]>
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" |
faf07963 | 25 | #include "qemu-common.h" |
737e150e | 26 | #include "block/block_int.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" |
585f8587 FB |
38 | |
39 | /* | |
40 | Differences with QCOW: | |
41 | ||
42 | - Support for multiple incremental snapshots. | |
43 | - Memory management by reference counts. | |
44 | - Clusters which have a reference count of one have the bit | |
45 | QCOW_OFLAG_COPIED to optimize write performance. | |
5fafdf24 | 46 | - Size of compressed clusters is stored in sectors to reduce bit usage |
585f8587 FB |
47 | in the cluster offsets. |
48 | - Support for storing additional data (such as the VM state) in the | |
3b46e624 | 49 | snapshots. |
585f8587 FB |
50 | - If a backing store is used, the cluster size is not constrained |
51 | (could be backported to QCOW). | |
52 | - L2 tables have always a size of one cluster. | |
53 | */ | |
54 | ||
9b80ddf3 AL |
55 | |
56 | typedef struct { | |
57 | uint32_t magic; | |
58 | uint32_t len; | |
c4217f64 | 59 | } QEMU_PACKED QCowExtension; |
21d82ac9 | 60 | |
7c80ab3f JS |
61 | #define QCOW2_EXT_MAGIC_END 0 |
62 | #define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA | |
cfcc4c62 | 63 | #define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 |
9b80ddf3 | 64 | |
7c80ab3f | 65 | static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) |
585f8587 FB |
66 | { |
67 | const QCowHeader *cow_header = (const void *)buf; | |
3b46e624 | 68 | |
585f8587 FB |
69 | if (buf_size >= sizeof(QCowHeader) && |
70 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
6744cbab | 71 | be32_to_cpu(cow_header->version) >= 2) |
585f8587 FB |
72 | return 100; |
73 | else | |
74 | return 0; | |
75 | } | |
76 | ||
9b80ddf3 AL |
77 | |
78 | /* | |
79 | * read qcow2 extension and fill bs | |
80 | * start reading from start_offset | |
81 | * finish reading upon magic of value 0 or when end_offset reached | |
82 | * unknown magic is skipped (future extension this version knows nothing about) | |
83 | * return 0 upon success, non-0 otherwise | |
84 | */ | |
7c80ab3f | 85 | static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
3ef6c40a HR |
86 | uint64_t end_offset, void **p_feature_table, |
87 | Error **errp) | |
9b80ddf3 | 88 | { |
ff99129a | 89 | BDRVQcow2State *s = bs->opaque; |
9b80ddf3 AL |
90 | QCowExtension ext; |
91 | uint64_t offset; | |
75bab85c | 92 | int ret; |
9b80ddf3 AL |
93 | |
94 | #ifdef DEBUG_EXT | |
7c80ab3f | 95 | printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); |
9b80ddf3 AL |
96 | #endif |
97 | offset = start_offset; | |
98 | while (offset < end_offset) { | |
99 | ||
100 | #ifdef DEBUG_EXT | |
101 | /* Sanity check */ | |
102 | if (offset > s->cluster_size) | |
7c80ab3f | 103 | printf("qcow2_read_extension: suspicious offset %lu\n", offset); |
9b80ddf3 | 104 | |
9b2260cb | 105 | printf("attempting to read extended header in offset %lu\n", offset); |
9b80ddf3 AL |
106 | #endif |
107 | ||
9a4f4c31 | 108 | ret = bdrv_pread(bs->file->bs, offset, &ext, sizeof(ext)); |
3ef6c40a HR |
109 | if (ret < 0) { |
110 | error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: " | |
111 | "pread fail from offset %" PRIu64, offset); | |
9b80ddf3 AL |
112 | return 1; |
113 | } | |
114 | be32_to_cpus(&ext.magic); | |
115 | be32_to_cpus(&ext.len); | |
116 | offset += sizeof(ext); | |
117 | #ifdef DEBUG_EXT | |
118 | printf("ext.magic = 0x%x\n", ext.magic); | |
119 | #endif | |
2ebafc85 | 120 | if (offset > end_offset || ext.len > end_offset - offset) { |
3ef6c40a | 121 | error_setg(errp, "Header extension too large"); |
64ca6aee KW |
122 | return -EINVAL; |
123 | } | |
124 | ||
9b80ddf3 | 125 | switch (ext.magic) { |
7c80ab3f | 126 | case QCOW2_EXT_MAGIC_END: |
9b80ddf3 | 127 | return 0; |
f965509c | 128 | |
7c80ab3f | 129 | case QCOW2_EXT_MAGIC_BACKING_FORMAT: |
f965509c | 130 | if (ext.len >= sizeof(bs->backing_format)) { |
521b2b5d HR |
131 | error_setg(errp, "ERROR: ext_backing_format: len=%" PRIu32 |
132 | " too large (>=%zu)", ext.len, | |
133 | sizeof(bs->backing_format)); | |
f965509c AL |
134 | return 2; |
135 | } | |
9a4f4c31 | 136 | ret = bdrv_pread(bs->file->bs, offset, bs->backing_format, ext.len); |
3ef6c40a HR |
137 | if (ret < 0) { |
138 | error_setg_errno(errp, -ret, "ERROR: ext_backing_format: " | |
139 | "Could not read format name"); | |
f965509c | 140 | return 3; |
3ef6c40a | 141 | } |
f965509c | 142 | bs->backing_format[ext.len] = '\0'; |
e4603fe1 | 143 | s->image_backing_format = g_strdup(bs->backing_format); |
f965509c AL |
144 | #ifdef DEBUG_EXT |
145 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
146 | #endif | |
f965509c AL |
147 | break; |
148 | ||
cfcc4c62 KW |
149 | case QCOW2_EXT_MAGIC_FEATURE_TABLE: |
150 | if (p_feature_table != NULL) { | |
151 | void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); | |
9a4f4c31 | 152 | ret = bdrv_pread(bs->file->bs, offset , feature_table, ext.len); |
cfcc4c62 | 153 | if (ret < 0) { |
3ef6c40a HR |
154 | error_setg_errno(errp, -ret, "ERROR: ext_feature_table: " |
155 | "Could not read table"); | |
cfcc4c62 KW |
156 | return ret; |
157 | } | |
158 | ||
159 | *p_feature_table = feature_table; | |
160 | } | |
161 | break; | |
162 | ||
9b80ddf3 | 163 | default: |
75bab85c KW |
164 | /* unknown magic - save it in case we need to rewrite the header */ |
165 | { | |
166 | Qcow2UnknownHeaderExtension *uext; | |
167 | ||
168 | uext = g_malloc0(sizeof(*uext) + ext.len); | |
169 | uext->magic = ext.magic; | |
170 | uext->len = ext.len; | |
171 | QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); | |
172 | ||
9a4f4c31 | 173 | ret = bdrv_pread(bs->file->bs, offset , uext->data, uext->len); |
75bab85c | 174 | if (ret < 0) { |
3ef6c40a HR |
175 | error_setg_errno(errp, -ret, "ERROR: unknown extension: " |
176 | "Could not read data"); | |
75bab85c KW |
177 | return ret; |
178 | } | |
75bab85c | 179 | } |
9b80ddf3 AL |
180 | break; |
181 | } | |
fd29b4bb KW |
182 | |
183 | offset += ((ext.len + 7) & ~7); | |
9b80ddf3 AL |
184 | } |
185 | ||
186 | return 0; | |
187 | } | |
188 | ||
75bab85c KW |
189 | static void cleanup_unknown_header_ext(BlockDriverState *bs) |
190 | { | |
ff99129a | 191 | BDRVQcow2State *s = bs->opaque; |
75bab85c KW |
192 | Qcow2UnknownHeaderExtension *uext, *next; |
193 | ||
194 | QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { | |
195 | QLIST_REMOVE(uext, next); | |
196 | g_free(uext); | |
197 | } | |
198 | } | |
9b80ddf3 | 199 | |
3ef6c40a HR |
200 | static void GCC_FMT_ATTR(3, 4) report_unsupported(BlockDriverState *bs, |
201 | Error **errp, const char *fmt, ...) | |
6744cbab KW |
202 | { |
203 | char msg[64]; | |
204 | va_list ap; | |
205 | ||
206 | va_start(ap, fmt); | |
207 | vsnprintf(msg, sizeof(msg), fmt, ap); | |
208 | va_end(ap); | |
209 | ||
c6bd8c70 MA |
210 | error_setg(errp, QERR_UNKNOWN_BLOCK_FORMAT_FEATURE, |
211 | bdrv_get_device_or_node_name(bs), "qcow2", msg); | |
6744cbab KW |
212 | } |
213 | ||
cfcc4c62 | 214 | static void report_unsupported_feature(BlockDriverState *bs, |
3ef6c40a | 215 | Error **errp, Qcow2Feature *table, uint64_t mask) |
cfcc4c62 | 216 | { |
12ac6d3d KW |
217 | char *features = g_strdup(""); |
218 | char *old; | |
219 | ||
cfcc4c62 KW |
220 | while (table && table->name[0] != '\0') { |
221 | if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) { | |
12ac6d3d KW |
222 | if (mask & (1ULL << table->bit)) { |
223 | old = features; | |
224 | features = g_strdup_printf("%s%s%.46s", old, *old ? ", " : "", | |
225 | table->name); | |
226 | g_free(old); | |
227 | mask &= ~(1ULL << table->bit); | |
cfcc4c62 KW |
228 | } |
229 | } | |
230 | table++; | |
231 | } | |
232 | ||
233 | if (mask) { | |
12ac6d3d KW |
234 | old = features; |
235 | features = g_strdup_printf("%s%sUnknown incompatible feature: %" PRIx64, | |
236 | old, *old ? ", " : "", mask); | |
237 | g_free(old); | |
cfcc4c62 | 238 | } |
12ac6d3d KW |
239 | |
240 | report_unsupported(bs, errp, "%s", features); | |
241 | g_free(features); | |
cfcc4c62 KW |
242 | } |
243 | ||
bfe8043e SH |
244 | /* |
245 | * Sets the dirty bit and flushes afterwards if necessary. | |
246 | * | |
247 | * The incompatible_features bit is only set if the image file header was | |
248 | * updated successfully. Therefore it is not required to check the return | |
249 | * value of this function. | |
250 | */ | |
280d3735 | 251 | int qcow2_mark_dirty(BlockDriverState *bs) |
bfe8043e | 252 | { |
ff99129a | 253 | BDRVQcow2State *s = bs->opaque; |
bfe8043e SH |
254 | uint64_t val; |
255 | int ret; | |
256 | ||
257 | assert(s->qcow_version >= 3); | |
258 | ||
259 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
260 | return 0; /* already dirty */ | |
261 | } | |
262 | ||
263 | val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); | |
9a4f4c31 | 264 | ret = bdrv_pwrite(bs->file->bs, offsetof(QCowHeader, incompatible_features), |
bfe8043e SH |
265 | &val, sizeof(val)); |
266 | if (ret < 0) { | |
267 | return ret; | |
268 | } | |
9a4f4c31 | 269 | ret = bdrv_flush(bs->file->bs); |
bfe8043e SH |
270 | if (ret < 0) { |
271 | return ret; | |
272 | } | |
273 | ||
274 | /* Only treat image as dirty if the header was updated successfully */ | |
275 | s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; | |
276 | return 0; | |
277 | } | |
278 | ||
c61d0004 SH |
279 | /* |
280 | * Clears the dirty bit and flushes before if necessary. Only call this | |
281 | * function when there are no pending requests, it does not guard against | |
282 | * concurrent requests dirtying the image. | |
283 | */ | |
284 | static int qcow2_mark_clean(BlockDriverState *bs) | |
285 | { | |
ff99129a | 286 | BDRVQcow2State *s = bs->opaque; |
c61d0004 SH |
287 | |
288 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
4c2e5f8f KW |
289 | int ret; |
290 | ||
291 | s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; | |
292 | ||
293 | ret = bdrv_flush(bs); | |
c61d0004 SH |
294 | if (ret < 0) { |
295 | return ret; | |
296 | } | |
297 | ||
c61d0004 SH |
298 | return qcow2_update_header(bs); |
299 | } | |
300 | return 0; | |
301 | } | |
302 | ||
69c98726 HR |
303 | /* |
304 | * Marks the image as corrupt. | |
305 | */ | |
306 | int qcow2_mark_corrupt(BlockDriverState *bs) | |
307 | { | |
ff99129a | 308 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
309 | |
310 | s->incompatible_features |= QCOW2_INCOMPAT_CORRUPT; | |
311 | return qcow2_update_header(bs); | |
312 | } | |
313 | ||
314 | /* | |
315 | * Marks the image as consistent, i.e., unsets the corrupt bit, and flushes | |
316 | * before if necessary. | |
317 | */ | |
318 | int qcow2_mark_consistent(BlockDriverState *bs) | |
319 | { | |
ff99129a | 320 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
321 | |
322 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { | |
323 | int ret = bdrv_flush(bs); | |
324 | if (ret < 0) { | |
325 | return ret; | |
326 | } | |
327 | ||
328 | s->incompatible_features &= ~QCOW2_INCOMPAT_CORRUPT; | |
329 | return qcow2_update_header(bs); | |
330 | } | |
331 | return 0; | |
332 | } | |
333 | ||
acbe5982 SH |
334 | static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result, |
335 | BdrvCheckMode fix) | |
336 | { | |
337 | int ret = qcow2_check_refcounts(bs, result, fix); | |
338 | if (ret < 0) { | |
339 | return ret; | |
340 | } | |
341 | ||
342 | if (fix && result->check_errors == 0 && result->corruptions == 0) { | |
24530f3e HR |
343 | ret = qcow2_mark_clean(bs); |
344 | if (ret < 0) { | |
345 | return ret; | |
346 | } | |
347 | return qcow2_mark_consistent(bs); | |
