header extensions */ |
1960 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
756e6736 KW |
1961 | if (ret < 0) { |
1962 | goto fail; | |
1963 | } | |
1964 | ||
e24e49e6 KW |
1965 | buf += ret; |
1966 | buflen -= ret; | |
756e6736 | 1967 | |
e24e49e6 | 1968 | /* Backing file name */ |
e4603fe1 KW |
1969 | if (s->image_backing_file) { |
1970 | size_t backing_file_len = strlen(s->image_backing_file); | |
e24e49e6 KW |
1971 | |
1972 | if (buflen < backing_file_len) { | |
1973 | ret = -ENOSPC; | |
1974 | goto fail; | |
1975 | } | |
1976 | ||
00ea1881 | 1977 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ |
e4603fe1 | 1978 | strncpy(buf, s->image_backing_file, buflen); |
e24e49e6 KW |
1979 | |
1980 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
1981 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
756e6736 KW |
1982 | } |
1983 | ||
e24e49e6 | 1984 | /* Write the new header */ |
d9ca2ea2 | 1985 | ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size); |
756e6736 KW |
1986 | if (ret < 0) { |
1987 | goto fail; | |
1988 | } | |
1989 | ||
1990 | ret = 0; | |
1991 | fail: | |
e24e49e6 | 1992 | qemu_vfree(header); |
756e6736 KW |
1993 | return ret; |
1994 | } | |
1995 | ||
1996 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
1997 | const char *backing_file, const char *backing_fmt) | |
1998 | { | |
ff99129a | 1999 | BDRVQcow2State *s = bs->opaque; |
e4603fe1 | 2000 | |
4e876bcf HR |
2001 | if (backing_file && strlen(backing_file) > 1023) { |
2002 | return -EINVAL; | |
2003 | } | |
2004 | ||
e24e49e6 KW |
2005 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
2006 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
2007 | ||
e4603fe1 KW |
2008 | g_free(s->image_backing_file); |
2009 | g_free(s->image_backing_format); | |
2010 | ||
2011 | s->image_backing_file = backing_file ? g_strdup(bs->backing_file) : NULL; | |
2012 | s->image_backing_format = backing_fmt ? g_strdup(bs->backing_format) : NULL; | |
2013 | ||
e24e49e6 | 2014 | return qcow2_update_header(bs); |
756e6736 KW |
2015 | } |
2016 | ||
a35e1c17 KW |
2017 | static int preallocate(BlockDriverState *bs) |
2018 | { | |
d46a0bb2 | 2019 | uint64_t bytes; |
a35e1c17 | 2020 | uint64_t offset; |
060bee89 | 2021 | uint64_t host_offset = 0; |
d46a0bb2 | 2022 | unsigned int cur_bytes; |
148da7ea | 2023 | int ret; |
f50f88b9 | 2024 | QCowL2Meta *meta; |
a35e1c17 | 2025 | |
d46a0bb2 | 2026 | bytes = bdrv_getlength(bs); |
a35e1c17 KW |
2027 | offset = 0; |
2028 | ||
d46a0bb2 KW |
2029 | while (bytes) { |
2030 | cur_bytes = MIN(bytes, INT_MAX); | |
2031 | ret = qcow2_alloc_cluster_offset(bs, offset, &cur_bytes, | |
060bee89 | 2032 | &host_offset, &meta); |
148da7ea | 2033 | if (ret < 0) { |
19dbcbf7 | 2034 | return ret; |
a35e1c17 KW |
2035 | } |
2036 | ||
c792707f SH |
2037 | while (meta) { |
2038 | QCowL2Meta *next = meta->next; | |
2039 | ||
7c2bbf4a HT |
2040 | ret = qcow2_alloc_cluster_link_l2(bs, meta); |
2041 | if (ret < 0) { | |
2042 | qcow2_free_any_clusters(bs, meta->alloc_offset, | |
2043 | meta->nb_clusters, QCOW2_DISCARD_NEVER); | |
2044 | return ret; | |
2045 | } | |
2046 | ||
2047 | /* There are no dependent requests, but we need to remove our | |
2048 | * request from the list of in-flight requests */ | |
4e95314e | 2049 | QLIST_REMOVE(meta, next_in_flight); |
c792707f SH |
2050 | |
2051 | g_free(meta); | |
2052 | meta = next; | |
f50f88b9 | 2053 | } |
f214978a | 2054 | |
a35e1c17 KW |
2055 | /* TODO Preallocate data if requested */ |
2056 | ||
d46a0bb2 KW |
2057 | bytes -= cur_bytes; |
2058 | offset += cur_bytes; | |
a35e1c17 KW |
2059 | } |
2060 | ||
2061 | /* | |
2062 | * It is expected that the image file is large enough to actually contain | |
2063 | * all of the allocated clusters (otherwise we get failing reads after | |
2064 | * EOF). Extend the image to the last allocated sector. | |
2065 | */ | |
060bee89 | 2066 | if (host_offset != 0) { |
d46a0bb2 | 2067 | uint8_t data = 0; |
d9ca2ea2 | 2068 | ret = bdrv_pwrite(bs->file, (host_offset + cur_bytes) - 1, |
d46a0bb2 | 2069 | &data, 1); |
19dbcbf7 KW |
2070 | if (ret < 0) { |
2071 | return ret; | |
2072 | } | |
a35e1c17 KW |
2073 | } |
2074 | ||
2075 | return 0; | |
2076 | } | |
2077 | ||
7c80ab3f JS |
2078 | static int qcow2_create2(const char *filename, int64_t total_size, |
2079 | const char *backing_file, const char *backing_format, | |
ffeaac9b | 2080 | int flags, size_t cluster_size, PreallocMode prealloc, |
bd4b167f | 2081 | QemuOpts *opts, int version, int refcount_order, |
3ef6c40a | 2082 | Error **errp) |
a9420734 | 2083 | { |
a9420734 | 2084 | int cluster_bits; |
e6641719 HR |
2085 | QDict *options; |
2086 | ||
2087 | /* Calculate cluster_bits */ | |
786a4ea8 | 2088 | cluster_bits = ctz32(cluster_size); |
a9420734 KW |
2089 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || |
2090 | (1 << cluster_bits) != cluster_size) | |
2091 | { | |
3ef6c40a HR |
2092 | error_setg(errp, "Cluster size must be a power of two between %d and " |
2093 | "%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); | |
a9420734 KW |
2094 | return -EINVAL; |
2095 | } | |
2096 | ||
2097 | /* | |
2098 | * Open the image file and write a minimal qcow2 header. | |
2099 | * | |
2100 | * We keep things simple and start with a zero-sized image. We also | |
2101 | * do without refcount blocks or a L1 table for now. We'll fix the | |
2102 | * inconsistency later. | |
2103 | * | |
2104 | * We do need a refcount table because growing the refcount table means | |
2105 | * allocating two new refcount blocks - the seconds of which would be at | |
2106 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
2107 | * size for any qcow2 image. | |
2108 | */ | |
23588797 | 2109 | BlockBackend *blk; |
f8413b3c | 2110 | QCowHeader *header; |
b106ad91 | 2111 | uint64_t* refcount_table; |
3ef6c40a | 2112 | Error *local_err = NULL; |
a9420734 KW |
2113 | int ret; |
2114 | ||
0e4271b7 | 2115 | if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) { |
bd4b167f HR |
2116 | /* Note: The following calculation does not need to be exact; if it is a |
2117 | * bit off, either some bytes will be "leaked" (which is fine) or we | |
2118 | * will need to increase the file size by some bytes (which is fine, | |
2119 | * too, as long as the bulk is allocated here). Therefore, using | |
2120 | * floating point arithmetic is fine. */ | |
0e4271b7 HT |
2121 | int64_t meta_size = 0; |
2122 | uint64_t nreftablee, nrefblocke, nl1e, nl2e; | |
2123 | int64_t aligned_total_size = align_offset(total_size, cluster_size); | |
bd4b167f HR |
2124 | int refblock_bits, refblock_size; |
2125 | /* refcount entry size in bytes */ | |
2126 | double rces = (1 << refcount_order) / 8.; | |
2127 | ||
2128 | /* see qcow2_open() */ | |
2129 | refblock_bits = cluster_bits - (refcount_order - 3); | |
2130 | refblock_size = 1 << refblock_bits; | |
0e4271b7 HT |
2131 | |
2132 | /* header: 1 cluster */ | |
2133 | meta_size += cluster_size; | |
2134 | ||
2135 | /* total size of L2 tables */ | |
2136 | nl2e = aligned_total_size / cluster_size; | |
2137 | nl2e = align_offset(nl2e, cluster_size / sizeof(uint64_t)); | |
2138 | meta_size += nl2e * sizeof(uint64_t); | |
2139 | ||
2140 | /* total size of L1 tables */ | |
2141 | nl1e = nl2e * sizeof(uint64_t) / cluster_size; | |
2142 | nl1e = align_offset(nl1e, cluster_size / sizeof(uint64_t)); | |
2143 | meta_size += nl1e * sizeof(uint64_t); | |
2144 | ||
2145 | /* total size of refcount blocks | |
2146 | * | |