acbe5982 SH |
348 | } |
349 | return ret; | |
350 | } | |
351 | ||
8c7de283 KW |
352 | static int validate_table_offset(BlockDriverState *bs, uint64_t offset, |
353 | uint64_t entries, size_t entry_len) | |
354 | { | |
ff99129a | 355 | BDRVQcow2State *s = bs->opaque; |
8c7de283 KW |
356 | uint64_t size; |
357 | ||
358 | /* Use signed INT64_MAX as the maximum even for uint64_t header fields, | |
359 | * because values will be passed to qemu functions taking int64_t. */ | |
360 | if (entries > INT64_MAX / entry_len) { | |
361 | return -EINVAL; | |
362 | } | |
363 | ||
364 | size = entries * entry_len; | |
365 | ||
366 | if (INT64_MAX - size < offset) { | |
367 | return -EINVAL; | |
368 | } | |
369 | ||
370 | /* Tables must be cluster aligned */ | |
371 | if (offset & (s->cluster_size - 1)) { | |
372 | return -EINVAL; | |
373 | } | |
374 | ||
375 | return 0; | |
376 | } | |
377 | ||
74c4510a KW |
378 | static QemuOptsList qcow2_runtime_opts = { |
379 | .name = "qcow2", | |
380 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head), | |
381 | .desc = { | |
382 | { | |
64aa99d3 | 383 | .name = QCOW2_OPT_LAZY_REFCOUNTS, |
74c4510a KW |
384 | .type = QEMU_OPT_BOOL, |
385 | .help = "Postpone refcount updates", | |
386 | }, | |
67af674e KW |
387 | { |
388 | .name = QCOW2_OPT_DISCARD_REQUEST, | |
389 | .type = QEMU_OPT_BOOL, | |
390 | .help = "Pass guest discard requests to the layer below", | |
391 | }, | |
392 | { | |
393 | .name = QCOW2_OPT_DISCARD_SNAPSHOT, | |
394 | .type = QEMU_OPT_BOOL, | |
395 | .help = "Generate discard requests when snapshot related space " | |
396 | "is freed", | |
397 | }, | |
398 | { | |
399 | .name = QCOW2_OPT_DISCARD_OTHER, | |
400 | .type = QEMU_OPT_BOOL, | |
401 | .help = "Generate discard requests when other clusters are freed", | |
402 | }, | |
05de7e86 HR |
403 | { |
404 | .name = QCOW2_OPT_OVERLAP, | |
405 | .type = QEMU_OPT_STRING, | |
406 | .help = "Selects which overlap checks to perform from a range of " | |
407 | "templates (none, constant, cached, all)", | |
408 | }, | |
ee42b5ce HR |
409 | { |
410 | .name = QCOW2_OPT_OVERLAP_TEMPLATE, | |
411 | .type = QEMU_OPT_STRING, | |
412 | .help = "Selects which overlap checks to perform from a range of " | |
413 | "templates (none, constant, cached, all)", | |
414 | }, | |
05de7e86 HR |
415 | { |
416 | .name = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
417 | .type = QEMU_OPT_BOOL, | |
418 | .help = "Check for unintended writes into the main qcow2 header", | |
419 | }, | |
420 | { | |
421 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
422 | .type = QEMU_OPT_BOOL, | |
423 | .help = "Check for unintended writes into the active L1 table", | |
424 | }, | |
425 | { | |
426 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
427 | .type = QEMU_OPT_BOOL, | |
428 | .help = "Check for unintended writes into an active L2 table", | |
429 | }, | |
430 | { | |
431 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
432 | .type = QEMU_OPT_BOOL, | |
433 | .help = "Check for unintended writes into the refcount table", | |
434 | }, | |
435 | { | |
436 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
437 | .type = QEMU_OPT_BOOL, | |
438 | .help = "Check for unintended writes into a refcount block", | |
439 | }, | |
440 | { | |
441 | .name = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
442 | .type = QEMU_OPT_BOOL, | |
443 | .help = "Check for unintended writes into the snapshot table", | |
444 | }, | |
445 | { | |
446 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
447 | .type = QEMU_OPT_BOOL, | |
448 | .help = "Check for unintended writes into an inactive L1 table", | |
449 | }, | |
450 | { | |
451 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
452 | .type = QEMU_OPT_BOOL, | |
453 | .help = "Check for unintended writes into an inactive L2 table", | |
454 | }, | |
6c1c8d5d HR |
455 | { |
456 | .name = QCOW2_OPT_CACHE_SIZE, | |
457 | .type = QEMU_OPT_SIZE, | |
458 | .help = "Maximum combined metadata (L2 tables and refcount blocks) " | |
459 | "cache size", | |
460 | }, | |
461 | { | |
462 | .name = QCOW2_OPT_L2_CACHE_SIZE, | |
463 | .type = QEMU_OPT_SIZE, | |
464 | .help = "Maximum L2 table cache size", | |
465 | }, | |
466 | { | |
467 | .name = QCOW2_OPT_REFCOUNT_CACHE_SIZE, | |
468 | .type = QEMU_OPT_SIZE, | |
469 | .help = "Maximum refcount block cache size", | |
470 | }, | |
279621c0 AG |
471 | { |
472 | .name = QCOW2_OPT_CACHE_CLEAN_INTERVAL, | |
473 | .type = QEMU_OPT_NUMBER, | |
474 | .help = "Clean unused cache entries after this time (in seconds)", | |
475 | }, | |
74c4510a KW |
476 | { /* end of list */ } |
477 | }, | |
478 | }; | |
479 | ||
4092e99d HR |
480 | static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = { |
481 | [QCOW2_OL_MAIN_HEADER_BITNR] = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
482 | [QCOW2_OL_ACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
483 | [QCOW2_OL_ACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
484 | [QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
485 | [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
486 | [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
487 | [QCOW2_OL_INACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
488 | [QCOW2_OL_INACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
489 | }; | |
490 | ||
279621c0 AG |
491 | static void cache_clean_timer_cb(void *opaque) |
492 | { | |
493 | BlockDriverState *bs = opaque; | |
ff99129a | 494 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
495 | qcow2_cache_clean_unused(bs, s->l2_table_cache); |
496 | qcow2_cache_clean_unused(bs, s->refcount_block_cache); | |
497 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
498 | (int64_t) s->cache_clean_interval * 1000); | |
499 | } | |
500 | ||
501 | static void cache_clean_timer_init(BlockDriverState *bs, AioContext *context) | |
502 | { | |
ff99129a | 503 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
504 | if (s->cache_clean_interval > 0) { |
505 | s->cache_clean_timer = aio_timer_new(context, QEMU_CLOCK_VIRTUAL, | |
506 | SCALE_MS, cache_clean_timer_cb, | |
507 | bs); | |
508 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
509 | (int64_t) s->cache_clean_interval * 1000); | |
510 | } | |
511 | } | |
512 | ||
513 | static void cache_clean_timer_del(BlockDriverState *bs) | |
514 | { | |
ff99129a | 515 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
516 | if (s->cache_clean_timer) { |
517 | timer_del(s->cache_clean_timer); | |
518 | timer_free(s->cache_clean_timer); | |
519 | s->cache_clean_timer = NULL; | |
520 | } | |
521 | } | |
522 | ||
523 | static void qcow2_detach_aio_context(BlockDriverState *bs) | |
524 | { | |
525 | cache_clean_timer_del(bs); | |
526 | } | |
527 | ||
528 | static void qcow2_attach_aio_context(BlockDriverState *bs, | |
529 | AioContext *new_context) | |
530 | { | |
531 | cache_clean_timer_init(bs, new_context); | |
532 | } | |
533 | ||
bc85ef26 HR |
534 | static void read_cache_sizes(BlockDriverState *bs, QemuOpts *opts, |
535 | uint64_t *l2_cache_size, | |
6c1c8d5d HR |
536 | uint64_t *refcount_cache_size, Error **errp) |
537 | { | |
ff99129a | 538 | BDRVQcow2State *s = bs->opaque; |
6c1c8d5d HR |
539 | uint64_t combined_cache_size; |
540 | bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set; | |
541 | ||
542 | combined_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_CACHE_SIZE); | |
543 | l2_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_SIZE); | |
544 | refcount_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
545 | ||
546 | combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0); | |
547 | *l2_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE, 0); | |
548 | *refcount_cache_size = qemu_opt_get_size(opts, | |
549 | QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0); | |
550 | ||
551 | if (combined_cache_size_set) { | |
552 | if (l2_cache_size_set && refcount_cache_size_set) { | |
553 | error_setg(errp, QCOW2_OPT_CACHE_SIZE ", " QCOW2_OPT_L2_CACHE_SIZE | |
554 | " and " QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not be set " | |
555 | "the same time"); | |
556 | return; | |
557 | } else if (*l2_cache_size > combined_cache_size) { | |
558 | error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed " | |
559 | QCOW2_OPT_CACHE_SIZE); | |
560 | return; | |
561 | } else if (*refcount_cache_size > combined_cache_size) { | |
562 | error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed " | |
563 | QCOW2_OPT_CACHE_SIZE); | |
564 | return; | |
565 | } | |
566 | ||
567 | if (l2_cache_size_set) { | |
568 | *refcount_cache_size = combined_cache_size - *l2_cache_size; | |
569 | } else if (refcount_cache_size_set) { | |
570 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
571 | } else { | |
572 | *refcount_cache_size = combined_cache_size | |
573 | / (DEFAULT_L2_REFCOUNT_SIZE_RATIO + 1); | |
574 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
575 | } | |
576 | } else { | |
577 | if (!l2_cache_size_set && !refcount_cache_size_set) { | |
bc85ef26 HR |
578 | *l2_cache_size = MAX(DEFAULT_L2_CACHE_BYTE_SIZE, |
579 | (uint64_t)DEFAULT_L2_CACHE_CLUSTERS | |
580 | * s->cluster_size); | |
6c1c8d5d HR |
581 | *refcount_cache_size = *l2_cache_size |
582 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
583 | } else if (!l2_cache_size_set) { | |
584 | *l2_cache_size = *refcount_cache_size | |
585 | * DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
586 | } else if (!refcount_cache_size_set) { | |
587 | *refcount_cache_size = *l2_cache_size | |
588 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
589 | } | |
590 | } | |
591 | } | |
592 | ||
ee55b173 KW |
593 | typedef struct Qcow2ReopenState { |
594 | Qcow2Cache *l2_table_cache; | |
595 | Qcow2Cache *refcount_block_cache; | |
596 | bool use_lazy_refcounts; | |
597 | int overlap_check; | |
598 | bool discard_passthrough[QCOW2_DISCARD_MAX]; | |
599 | uint64_t cache_clean_interval; | |
600 | } Qcow2ReopenState; | |
601 | ||
602 | static int qcow2_update_options_prepare(BlockDriverState *bs, | |
603 | Qcow2ReopenState *r, | |
604 | QDict *options, int flags, | |
605 | Error **errp) | |
4c75d1a1 KW |
606 | { |
607 | BDRVQcow2State *s = bs->opaque; | |
94edf3fb | 608 | QemuOpts *opts = NULL; |
4c75d1a1 KW |
609 | const char *opt_overlap_check, *opt_overlap_check_template; |
610 | int overlap_check_template = 0; | |
94edf3fb | 611 | uint64_t l2_cache_size, refcount_cache_size; |
4c75d1a1 | 612 | int i; |
94edf3fb | 613 | Error *local_err = NULL; |
4c75d1a1 KW |
614 | int ret; |
615 | ||
94edf3fb KW |
616 | opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort); |
617 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
618 | if (local_err) { | |
619 | error_propagate(errp, local_err); | |
620 | ret = -EINVAL; | |
621 | goto fail; | |
622 | } | |
623 | ||
624 | /* get L2 table/refcount block cache size from command line options */ | |
625 | read_cache_sizes(bs, opts, &l2_cache_size, &refcount_cache_size, | |
626 | &local_err); | |
627 | if (local_err) { | |
628 | error_propagate(errp, local_err); | |
629 | ret = -EINVAL; | |
630 | goto fail; | |
631 | } | |
632 | ||
633 | l2_cache_size /= s->cluster_size; | |
634 | if (l2_cache_size < MIN_L2_CACHE_SIZE) { | |
635 | l2_cache_size = MIN_L2_CACHE_SIZE; | |
636 | } | |
637 | if (l2_cache_size > INT_MAX) { | |
638 | error_setg(errp, "L2 cache size too big"); | |
639 | ret = -EINVAL; | |
640 | goto fail; | |
641 | } | |
642 | ||
643 | refcount_cache_size /= s->cluster_size; | |
644 | if (refcount_cache_size < MIN_REFCOUNT_CACHE_SIZE) { | |