2147 | * note: every host cluster is reference-counted, including metadata | |
2148 | * (even refcount blocks are recursively included). | |
2149 | * Let: | |
2150 | * a = total_size (this is the guest disk size) | |
2151 | * m = meta size not including refcount blocks and refcount tables | |
2152 | * c = cluster size | |
2153 | * y1 = number of refcount blocks entries | |
2154 | * y2 = meta size including everything | |
bd4b167f | 2155 | * rces = refcount entry size in bytes |
0e4271b7 HT |
2156 | * then, |
2157 | * y1 = (y2 + a)/c | |
bd4b167f | 2158 | * y2 = y1 * rces + y1 * rces * sizeof(u64) / c + m |
0e4271b7 | 2159 | * we can get y1: |
bd4b167f | 2160 | * y1 = (a + m) / (c - rces - rces * sizeof(u64) / c) |
0e4271b7 | 2161 | */ |
bd4b167f HR |
2162 | nrefblocke = (aligned_total_size + meta_size + cluster_size) |
2163 | / (cluster_size - rces - rces * sizeof(uint64_t) | |
2164 | / cluster_size); | |
2165 | meta_size += DIV_ROUND_UP(nrefblocke, refblock_size) * cluster_size; | |
0e4271b7 HT |
2166 | |
2167 | /* total size of refcount tables */ | |
bd4b167f | 2168 | nreftablee = nrefblocke / refblock_size; |
0e4271b7 HT |
2169 | nreftablee = align_offset(nreftablee, cluster_size / sizeof(uint64_t)); |
2170 | meta_size += nreftablee * sizeof(uint64_t); | |
2171 | ||
2172 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, | |
39101f25 | 2173 | aligned_total_size + meta_size, &error_abort); |
f43e47db MA |
2174 | qemu_opt_set(opts, BLOCK_OPT_PREALLOC, PreallocMode_lookup[prealloc], |
2175 | &error_abort); | |
0e4271b7 HT |
2176 | } |
2177 | ||
c282e1fd | 2178 | ret = bdrv_create_file(filename, opts, &local_err); |
a9420734 | 2179 | if (ret < 0) { |
3ef6c40a | 2180 | error_propagate(errp, local_err); |
a9420734 KW |
2181 | return ret; |
2182 | } | |
2183 | ||
efaa7c4e | 2184 | blk = blk_new_open(filename, NULL, NULL, |
72e775c7 | 2185 | BDRV_O_RDWR | BDRV_O_PROTOCOL, &local_err); |
23588797 | 2186 | if (blk == NULL) { |
3ef6c40a | 2187 | error_propagate(errp, local_err); |
23588797 | 2188 | return -EIO; |
a9420734 KW |
2189 | } |
2190 | ||
23588797 KW |
2191 | blk_set_allow_write_beyond_eof(blk, true); |
2192 | ||
a9420734 | 2193 | /* Write the header */ |
f8413b3c KW |
2194 | QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header)); |
2195 | header = g_malloc0(cluster_size); | |
2196 | *header = (QCowHeader) { | |
2197 | .magic = cpu_to_be32(QCOW_MAGIC), | |
2198 | .version = cpu_to_be32(version), | |
2199 | .cluster_bits = cpu_to_be32(cluster_bits), | |
2200 | .size = cpu_to_be64(0), | |
2201 | .l1_table_offset = cpu_to_be64(0), | |
2202 | .l1_size = cpu_to_be32(0), | |
2203 | .refcount_table_offset = cpu_to_be64(cluster_size), | |
2204 | .refcount_table_clusters = cpu_to_be32(1), | |
bd4b167f | 2205 | .refcount_order = cpu_to_be32(refcount_order), |
f8413b3c KW |
2206 | .header_length = cpu_to_be32(sizeof(*header)), |
2207 | }; | |
a9420734 KW |
2208 | |
2209 | if (flags & BLOCK_FLAG_ENCRYPT) { | |
f8413b3c | 2210 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_AES); |
a9420734 | 2211 | } else { |
f8413b3c | 2212 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); |
a9420734 KW |
2213 | } |
2214 | ||
bfe8043e | 2215 | if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) { |
f8413b3c | 2216 | header->compatible_features |= |
bfe8043e SH |
2217 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); |
2218 | } | |
2219 | ||
8341f00d | 2220 | ret = blk_pwrite(blk, 0, header, cluster_size, 0); |
f8413b3c | 2221 | g_free(header); |
a9420734 | 2222 | if (ret < 0) { |
3ef6c40a | 2223 | error_setg_errno(errp, -ret, "Could not write qcow2 header"); |
a9420734 KW |
2224 | goto out; |
2225 | } | |
2226 | ||
b106ad91 KW |
2227 | /* Write a refcount table with one refcount block */ |
2228 | refcount_table = g_malloc0(2 * cluster_size); | |
2229 | refcount_table[0] = cpu_to_be64(2 * cluster_size); | |
8341f00d | 2230 | ret = blk_pwrite(blk, cluster_size, refcount_table, 2 * cluster_size, 0); |
7267c094 | 2231 | g_free(refcount_table); |
a9420734 KW |
2232 | |
2233 | if (ret < 0) { | |
3ef6c40a | 2234 | error_setg_errno(errp, -ret, "Could not write refcount table"); |
a9420734 KW |
2235 | goto out; |
2236 | } | |
2237 | ||
23588797 KW |
2238 | blk_unref(blk); |
2239 | blk = NULL; | |
a9420734 KW |
2240 | |
2241 | /* | |
2242 | * And now open the image and make it consistent first (i.e. increase the | |
2243 | * refcount of the cluster that is occupied by the header and the refcount | |
2244 | * table) | |
2245 | */ | |
e6641719 HR |
2246 | options = qdict_new(); |
2247 | qdict_put(options, "driver", qstring_from_str("qcow2")); | |
efaa7c4e | 2248 | blk = blk_new_open(filename, NULL, options, |
72e775c7 | 2249 | BDRV_O_RDWR | BDRV_O_NO_FLUSH, &local_err); |
23588797 | 2250 | if (blk == NULL) { |
3ef6c40a | 2251 | error_propagate(errp, local_err); |
23588797 | 2252 | ret = -EIO; |
a9420734 KW |
2253 | goto out; |
2254 | } | |
2255 | ||
23588797 | 2256 | ret = qcow2_alloc_clusters(blk_bs(blk), 3 * cluster_size); |
a9420734 | 2257 | if (ret < 0) { |
3ef6c40a HR |
2258 | error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 " |
2259 | "header and refcount table"); | |
a9420734 KW |
2260 | goto out; |
2261 | ||
2262 | } else if (ret != 0) { | |
2263 | error_report("Huh, first cluster in empty image is already in use?"); | |
2264 | abort(); | |
2265 | } | |
2266 | ||
b527c9b3 | 2267 | /* Create a full header (including things like feature table) */ |
23588797 | 2268 | ret = qcow2_update_header(blk_bs(blk)); |
b527c9b3 KW |
2269 | if (ret < 0) { |
2270 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); | |
2271 | goto out; | |
2272 | } | |
2273 | ||
a9420734 | 2274 | /* Okay, now that we have a valid image, let's give it the right size */ |
23588797 | 2275 | ret = blk_truncate(blk, total_size); |
a9420734 | 2276 | if (ret < 0) { |
3ef6c40a | 2277 | error_setg_errno(errp, -ret, "Could not resize image"); |
a9420734 KW |
2278 | goto out; |
2279 | } | |
2280 | ||
2281 | /* Want a backing file? There you go.*/ | |
2282 | if (backing_file) { | |
23588797 | 2283 | ret = bdrv_change_backing_file(blk_bs(blk), backing_file, backing_format); |
a9420734 | 2284 | if (ret < 0) { |
3ef6c40a HR |
2285 | error_setg_errno(errp, -ret, "Could not assign backing file '%s' " |
2286 | "with format '%s'", backing_file, backing_format); | |
a9420734 KW |
2287 | goto out; |
2288 | } | |
2289 | } | |
2290 | ||
2291 | /* And if we're supposed to preallocate metadata, do that now */ | |
0e4271b7 | 2292 | if (prealloc != PREALLOC_MODE_OFF) { |
23588797 | 2293 | BDRVQcow2State *s = blk_bs(blk)->opaque; |
15552c4a | 2294 | qemu_co_mutex_lock(&s->lock); |
23588797 | 2295 | ret = preallocate(blk_bs(blk)); |
15552c4a | 2296 | qemu_co_mutex_unlock(&s->lock); |
a9420734 | 2297 | if (ret < 0) { |
3ef6c40a | 2298 | error_setg_errno(errp, -ret, "Could not preallocate metadata"); |
a9420734 KW |
2299 | goto out; |
2300 | } | |
2301 | } | |
2302 | ||
23588797 KW |
2303 | blk_unref(blk); |
2304 | blk = NULL; | |
ba2ab2f2 HR |
2305 | |
2306 | /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning */ | |
e6641719 HR |
2307 | options = qdict_new(); |
2308 | qdict_put(options, "driver", qstring_from_str("qcow2")); | |
efaa7c4e | 2309 | blk = blk_new_open(filename, NULL, options, |
72e775c7 | 2310 | BDRV_O_RDWR | BDRV_O_NO_BACKING, &local_err); |
23588797 | 2311 | if (blk == NULL) { |
ba2ab2f2 | 2312 | error_propagate(errp, local_err); |
23588797 | 2313 | ret = -EIO; |