645 | refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE; | |
646 | } | |
647 | if (refcount_cache_size > INT_MAX) { | |
648 | error_setg(errp, "Refcount cache size too big"); | |
649 | ret = -EINVAL; | |
650 | goto fail; | |
651 | } | |
652 | ||
5b0959a7 KW |
653 | /* alloc new L2 table/refcount block cache, flush old one */ |
654 | if (s->l2_table_cache) { | |
655 | ret = qcow2_cache_flush(bs, s->l2_table_cache); | |
656 | if (ret) { | |
657 | error_setg_errno(errp, -ret, "Failed to flush the L2 table cache"); | |
658 | goto fail; | |
659 | } | |
660 | } | |
661 | ||
662 | if (s->refcount_block_cache) { | |
663 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
664 | if (ret) { | |
665 | error_setg_errno(errp, -ret, | |
666 | "Failed to flush the refcount block cache"); | |
667 | goto fail; | |
668 | } | |
669 | } | |
670 | ||
ee55b173 KW |
671 | r->l2_table_cache = qcow2_cache_create(bs, l2_cache_size); |
672 | r->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size); | |
673 | if (r->l2_table_cache == NULL || r->refcount_block_cache == NULL) { | |
94edf3fb KW |
674 | error_setg(errp, "Could not allocate metadata caches"); |
675 | ret = -ENOMEM; | |
676 | goto fail; | |
677 | } | |
678 | ||
679 | /* New interval for cache cleanup timer */ | |
ee55b173 | 680 | r->cache_clean_interval = |
5b0959a7 KW |
681 | qemu_opt_get_number(opts, QCOW2_OPT_CACHE_CLEAN_INTERVAL, |
682 | s->cache_clean_interval); | |
ee55b173 | 683 | if (r->cache_clean_interval > UINT_MAX) { |
94edf3fb KW |
684 | error_setg(errp, "Cache clean interval too big"); |
685 | ret = -EINVAL; | |
686 | goto fail; | |
687 | } | |
94edf3fb | 688 | |
5b0959a7 | 689 | /* lazy-refcounts; flush if going from enabled to disabled */ |
ee55b173 | 690 | r->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, |
4c75d1a1 | 691 | (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); |
ee55b173 | 692 | if (r->use_lazy_refcounts && s->qcow_version < 3) { |
007dbc39 KW |
693 | error_setg(errp, "Lazy refcounts require a qcow2 image with at least " |
694 | "qemu 1.1 compatibility level"); | |
695 | ret = -EINVAL; | |
696 | goto fail; | |
697 | } | |
4c75d1a1 | 698 | |
5b0959a7 KW |
699 | if (s->use_lazy_refcounts && !r->use_lazy_refcounts) { |
700 | ret = qcow2_mark_clean(bs); | |
701 | if (ret < 0) { | |
702 | error_setg_errno(errp, -ret, "Failed to disable lazy refcounts"); | |
703 | goto fail; | |
704 | } | |
705 | } | |
706 | ||
007dbc39 | 707 | /* Overlap check options */ |
4c75d1a1 KW |
708 | opt_overlap_check = qemu_opt_get(opts, QCOW2_OPT_OVERLAP); |
709 | opt_overlap_check_template = qemu_opt_get(opts, QCOW2_OPT_OVERLAP_TEMPLATE); | |
710 | if (opt_overlap_check_template && opt_overlap_check && | |
711 | strcmp(opt_overlap_check_template, opt_overlap_check)) | |
712 | { | |
713 | error_setg(errp, "Conflicting values for qcow2 options '" | |
714 | QCOW2_OPT_OVERLAP "' ('%s') and '" QCOW2_OPT_OVERLAP_TEMPLATE | |
715 | "' ('%s')", opt_overlap_check, opt_overlap_check_template); | |
716 | ret = -EINVAL; | |
717 | goto fail; | |
718 | } | |
719 | if (!opt_overlap_check) { | |
720 | opt_overlap_check = opt_overlap_check_template ?: "cached"; | |
721 | } | |
722 | ||
723 | if (!strcmp(opt_overlap_check, "none")) { | |
724 | overlap_check_template = 0; | |
725 | } else if (!strcmp(opt_overlap_check, "constant")) { | |
726 | overlap_check_template = QCOW2_OL_CONSTANT; | |
727 | } else if (!strcmp(opt_overlap_check, "cached")) { | |
728 | overlap_check_template = QCOW2_OL_CACHED; | |
729 | } else if (!strcmp(opt_overlap_check, "all")) { | |
730 | overlap_check_template = QCOW2_OL_ALL; | |
731 | } else { | |
732 | error_setg(errp, "Unsupported value '%s' for qcow2 option " | |
733 | "'overlap-check'. Allowed are any of the following: " | |
734 | "none, constant, cached, all", opt_overlap_check); | |
735 | ret = -EINVAL; | |
736 | goto fail; | |
737 | } | |
738 | ||
ee55b173 | 739 | r->overlap_check = 0; |
4c75d1a1 KW |
740 | for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) { |
741 | /* overlap-check defines a template bitmask, but every flag may be | |
742 | * overwritten through the associated boolean option */ | |
ee55b173 | 743 | r->overlap_check |= |
4c75d1a1 KW |
744 | qemu_opt_get_bool(opts, overlap_bool_option_names[i], |
745 | overlap_check_template & (1 << i)) << i; | |
746 | } | |
747 | ||
ee55b173 KW |
748 | r->discard_passthrough[QCOW2_DISCARD_NEVER] = false; |
749 | r->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true; | |
750 | r->discard_passthrough[QCOW2_DISCARD_REQUEST] = | |
007dbc39 KW |
751 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST, |
752 | flags & BDRV_O_UNMAP); | |
ee55b173 | 753 | r->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] = |
007dbc39 | 754 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true); |
ee55b173 | 755 | r->discard_passthrough[QCOW2_DISCARD_OTHER] = |
007dbc39 KW |
756 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false); |
757 | ||
4c75d1a1 KW |
758 | ret = 0; |
759 | fail: | |
94edf3fb KW |
760 | qemu_opts_del(opts); |
761 | opts = NULL; | |
ee55b173 KW |
762 | return ret; |
763 | } | |
764 | ||
765 | static void qcow2_update_options_commit(BlockDriverState *bs, | |
766 | Qcow2ReopenState *r) | |
767 | { | |
768 | BDRVQcow2State *s = bs->opaque; | |
769 | int i; | |
770 | ||
5b0959a7 KW |
771 | if (s->l2_table_cache) { |
772 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
773 | } | |
774 | if (s->refcount_block_cache) { | |
775 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
776 | } | |
ee55b173 KW |
777 | s->l2_table_cache = r->l2_table_cache; |
778 | s->refcount_block_cache = r->refcount_block_cache; | |
779 | ||
780 | s->overlap_check = r->overlap_check; | |
781 | s->use_lazy_refcounts = r->use_lazy_refcounts; | |
782 | ||
783 | for (i = 0; i < QCOW2_DISCARD_MAX; i++) { | |
784 | s->discard_passthrough[i] = r->discard_passthrough[i]; | |
785 | } | |
786 | ||
5b0959a7 KW |
787 | if (s->cache_clean_interval != r->cache_clean_interval) { |
788 | cache_clean_timer_del(bs); | |
789 | s->cache_clean_interval = r->cache_clean_interval; | |
790 | cache_clean_timer_init(bs, bdrv_get_aio_context(bs)); | |
791 | } | |
ee55b173 KW |
792 | } |
793 | ||
794 | static void qcow2_update_options_abort(BlockDriverState *bs, | |
795 | Qcow2ReopenState *r) | |
796 | { | |
797 | if (r->l2_table_cache) { | |
798 | qcow2_cache_destroy(bs, r->l2_table_cache); | |
799 | } | |
800 | if (r->refcount_block_cache) { | |
801 | qcow2_cache_destroy(bs, r->refcount_block_cache); | |
802 | } | |
803 | } | |
804 | ||
805 | static int qcow2_update_options(BlockDriverState *bs, QDict *options, | |
806 | int flags, Error **errp) | |
807 | { | |
808 | Qcow2ReopenState r = {}; | |
809 | int ret; | |
810 | ||
811 | ret = qcow2_update_options_prepare(bs, &r, options, flags, errp); | |
812 | if (ret >= 0) { | |
813 | qcow2_update_options_commit(bs, &r); | |
814 | } else { | |
815 | qcow2_update_options_abort(bs, &r); | |
816 | } | |
94edf3fb | 817 | |
4c75d1a1 KW |
818 | return ret; |
819 | } | |
820 | ||
015a1036 HR |
821 | static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, |
822 | Error **errp) | |
585f8587 | 823 | { |
ff99129a | 824 | BDRVQcow2State *s = bs->opaque; |
6d33e8e7 KW |
825 | unsigned int len, i; |
826 | int ret = 0; | |
585f8587 | 827 | QCowHeader header; |
74c4510a | 828 | Error *local_err = NULL; |
9b80ddf3 | 829 | uint64_t ext_end; |
2cf7cfa1 | 830 | uint64_t l1_vm_state_index; |
585f8587 | 831 | |
9a4f4c31 | 832 | ret = bdrv_pread(bs->file->bs, 0, &header, sizeof(header)); |
6d85a57e | 833 | if (ret < 0) { |
3ef6c40a | 834 | error_setg_errno(errp, -ret, "Could not read qcow2 header"); |
585f8587 | 835 | goto fail; |
6d85a57e | 836 | } |
585f8587 FB |
837 | be32_to_cpus(&header.magic); |
838 | be32_to_cpus(&header.version); | |
839 | be64_to_cpus(&header.backing_file_offset); | |
840 | be32_to_cpus(&header.backing_file_size); | |
841 | be64_to_cpus(&header.size); | |
842 | be32_to_cpus(&header.cluster_bits); | |
843 | be32_to_cpus(&header.crypt_method); | |
844 | be64_to_cpus(&header.l1_table_offset); | |
845 | be32_to_cpus(&header.l1_size); | |
846 | be64_to_cpus(&header.refcount_table_offset); | |
847 | be32_to_cpus(&header.refcount_table_clusters); | |
848 | be64_to_cpus(&header.snapshots_offset); | |
849 | be32_to_cpus(&header.nb_snapshots); | |
3b46e624 | 850 | |
e8cdcec1 | 851 | if (header.magic != QCOW_MAGIC) { |
3ef6c40a | 852 | error_setg(errp, "Image is not in qcow2 format"); |
76abe407 | 853 | ret = -EINVAL; |
585f8587 | 854 | goto fail; |
6d85a57e | 855 | } |
6744cbab | 856 | if (header.version < 2 || header.version > 3) { |
521b2b5d | 857 | report_unsupported(bs, errp, "QCOW version %" PRIu32, header.version); |
6744cbab KW |
858 | ret = -ENOTSUP; |
859 | goto fail; | |
860 | } | |
861 | ||
862 | s->qcow_version = header.version; | |
863 | ||
24342f2c KW |
864 | /* Initialise cluster size */ |
865 | if (header.cluster_bits < MIN_CLUSTER_BITS || | |
866 | header.cluster_bits > MAX_CLUSTER_BITS) { | |
521b2b5d HR |
867 | error_setg(errp, "Unsupported cluster size: 2^%" PRIu32, |
868 | header.cluster_bits); | |
24342f2c KW |
869 | ret = -EINVAL; |
870 | goto fail; | |
871 | } | |
872 | ||
873 | s->cluster_bits = header.cluster_bits; | |
874 | s->cluster_size = 1 << s->cluster_bits; | |
875 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
876 | ||
6744cbab KW |
877 | /* Initialise version 3 header fields */ |
878 | if (header.version == 2) { | |
879 | header.incompatible_features = 0; | |
880 | header.compatible_features = 0; | |
881 | header.autoclear_features = 0; | |
882 | header.refcount_order = 4; | |
883 | header.header_length = 72; | |
884 | } else { | |
885 | be64_to_cpus(&header.incompatible_features); | |
886 | be64_to_cpus(&header.compatible_features); | |
887 | be64_to_cpus(&header.autoclear_features); | |
888 | be32_to_cpus(&header.refcount_order); | |
889 | be32_to_cpus(&header.header_length); | |
24342f2c KW |
890 | |
891 | if (header.header_length < 104) { | |
892 | error_setg(errp, "qcow2 header too short"); | |
893 | ret = -EINVAL; | |
894 | goto fail; | |
895 | } | |
896 | } | |
897 | ||
898 | if (header.header_length > s->cluster_size) { | |
899 | error_setg(errp, "qcow2 header exceeds cluster size"); | |
900 | ret = -EINVAL; | |
901 | goto fail; | |
6744cbab KW |
902 | } |
903 | ||
904 | if (header.header_length > sizeof(header)) { | |
905 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
906 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
9a4f4c31 | 907 | ret = bdrv_pread(bs->file->bs, sizeof(header), s->unknown_header_fields, |
6744cbab KW |
908 | s->unknown_header_fields_size); |
909 | if (ret < 0) { | |
3ef6c40a HR |
910 | error_setg_errno(errp, -ret, "Could not read unknown qcow2 header " |
911 | "fields"); | |
6744cbab KW |
912 | goto fail; |
913 | } | |
914 | } | |
915 | ||
a1b3955c KW |
916 | if (header.backing_file_offset > s->cluster_size) { |
917 | error_setg(errp, "Invalid backing file offset"); | |
918 | ret = -EINVAL; | |
919 | goto fail; | |
920 | } | |
921 | ||
cfcc4c62 KW |
922 | if (header.backing_file_offset) { |
923 | ext_end = header.backing_file_offset; | |
924 | } else { | |
925 | ext_end = 1 << header.cluster_bits; | |
926 | } | |
927 | ||
6744cbab KW |
928 | /* Handle feature bits */ |
929 | s->incompatible_features = header.incompatible_features; | |
930 | s->compatible_features = header.compatible_features; | |
931 | s->autoclear_features = header.autoclear_features; | |
932 | ||
c61d0004 | 933 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { |
cfcc4c62 KW |
934 | void *feature_table = NULL; |
935 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
3ef6c40a HR |
936 | &feature_table, NULL); |
937 | report_unsupported_feature(bs, errp, feature_table, | |
c61d0004 SH |
938 | s->incompatible_features & |
939 | ~QCOW2_INCOMPAT_MASK); | |
6744cbab | 940 | ret = -ENOTSUP; |
c5a33ee9 | 941 | g_free(feature_table); |
6744cbab KW |
942 | goto fail; |
943 | } | |
944 | ||
69c98726 HR |
945 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { |
946 | /* Corrupt images may not be written to unless they are being repaired | |
947 | */ | |
948 | if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) { | |
3ef6c40a HR |
949 | error_setg(errp, "qcow2: Image is corrupt; cannot be opened " |
950 | "read/write"); | |
69c98726 HR |
951 | ret = -EACCES; |
952 | goto fail; | |
953 | } | |
954 | } | |
955 | ||
6744cbab | 956 | /* Check support for various header values */ |
b72faf9f HR |
957 | if (header.refcount_order > 6) { |
958 | error_setg(errp, "Reference count entry width too large; may not " | |
959 | "exceed 64 bits"); | |
960 | ret = -EINVAL; | |
e8cdcec1 KW |
961 | goto fail; |
962 | } | |
b6481f37 | 963 | s->refcount_order = header.refcount_order; |
346a53df HR |
964 | s->refcount_bits = 1 << s->refcount_order; |
965 | s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1); | |
966 | s->refcount_max += s->refcount_max - 1; | |
6744cbab | 967 | |
6d85a57e | 968 | if (header.crypt_method > QCOW_CRYPT_AES) { |
521b2b5d | 969 | error_setg(errp, "Unsupported encryption method: %" PRIu32, |
3ef6c40a | 970 | header.crypt_method); |
6d85a57e | 971 | ret = -EINVAL; |
585f8587 | 972 | goto fail; |
6d85a57e | 973 | } |
f6fa64f6 DB |
974 | if (!qcrypto_cipher_supports(QCRYPTO_CIPHER_ALG_AES_128)) { |
975 | error_setg(errp, "AES cipher not available"); | |
976 | ret = -EINVAL; | |
977 | goto fail; | |
978 | } | |
585f8587 | 979 | s->crypt_method_header = header.crypt_method; |
6d85a57e | 980 | if (s->crypt_method_header) { |
585f8587 | 981 | bs->encrypted = 1; |
6d85a57e | 982 | } |
24342f2c | 983 | |
585f8587 FB |
984 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
985 | s->l2_size = 1 << s->l2_bits; | |
1d13d654 HR |
986 | /* 2^(s->refcount_order - 3) is the refcount width in bytes */ |
987 | s->refcount_block_bits = s->cluster_bits - (s->refcount_order - 3); | |
988 | s->refcount_block_size = 1 << s->refcount_block_bits; | |
585f8587 FB |
989 | bs->total_sectors = header.size / 512; |
990 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
991 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
992 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
5dab2fad | 993 | |
585f8587 | 994 | s->refcount_table_offset = header.refcount_table_offset; |
5fafdf24 | 995 | s->refcount_table_size = |
585f8587 FB |
996 | header.refcount_table_clusters << (s->cluster_bits - 3); |
997 | ||
2b5d5953 | 998 | if (header.refcount_table_clusters > qcow2_max_refcount_clusters(s)) { |
5dab2fad KW |
999 | error_setg(errp, "Reference count table too large"); |
1000 | ret = -EINVAL; | |
1001 | goto fail; | |
1002 | } | |
1003 | ||
8c7de283 KW |
1004 | ret = validate_table_offset(bs, s->refcount_table_offset, |
1005 | s->refcount_table_size, sizeof(uint64_t)); | |
1006 | if (ret < 0) { | |
1007 | error_setg(errp, "Invalid reference count table offset"); | |
1008 | goto fail; | |
1009 | } | |
1010 | ||
ce48f2f4 KW |
1011 | /* Snapshot table offset/length */ |
1012 | if (header.nb_snapshots > QCOW_MAX_SNAPSHOTS) { | |
1013 | error_setg(errp, "Too many snapshots"); | |
1014 | ret = -EINVAL; | |
1015 | goto fail; | |
1016 | } | |
1017 | ||
1018 | ret = validate_table_offset(bs, header.snapshots_offset, | |
1019 | header.nb_snapshots, | |
1020 | sizeof(QCowSnapshotHeader)); | |
1021 | if (ret < 0) { | |
1022 | error_setg(errp, "Invalid snapshot table offset"); | |
1023 | goto fail; | |
1024 | } | |
1025 | ||
585f8587 | 1026 | /* read the level 1 table */ |
87b86e7e | 1027 | if (header.l1_size > QCOW_MAX_L1_SIZE / sizeof(uint64_t)) { |
2d51c32c KW |
1028 | error_setg(errp, "Active L1 table too large"); |
1029 | ret = -EFBIG; | |
1030 | goto fail; | |
1031 | } | |
585f8587 | 1032 | s->l1_size = header.l1_size; |
2cf7cfa1 KW |
1033 | |
1034 | l1_vm_state_index = size_to_l1(s, header.size); | |
1035 | if (l1_vm_state_index > INT_MAX) { | |
3ef6c40a | 1036 | error_setg(errp, "Image is too big"); |
2cf7cfa1 KW |
1037 | ret = -EFBIG; |
1038 | goto fail; | |
1039 | } | |
1040 | s->l1_vm_state_index = l1_vm_state_index; | |
1041 | ||
585f8587 FB |
1042 | /* the L1 table must contain at least enough entries to put |
1043 | header.size bytes */ | |
6d85a57e | 1044 | if (s->l1_size < s->l1_vm_state_index) { |
3ef6c40a | 1045 | error_setg(errp, "L1 table is too small"); |
6d85a57e | 1046 | ret = -EINVAL; |
585f8587 | 1047 | goto fail; |
6d85a57e | 1048 | } |
2d51c32c KW |
1049 | |
1050 | ret = validate_table_offset(bs, header.l1_table_offset, | |
1051 | header.l1_size, sizeof(uint64_t)); | |
1052 | if (ret < 0) { | |
1053 | error_setg(errp, "Invalid L1 table offset"); | |
1054 | goto fail; | |
1055 | } | |
585f8587 | 1056 | s->l1_table_offset = header.l1_table_offset; |
2d51c32c KW |
1057 | |
1058 | ||
d191d12d | 1059 | if (s->l1_size > 0) { |
9a4f4c31 | 1060 | s->l1_table = qemu_try_blockalign(bs->file->bs, |
d191d12d | 1061 | align_offset(s->l1_size * sizeof(uint64_t), 512)); |
de82815d KW |
1062 | if (s->l1_table == NULL) { |
1063 | error_setg(errp, "Could not allocate L1 table"); | |
1064 | ret = -ENOMEM; | |
1065 | goto fail; | |
1066 | } | |
9a4f4c31 | 1067 | ret = bdrv_pread(bs->file->bs, s->l1_table_offset, s->l1_table, |
6d85a57e JS |
1068 | s->l1_size * sizeof(uint64_t)); |
1069 | if (ret < 0) { | |
3ef6c40a | 1070 | error_setg_errno(errp, -ret, "Could not read L1 table"); |
d191d12d | 1071 | goto fail; |
6d85a57e | 1072 | } |
d191d12d SW |
1073 | for(i = 0;i < s->l1_size; i++) { |
1074 | be64_to_cpus(&s->l1_table[i]); | |
1075 | } | |
585f8587 | 1076 | } |
29c1a730 | 1077 | |
94edf3fb KW |
1078 | /* Parse driver-specific options */ |
1079 | ret = qcow2_update_options(bs, options, flags, errp); | |
90efa0ea KW |
1080 | if (ret < 0) { |
1081 | goto fail; | |
1082 | } | |
1083 | ||
7267c094 | 1084 | s->cluster_cache = g_malloc(s->cluster_size); |
585f8587 | 1085 | /* one more sector for decompressed data alignment */ |
9a4f4c31 | 1086 | s->cluster_data = qemu_try_blockalign(bs->file->bs, QCOW_MAX_CRYPT_CLUSTERS |
de82815d KW |
1087 | * s->cluster_size + 512); |
1088 | if (s->cluster_data == NULL) { | |
1089 | error_setg(errp, "Could not allocate temporary cluster buffer"); | |
1090 | ret = -ENOMEM; | |
1091 | goto fail; | |
1092 | } | |
1093 | ||
585f8587 | 1094 | s->cluster_cache_offset = -1; |
06d9260f | 1095 | s->flags = flags; |
3b46e624 | 1096 | |
6d85a57e JS |
1097 | ret = qcow2_refcount_init(bs); |
1098 | if (ret != 0) { | |
3ef6c40a | 1099 | error_setg_errno(errp, -ret, "Could not initialize refcount handling"); |
585f8587 | 1100 | goto fail; |
6d85a57e | 1101 | } |
585f8587 | 1102 | |
72cf2d4f | 1103 | QLIST_INIT(&s->cluster_allocs); |
0b919fae | 1104 | QTAILQ_INIT(&s->discards); |
f214978a | 1105 | |
9b80ddf3 | 1106 | /* read qcow2 extensions */ |
3ef6c40a HR |
1107 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL, |
1108 | &local_err)) { | |
1109 | error_propagate(errp, local_err); | |
6d85a57e | 1110 | ret = -EINVAL; |
9b80ddf3 | 1111 | goto fail; |
6d85a57e | 1112 | } |
9b80ddf3 | 1113 | |
585f8587 FB |
1114 | /* read the backing file name */ |
1115 | if (header.backing_file_offset != 0) { | |
1116 | len = header.backing_file_size; | |
9a29e18f | 1117 | if (len > MIN(1023, s->cluster_size - header.backing_file_offset) || |
e729fa6a | 1118 | len >= sizeof(bs->backing_file)) { |
6d33e8e7 KW |
1119 | error_setg(errp, "Backing file name too long"); |
1120 | ret = -EINVAL; | |
1121 | goto fail; | |
6d85a57e | 1122 | } |
9a4f4c31 | 1123 | ret = bdrv_pread(bs->file->bs, header.backing_file_offset, |
6d85a57e JS |
1124 | bs->backing_file, len); |
1125 | if (ret < 0) { | |
3ef6c40a | 1126 | error_setg_errno(errp, -ret, "Could not read backing file name"); |
585f8587 | 1127 | goto fail; |
6d85a57e | 1128 | } |
585f8587 | 1129 | bs->backing_file[len] = '\0'; |
e4603fe1 | 1130 | s->image_backing_file = g_strdup(bs->backing_file); |
585f8587 | 1131 | } |
42deb29f | 1132 | |
11b128f4 KW |
1133 | /* Internal snapshots */ |
1134 | s->snapshots_offset = header.snapshots_offset; | |
1135 | s->nb_snapshots = header.nb_snapshots; | |
1136 | ||
42deb29f KW |
1137 | ret = qcow2_read_snapshots(bs); |
1138 | if (ret < 0) { | |
3ef6c40a | 1139 | error_setg_errno(errp, -ret, "Could not read snapshots"); |
585f8587 | 1140 | goto fail; |
6d85a57e | 1141 | } |
585f8587 | 1142 | |
af7b708d | 1143 | /* Clear unknown autoclear feature bits */ |
27eb6c09 | 1144 | if (!bs->read_only && !(flags & BDRV_O_INCOMING) && s->autoclear_features) { |
af7b708d SH |
1145 | s->autoclear_features = 0; |
1146 | ret = qcow2_update_header(bs); | |
1147 | if (ret < 0) { | |
3ef6c40a | 1148 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); |
af7b708d SH |
1149 | goto fail; |
1150 | } | |
1151 | } | |
1152 | ||
68d100e9 KW |
1153 | /* Initialise locks */ |
1154 | qemu_co_mutex_init(&s->lock); | |
1155 | ||
c61d0004 | 1156 | /* Repair image if dirty */ |
27eb6c09 | 1157 | if (!(flags & (BDRV_O_CHECK | BDRV_O_INCOMING)) && !bs->read_only && |
058f8f16 | 1158 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { |
c61d0004 SH |
1159 | BdrvCheckResult result = {0}; |
1160 | ||
5b84106b | 1161 | ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS); |
c61d0004 | 1162 | if (ret < 0) { |
3ef6c40a | 1163 | error_setg_errno(errp, -ret, "Could not repair dirty image"); |
c61d0004 SH |
1164 | goto fail; |
1165 | } | |
1166 | } | |
1167 | ||
585f8587 | 1168 | #ifdef DEBUG_ALLOC |
6cbc3031 PH |
1169 | { |
1170 | BdrvCheckResult result = {0}; | |
b35278f7 | 1171 | qcow2_check_refcounts(bs, &result, 0); |
6cbc3031 | 1172 | } |
585f8587 | 1173 | #endif |
6d85a57e | 1174 | return ret; |
585f8587 FB |
1175 | |
1176 | fail: | |
6744cbab | 1177 | g_free(s->unknown_header_fields); |
75bab85c | 1178 | cleanup_unknown_header_ext(bs); |
ed6ccf0f KW |
1179 | qcow2_free_snapshots(bs); |
1180 | qcow2_refcount_close(bs); | |
de82815d | 1181 | qemu_vfree(s->l1_table); |
cf93980e HR |
1182 | /* else pre-write overlap checks in cache_destroy may crash */ |
1183 | s->l1_table = NULL; | |
279621c0 | 1184 | cache_clean_timer_del(bs); |
29c1a730 KW |
1185 | if (s->l2_table_cache) { |
1186 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
1187 | } | |
c5a33ee9 PJ |
1188 | if (s->refcount_block_cache) { |
1189 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1190 | } | |
7267c094 | 1191 | g_free(s->cluster_cache); |
dea43a65 | 1192 | qemu_vfree(s->cluster_data); |
6d85a57e | 1193 | return ret; |
585f8587 FB |
1194 | } |
1195 | ||
3baca891 | 1196 | static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd | 1197 | { |
ff99129a | 1198 | BDRVQcow2State *s = bs->opaque; |
d34682cd KW |
1199 | |
1200 | bs->bl.write_zeroes_alignment = s->cluster_sectors; | |
d34682cd KW |
1201 | } |
1202 | ||
7c80ab3f | 1203 | static int qcow2_set_key(BlockDriverState *bs, const char *key) |
585f8587 | 1204 | { |
ff99129a | 1205 | BDRVQcow2State *s = bs->opaque; |
585f8587 FB |
1206 | uint8_t keybuf[16]; |
1207 | int len, i; | |
f6fa64f6 | 1208 | Error *err = NULL; |
3b46e624 | 1209 | |
585f8587 FB |
1210 | memset(keybuf, 0, 16); |
1211 | len = strlen(key); | |