ba2ab2f2 HR |
2314 | goto out; |
2315 | } | |
2316 | ||
a9420734 KW |
2317 | ret = 0; |
2318 | out: | |
23588797 KW |
2319 | if (blk) { |
2320 | blk_unref(blk); | |
f67503e5 | 2321 | } |
a9420734 KW |
2322 | return ret; |
2323 | } | |
de5f3f40 | 2324 | |
1bd0e2d1 | 2325 | static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) |
de5f3f40 | 2326 | { |
1bd0e2d1 CL |
2327 | char *backing_file = NULL; |
2328 | char *backing_fmt = NULL; | |
2329 | char *buf = NULL; | |
180e9526 | 2330 | uint64_t size = 0; |
de5f3f40 | 2331 | int flags = 0; |
99cce9fa | 2332 | size_t cluster_size = DEFAULT_CLUSTER_SIZE; |
ffeaac9b | 2333 | PreallocMode prealloc; |
8ad1898c | 2334 | int version = 3; |
bd4b167f HR |
2335 | uint64_t refcount_bits = 16; |
2336 | int refcount_order; | |
3ef6c40a HR |
2337 | Error *local_err = NULL; |
2338 | int ret; | |
de5f3f40 KW |
2339 | |
2340 | /* Read out options */ | |
180e9526 HT |
2341 | size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
2342 | BDRV_SECTOR_SIZE); | |
1bd0e2d1 CL |
2343 | backing_file = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE); |
2344 | backing_fmt = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FMT); | |
2345 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_ENCRYPT, false)) { | |
2346 | flags |= BLOCK_FLAG_ENCRYPT; | |
2347 | } | |
2348 | cluster_size = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, | |
2349 | DEFAULT_CLUSTER_SIZE); | |
2350 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
ffeaac9b | 2351 | prealloc = qapi_enum_parse(PreallocMode_lookup, buf, |
7fb1cf16 | 2352 | PREALLOC_MODE__MAX, PREALLOC_MODE_OFF, |
ffeaac9b HT |
2353 | &local_err); |
2354 | if (local_err) { | |
2355 | error_propagate(errp, local_err); | |
1bd0e2d1 CL |
2356 | ret = -EINVAL; |
2357 | goto finish; | |
2358 | } | |
2359 | g_free(buf); | |
2360 | buf = qemu_opt_get_del(opts, BLOCK_OPT_COMPAT_LEVEL); | |
2361 | if (!buf) { | |
2362 | /* keep the default */ | |
2363 | } else if (!strcmp(buf, "0.10")) { | |
2364 | version = 2; | |
2365 | } else if (!strcmp(buf, "1.1")) { | |
2366 | version = 3; | |
2367 | } else { | |
2368 | error_setg(errp, "Invalid compatibility level: '%s'", buf); | |
2369 | ret = -EINVAL; | |
2370 | goto finish; | |
2371 | } | |
2372 | ||
2373 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_LAZY_REFCOUNTS, false)) { | |
2374 | flags |= BLOCK_FLAG_LAZY_REFCOUNTS; | |
de5f3f40 KW |
2375 | } |
2376 | ||
ffeaac9b | 2377 | if (backing_file && prealloc != PREALLOC_MODE_OFF) { |
3ef6c40a HR |
2378 | error_setg(errp, "Backing file and preallocation cannot be used at " |
2379 | "the same time"); | |
1bd0e2d1 CL |
2380 | ret = -EINVAL; |
2381 | goto finish; | |
de5f3f40 KW |
2382 | } |
2383 | ||
bfe8043e | 2384 | if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { |
3ef6c40a HR |
2385 | error_setg(errp, "Lazy refcounts only supported with compatibility " |
2386 | "level 1.1 and above (use compat=1.1 or greater)"); | |
1bd0e2d1 CL |
2387 | ret = -EINVAL; |
2388 | goto finish; | |
bfe8043e SH |
2389 | } |
2390 | ||
06d05fa7 HR |
2391 | refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS, |
2392 | refcount_bits); | |
2393 | if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) { | |
2394 | error_setg(errp, "Refcount width must be a power of two and may not " | |
2395 | "exceed 64 bits"); | |
2396 | ret = -EINVAL; | |
2397 | goto finish; | |
2398 | } | |
2399 | ||
bd4b167f HR |
2400 | if (version < 3 && refcount_bits != 16) { |
2401 | error_setg(errp, "Different refcount widths than 16 bits require " | |
2402 | "compatibility level 1.1 or above (use compat=1.1 or " | |
2403 | "greater)"); | |
2404 | ret = -EINVAL; | |
2405 | goto finish; | |
2406 | } | |
2407 | ||
786a4ea8 | 2408 | refcount_order = ctz32(refcount_bits); |
bd4b167f | 2409 | |
180e9526 | 2410 | ret = qcow2_create2(filename, size, backing_file, backing_fmt, flags, |
bd4b167f HR |
2411 | cluster_size, prealloc, opts, version, refcount_order, |
2412 | &local_err); | |
621ff94d | 2413 | error_propagate(errp, local_err); |
1bd0e2d1 CL |
2414 | |
2415 | finish: | |
2416 | g_free(backing_file); | |
2417 | g_free(backing_fmt); | |
2418 | g_free(buf); | |
3ef6c40a | 2419 | return ret; |
de5f3f40 KW |
2420 | } |
2421 | ||
2928abce | 2422 | |
ebb718a5 EB |
2423 | static bool is_zero_sectors(BlockDriverState *bs, int64_t start, |
2424 | uint32_t count) | |
2928abce | 2425 | { |
2928abce DL |
2426 | int nr; |
2427 | BlockDriverState *file; | |
ebb718a5 | 2428 | int64_t res; |
2928abce | 2429 | |
ebb718a5 EB |
2430 | if (!count) { |
2431 | return true; | |
2432 | } | |
2433 | res = bdrv_get_block_status_above(bs, NULL, start, count, | |
2434 | &nr, &file); | |
2435 | return res >= 0 && (res & BDRV_BLOCK_ZERO) && nr == count; | |
2928abce DL |
2436 | } |
2437 | ||
5544b59f EB |
2438 | static coroutine_fn int qcow2_co_pwrite_zeroes(BlockDriverState *bs, |
2439 | int64_t offset, int count, BdrvRequestFlags flags) | |
621f0589 KW |
2440 | { |
2441 | int ret; | |
ff99129a | 2442 | BDRVQcow2State *s = bs->opaque; |
621f0589 | 2443 | |
5544b59f EB |
2444 | uint32_t head = offset % s->cluster_size; |
2445 | uint32_t tail = (offset + count) % s->cluster_size; | |
2928abce | 2446 | |
5544b59f | 2447 | trace_qcow2_pwrite_zeroes_start_req(qemu_coroutine_self(), offset, count); |
5a64e942 | 2448 | |
ebb718a5 | 2449 | if (head || tail) { |
5544b59f | 2450 | int64_t cl_start = (offset - head) >> BDRV_SECTOR_BITS; |
ebb718a5 | 2451 | uint64_t off; |
ecfe1863 | 2452 | unsigned int nr; |
2928abce | 2453 | |
5544b59f | 2454 | assert(head + count <= s->cluster_size); |
2928abce | 2455 | |
ebb718a5 | 2456 | /* check whether remainder of cluster already reads as zero */ |
5544b59f EB |
2457 | if (!(is_zero_sectors(bs, cl_start, |
2458 | DIV_ROUND_UP(head, BDRV_SECTOR_SIZE)) && | |
2459 | is_zero_sectors(bs, (offset + count) >> BDRV_SECTOR_BITS, | |
2460 | DIV_ROUND_UP(-tail & (s->cluster_size - 1), | |
2461 | BDRV_SECTOR_SIZE)))) { | |
2928abce DL |
2462 | return -ENOTSUP; |
2463 | } | |
2464 | ||
2928abce DL |
2465 | qemu_co_mutex_lock(&s->lock); |
2466 | /* We can have new write after previous check */ | |
5544b59f EB |
2467 | offset = cl_start << BDRV_SECTOR_BITS; |
2468 | count = s->cluster_size; | |
ecfe1863 | 2469 | nr = s->cluster_size; |
5544b59f | 2470 | ret = qcow2_get_cluster_offset(bs, offset, &nr, &off); |
ebb718a5 | 2471 | if (ret != QCOW2_CLUSTER_UNALLOCATED && ret != QCOW2_CLUSTER_ZERO) { |
2928abce DL |
2472 | qemu_co_mutex_unlock(&s->lock); |
2473 | return -ENOTSUP; | |
2474 | } | |
2475 | } else { | |
2476 | qemu_co_mutex_lock(&s->lock); | |
621f0589 KW |
2477 | } |
2478 | ||
5544b59f | 2479 | trace_qcow2_pwrite_zeroes(qemu_coroutine_self(), offset, count); |
5a64e942 | 2480 | |
621f0589 | 2481 | /* Whatever is left can use real zero clusters */ |
170f4b2e | 2482 | ret = qcow2_zero_clusters(bs, offset, count >> BDRV_SECTOR_BITS, flags); |
621f0589 KW |
2483 | qemu_co_mutex_unlock(&s->lock); |
2484 | ||
2485 | return ret; | |
2486 | } | |
2487 | ||
82e8a788 EB |
2488 | static coroutine_fn int qcow2_co_pdiscard(BlockDriverState *bs, |
2489 | int64_t offset, int count) | |
5ea929e3 | 2490 | { |
6db39ae2 | 2491 | int ret; |
ff99129a | 2492 | BDRVQcow2State *s = bs->opaque; |
6db39ae2 PB |
2493 | |
2494 | qemu_co_mutex_lock(&s->lock); | |
82e8a788 EB |
2495 | ret = qcow2_discard_clusters(bs, offset, count >> BDRV_SECTOR_BITS, |
2496 | QCOW2_DISCARD_REQUEST, false); | |
6db39ae2 PB |