1212 | if (len > 16) | |
1213 | len = 16; | |
1214 | /* XXX: we could compress the chars to 7 bits to increase | |
1215 | entropy */ | |
1216 | for(i = 0;i < len;i++) { | |
1217 | keybuf[i] = key[i]; | |
1218 | } | |
8336aafa | 1219 | assert(bs->encrypted); |
585f8587 | 1220 | |
f6fa64f6 DB |
1221 | qcrypto_cipher_free(s->cipher); |
1222 | s->cipher = qcrypto_cipher_new( | |
1223 | QCRYPTO_CIPHER_ALG_AES_128, | |
1224 | QCRYPTO_CIPHER_MODE_CBC, | |
1225 | keybuf, G_N_ELEMENTS(keybuf), | |
1226 | &err); | |
1227 | ||
1228 | if (!s->cipher) { | |
1229 | /* XXX would be nice if errors in this method could | |
1230 | * be properly propagate to the caller. Would need | |
1231 | * the bdrv_set_key() API signature to be fixed. */ | |
1232 | error_free(err); | |
585f8587 | 1233 | return -1; |
585f8587 | 1234 | } |
585f8587 FB |
1235 | return 0; |
1236 | } | |
1237 | ||
21d82ac9 JC |
1238 | static int qcow2_reopen_prepare(BDRVReopenState *state, |
1239 | BlockReopenQueue *queue, Error **errp) | |
1240 | { | |
5b0959a7 | 1241 | Qcow2ReopenState *r; |
4c2e5f8f KW |
1242 | int ret; |
1243 | ||
5b0959a7 KW |
1244 | r = g_new0(Qcow2ReopenState, 1); |
1245 | state->opaque = r; | |
1246 | ||
1247 | ret = qcow2_update_options_prepare(state->bs, r, state->options, | |
1248 | state->flags, errp); | |
1249 | if (ret < 0) { | |
1250 | goto fail; | |
1251 | } | |
1252 | ||
1253 | /* We need to write out any unwritten data if we reopen read-only. */ | |
4c2e5f8f KW |
1254 | if ((state->flags & BDRV_O_RDWR) == 0) { |
1255 | ret = bdrv_flush(state->bs); | |
1256 | if (ret < 0) { | |
5b0959a7 | 1257 | goto fail; |
4c2e5f8f KW |
1258 | } |
1259 | ||
1260 | ret = qcow2_mark_clean(state->bs); | |
1261 | if (ret < 0) { | |
5b0959a7 | 1262 | goto fail; |
4c2e5f8f KW |
1263 | } |
1264 | } | |
1265 | ||
21d82ac9 | 1266 | return 0; |
5b0959a7 KW |
1267 | |
1268 | fail: | |
1269 | qcow2_update_options_abort(state->bs, r); | |
1270 | g_free(r); | |
1271 | return ret; | |
1272 | } | |
1273 | ||
1274 | static void qcow2_reopen_commit(BDRVReopenState *state) | |
1275 | { | |
1276 | qcow2_update_options_commit(state->bs, state->opaque); | |
1277 | g_free(state->opaque); | |
1278 | } | |
1279 | ||
1280 | static void qcow2_reopen_abort(BDRVReopenState *state) | |
1281 | { | |
1282 | qcow2_update_options_abort(state->bs, state->opaque); | |
1283 | g_free(state->opaque); | |
21d82ac9 JC |
1284 | } |
1285 | ||
5365f44d KW |
1286 | static void qcow2_join_options(QDict *options, QDict *old_options) |
1287 | { | |
1288 | bool has_new_overlap_template = | |
1289 | qdict_haskey(options, QCOW2_OPT_OVERLAP) || | |
1290 | qdict_haskey(options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1291 | bool has_new_total_cache_size = | |
1292 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE); | |
1293 | bool has_all_cache_options; | |
1294 | ||
1295 | /* New overlap template overrides all old overlap options */ | |
1296 | if (has_new_overlap_template) { | |
1297 | qdict_del(old_options, QCOW2_OPT_OVERLAP); | |
1298 | qdict_del(old_options, QCOW2_OPT_OVERLAP_TEMPLATE); | |
1299 | qdict_del(old_options, QCOW2_OPT_OVERLAP_MAIN_HEADER); | |
1300 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L1); | |
1301 | qdict_del(old_options, QCOW2_OPT_OVERLAP_ACTIVE_L2); | |
1302 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_TABLE); | |
1303 | qdict_del(old_options, QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK); | |
1304 | qdict_del(old_options, QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE); | |
1305 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L1); | |
1306 | qdict_del(old_options, QCOW2_OPT_OVERLAP_INACTIVE_L2); | |
1307 | } | |
1308 | ||
1309 | /* New total cache size overrides all old options */ | |
1310 | if (qdict_haskey(options, QCOW2_OPT_CACHE_SIZE)) { | |
1311 | qdict_del(old_options, QCOW2_OPT_L2_CACHE_SIZE); | |
1312 | qdict_del(old_options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1313 | } | |
1314 | ||
1315 | qdict_join(options, old_options, false); | |
1316 | ||
1317 | /* | |
1318 | * If after merging all cache size options are set, an old total size is | |
1319 | * overwritten. Do keep all options, however, if all three are new. The | |
1320 | * resulting error message is what we want to happen. | |
1321 | */ | |
1322 | has_all_cache_options = | |
1323 | qdict_haskey(options, QCOW2_OPT_CACHE_SIZE) || | |
1324 | qdict_haskey(options, QCOW2_OPT_L2_CACHE_SIZE) || | |
1325 | qdict_haskey(options, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
1326 | ||
1327 | if (has_all_cache_options && !has_new_total_cache_size) { | |
1328 | qdict_del(options, QCOW2_OPT_CACHE_SIZE); | |
1329 | } | |
1330 | } | |
1331 | ||
b6b8a333 | 1332 | static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs, |
f8a2e5e3 | 1333 | int64_t sector_num, int nb_sectors, int *pnum) |
585f8587 | 1334 | { |
ff99129a | 1335 | BDRVQcow2State *s = bs->opaque; |
585f8587 | 1336 | uint64_t cluster_offset; |
4bc74be9 PB |
1337 | int index_in_cluster, ret; |
1338 | int64_t status = 0; | |
585f8587 | 1339 | |
095a9c58 | 1340 | *pnum = nb_sectors; |
f8a2e5e3 | 1341 | qemu_co_mutex_lock(&s->lock); |
1c46efaa | 1342 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, pnum, &cluster_offset); |
f8a2e5e3 | 1343 | qemu_co_mutex_unlock(&s->lock); |
1c46efaa | 1344 | if (ret < 0) { |
d663640c | 1345 | return ret; |
1c46efaa | 1346 | } |
095a9c58 | 1347 | |
4bc74be9 | 1348 | if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED && |
f6fa64f6 | 1349 | !s->cipher) { |
4bc74be9 PB |
1350 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
1351 | cluster_offset |= (index_in_cluster << BDRV_SECTOR_BITS); | |
1352 | status |= BDRV_BLOCK_OFFSET_VALID | cluster_offset; | |
1353 | } | |
1354 | if (ret == QCOW2_CLUSTER_ZERO) { | |
1355 | status |= BDRV_BLOCK_ZERO; | |
1356 | } else if (ret != QCOW2_CLUSTER_UNALLOCATED) { | |
1357 | status |= BDRV_BLOCK_DATA; | |
1358 | } | |
1359 | return status; | |
585f8587 FB |
1360 | } |
1361 | ||
a9465922 | 1362 | /* handle reading after the end of the backing file */ |
bd28f835 KW |
1363 | int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, |
1364 | int64_t sector_num, int nb_sectors) | |
a9465922 FB |
1365 | { |
1366 | int n1; | |
1367 | if ((sector_num + nb_sectors) <= bs->total_sectors) | |
1368 | return nb_sectors; | |
1369 | if (sector_num >= bs->total_sectors) | |
1370 | n1 = 0; | |
1371 | else | |
1372 | n1 = bs->total_sectors - sector_num; | |
bd28f835 | 1373 | |
3d9b4925 | 1374 | qemu_iovec_memset(qiov, 512 * n1, 0, 512 * (nb_sectors - n1)); |
bd28f835 | 1375 | |
a9465922 FB |
1376 | return n1; |
1377 | } | |
1378 | ||
a968168c | 1379 | static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num, |
3fc48d09 | 1380 | int remaining_sectors, QEMUIOVector *qiov) |
585f8587 | 1381 | { |
ff99129a | 1382 | BDRVQcow2State *s = bs->opaque; |
a9465922 | 1383 | int index_in_cluster, n1; |
68d100e9 | 1384 | int ret; |
faf575c1 | 1385 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 1386 | uint64_t cluster_offset = 0; |
3fc48d09 FZ |
1387 | uint64_t bytes_done = 0; |
1388 | QEMUIOVector hd_qiov; | |
1389 | uint8_t *cluster_data = NULL; | |
585f8587 | 1390 | |
3fc48d09 FZ |
1391 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1392 | ||
1393 | qemu_co_mutex_lock(&s->lock); | |
1394 | ||
1395 | while (remaining_sectors != 0) { | |
bd28f835 | 1396 | |
5ebaa27e | 1397 | /* prepare next request */ |
3fc48d09 | 1398 | cur_nr_sectors = remaining_sectors; |
f6fa64f6 | 1399 | if (s->cipher) { |
5ebaa27e FZ |
1400 | cur_nr_sectors = MIN(cur_nr_sectors, |
1401 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
585f8587 | 1402 | } |
5ebaa27e | 1403 | |
3fc48d09 | 1404 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, |
5ebaa27e | 1405 | &cur_nr_sectors, &cluster_offset); |
8af36488 | 1406 | if (ret < 0) { |
3fc48d09 | 1407 | goto fail; |
8af36488 | 1408 | } |
bd28f835 | 1409 | |
3fc48d09 | 1410 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
c87c0672 | 1411 | |
3fc48d09 | 1412 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 1413 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e FZ |
1414 | cur_nr_sectors * 512); |
1415 | ||
68d000a3 KW |
1416 | switch (ret) { |
1417 | case QCOW2_CLUSTER_UNALLOCATED: | |
5ebaa27e | 1418 | |
760e0063 | 1419 | if (bs->backing) { |
5ebaa27e | 1420 | /* read from the base image */ |
760e0063 | 1421 | n1 = qcow2_backing_read1(bs->backing->bs, &hd_qiov, |
3fc48d09 | 1422 | sector_num, cur_nr_sectors); |
5ebaa27e | 1423 | if (n1 > 0) { |
44deba5a KW |
1424 | QEMUIOVector local_qiov; |
1425 | ||
1426 | qemu_iovec_init(&local_qiov, hd_qiov.niov); | |
1427 | qemu_iovec_concat(&local_qiov, &hd_qiov, 0, | |
1428 | n1 * BDRV_SECTOR_SIZE); | |
1429 | ||
5ebaa27e FZ |
1430 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); |
1431 | qemu_co_mutex_unlock(&s->lock); | |
760e0063 | 1432 | ret = bdrv_co_readv(bs->backing->bs, sector_num, |
44deba5a | 1433 | n1, &local_qiov); |
5ebaa27e | 1434 | qemu_co_mutex_lock(&s->lock); |
44deba5a KW |
1435 | |
1436 | qemu_iovec_destroy(&local_qiov); | |
1437 | ||
5ebaa27e | 1438 | if (ret < 0) { |
3fc48d09 | 1439 | goto fail; |
5ebaa27e FZ |
1440 | } |
1441 | } | |
1442 | } else { | |
1443 | /* Note: in this case, no need to wait */ | |
3d9b4925 | 1444 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
5ebaa27e | 1445 | } |
68d000a3 KW |
1446 | break; |
1447 | ||
6377af48 | 1448 | case QCOW2_CLUSTER_ZERO: |
3d9b4925 | 1449 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
6377af48 KW |
1450 | break; |
1451 | ||
68d000a3 | 1452 | case QCOW2_CLUSTER_COMPRESSED: |
5ebaa27e FZ |
1453 | /* add AIO support for compressed blocks ? */ |
1454 | ret = qcow2_decompress_cluster(bs, cluster_offset); | |
1455 | if (ret < 0) { | |
3fc48d09 | 1456 | goto fail; |
bd28f835 KW |
1457 | } |
1458 | ||
03396148 | 1459 | qemu_iovec_from_buf(&hd_qiov, 0, |
5ebaa27e | 1460 | s->cluster_cache + index_in_cluster * 512, |
faf575c1 | 1461 | 512 * cur_nr_sectors); |
68d000a3 KW |
1462 | break; |
1463 | ||
1464 | case QCOW2_CLUSTER_NORMAL: | |
5ebaa27e | 1465 | if ((cluster_offset & 511) != 0) { |
3fc48d09 FZ |
1466 | ret = -EIO; |
1467 | goto fail; | |
5ebaa27e | 1468 | } |
bd28f835 | 1469 | |
8336aafa | 1470 | if (bs->encrypted) { |
f6fa64f6 | 1471 | assert(s->cipher); |
8336aafa | 1472 | |
5ebaa27e FZ |
1473 | /* |
1474 | * For encrypted images, read everything into a temporary | |
1475 | * contiguous buffer on which the AES functions can work. | |
1476 | */ | |
3fc48d09 FZ |
1477 | if (!cluster_data) { |
1478 | cluster_data = | |
9a4f4c31 KW |
1479 | qemu_try_blockalign(bs->file->bs, |
1480 | QCOW_MAX_CRYPT_CLUSTERS | |
1481 | * s->cluster_size); | |
de82815d KW |
1482 | if (cluster_data == NULL) { |
1483 | ret = -ENOMEM; | |
1484 | goto fail; | |
1485 | } | |
5ebaa27e FZ |
1486 | } |
1487 | ||
1488 | assert(cur_nr_sectors <= | |
1489 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
3fc48d09 FZ |
1490 | qemu_iovec_reset(&hd_qiov); |
1491 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
1492 | 512 * cur_nr_sectors); |
1493 | } | |
1494 | ||
1495 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
1496 | qemu_co_mutex_unlock(&s->lock); | |
9a4f4c31 | 1497 | ret = bdrv_co_readv(bs->file->bs, |
5ebaa27e | 1498 | (cluster_offset >> 9) + index_in_cluster, |
3fc48d09 | 1499 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
1500 | qemu_co_mutex_lock(&s->lock); |
1501 | if (ret < 0) { | |
3fc48d09 | 1502 | goto fail; |
5ebaa27e | 1503 | } |
8336aafa | 1504 | if (bs->encrypted) { |
f6fa64f6 DB |