2497 | qemu_co_mutex_unlock(&s->lock); |
2498 | return ret; | |
5ea929e3 KW |
2499 | } |
2500 | ||
419b19d9 SH |
2501 | static int qcow2_truncate(BlockDriverState *bs, int64_t offset) |
2502 | { | |
ff99129a | 2503 | BDRVQcow2State *s = bs->opaque; |
2cf7cfa1 KW |
2504 | int64_t new_l1_size; |
2505 | int ret; | |
419b19d9 SH |
2506 | |
2507 | if (offset & 511) { | |
259b2173 | 2508 | error_report("The new size must be a multiple of 512"); |
419b19d9 SH |
2509 | return -EINVAL; |
2510 | } | |
2511 | ||
2512 | /* cannot proceed if image has snapshots */ | |
2513 | if (s->nb_snapshots) { | |
259b2173 | 2514 | error_report("Can't resize an image which has snapshots"); |
419b19d9 SH |
2515 | return -ENOTSUP; |
2516 | } | |
2517 | ||
2518 | /* shrinking is currently not supported */ | |
2519 | if (offset < bs->total_sectors * 512) { | |
259b2173 | 2520 | error_report("qcow2 doesn't support shrinking images yet"); |
419b19d9 SH |
2521 | return -ENOTSUP; |
2522 | } | |
2523 | ||
2524 | new_l1_size = size_to_l1(s, offset); | |
72893756 | 2525 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); |
419b19d9 SH |
2526 | if (ret < 0) { |
2527 | return ret; | |
2528 | } | |
2529 | ||
2530 | /* write updated header.size */ | |
2531 | offset = cpu_to_be64(offset); | |
d9ca2ea2 | 2532 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), |
8b3b7206 | 2533 | &offset, sizeof(uint64_t)); |
419b19d9 SH |
2534 | if (ret < 0) { |
2535 | return ret; | |
2536 | } | |
2537 | ||
2538 | s->l1_vm_state_index = new_l1_size; | |
2539 | return 0; | |
2540 | } | |
2541 | ||
20d97356 BS |
2542 | /* XXX: put compressed sectors first, then all the cluster aligned |
2543 | tables to avoid losing bytes in alignment */ | |
fcccefc5 PB |
2544 | static coroutine_fn int |
2545 | qcow2_co_pwritev_compressed(BlockDriverState *bs, uint64_t offset, | |
2546 | uint64_t bytes, QEMUIOVector *qiov) | |
20d97356 | 2547 | { |
ff99129a | 2548 | BDRVQcow2State *s = bs->opaque; |
fcccefc5 PB |
2549 | QEMUIOVector hd_qiov; |
2550 | struct iovec iov; | |
20d97356 BS |
2551 | z_stream strm; |
2552 | int ret, out_len; | |
fcccefc5 | 2553 | uint8_t *buf, *out_buf; |
20d97356 BS |
2554 | uint64_t cluster_offset; |
2555 | ||
fcccefc5 | 2556 | if (bytes == 0) { |
20d97356 BS |
2557 | /* align end of file to a sector boundary to ease reading with |
2558 | sector based I/Os */ | |
9a4f4c31 KW |
2559 | cluster_offset = bdrv_getlength(bs->file->bs); |
2560 | return bdrv_truncate(bs->file->bs, cluster_offset); | |
20d97356 BS |
2561 | } |
2562 | ||
a2c0ca6f | 2563 | buf = qemu_blockalign(bs, s->cluster_size); |
fcccefc5 | 2564 | if (bytes != s->cluster_size) { |
a2c0ca6f PB |
2565 | if (bytes > s->cluster_size || |
2566 | offset + bytes != bs->total_sectors << BDRV_SECTOR_BITS) | |
fcccefc5 | 2567 | { |
a2c0ca6f PB |
2568 | qemu_vfree(buf); |
2569 | return -EINVAL; | |
f4d38bef | 2570 | } |
a2c0ca6f PB |
2571 | /* Zero-pad last write if image size is not cluster aligned */ |
2572 | memset(buf + bytes, 0, s->cluster_size - bytes); | |
f4d38bef | 2573 | } |
8b2bd093 | 2574 | qemu_iovec_to_buf(qiov, 0, buf, bytes); |
20d97356 | 2575 | |
ebf7bba0 | 2576 | out_buf = g_malloc(s->cluster_size); |
20d97356 BS |
2577 | |
2578 | /* best compression, small window, no zlib header */ | |
2579 | memset(&strm, 0, sizeof(strm)); | |
2580 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
2581 | Z_DEFLATED, -12, | |
2582 | 9, Z_DEFAULT_STRATEGY); | |
2583 | if (ret != 0) { | |
8f1efd00 KW |
2584 | ret = -EINVAL; |
2585 | goto fail; | |
20d97356 BS |
2586 | } |
2587 | ||
2588 | strm.avail_in = s->cluster_size; | |
2589 | strm.next_in = (uint8_t *)buf; | |
2590 | strm.avail_out = s->cluster_size; | |
2591 | strm.next_out = out_buf; | |
2592 | ||
2593 | ret = deflate(&strm, Z_FINISH); | |
2594 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
20d97356 | 2595 | deflateEnd(&strm); |
8f1efd00 KW |
2596 | ret = -EINVAL; |
2597 | goto fail; | |
20d97356 BS |
2598 | } |
2599 | out_len = strm.next_out - out_buf; | |
2600 | ||
2601 | deflateEnd(&strm); | |
2602 | ||
2603 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
2604 | /* could not compress: write normal cluster */ | |
fcccefc5 | 2605 | ret = qcow2_co_pwritev(bs, offset, bytes, qiov, 0); |
8f1efd00 KW |
2606 | if (ret < 0) { |
2607 | goto fail; | |
2608 | } | |
fcccefc5 PB |
2609 | goto success; |
2610 | } | |
cf93980e | 2611 | |
fcccefc5 PB |
2612 | qemu_co_mutex_lock(&s->lock); |
2613 | cluster_offset = | |
2614 | qcow2_alloc_compressed_cluster_offset(bs, offset, out_len); | |
2615 | if (!cluster_offset) { | |
2616 | qemu_co_mutex_unlock(&s->lock); | |
2617 | ret = -EIO; | |
2618 | goto fail; | |
2619 | } | |
2620 | cluster_offset &= s->cluster_offset_mask; | |
cf93980e | 2621 | |
fcccefc5 PB |
2622 | ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len); |
2623 | qemu_co_mutex_unlock(&s->lock); | |
2624 | if (ret < 0) { | |
2625 | goto fail; | |
20d97356 BS |
2626 | } |
2627 | ||
fcccefc5 PB |
2628 | iov = (struct iovec) { |
2629 | .iov_base = out_buf, | |
2630 | .iov_len = out_len, | |
2631 | }; | |
2632 | qemu_iovec_init_external(&hd_qiov, &iov, 1); | |
2633 | ||
2634 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); | |
2635 | ret = bdrv_co_pwritev(bs->file, cluster_offset, out_len, &hd_qiov, 0); | |
2636 | if (ret < 0) { | |
2637 | goto fail; | |
2638 | } | |
2639 | success: | |
8f1efd00 KW |
2640 | ret = 0; |
2641 | fail: | |
fcccefc5 | 2642 | qemu_vfree(buf); |
7267c094 | 2643 | g_free(out_buf); |
8f1efd00 | 2644 | return ret; |
20d97356 BS |
2645 | } |
2646 | ||
94054183 HR |
2647 | static int make_completely_empty(BlockDriverState *bs) |
2648 | { | |
ff99129a | 2649 | BDRVQcow2State *s = bs->opaque; |
94054183 HR |
2650 | int ret, l1_clusters; |
2651 | int64_t offset; | |
2652 | uint64_t *new_reftable = NULL; | |
2653 | uint64_t rt_entry, l1_size2; | |
2654 | struct { | |
2655 | uint64_t l1_offset; | |
2656 | uint64_t reftable_offset; | |
2657 | uint32_t reftable_clusters; | |
2658 | } QEMU_PACKED l1_ofs_rt_ofs_cls; | |
2659 | ||
2660 | ret = qcow2_cache_empty(bs, s->l2_table_cache); | |
2661 | if (ret < 0) { | |
2662 | goto fail; | |
2663 | } | |
2664 | ||
2665 | ret = qcow2_cache_empty(bs, s->refcount_block_cache); | |
2666 | if (ret < 0) { | |
2667 | goto fail; | |
2668 | } | |
2669 | ||
2670 | /* Refcounts will be broken utterly */ | |
2671 | ret = qcow2_mark_dirty(bs); | |
2672 | if (ret < 0) { | |
2673 | goto fail; | |
2674 | } | |
2675 | ||
2676 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
2677 | ||
2678 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
2679 | l1_size2 = (uint64_t)s->l1_size * sizeof(uint64_t); | |
2680 | ||
2681 | /* After this call, neither the in-memory nor the on-disk refcount | |
2682 | * information accurately describe the actual references */ | |
2683 | ||
720ff280 | 2684 | ret = bdrv_pwrite_zeroes(bs->file, s->l1_table_offset, |
74021bc4 | 2685 | l1_clusters * s->cluster_size, 0); |
94054183 HR |
2686 | if (ret < 0) { |
2687 | goto fail_broken_refcounts; | |
2688 | } | |
2689 | memset(s->l1_table, 0, l1_size2); | |
2690 | ||
2691 | BLKDBG_EVENT(bs->file, BLKDBG_EMPTY_IMAGE_PREPARE); | |
2692 | ||
2693 | /* Overwrite enough clusters at the beginning of the sectors to place | |
2694 | * the refcount table, a refcount block and the L1 table in; this may | |
2695 | * overwrite parts of the existing refcount and L1 table, which is not | |
2696 | * an issue because the dirty flag is set, complete data loss is in fact | |