1505 | assert(s->cipher); |
1506 | Error *err = NULL; | |
1507 | if (qcow2_encrypt_sectors(s, sector_num, cluster_data, | |
1508 | cluster_data, cur_nr_sectors, false, | |
1509 | &err) < 0) { | |
1510 | error_free(err); | |
1511 | ret = -EIO; | |
1512 | goto fail; | |
1513 | } | |
03396148 MT |
1514 | qemu_iovec_from_buf(qiov, bytes_done, |
1515 | cluster_data, 512 * cur_nr_sectors); | |
5ebaa27e | 1516 | } |
68d000a3 KW |
1517 | break; |
1518 | ||
1519 | default: | |
1520 | g_assert_not_reached(); | |
1521 | ret = -EIO; | |
1522 | goto fail; | |
faf575c1 | 1523 | } |
f141eafe | 1524 | |
3fc48d09 FZ |
1525 | remaining_sectors -= cur_nr_sectors; |
1526 | sector_num += cur_nr_sectors; | |
1527 | bytes_done += cur_nr_sectors * 512; | |
5ebaa27e | 1528 | } |
3fc48d09 | 1529 | ret = 0; |
faf575c1 | 1530 | |
3fc48d09 | 1531 | fail: |
68d100e9 | 1532 | qemu_co_mutex_unlock(&s->lock); |
42496d62 | 1533 | |
3fc48d09 | 1534 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1535 | qemu_vfree(cluster_data); |
68d100e9 KW |
1536 | |
1537 | return ret; | |
585f8587 FB |
1538 | } |
1539 | ||
a968168c | 1540 | static coroutine_fn int qcow2_co_writev(BlockDriverState *bs, |
3fc48d09 FZ |
1541 | int64_t sector_num, |
1542 | int remaining_sectors, | |
1543 | QEMUIOVector *qiov) | |
585f8587 | 1544 | { |
ff99129a | 1545 | BDRVQcow2State *s = bs->opaque; |
585f8587 | 1546 | int index_in_cluster; |
68d100e9 | 1547 | int ret; |
faf575c1 | 1548 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 1549 | uint64_t cluster_offset; |
3fc48d09 FZ |
1550 | QEMUIOVector hd_qiov; |
1551 | uint64_t bytes_done = 0; | |
1552 | uint8_t *cluster_data = NULL; | |
8d2497c3 | 1553 | QCowL2Meta *l2meta = NULL; |
c2271403 | 1554 | |
3cce16f4 KW |
1555 | trace_qcow2_writev_start_req(qemu_coroutine_self(), sector_num, |
1556 | remaining_sectors); | |
1557 | ||
3fc48d09 FZ |
1558 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1559 | ||
1560 | s->cluster_cache_offset = -1; /* disable compressed cache */ | |
3b46e624 | 1561 | |
3fc48d09 FZ |
1562 | qemu_co_mutex_lock(&s->lock); |
1563 | ||
1564 | while (remaining_sectors != 0) { | |
1565 | ||
f50f88b9 | 1566 | l2meta = NULL; |
cf5c1a23 | 1567 | |
3cce16f4 | 1568 | trace_qcow2_writev_start_part(qemu_coroutine_self()); |
3fc48d09 | 1569 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
16f0587e | 1570 | cur_nr_sectors = remaining_sectors; |
8336aafa | 1571 | if (bs->encrypted && |
16f0587e HT |
1572 | cur_nr_sectors > |
1573 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster) { | |
1574 | cur_nr_sectors = | |
1575 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster; | |
5ebaa27e | 1576 | } |
095a9c58 | 1577 | |
3fc48d09 | 1578 | ret = qcow2_alloc_cluster_offset(bs, sector_num << 9, |
16f0587e | 1579 | &cur_nr_sectors, &cluster_offset, &l2meta); |
5ebaa27e | 1580 | if (ret < 0) { |
3fc48d09 | 1581 | goto fail; |
5ebaa27e | 1582 | } |
148da7ea | 1583 | |
5ebaa27e | 1584 | assert((cluster_offset & 511) == 0); |
148da7ea | 1585 | |
3fc48d09 | 1586 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 1587 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e | 1588 | cur_nr_sectors * 512); |
6f5f060b | 1589 | |
8336aafa | 1590 | if (bs->encrypted) { |
f6fa64f6 DB |
1591 | Error *err = NULL; |
1592 | assert(s->cipher); | |
3fc48d09 | 1593 | if (!cluster_data) { |
9a4f4c31 | 1594 | cluster_data = qemu_try_blockalign(bs->file->bs, |
de82815d KW |
1595 | QCOW_MAX_CRYPT_CLUSTERS |
1596 | * s->cluster_size); | |
1597 | if (cluster_data == NULL) { | |
1598 | ret = -ENOMEM; | |
1599 | goto fail; | |
1600 | } | |
5ebaa27e | 1601 | } |
6f5f060b | 1602 | |
3fc48d09 | 1603 | assert(hd_qiov.size <= |
5ebaa27e | 1604 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
d5e6b161 | 1605 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); |
6f5f060b | 1606 | |
f6fa64f6 DB |
1607 | if (qcow2_encrypt_sectors(s, sector_num, cluster_data, |
1608 | cluster_data, cur_nr_sectors, | |
1609 | true, &err) < 0) { | |
1610 | error_free(err); | |
1611 | ret = -EIO; | |
1612 | goto fail; | |
1613 | } | |
6f5f060b | 1614 | |
3fc48d09 FZ |
1615 | qemu_iovec_reset(&hd_qiov); |
1616 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
1617 | cur_nr_sectors * 512); |
1618 | } | |
6f5f060b | 1619 | |
231bb267 | 1620 | ret = qcow2_pre_write_overlap_check(bs, 0, |
cf93980e HR |
1621 | cluster_offset + index_in_cluster * BDRV_SECTOR_SIZE, |
1622 | cur_nr_sectors * BDRV_SECTOR_SIZE); | |
1623 | if (ret < 0) { | |
1624 | goto fail; | |
1625 | } | |
1626 | ||
5ebaa27e | 1627 | qemu_co_mutex_unlock(&s->lock); |
67a7a0eb | 1628 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); |
3cce16f4 KW |
1629 | trace_qcow2_writev_data(qemu_coroutine_self(), |
1630 | (cluster_offset >> 9) + index_in_cluster); | |
9a4f4c31 | 1631 | ret = bdrv_co_writev(bs->file->bs, |
5ebaa27e | 1632 | (cluster_offset >> 9) + index_in_cluster, |
3fc48d09 | 1633 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
1634 | qemu_co_mutex_lock(&s->lock); |
1635 | if (ret < 0) { | |
3fc48d09 | 1636 | goto fail; |
5ebaa27e | 1637 | } |
f141eafe | 1638 | |
88c6588c KW |
1639 | while (l2meta != NULL) { |
1640 | QCowL2Meta *next; | |
1641 | ||
f50f88b9 KW |
1642 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); |
1643 | if (ret < 0) { | |
1644 | goto fail; | |
1645 | } | |
faf575c1 | 1646 | |
4e95314e KW |
1647 | /* Take the request off the list of running requests */ |
1648 | if (l2meta->nb_clusters != 0) { | |
1649 | QLIST_REMOVE(l2meta, next_in_flight); | |
1650 | } | |
1651 | ||
4e95314e | 1652 | qemu_co_queue_restart_all(&l2meta->dependent_requests); |
4e95314e | 1653 | |
88c6588c | 1654 | next = l2meta->next; |
f50f88b9 | 1655 | g_free(l2meta); |
88c6588c | 1656 | l2meta = next; |
f50f88b9 | 1657 | } |
0fa9131a | 1658 | |
3fc48d09 FZ |
1659 | remaining_sectors -= cur_nr_sectors; |
1660 | sector_num += cur_nr_sectors; | |
1661 | bytes_done += cur_nr_sectors * 512; | |
3cce16f4 | 1662 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_nr_sectors); |
5ebaa27e | 1663 | } |
3fc48d09 | 1664 | ret = 0; |
faf575c1 | 1665 | |
3fc48d09 | 1666 | fail: |
4e95314e KW |
1667 | qemu_co_mutex_unlock(&s->lock); |
1668 | ||
88c6588c KW |
1669 | while (l2meta != NULL) { |
1670 | QCowL2Meta *next; | |
1671 | ||
4e95314e KW |
1672 | if (l2meta->nb_clusters != 0) { |
1673 | QLIST_REMOVE(l2meta, next_in_flight); | |
1674 | } | |
1675 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
88c6588c KW |
1676 | |
1677 | next = l2meta->next; | |
cf5c1a23 | 1678 | g_free(l2meta); |
88c6588c | 1679 | l2meta = next; |
cf5c1a23 | 1680 | } |
0fa9131a | 1681 | |
3fc48d09 | 1682 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1683 | qemu_vfree(cluster_data); |
3cce16f4 | 1684 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
42496d62 | 1685 | |
68d100e9 | 1686 | return ret; |
585f8587 FB |
1687 | } |
1688 | ||
7c80ab3f | 1689 | static void qcow2_close(BlockDriverState *bs) |
585f8587 | 1690 | { |
ff99129a | 1691 | BDRVQcow2State *s = bs->opaque; |
de82815d | 1692 | qemu_vfree(s->l1_table); |
cf93980e HR |
1693 | /* else pre-write overlap checks in cache_destroy may crash */ |
1694 | s->l1_table = NULL; | |
29c1a730 | 1695 | |
27eb6c09 | 1696 | if (!(bs->open_flags & BDRV_O_INCOMING)) { |
3b5e14c7 | 1697 | int ret1, ret2; |
29c1a730 | 1698 | |
3b5e14c7 HR |
1699 | ret1 = qcow2_cache_flush(bs, s->l2_table_cache); |
1700 | ret2 = qcow2_cache_flush(bs, s->refcount_block_cache); | |
1701 | ||
1702 | if (ret1) { | |
1703 | error_report("Failed to flush the L2 table cache: %s", | |
1704 | strerror(-ret1)); | |
1705 | } | |
1706 | if (ret2) { | |
1707 | error_report("Failed to flush the refcount block cache: %s", | |
1708 | strerror(-ret2)); | |
1709 | } | |
1710 | ||
1711 | if (!ret1 && !ret2) { | |
1712 | qcow2_mark_clean(bs); | |
1713 | } | |
27eb6c09 | 1714 | } |
c61d0004 | 1715 | |
279621c0 | 1716 | cache_clean_timer_del(bs); |
29c1a730 KW |
1717 | qcow2_cache_destroy(bs, s->l2_table_cache); |
1718 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1719 | ||
f6fa64f6 DB |
1720 | qcrypto_cipher_free(s->cipher); |
1721 | s->cipher = NULL; | |
1722 | ||
6744cbab | 1723 | g_free(s->unknown_header_fields); |
75bab85c | 1724 | cleanup_unknown_header_ext(bs); |
6744cbab | 1725 | |
e4603fe1 KW |
1726 | g_free(s->image_backing_file); |
1727 | g_free(s->image_backing_format); | |
1728 | ||
7267c094 | 1729 | g_free(s->cluster_cache); |
dea43a65 | 1730 | qemu_vfree(s->cluster_data); |
ed6ccf0f | 1731 | qcow2_refcount_close(bs); |
28c1202b | 1732 | qcow2_free_snapshots(bs); |
585f8587 FB |
1733 | } |
1734 | ||
5a8a30db | 1735 | static void qcow2_invalidate_cache(BlockDriverState *bs, Error **errp) |
06d9260f | 1736 | { |
ff99129a | 1737 | BDRVQcow2State *s = bs->opaque; |
06d9260f | 1738 | int flags = s->flags; |
f6fa64f6 | 1739 | QCryptoCipher *cipher = NULL; |
acdfb480 | 1740 | QDict *options; |
5a8a30db KW |
1741 | Error *local_err = NULL; |
1742 | int ret; | |
06d9260f AL |
1743 | |
1744 | /* | |
1745 | * Backing files are read-only which makes all of their metadata immutable, | |
1746 | * that means we don't have to worry about reopening them here. | |
1747 | */ | |
1748 | ||
f6fa64f6 DB |
1749 | cipher = s->cipher; |
1750 | s->cipher = NULL; | |
06d9260f AL |
1751 | |
1752 | qcow2_close(bs); | |
1753 | ||
9a4f4c31 | 1754 | bdrv_invalidate_cache(bs->file->bs, &local_err); |
5a8a30db KW |
1755 | if (local_err) { |
1756 | error_propagate(errp, local_err); | |
1757 | return; | |
1758 | } | |
3456a8d1 | 1759 | |
ff99129a | 1760 | memset(s, 0, sizeof(BDRVQcow2State)); |
d475e5ac | 1761 | options = qdict_clone_shallow(bs->options); |
5a8a30db KW |
1762 | |
1763 | ret = qcow2_open(bs, options, flags, &local_err); | |
a1904e48 | 1764 | QDECREF(options); |
5a8a30db | 1765 | if (local_err) { |
e43bfd9c MA |
1766 | error_propagate(errp, local_err); |
1767 | error_prepend(errp, "Could not reopen qcow2 layer: "); | |
5a8a30db KW |
1768 | return; |
1769 | } else if (ret < 0) { | |
1770 | error_setg_errno(errp, -ret, "Could not reopen qcow2 layer"); | |
1771 | return; | |
1772 | } | |
acdfb480 | 1773 | |
f6fa64f6 | 1774 | s->cipher = cipher; |
06d9260f AL |
1775 | } |
1776 | ||
e24e49e6 KW |
1777 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
1778 | size_t len, size_t buflen) | |
1779 | { | |
1780 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
1781 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
1782 | ||
1783 | if (buflen < ext_len) { | |
1784 | return -ENOSPC; | |
1785 | } | |
1786 | ||
1787 | *ext_backing_fmt = (QCowExtension) { | |
1788 | .magic = cpu_to_be32(magic), | |
1789 | .len = cpu_to_be32(len), | |
1790 | }; | |
1791 | memcpy(buf + sizeof(QCowExtension), s, len); | |
1792 | ||
1793 | return ext_len; | |
1794 | } | |
1795 | ||
756e6736 | 1796 | /* |
e24e49e6 KW |
1797 | * Updates the qcow2 header, including the variable length parts of it, i.e. |
1798 | * the backing file name and all extensions. qcow2 was not designed to allow | |
1799 | * such changes, so if we run out of space (we can only use the first cluster) | |
1800 | * this function may fail. | |
756e6736 KW |
1801 | * |
1802 | * Returns 0 on success, -errno in error cases. | |
1803 | */ | |
e24e49e6 | 1804 | int qcow2_update_header(BlockDriverState *bs) |
756e6736 | 1805 | { |
ff99129a | 1806 | BDRVQcow2State *s = bs->opaque; |
e24e49e6 KW |
1807 | QCowHeader *header; |
1808 | char *buf; | |
1809 | size_t buflen = s->cluster_size; | |
756e6736 | 1810 | int ret; |
e24e49e6 KW |
1811 | uint64_t total_size; |
1812 | uint32_t refcount_table_clusters; | |