2697 | * desired and partial data loss is consequently fine as well */ | |
720ff280 | 2698 | ret = bdrv_pwrite_zeroes(bs->file, s->cluster_size, |
74021bc4 | 2699 | (2 + l1_clusters) * s->cluster_size, 0); |
94054183 HR |
2700 | /* This call (even if it failed overall) may have overwritten on-disk |
2701 | * refcount structures; in that case, the in-memory refcount information | |
2702 | * will probably differ from the on-disk information which makes the BDS | |
2703 | * unusable */ | |
2704 | if (ret < 0) { | |
2705 | goto fail_broken_refcounts; | |
2706 | } | |
2707 | ||
2708 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
2709 | BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE); | |
2710 | ||
2711 | /* "Create" an empty reftable (one cluster) directly after the image | |
2712 | * header and an empty L1 table three clusters after the image header; | |
2713 | * the cluster between those two will be used as the first refblock */ | |
f1f7a1dd PM |
2714 | l1_ofs_rt_ofs_cls.l1_offset = cpu_to_be64(3 * s->cluster_size); |
2715 | l1_ofs_rt_ofs_cls.reftable_offset = cpu_to_be64(s->cluster_size); | |
2716 | l1_ofs_rt_ofs_cls.reftable_clusters = cpu_to_be32(1); | |
d9ca2ea2 | 2717 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, l1_table_offset), |
94054183 HR |
2718 | &l1_ofs_rt_ofs_cls, sizeof(l1_ofs_rt_ofs_cls)); |
2719 | if (ret < 0) { | |
2720 | goto fail_broken_refcounts; | |
2721 | } | |
2722 | ||
2723 | s->l1_table_offset = 3 * s->cluster_size; | |
2724 | ||
2725 | new_reftable = g_try_new0(uint64_t, s->cluster_size / sizeof(uint64_t)); | |
2726 | if (!new_reftable) { | |
2727 | ret = -ENOMEM; | |
2728 | goto fail_broken_refcounts; | |
2729 | } | |
2730 | ||
2731 | s->refcount_table_offset = s->cluster_size; | |
2732 | s->refcount_table_size = s->cluster_size / sizeof(uint64_t); | |
2733 | ||
2734 | g_free(s->refcount_table); | |
2735 | s->refcount_table = new_reftable; | |
2736 | new_reftable = NULL; | |
2737 | ||
2738 | /* Now the in-memory refcount information again corresponds to the on-disk | |
2739 | * information (reftable is empty and no refblocks (the refblock cache is | |
2740 | * empty)); however, this means some clusters (e.g. the image header) are | |
2741 | * referenced, but not refcounted, but the normal qcow2 code assumes that | |
2742 | * the in-memory information is always correct */ | |
2743 | ||
2744 | BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC); | |
2745 | ||
2746 | /* Enter the first refblock into the reftable */ | |
2747 | rt_entry = cpu_to_be64(2 * s->cluster_size); | |
d9ca2ea2 | 2748 | ret = bdrv_pwrite_sync(bs->file, s->cluster_size, |
94054183 HR |
2749 | &rt_entry, sizeof(rt_entry)); |
2750 | if (ret < 0) { | |
2751 | goto fail_broken_refcounts; | |
2752 | } | |
2753 | s->refcount_table[0] = 2 * s->cluster_size; | |
2754 | ||
2755 | s->free_cluster_index = 0; | |
2756 | assert(3 + l1_clusters <= s->refcount_block_size); | |
2757 | offset = qcow2_alloc_clusters(bs, 3 * s->cluster_size + l1_size2); | |
2758 | if (offset < 0) { | |
2759 | ret = offset; | |
2760 | goto fail_broken_refcounts; | |
2761 | } else if (offset > 0) { | |
2762 | error_report("First cluster in emptied image is in use"); | |
2763 | abort(); | |
2764 | } | |
2765 | ||
2766 | /* Now finally the in-memory information corresponds to the on-disk | |
2767 | * structures and is correct */ | |
2768 | ret = qcow2_mark_clean(bs); | |
2769 | if (ret < 0) { | |
2770 | goto fail; | |
2771 | } | |
2772 | ||
9a4f4c31 | 2773 | ret = bdrv_truncate(bs->file->bs, (3 + l1_clusters) * s->cluster_size); |
94054183 HR |
2774 | if (ret < 0) { |
2775 | goto fail; | |
2776 | } | |
2777 | ||
2778 | return 0; | |
2779 | ||
2780 | fail_broken_refcounts: | |
2781 | /* The BDS is unusable at this point. If we wanted to make it usable, we | |
2782 | * would have to call qcow2_refcount_close(), qcow2_refcount_init(), | |
2783 | * qcow2_check_refcounts(), qcow2_refcount_close() and qcow2_refcount_init() | |
2784 | * again. However, because the functions which could have caused this error | |
2785 | * path to be taken are used by those functions as well, it's very likely | |
2786 | * that that sequence will fail as well. Therefore, just eject the BDS. */ | |
2787 | bs->drv = NULL; | |
2788 | ||
2789 | fail: | |
2790 | g_free(new_reftable); | |
2791 | return ret; | |
2792 | } | |
2793 | ||
491d27e2 HR |
2794 | static int qcow2_make_empty(BlockDriverState *bs) |
2795 | { | |
ff99129a | 2796 | BDRVQcow2State *s = bs->opaque; |
491d27e2 HR |
2797 | uint64_t start_sector; |
2798 | int sector_step = INT_MAX / BDRV_SECTOR_SIZE; | |
94054183 HR |
2799 | int l1_clusters, ret = 0; |
2800 | ||
2801 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
2802 | ||
2803 | if (s->qcow_version >= 3 && !s->snapshots && | |
2804 | 3 + l1_clusters <= s->refcount_block_size) { | |
2805 | /* The following function only works for qcow2 v3 images (it requires | |
2806 | * the dirty flag) and only as long as there are no snapshots (because | |
2807 | * it completely empties the image). Furthermore, the L1 table and three | |
2808 | * additional clusters (image header, refcount table, one refcount | |
2809 | * block) have to fit inside one refcount block. */ | |
2810 | return make_completely_empty(bs); | |
2811 | } | |
491d27e2 | 2812 | |
94054183 HR |
2813 | /* This fallback code simply discards every active cluster; this is slow, |
2814 | * but works in all cases */ | |
491d27e2 HR |
2815 | for (start_sector = 0; start_sector < bs->total_sectors; |
2816 | start_sector += sector_step) | |
2817 | { | |
2818 | /* As this function is generally used after committing an external | |
2819 | * snapshot, QCOW2_DISCARD_SNAPSHOT seems appropriate. Also, the | |
2820 | * default action for this kind of discard is to pass the discard, | |
2821 | * which will ideally result in an actually smaller image file, as | |
2822 | * is probably desired. */ | |
2823 | ret = qcow2_discard_clusters(bs, start_sector * BDRV_SECTOR_SIZE, | |
2824 | MIN(sector_step, | |
2825 | bs->total_sectors - start_sector), | |
2826 | QCOW2_DISCARD_SNAPSHOT, true); | |
2827 | if (ret < 0) { | |
2828 | break; | |
2829 | } | |
2830 | } | |
2831 | ||
2832 | return ret; | |
2833 | } | |
2834 | ||
a968168c | 2835 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
20d97356 | 2836 | { |
ff99129a | 2837 | BDRVQcow2State *s = bs->opaque; |
29c1a730 KW |
2838 | int ret; |
2839 | ||
8b94ff85 | 2840 | qemu_co_mutex_lock(&s->lock); |
f3c3b87d | 2841 | ret = qcow2_cache_write(bs, s->l2_table_cache); |
29c1a730 | 2842 | if (ret < 0) { |
c95de7e2 | 2843 | qemu_co_mutex_unlock(&s->lock); |
8b94ff85 | 2844 | return ret; |
29c1a730 KW |
2845 | } |
2846 | ||
bfe8043e | 2847 | if (qcow2_need_accurate_refcounts(s)) { |
f3c3b87d | 2848 | ret = qcow2_cache_write(bs, s->refcount_block_cache); |
bfe8043e SH |
2849 | if (ret < 0) { |
2850 | qemu_co_mutex_unlock(&s->lock); | |
2851 | return ret; | |
2852 | } | |
29c1a730 | 2853 | } |
8b94ff85 | 2854 | qemu_co_mutex_unlock(&s->lock); |
29c1a730 | 2855 | |
eb489bb1 KW |
2856 | return 0; |
2857 | } | |
2858 | ||
7c80ab3f | 2859 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
20d97356 | 2860 | { |
ff99129a | 2861 | BDRVQcow2State *s = bs->opaque; |
95de6d70 PB |
2862 | bdi->unallocated_blocks_are_zero = true; |
2863 | bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3); | |
20d97356 | 2864 | bdi->cluster_size = s->cluster_size; |