6744cbab | 1813 | size_t header_length; |
75bab85c | 1814 | Qcow2UnknownHeaderExtension *uext; |
756e6736 | 1815 | |
e24e49e6 | 1816 | buf = qemu_blockalign(bs, buflen); |
756e6736 | 1817 | |
e24e49e6 KW |
1818 | /* Header structure */ |
1819 | header = (QCowHeader*) buf; | |
756e6736 | 1820 | |
e24e49e6 KW |
1821 | if (buflen < sizeof(*header)) { |
1822 | ret = -ENOSPC; | |
1823 | goto fail; | |
756e6736 KW |
1824 | } |
1825 | ||
6744cbab | 1826 | header_length = sizeof(*header) + s->unknown_header_fields_size; |
e24e49e6 KW |
1827 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
1828 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
1829 | ||
1830 | *header = (QCowHeader) { | |
6744cbab | 1831 | /* Version 2 fields */ |
e24e49e6 | 1832 | .magic = cpu_to_be32(QCOW_MAGIC), |
6744cbab | 1833 | .version = cpu_to_be32(s->qcow_version), |
e24e49e6 KW |
1834 | .backing_file_offset = 0, |
1835 | .backing_file_size = 0, | |
1836 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
1837 | .size = cpu_to_be64(total_size), | |
1838 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
1839 | .l1_size = cpu_to_be32(s->l1_size), | |
1840 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
1841 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
1842 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
1843 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
1844 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
6744cbab KW |
1845 | |
1846 | /* Version 3 fields */ | |
1847 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
1848 | .compatible_features = cpu_to_be64(s->compatible_features), | |
1849 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
b6481f37 | 1850 | .refcount_order = cpu_to_be32(s->refcount_order), |
6744cbab | 1851 | .header_length = cpu_to_be32(header_length), |
e24e49e6 | 1852 | }; |
756e6736 | 1853 | |
6744cbab KW |
1854 | /* For older versions, write a shorter header */ |
1855 | switch (s->qcow_version) { | |
1856 | case 2: | |
1857 | ret = offsetof(QCowHeader, incompatible_features); | |
1858 | break; | |
1859 | case 3: | |
1860 | ret = sizeof(*header); | |
1861 | break; | |
1862 | default: | |
b6c14762 JM |
1863 | ret = -EINVAL; |
1864 | goto fail; | |
6744cbab KW |
1865 | } |
1866 | ||
1867 | buf += ret; | |
1868 | buflen -= ret; | |
1869 | memset(buf, 0, buflen); | |
1870 | ||
1871 | /* Preserve any unknown field in the header */ | |
1872 | if (s->unknown_header_fields_size) { | |
1873 | if (buflen < s->unknown_header_fields_size) { | |
1874 | ret = -ENOSPC; | |
1875 | goto fail; | |
1876 | } | |
1877 | ||
1878 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
1879 | buf += s->unknown_header_fields_size; | |
1880 | buflen -= s->unknown_header_fields_size; | |
1881 | } | |
756e6736 | 1882 | |
e24e49e6 | 1883 | /* Backing file format header extension */ |
e4603fe1 | 1884 | if (s->image_backing_format) { |
e24e49e6 | 1885 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, |
e4603fe1 KW |
1886 | s->image_backing_format, |
1887 | strlen(s->image_backing_format), | |
e24e49e6 KW |
1888 | buflen); |
1889 | if (ret < 0) { | |
1890 | goto fail; | |
756e6736 KW |
1891 | } |
1892 | ||
e24e49e6 KW |
1893 | buf += ret; |
1894 | buflen -= ret; | |
756e6736 KW |
1895 | } |
1896 | ||
cfcc4c62 KW |
1897 | /* Feature table */ |
1898 | Qcow2Feature features[] = { | |
c61d0004 SH |
1899 | { |
1900 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1901 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
1902 | .name = "dirty bit", | |
1903 | }, | |
69c98726 HR |
1904 | { |
1905 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1906 | .bit = QCOW2_INCOMPAT_CORRUPT_BITNR, | |
1907 | .name = "corrupt bit", | |
1908 | }, | |
bfe8043e SH |
1909 | { |
1910 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
1911 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
1912 | .name = "lazy refcounts", | |
1913 | }, | |
cfcc4c62 KW |
1914 | }; |
1915 | ||
1916 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
1917 | features, sizeof(features), buflen); | |
1918 | if (ret < 0) { | |
1919 | goto fail; | |
1920 | } | |
1921 | buf += ret; | |
1922 | buflen -= ret; | |
1923 | ||
75bab85c KW |
1924 | /* Keep unknown header extensions */ |
1925 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
1926 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
1927 | if (ret < 0) { | |
1928 | goto fail; | |
1929 | } | |
1930 | ||
1931 | buf += ret; | |
1932 | buflen -= ret; | |
1933 | } | |
1934 | ||
e24e49e6 KW |
1935 | /* End of header extensions */ |
1936 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
756e6736 KW |
1937 | if (ret < 0) { |
1938 | goto fail; | |
1939 | } | |
1940 | ||
e24e49e6 KW |
1941 | buf += ret; |
1942 | buflen -= ret; | |
756e6736 | 1943 | |
e24e49e6 | 1944 | /* Backing file name */ |
e4603fe1 KW |
1945 | if (s->image_backing_file) { |
1946 | size_t backing_file_len = strlen(s->image_backing_file); | |
e24e49e6 KW |
1947 | |
1948 | if (buflen < backing_file_len) { | |
1949 | ret = -ENOSPC; | |
1950 | goto fail; | |
1951 | } | |
1952 | ||
00ea1881 | 1953 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ |
e4603fe1 | 1954 | strncpy(buf, s->image_backing_file, buflen); |
e24e49e6 KW |
1955 | |
1956 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
1957 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
756e6736 KW |
1958 | } |
1959 | ||
e24e49e6 | 1960 | /* Write the new header */ |
9a4f4c31 | 1961 | ret = bdrv_pwrite(bs->file->bs, 0, header, s->cluster_size); |
756e6736 KW |
1962 | if (ret < 0) { |
1963 | goto fail; | |
1964 | } | |
1965 | ||
1966 | ret = 0; | |
1967 | fail: | |
e24e49e6 | 1968 | qemu_vfree(header); |
756e6736 KW |
1969 | return ret; |
1970 | } | |
1971 | ||
1972 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
1973 | const char *backing_file, const char *backing_fmt) | |
1974 | { | |
ff99129a | 1975 | BDRVQcow2State *s = bs->opaque; |
e4603fe1 | 1976 | |
e24e49e6 KW |
1977 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
1978 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1979 | ||
e4603fe1 KW |
1980 | g_free(s->image_backing_file); |
1981 | g_free(s->image_backing_format); | |
1982 | ||
1983 | s->image_backing_file = backing_file ? g_strdup(bs->backing_file) : NULL; | |
1984 | s->image_backing_format = backing_fmt ? g_strdup(bs->backing_format) : NULL; | |
1985 | ||
e24e49e6 | 1986 | return qcow2_update_header(bs); |
756e6736 KW |
1987 | } |
1988 | ||
a35e1c17 KW |
1989 | static int preallocate(BlockDriverState *bs) |
1990 | { | |
a35e1c17 KW |
1991 | uint64_t nb_sectors; |
1992 | uint64_t offset; | |
060bee89 | 1993 | uint64_t host_offset = 0; |
a35e1c17 | 1994 | int num; |
148da7ea | 1995 | int ret; |
f50f88b9 | 1996 | QCowL2Meta *meta; |
a35e1c17 | 1997 | |
57322b78 | 1998 | nb_sectors = bdrv_nb_sectors(bs); |
a35e1c17 KW |
1999 | offset = 0; |
2000 | ||
2001 | while (nb_sectors) { | |
7c2bbf4a | 2002 | num = MIN(nb_sectors, INT_MAX >> BDRV_SECTOR_BITS); |
16f0587e | 2003 | ret = qcow2_alloc_cluster_offset(bs, offset, &num, |
060bee89 | 2004 | &host_offset, &meta); |
148da7ea | 2005 | if (ret < 0) { |
19dbcbf7 | 2006 | return ret; |
a35e1c17 KW |
2007 | } |
2008 | ||
c792707f SH |
2009 | while (meta) { |
2010 | QCowL2Meta *next = meta->next; | |
2011 | ||
7c2bbf4a HT |
2012 | ret = qcow2_alloc_cluster_link_l2(bs, meta); |
2013 | if (ret < 0) { | |
2014 | qcow2_free_any_clusters(bs, meta->alloc_offset, | |
2015 | meta->nb_clusters, QCOW2_DISCARD_NEVER); | |
2016 | return ret; | |
2017 | } | |
2018 | ||
2019 | /* There are no dependent requests, but we need to remove our | |
2020 | * request from the list of in-flight requests */ | |
4e95314e | 2021 | QLIST_REMOVE(meta, next_in_flight); |
c792707f SH |
2022 | |
2023 | g_free(meta); | |
2024 | meta = next; | |
f50f88b9 | 2025 | } |
f214978a | 2026 | |
a35e1c17 KW |
2027 | /* TODO Preallocate data if requested */ |
2028 | ||
2029 | nb_sectors -= num; | |
7c2bbf4a | 2030 | offset += num << BDRV_SECTOR_BITS; |
a35e1c17 KW |
2031 | } |
2032 | ||
2033 | /* | |
2034 | * It is expected that the image file is large enough to actually contain | |
2035 | * all of the allocated clusters (otherwise we get failing reads after | |
2036 | * EOF). Extend the image to the last allocated sector. | |
2037 | */ | |
060bee89 | 2038 | if (host_offset != 0) { |
7c2bbf4a HT |
2039 | uint8_t buf[BDRV_SECTOR_SIZE]; |
2040 | memset(buf, 0, BDRV_SECTOR_SIZE); | |
9a4f4c31 KW |
2041 | ret = bdrv_write(bs->file->bs, |
2042 | (host_offset >> BDRV_SECTOR_BITS) + num - 1, | |
7c2bbf4a | 2043 | buf, 1); |
19dbcbf7 KW |
2044 | if (ret < 0) { |
2045 | return ret; | |
2046 | } | |
a35e1c17 KW |
2047 | } |
2048 | ||
2049 | return 0; | |
2050 | } | |
2051 | ||
7c80ab3f JS |
2052 | static int qcow2_create2(const char *filename, int64_t total_size, |
2053 | const char *backing_file, const char *backing_format, | |
ffeaac9b | 2054 | int flags, size_t cluster_size, PreallocMode prealloc, |
bd4b167f | 2055 | QemuOpts *opts, int version, int refcount_order, |
3ef6c40a | 2056 | Error **errp) |
a9420734 | 2057 | { |
a9420734 | 2058 | int cluster_bits; |
e6641719 HR |
2059 | QDict *options; |
2060 | ||
2061 | /* Calculate cluster_bits */ | |
786a4ea8 | 2062 | cluster_bits = ctz32(cluster_size); |
a9420734 KW |
2063 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || |
2064 | (1 << cluster_bits) != cluster_size) | |
2065 | { | |
3ef6c40a HR |
2066 | error_setg(errp, "Cluster size must be a power of two between %d and " |
2067 | "%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); | |
a9420734 KW |
2068 | return -EINVAL; |
2069 | } | |
2070 | ||
2071 | /* | |
2072 | * Open the image file and write a minimal qcow2 header. | |
2073 | * | |
2074 | * We keep things simple and start with a zero-sized image. We also | |
2075 | * do without refcount blocks or a L1 table for now. We'll fix the | |
2076 | * inconsistency later. | |
2077 | * | |
2078 | * We do need a refcount table because growing the refcount table means | |
2079 | * allocating two new refcount blocks - the seconds of which would be at | |
2080 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
2081 | * size for any qcow2 image. | |
2082 | */ | |
2083 | BlockDriverState* bs; | |
f8413b3c | 2084 | QCowHeader *header; |
b106ad91 | 2085 | uint64_t* refcount_table; |
3ef6c40a | 2086 | Error *local_err = NULL; |
a9420734 KW |
2087 | int ret; |
2088 | ||
0e4271b7 | 2089 | if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) { |
bd4b167f HR |
2090 | /* Note: The following calculation does not need to be exact; if it is a |
2091 | * bit off, either some bytes will be "leaked" (which is fine) or we | |
2092 | * will need to increase the file size by some bytes (which is fine, | |
2093 | * too, as long as the bulk is allocated here). Therefore, using | |
2094 | * floating point arithmetic is fine. */ | |
0e4271b7 HT |
2095 | int64_t meta_size = 0; |
2096 | uint64_t nreftablee, nrefblocke, nl1e, nl2e; | |
2097 | int64_t aligned_total_size = align_offset(total_size, cluster_size); | |
bd4b167f HR |
2098 | int refblock_bits, refblock_size; |
2099 | /* refcount entry size in bytes */ | |
2100 | double rces = (1 << refcount_order) / 8.; | |
2101 | ||
2102 | /* see qcow2_open() */ | |
2103 | refblock_bits = cluster_bits - (refcount_order - 3); | |
2104 | refblock_size = 1 << refblock_bits; | |
0e4271b7 HT |
2105 | |
2106 | /* header: 1 cluster */ | |
2107 | meta_size += cluster_size; | |
2108 | ||
2109 | /* total size of L2 tables */ | |
2110 | nl2e = aligned_total_size / cluster_size; | |