7c80ab3f | 2865 | bdi->vm_state_offset = qcow2_vm_state_offset(s); |
20d97356 BS |
2866 | return 0; |
2867 | } | |
2868 | ||
37764dfb HR |
2869 | static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs) |
2870 | { | |
ff99129a | 2871 | BDRVQcow2State *s = bs->opaque; |
37764dfb HR |
2872 | ImageInfoSpecific *spec_info = g_new(ImageInfoSpecific, 1); |
2873 | ||
2874 | *spec_info = (ImageInfoSpecific){ | |
6a8f9661 | 2875 | .type = IMAGE_INFO_SPECIFIC_KIND_QCOW2, |
32bafa8f | 2876 | .u.qcow2.data = g_new(ImageInfoSpecificQCow2, 1), |
37764dfb HR |
2877 | }; |
2878 | if (s->qcow_version == 2) { | |
32bafa8f | 2879 | *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){ |
0709c5a1 HR |
2880 | .compat = g_strdup("0.10"), |
2881 | .refcount_bits = s->refcount_bits, | |
37764dfb HR |
2882 | }; |
2883 | } else if (s->qcow_version == 3) { | |
32bafa8f | 2884 | *spec_info->u.qcow2.data = (ImageInfoSpecificQCow2){ |
37764dfb HR |
2885 | .compat = g_strdup("1.1"), |
2886 | .lazy_refcounts = s->compatible_features & | |
2887 | QCOW2_COMPAT_LAZY_REFCOUNTS, | |
2888 | .has_lazy_refcounts = true, | |
9009b196 HR |
2889 | .corrupt = s->incompatible_features & |
2890 | QCOW2_INCOMPAT_CORRUPT, | |
2891 | .has_corrupt = true, | |
0709c5a1 | 2892 | .refcount_bits = s->refcount_bits, |
37764dfb | 2893 | }; |
b1fc8f93 DL |
2894 | } else { |
2895 | /* if this assertion fails, this probably means a new version was | |
2896 | * added without having it covered here */ | |
2897 | assert(false); | |
37764dfb HR |
2898 | } |
2899 | ||
2900 | return spec_info; | |
2901 | } | |
2902 | ||
20d97356 BS |
2903 | #if 0 |
2904 | static void dump_refcounts(BlockDriverState *bs) | |
2905 | { | |
ff99129a | 2906 | BDRVQcow2State *s = bs->opaque; |
20d97356 BS |
2907 | int64_t nb_clusters, k, k1, size; |
2908 | int refcount; | |
2909 | ||
9a4f4c31 | 2910 | size = bdrv_getlength(bs->file->bs); |
20d97356 BS |
2911 | nb_clusters = size_to_clusters(s, size); |
2912 | for(k = 0; k < nb_clusters;) { | |
2913 | k1 = k; | |
2914 | refcount = get_refcount(bs, k); | |
2915 | k++; | |
2916 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
2917 | k++; | |
0bfcd599 BS |
2918 | printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount, |
2919 | k - k1); | |
20d97356 BS |
2920 | } |
2921 | } | |
2922 | #endif | |
2923 | ||
cf8074b3 KW |
2924 | static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
2925 | int64_t pos) | |
20d97356 | 2926 | { |
ff99129a | 2927 | BDRVQcow2State *s = bs->opaque; |
20d97356 | 2928 | |
66f82cee | 2929 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
734a7758 KW |
2930 | return bs->drv->bdrv_co_pwritev(bs, qcow2_vm_state_offset(s) + pos, |
2931 | qiov->size, qiov, 0); | |
20d97356 BS |
2932 | } |
2933 | ||
5ddda0b8 KW |
2934 | static int qcow2_load_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
2935 | int64_t pos) | |
20d97356 | 2936 | { |
ff99129a | 2937 | BDRVQcow2State *s = bs->opaque; |
20d97356 | 2938 | |
66f82cee | 2939 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
734a7758 KW |
2940 | return bs->drv->bdrv_co_preadv(bs, qcow2_vm_state_offset(s) + pos, |
2941 | qiov->size, qiov, 0); | |
20d97356 BS |
2942 | } |
2943 | ||
9296b3ed HR |
2944 | /* |
2945 | * Downgrades an image's version. To achieve this, any incompatible features | |
2946 | * have to be removed. | |
2947 | */ | |
4057a2b2 | 2948 | static int qcow2_downgrade(BlockDriverState *bs, int target_version, |
8b13976d | 2949 | BlockDriverAmendStatusCB *status_cb, void *cb_opaque) |
9296b3ed | 2950 | { |
ff99129a | 2951 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
2952 | int current_version = s->qcow_version; |
2953 | int ret; | |
2954 | ||
2955 | if (target_version == current_version) { | |
2956 | return 0; | |
2957 | } else if (target_version > current_version) { | |
2958 | return -EINVAL; | |
2959 | } else if (target_version != 2) { | |
2960 | return -EINVAL; | |
2961 | } | |
2962 | ||
2963 | if (s->refcount_order != 4) { | |
61ce55fc | 2964 | error_report("compat=0.10 requires refcount_bits=16"); |
9296b3ed HR |
2965 | return -ENOTSUP; |
2966 | } | |
2967 | ||
2968 | /* clear incompatible features */ | |
2969 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
2970 | ret = qcow2_mark_clean(bs); | |
2971 | if (ret < 0) { | |
2972 | return ret; | |
2973 | } | |
2974 | } | |
2975 | ||
2976 | /* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in | |
2977 | * the first place; if that happens nonetheless, returning -ENOTSUP is the | |
2978 | * best thing to do anyway */ | |
2979 | ||
2980 | if (s->incompatible_features) { | |
2981 | return -ENOTSUP; | |
2982 | } | |
2983 | ||
2984 | /* since we can ignore compatible features, we can set them to 0 as well */ | |
2985 | s->compatible_features = 0; | |
2986 | /* if lazy refcounts have been used, they have already been fixed through | |
2987 | * clearing the dirty flag */ | |
2988 | ||
2989 | /* clearing autoclear features is trivial */ | |
2990 | s->autoclear_features = 0; | |
2991 | ||
8b13976d | 2992 | ret = qcow2_expand_zero_clusters(bs, status_cb, cb_opaque); |
9296b3ed HR |
2993 | if (ret < 0) { |
2994 | return ret; | |
2995 | } | |
2996 | ||
2997 | s->qcow_version = target_version; | |
2998 | ret = qcow2_update_header(bs); | |
2999 | if (ret < 0) { | |
3000 | s->qcow_version = current_version; | |
3001 | return ret; | |
3002 | } | |
3003 | return 0; | |
3004 | } | |
3005 | ||
c293a809 HR |
3006 | typedef enum Qcow2AmendOperation { |
3007 | /* This is the value Qcow2AmendHelperCBInfo::last_operation will be | |
3008 | * statically initialized to so that the helper CB can discern the first | |
3009 | * invocation from an operation change */ | |
3010 | QCOW2_NO_OPERATION = 0, | |
3011 | ||
61ce55fc | 3012 | QCOW2_CHANGING_REFCOUNT_ORDER, |
c293a809 HR |
3013 | QCOW2_DOWNGRADING, |
3014 | } Qcow2AmendOperation; | |
3015 | ||
3016 | typedef struct Qcow2AmendHelperCBInfo { | |
3017 | /* The code coordinating the amend operations should only modify | |
3018 | * these four fields; the rest will be managed by the CB */ | |
3019 | BlockDriverAmendStatusCB *original_status_cb; | |
3020 | void *original_cb_opaque; | |
3021 | ||
3022 | Qcow2AmendOperation current_operation; | |
3023 | ||
3024 | /* Total number of operations to perform (only set once) */ | |
3025 | int total_operations; | |
3026 | ||
3027 | /* The following fields are managed by the CB */ | |
3028 | ||
3029 | /* Number of operations completed */ | |
3030 | int operations_completed; | |
3031 | ||
3032 | /* Cumulative offset of all completed operations */ | |
3033 | int64_t offset_completed; | |
3034 | ||
3035 | Qcow2AmendOperation last_operation; | |
3036 | int64_t last_work_size; | |
3037 | } Qcow2AmendHelperCBInfo; | |
3038 | ||
3039 | static void qcow2_amend_helper_cb(BlockDriverState *bs, | |
3040 | int64_t operation_offset, | |
3041 | int64_t operation_work_size, void *opaque) | |
3042 | { | |
3043 | Qcow2AmendHelperCBInfo *info = opaque; | |
3044 | int64_t current_work_size; | |
3045 | int64_t projected_work_size; | |
3046 | ||
3047 | if (info->current_operation != info->last_operation) { | |
3048 | if (info->last_operation != QCOW2_NO_OPERATION) { | |
3049 | info->offset_completed += info->last_work_size; | |
3050 | info->operations_completed++; | |
3051 | } | |
3052 | ||
3053 | info->last_operation = info->current_operation; | |
3054 | } | |
3055 | ||
3056 | assert(info->total_operations > 0); | |
3057 | assert(info->operations_completed < info->total_operations); | |
3058 | ||
3059 | info->last_work_size = operation_work_size; | |
3060 | ||
3061 | current_work_size = info->offset_completed + operation_work_size; | |
3062 | ||
3063 | /* current_work_size is the total work size for (operations_completed + 1) | |
3064 | * operations (which includes this one), so multiply it by the number of | |
3065 | * operations not covered and divide it by the number of operations | |
3066 | * covered to get a projection for the operations not covered */ | |
3067 | projected_work_size = current_work_size * (info->total_operations - | |
3068 | info->operations_completed - 1) | |
3069 | / (info->operations_completed + 1); | |
3070 | ||
3071 | info->original_status_cb(bs, info->offset_completed + operation_offset, | |
3072 | current_work_size + projected_work_size, | |
3073 | info->original_cb_opaque); | |
3074 | } | |
3075 | ||
77485434 | 3076 | static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts, |
8b13976d HR |
3077 | BlockDriverAmendStatusCB *status_cb, |
3078 | void *cb_opaque) | |
9296b3ed | 3079 | { |
ff99129a | 3080 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
3081 | int old_version = s->qcow_version, new_version = old_version; |
3082 | uint64_t new_size = 0; | |
3083 | const char *backing_file = NULL, *backing_format = NULL; | |
3084 | bool lazy_refcounts = s->use_lazy_refcounts; | |
1bd0e2d1 CL |
3085 | const char *compat = NULL; |
3086 | uint64_t cluster_size = s->cluster_size; | |
3087 | bool encrypt; | |
61ce55fc | 3088 | int refcount_bits = s->refcount_bits; |
9296b3ed | 3089 | int ret; |
1bd0e2d1 | 3090 | QemuOptDesc *desc = opts->list->desc; |
c293a809 | 3091 | Qcow2AmendHelperCBInfo helper_cb_info; |
9296b3ed | 3092 | |
1bd0e2d1 CL |
3093 | while (desc && desc->name) { |
3094 | if (!qemu_opt_find(opts, desc->name)) { | |
9296b3ed | 3095 | /* only change explicitly defined options */ |
1bd0e2d1 | 3096 | desc++; |
9296b3ed HR |
3097 | continue; |
3098 | } | |
3099 | ||
8a17b83c HR |
3100 | if (!strcmp(desc->name, BLOCK_OPT_COMPAT_LEVEL)) { |
3101 | compat = qemu_opt_get(opts, BLOCK_OPT_COMPAT_LEVEL); | |
1bd0e2d1 | 3102 | if (!compat) { |
9296b3ed | 3103 | /* preserve default */ |
1bd0e2d1 | 3104 | } else if (!strcmp(compat, "0.10")) { |
9296b3ed | 3105 | new_version = 2; |
1bd0e2d1 | 3106 | } else if (!strcmp(compat, "1.1")) { |
9296b3ed HR |
3107 | new_version = 3; |
3108 | } else { | |
29d72431 | 3109 | error_report("Unknown compatibility level %s", compat); |
9296b3ed HR |
3110 | return -EINVAL; |
3111 | } | |
8a17b83c | 3112 | } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) { |
29d72431 | 3113 | error_report("Cannot change preallocation mode"); |
9296b3ed | 3114 | return -ENOTSUP; |
8a17b83c HR |
3115 | } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) { |
3116 | new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); | |
3117 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FILE)) { | |
3118 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); | |
3119 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FMT)) { | |
3120 | backing_format = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); | |
3121 | } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT)) { | |
3122 | encrypt = qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, | |
f6fa64f6 DB |
3123 | !!s->cipher); |
3124 | ||
3125 | if (encrypt != !!s->cipher) { | |
29d72431 | 3126 | error_report("Changing the encryption flag is not supported"); |
9296b3ed HR |
3127 | return -ENOTSUP; |
3128 | } | |
8a17b83c HR |
3129 | } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) { |
3130 | cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, | |
1bd0e2d1 CL |
3131 | cluster_size); |
3132 | if (cluster_size != s->cluster_size) { | |
29d72431 | 3133 | error_report("Changing the cluster size is not supported"); |
9296b3ed HR |
3134 | return -ENOTSUP; |
3135 | } | |
8a17b83c HR |
3136 | } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) { |
3137 | lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS, | |
1bd0e2d1 | 3138 | lazy_refcounts); |
06d05fa7 | 3139 | } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) { |
61ce55fc HR |
3140 | refcount_bits = qemu_opt_get_number(opts, BLOCK_OPT_REFCOUNT_BITS, |
3141 | refcount_bits); | |
3142 | ||
3143 | if (refcount_bits <= 0 || refcount_bits > 64 || | |
3144 | !is_power_of_2(refcount_bits)) | |
3145 | { | |
3146 | error_report("Refcount width must be a power of two and may " | |
3147 | "not exceed 64 bits"); | |
3148 | return -EINVAL; | |
3149 | } | |
9296b3ed | 3150 | } else { |
164e0f89 | 3151 | /* if this point is reached, this probably means a new option was |
9296b3ed | 3152 | * added without having it covered here */ |
164e0f89 | 3153 | abort(); |
9296b3ed | 3154 | } |
1bd0e2d1 CL |
3155 | |
3156 | desc++; | |
9296b3ed HR |
3157 | } |
3158 | ||
c293a809 HR |
3159 | helper_cb_info = (Qcow2AmendHelperCBInfo){ |
3160 | .original_status_cb = status_cb, | |
3161 | .original_cb_opaque = cb_opaque, | |
3162 | .total_operations = (new_version < old_version) | |
61ce55fc | 3163 | + (s->refcount_bits != refcount_bits) |
c293a809 HR |
3164 | }; |
3165 | ||
1038bbb8 HR |
3166 | /* Upgrade first (some features may require compat=1.1) */ |
3167 | if (new_version > old_version) { | |
3168 | s->qcow_version = new_version; | |
3169 | ret = qcow2_update_header(bs); | |
3170 | if (ret < 0) { | |
3171 | s->qcow_version = old_version; | |
3172 | return ret; | |
9296b3ed HR |
3173 | } |
3174 | } | |
3175 | ||
61ce55fc HR |
3176 | if (s->refcount_bits != refcount_bits) { |
3177 | int refcount_order = ctz32(refcount_bits); | |
3178 | Error *local_error = NULL; | |
3179 | ||
3180 | if (new_version < 3 && refcount_bits != 16) { | |
3181 | error_report("Different refcount widths than 16 bits require " | |
3182 | "compatibility level 1.1 or above (use compat=1.1 or " | |
3183 | "greater)"); | |
3184 | return -EINVAL; | |
3185 | } | |
3186 | ||
3187 | helper_cb_info.current_operation = QCOW2_CHANGING_REFCOUNT_ORDER; | |
3188 | ret = qcow2_change_refcount_order(bs, refcount_order, | |
3189 | &qcow2_amend_helper_cb, | |
3190 | &helper_cb_info, &local_error); | |
3191 | if (ret < 0) { | |
3192 | error_report_err(local_error); | |
3193 | return ret; | |
3194 | } | |
3195 | } | |
3196 | ||
9296b3ed | 3197 | if (backing_file || backing_format) { |
e4603fe1 KW |
3198 | ret = qcow2_change_backing_file(bs, |
3199 | backing_file ?: s->image_backing_file, | |
3200 | backing_format ?: s->image_backing_format); | |
9296b3ed HR |
3201 | if (ret < 0) { |
3202 | return ret; | |
3203 | } | |
3204 | } | |
3205 | ||
3206 | if (s->use_lazy_refcounts != lazy_refcounts) { | |
3207 | if (lazy_refcounts) { | |
1038bbb8 | 3208 | if (new_version < 3) { |
29d72431 HR |
3209 | error_report("Lazy refcounts only supported with compatibility " |
3210 | "level 1.1 and above (use compat=1.1 or greater)"); | |
9296b3ed HR |
3211 | return -EINVAL; |
3212 | } | |
3213 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3214 | ret = qcow2_update_header(bs); | |
3215 | if (ret < 0) { | |
3216 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3217 | return ret; | |
3218 | } | |
3219 | s->use_lazy_refcounts = true; | |
3220 | } else { | |
3221 | /* make image clean first */ | |
3222 | ret = qcow2_mark_clean(bs); | |
3223 | if (ret < 0) { | |
3224 | return ret; | |
3225 | } | |
3226 | /* now disallow lazy refcounts */ | |
3227 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3228 | ret = qcow2_update_header(bs); | |
3229 | if (ret < 0) { | |
3230 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3231 | return ret; | |
3232 | } | |
3233 | s->use_lazy_refcounts = false; | |
3234 | } | |
3235 | } | |
3236 | ||
3237 | if (new_size) { | |
3238 | ret = bdrv_truncate(bs, new_size); | |
3239 | if (ret < 0) { | |
3240 | return ret; | |
3241 | } | |
3242 | } | |
3243 | ||
1038bbb8 HR |
3244 | /* Downgrade last (so unsupported features can be removed before) */ |
3245 | if (new_version < old_version) { | |
c293a809 HR |
3246 | helper_cb_info.current_operation = QCOW2_DOWNGRADING; |
3247 | ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb, | |
3248 | &helper_cb_info); | |
1038bbb8 HR |
3249 | if (ret < 0) { |
3250 | return ret; | |
3251 | } | |
3252 | } | |
3253 | ||
9296b3ed HR |
3254 | return 0; |
3255 | } | |
3256 | ||
85186ebd HR |
3257 | /* |
3258 | * If offset or size are negative, respectively, they will not be included in | |
3259 | * the BLOCK_IMAGE_CORRUPTED event emitted. | |
3260 | * fatal will be ignored for read-only BDS; corruptions found there will always | |
3261 | * be considered non-fatal. | |
3262 | */ | |
3263 | void qcow2_signal_corruption(BlockDriverState *bs, bool fatal, int64_t offset, | |
3264 | int64_t size, const char *message_format, ...) | |
3265 | { | |
ff99129a | 3266 | BDRVQcow2State *s = bs->opaque; |
dc881b44 | 3267 | const char *node_name; |
85186ebd HR |
3268 | char *message; |
3269 | va_list ap; | |
3270 | ||
3271 | fatal = fatal && !bs->read_only; | |
3272 | ||
3273 | if (s->signaled_corruption && | |
3274 | (!fatal || (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT))) | |
3275 | { | |
3276 | return; | |
3277 | } | |
3278 | ||
3279 | va_start(ap, message_format); | |
3280 | message = g_strdup_vprintf(message_format, ap); | |
3281 | va_end(ap); | |
3282 | ||
3283 | if (fatal) { | |
3284 | fprintf(stderr, "qcow2: Marking image as corrupt: %s; further " | |
3285 | "corruption events will be suppressed\n", message); | |
3286 | } else { | |
3287 | fprintf(stderr, "qcow2: Image is corrupt: %s; further non-fatal " | |
3288 | "corruption events will be suppressed\n", message); | |
3289 | } | |
3290 | ||
dc881b44 AG |
3291 | node_name = bdrv_get_node_name(bs); |
3292 | qapi_event_send_block_image_corrupted(bdrv_get_device_name(bs), | |
3293 | *node_name != '\0', node_name, | |
3294 | message, offset >= 0, offset, | |
3295 | size >= 0, size, | |
85186ebd HR |
3296 | fatal, &error_abort); |
3297 | g_free(message); | |
3298 | ||
3299 | if (fatal) { | |
3300 | qcow2_mark_corrupt(bs); | |
3301 | bs->drv = NULL; /* make BDS unusable */ | |
3302 | } | |
3303 | ||
3304 | s->signaled_corruption = true; | |
3305 | } | |
3306 | ||
1bd0e2d1 CL |
3307 | static QemuOptsList qcow2_create_opts = { |
3308 | .name = "qcow2-create-opts", | |
3309 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_create_opts.head), | |
3310 | .desc = { | |
3311 | { | |
3312 | .name = BLOCK_OPT_SIZE, | |
3313 | .type = QEMU_OPT_SIZE, | |
3314 | .help = "Virtual disk size" | |
3315 | }, | |
3316 | { | |
3317 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
3318 | .type = QEMU_OPT_STRING, | |
3319 | .help = "Compatibility level (0.10 or 1.1)" | |
3320 | }, | |
3321 | { | |
3322 | .name = BLOCK_OPT_BACKING_FILE, | |
3323 | .type = QEMU_OPT_STRING, | |
3324 | .help = "File name of a base image" | |
3325 | }, | |
3326 | { | |
3327 | .name = BLOCK_OPT_BACKING_FMT, | |
3328 | .type = QEMU_OPT_STRING, | |
3329 | .help = "Image format of the base image" | |
3330 | }, | |
3331 | { | |
3332 | .name = BLOCK_OPT_ENCRYPT, | |
3333 | .type = QEMU_OPT_BOOL, | |
3334 | .help = "Encrypt the image", | |
3335 | .def_value_str = "off" | |
3336 | }, | |
3337 | { | |
3338 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
3339 | .type = QEMU_OPT_SIZE, | |
3340 | .help = "qcow2 cluster size", | |
3341 | .def_value_str = stringify(DEFAULT_CLUSTER_SIZE) | |
3342 | }, | |
3343 | { | |
3344 | .name = BLOCK_OPT_PREALLOC, | |
3345 | .type = QEMU_OPT_STRING, | |
0e4271b7 HT |
3346 | .help = "Preallocation mode (allowed values: off, metadata, " |
3347 | "falloc, full)" | |
1bd0e2d1 CL |
3348 | }, |
3349 | { | |
3350 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
3351 | .type = QEMU_OPT_BOOL, | |
3352 | .help = "Postpone refcount updates", | |
3353 | .def_value_str = "off" | |
3354 | }, | |
06d05fa7 HR |
3355 | { |
3356 | .name = BLOCK_OPT_REFCOUNT_BITS, | |
3357 | .type = QEMU_OPT_NUMBER, | |
3358 | .help = "Width of a reference count entry in bits", | |
3359 | .def_value_str = "16" | |
3360 | }, | |
1bd0e2d1 CL |
3361 | { /* end of list */ } |
3362 | } | |
20d97356 BS |
3363 | }; |
3364 | ||
5f535a94 | 3365 | BlockDriver bdrv_qcow2 = { |
7c80ab3f | 3366 | .format_name = "qcow2", |
ff99129a | 3367 | .instance_size = sizeof(BDRVQcow2State), |
7c80ab3f JS |
3368 | .bdrv_probe = qcow2_probe, |
3369 | .bdrv_open = qcow2_open, | |
3370 | .bdrv_close = qcow2_close, | |
21d82ac9 | 3371 | .bdrv_reopen_prepare = qcow2_reopen_prepare, |
5b0959a7 KW |
3372 | .bdrv_reopen_commit = qcow2_reopen_commit, |
3373 | .bdrv_reopen_abort = qcow2_reopen_abort, | |
5365f44d | 3374 | .bdrv_join_options = qcow2_join_options, |
c282e1fd | 3375 | .bdrv_create = qcow2_create, |
3ac21627 | 3376 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 3377 | .bdrv_co_get_block_status = qcow2_co_get_block_status, |
7c80ab3f | 3378 | .bdrv_set_key = qcow2_set_key, |
7c80ab3f | 3379 | |
ecfe1863 | 3380 | .bdrv_co_preadv = qcow2_co_preadv, |
d46a0bb2 | 3381 | .bdrv_co_pwritev = qcow2_co_pwritev, |
eb489bb1 | 3382 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
419b19d9 | 3383 | |
5544b59f | 3384 | .bdrv_co_pwrite_zeroes = qcow2_co_pwrite_zeroes, |
82e8a788 | 3385 | .bdrv_co_pdiscard = qcow2_co_pdiscard, |
419b19d9 | 3386 | .bdrv_truncate = qcow2_truncate, |
fcccefc5 | 3387 | .bdrv_co_pwritev_compressed = qcow2_co_pwritev_compressed, |
491d27e2 | 3388 | .bdrv_make_empty = qcow2_make_empty, |
20d97356 BS |
3389 | |
3390 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
3391 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
3392 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
3393 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1bd0e2d1 CL |
3394 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
3395 | .bdrv_get_info = qcow2_get_info, | |
37764dfb | 3396 | .bdrv_get_specific_info = qcow2_get_specific_info, |
20d97356 | 3397 | |
7c80ab3f JS |
3398 | .bdrv_save_vmstate = qcow2_save_vmstate, |
3399 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
20d97356 | 3400 | |
8ee79e70 | 3401 | .supports_backing = true, |
20d97356 BS |
3402 | .bdrv_change_backing_file = qcow2_change_backing_file, |
3403 | ||
d34682cd | 3404 | .bdrv_refresh_limits = qcow2_refresh_limits, |
06d9260f | 3405 | .bdrv_invalidate_cache = qcow2_invalidate_cache, |
ec6d8912 | 3406 | .bdrv_inactivate = qcow2_inactivate, |
06d9260f | 3407 | |
1bd0e2d1 CL |
3408 | .create_opts = &qcow2_create_opts, |
3409 | .bdrv_check = qcow2_check, | |
c282e1fd | 3410 | .bdrv_amend_options = qcow2_amend_options, |
279621c0 AG |
3411 | |
3412 | .bdrv_detach_aio_context = qcow2_detach_aio_context, | |
3413 | .bdrv_attach_aio_context = qcow2_attach_aio_context, | |
20d97356 BS |
3414 | }; |
3415 | ||
5efa9d5a AL |
3416 | static void bdrv_qcow2_init(void) |
3417 | { | |
3418 | bdrv_register(&bdrv_qcow2); | |
3419 | } | |
3420 | ||
3421 | block_init(bdrv_qcow2_init); |