2111 | nl2e = align_offset(nl2e, cluster_size / sizeof(uint64_t)); | |
2112 | meta_size += nl2e * sizeof(uint64_t); | |
2113 | ||
2114 | /* total size of L1 tables */ | |
2115 | nl1e = nl2e * sizeof(uint64_t) / cluster_size; | |
2116 | nl1e = align_offset(nl1e, cluster_size / sizeof(uint64_t)); | |
2117 | meta_size += nl1e * sizeof(uint64_t); | |
2118 | ||
2119 | /* total size of refcount blocks | |
2120 | * | |
2121 | * note: every host cluster is reference-counted, including metadata | |
2122 | * (even refcount blocks are recursively included). | |
2123 | * Let: | |
2124 | * a = total_size (this is the guest disk size) | |
2125 | * m = meta size not including refcount blocks and refcount tables | |
2126 | * c = cluster size | |
2127 | * y1 = number of refcount blocks entries | |
2128 | * y2 = meta size including everything | |
bd4b167f | 2129 | * rces = refcount entry size in bytes |
0e4271b7 HT |
2130 | * then, |
2131 | * y1 = (y2 + a)/c | |
bd4b167f | 2132 | * y2 = y1 * rces + y1 * rces * sizeof(u64) / c + m |
0e4271b7 | 2133 | * we can get y1: |
bd4b167f | 2134 | * y1 = (a + m) / (c - rces - rces * sizeof(u64) / c) |
0e4271b7 | 2135 | */ |
bd4b167f HR |
2136 | nrefblocke = (aligned_total_size + meta_size + cluster_size) |
2137 | / (cluster_size - rces - rces * sizeof(uint64_t) | |
2138 | / cluster_size); | |
2139 | meta_size += DIV_ROUND_UP(nrefblocke, refblock_size) * cluster_size; | |
0e4271b7 HT |
2140 | |
2141 | /* total size of refcount tables */ | |
bd4b167f | 2142 | nreftablee = nrefblocke / refblock_size; |
0e4271b7 HT |
2143 | nreftablee = align_offset(nreftablee, cluster_size / sizeof(uint64_t)); |
2144 | meta_size += nreftablee * sizeof(uint64_t); | |
2145 | ||
2146 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, | |
39101f25 | 2147 | aligned_total_size + meta_size, &error_abort); |
f43e47db MA |
2148 | qemu_opt_set(opts, BLOCK_OPT_PREALLOC, PreallocMode_lookup[prealloc], |
2149 | &error_abort); | |
0e4271b7 HT |
2150 | } |
2151 | ||
c282e1fd | 2152 | ret = bdrv_create_file(filename, opts, &local_err); |
a9420734 | 2153 | if (ret < 0) { |
3ef6c40a | 2154 | error_propagate(errp, local_err); |
a9420734 KW |
2155 | return ret; |
2156 | } | |
2157 | ||
2e40134b HR |
2158 | bs = NULL; |
2159 | ret = bdrv_open(&bs, filename, NULL, NULL, BDRV_O_RDWR | BDRV_O_PROTOCOL, | |
6ebf9aa2 | 2160 | &local_err); |
a9420734 | 2161 | if (ret < 0) { |
3ef6c40a | 2162 | error_propagate(errp, local_err); |
a9420734 KW |
2163 | return ret; |
2164 | } | |
2165 | ||
2166 | /* Write the header */ | |
f8413b3c KW |
2167 | QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header)); |
2168 | header = g_malloc0(cluster_size); | |
2169 | *header = (QCowHeader) { | |
2170 | .magic = cpu_to_be32(QCOW_MAGIC), | |
2171 | .version = cpu_to_be32(version), | |
2172 | .cluster_bits = cpu_to_be32(cluster_bits), | |
2173 | .size = cpu_to_be64(0), | |
2174 | .l1_table_offset = cpu_to_be64(0), | |
2175 | .l1_size = cpu_to_be32(0), | |
2176 | .refcount_table_offset = cpu_to_be64(cluster_size), | |
2177 | .refcount_table_clusters = cpu_to_be32(1), | |
bd4b167f | 2178 | .refcount_order = cpu_to_be32(refcount_order), |
f8413b3c KW |
2179 | .header_length = cpu_to_be32(sizeof(*header)), |
2180 | }; | |
a9420734 KW |
2181 | |
2182 | if (flags & BLOCK_FLAG_ENCRYPT) { | |
f8413b3c | 2183 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_AES); |
a9420734 | 2184 | } else { |
f8413b3c | 2185 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); |
a9420734 KW |
2186 | } |
2187 | ||
bfe8043e | 2188 | if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) { |
f8413b3c | 2189 | header->compatible_features |= |
bfe8043e SH |
2190 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); |
2191 | } | |
2192 | ||
f8413b3c KW |
2193 | ret = bdrv_pwrite(bs, 0, header, cluster_size); |
2194 | g_free(header); | |
a9420734 | 2195 | if (ret < 0) { |
3ef6c40a | 2196 | error_setg_errno(errp, -ret, "Could not write qcow2 header"); |
a9420734 KW |
2197 | goto out; |
2198 | } | |
2199 | ||
b106ad91 KW |
2200 | /* Write a refcount table with one refcount block */ |
2201 | refcount_table = g_malloc0(2 * cluster_size); | |
2202 | refcount_table[0] = cpu_to_be64(2 * cluster_size); | |
2203 | ret = bdrv_pwrite(bs, cluster_size, refcount_table, 2 * cluster_size); | |
7267c094 | 2204 | g_free(refcount_table); |
a9420734 KW |
2205 | |
2206 | if (ret < 0) { | |
3ef6c40a | 2207 | error_setg_errno(errp, -ret, "Could not write refcount table"); |
a9420734 KW |
2208 | goto out; |
2209 | } | |
2210 | ||
f67503e5 HR |
2211 | bdrv_unref(bs); |
2212 | bs = NULL; | |
a9420734 KW |
2213 | |
2214 | /* | |
2215 | * And now open the image and make it consistent first (i.e. increase the | |
2216 | * refcount of the cluster that is occupied by the header and the refcount | |
2217 | * table) | |
2218 | */ | |
e6641719 HR |
2219 | options = qdict_new(); |
2220 | qdict_put(options, "driver", qstring_from_str("qcow2")); | |
2221 | ret = bdrv_open(&bs, filename, NULL, options, | |
ef810437 | 2222 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_FLUSH, |
6ebf9aa2 | 2223 | &local_err); |
a9420734 | 2224 | if (ret < 0) { |
3ef6c40a | 2225 | error_propagate(errp, local_err); |
a9420734 KW |
2226 | goto out; |
2227 | } | |
2228 | ||
b106ad91 | 2229 | ret = qcow2_alloc_clusters(bs, 3 * cluster_size); |
a9420734 | 2230 | if (ret < 0) { |
3ef6c40a HR |
2231 | error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 " |
2232 | "header and refcount table"); | |
a9420734 KW |
2233 | goto out; |
2234 | ||
2235 | } else if (ret != 0) { | |
2236 | error_report("Huh, first cluster in empty image is already in use?"); | |
2237 | abort(); | |
2238 | } | |
2239 | ||
2240 | /* Okay, now that we have a valid image, let's give it the right size */ | |
180e9526 | 2241 | ret = bdrv_truncate(bs, total_size); |
a9420734 | 2242 | if (ret < 0) { |
3ef6c40a | 2243 | error_setg_errno(errp, -ret, "Could not resize image"); |
a9420734 KW |
2244 | goto out; |
2245 | } | |
2246 | ||
2247 | /* Want a backing file? There you go.*/ | |
2248 | if (backing_file) { | |
2249 | ret = bdrv_change_backing_file(bs, backing_file, backing_format); | |
2250 | if (ret < 0) { | |
3ef6c40a HR |
2251 | error_setg_errno(errp, -ret, "Could not assign backing file '%s' " |
2252 | "with format '%s'", backing_file, backing_format); | |
a9420734 KW |
2253 | goto out; |
2254 | } | |
2255 | } | |
2256 | ||
2257 | /* And if we're supposed to preallocate metadata, do that now */ | |
0e4271b7 | 2258 | if (prealloc != PREALLOC_MODE_OFF) { |
ff99129a | 2259 | BDRVQcow2State *s = bs->opaque; |
15552c4a | 2260 | qemu_co_mutex_lock(&s->lock); |
a9420734 | 2261 | ret = preallocate(bs); |
15552c4a | 2262 | qemu_co_mutex_unlock(&s->lock); |
a9420734 | 2263 | if (ret < 0) { |
3ef6c40a | 2264 | error_setg_errno(errp, -ret, "Could not preallocate metadata"); |
a9420734 KW |
2265 | goto out; |
2266 | } | |
2267 | } | |
2268 | ||
f67503e5 HR |
2269 | bdrv_unref(bs); |
2270 | bs = NULL; | |
ba2ab2f2 HR |
2271 | |
2272 | /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning */ | |
e6641719 HR |
2273 | options = qdict_new(); |
2274 | qdict_put(options, "driver", qstring_from_str("qcow2")); | |
2275 | ret = bdrv_open(&bs, filename, NULL, options, | |
c9fbb99d | 2276 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_BACKING, |
6ebf9aa2 | 2277 | &local_err); |
84d18f06 | 2278 | if (local_err) { |
ba2ab2f2 HR |
2279 | error_propagate(errp, local_err); |
2280 | goto out; | |
2281 | } | |
2282 | ||
a9420734 KW |
2283 | ret = 0; |
2284 | out: | |
f67503e5 HR |
2285 | if (bs) { |
2286 | bdrv_unref(bs); | |
2287 | } | |
a9420734 KW |
2288 | return ret; |
2289 | } | |
de5f3f40 | 2290 | |
1bd0e2d1 | 2291 | static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) |
de5f3f40 | 2292 | { |
1bd0e2d1 CL |
2293 | char *backing_file = NULL; |
2294 | char *backing_fmt = NULL; | |
2295 | char *buf = NULL; | |
180e9526 | 2296 | uint64_t size = 0; |
de5f3f40 | 2297 | int flags = 0; |
99cce9fa | 2298 | size_t cluster_size = DEFAULT_CLUSTER_SIZE; |
ffeaac9b | 2299 | PreallocMode prealloc; |
8ad1898c | 2300 | int version = 3; |
bd4b167f HR |
2301 | uint64_t refcount_bits = 16; |
2302 | int refcount_order; | |
3ef6c40a HR |
2303 | Error *local_err = NULL; |
2304 | int ret; | |
de5f3f40 KW |
2305 | |
2306 | /* Read out options */ | |
180e9526 HT |
2307 | size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
2308 | BDRV_SECTOR_SIZE); | |
1bd0e2d1 CL |
2309 | backing_file = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE); |
2310 | backing_fmt = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FMT); | |
2311 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_ENCRYPT, false)) { | |
2312 | flags |= BLOCK_FLAG_ENCRYPT; | |
2313 | } | |
2314 | cluster_size = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, | |
2315 | DEFAULT_CLUSTER_SIZE); | |
2316 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
ffeaac9b | 2317 | prealloc = qapi_enum_parse(PreallocMode_lookup, buf, |
7fb1cf16 | 2318 | PREALLOC_MODE__MAX, PREALLOC_MODE_OFF, |
ffeaac9b HT |
2319 | &local_err); |
2320 | if (local_err) { | |
2321 | error_propagate(errp, local_err); | |
1bd0e2d1 CL |
2322 | ret = -EINVAL; |
2323 | goto finish; | |
2324 | } | |
2325 | g_free(buf); | |
2326 | buf = qemu_opt_get_del(opts, BLOCK_OPT_COMPAT_LEVEL); | |
2327 | if (!buf) { | |
2328 | /* keep the default */ | |
2329 | } else if (!strcmp(buf, "0.10")) { | |
2330 | version = 2; | |
2331 | } else if (!strcmp(buf, "1.1")) { | |
2332 | version = 3; | |
2333 | } else { | |
2334 | error_setg(errp, "Invalid compatibility level: '%s'", buf); | |
2335 | ret = -EINVAL; | |
2336 | goto finish; | |
2337 | } | |
2338 | ||
2339 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_LAZY_REFCOUNTS, false)) { | |
2340 | flags |= BLOCK_FLAG_LAZY_REFCOUNTS; | |
de5f3f40 KW |
2341 | } |
2342 | ||
ffeaac9b | 2343 | if (backing_file && prealloc != PREALLOC_MODE_OFF) { |
3ef6c40a HR |
2344 | error_setg(errp, "Backing file and preallocation cannot be used at " |
2345 | "the same time"); | |
1bd0e2d1 CL |
2346 | ret = -EINVAL; |
2347 | goto finish; | |
de5f3f40 KW |
2348 | } |
2349 | ||
bfe8043e | 2350 | if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { |
3ef6c40a HR |
2351 | error_setg(errp, "Lazy refcounts only supported with compatibility " |
2352 | "level 1.1 and above (use compat=1.1 or greater)"); | |
1bd0e2d1 CL |
2353 | ret = -EINVAL; |
2354 | goto finish; | |
bfe8043e SH |
2355 | } |
2356 | ||
06d05fa7 HR |
2357 | refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS, |
2358 | refcount_bits); | |
2359 | if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) { | |
2360 | error_setg(errp, "Refcount width must be a power of two and may not " | |
2361 | "exceed 64 bits"); | |
2362 | ret = -EINVAL; | |
2363 | goto finish; | |
2364 | } | |
2365 | ||
bd4b167f HR |
2366 | if (version < 3 && refcount_bits != 16) { |
2367 | error_setg(errp, "Different refcount widths than 16 bits require " | |
2368 | "compatibility level 1.1 or above (use compat=1.1 or " | |
2369 | "greater)"); | |
2370 | ret = -EINVAL; | |
2371 | goto finish; | |
2372 | } | |
2373 | ||
786a4ea8 | 2374 | refcount_order = ctz32(refcount_bits); |
bd4b167f | 2375 | |
180e9526 | 2376 | ret = qcow2_create2(filename, size, backing_file, backing_fmt, flags, |
bd4b167f HR |
2377 | cluster_size, prealloc, opts, version, refcount_order, |
2378 | &local_err); | |
84d18f06 | 2379 | if (local_err) { |
3ef6c40a HR |
2380 | error_propagate(errp, local_err); |
2381 | } | |
1bd0e2d1 CL |
2382 | |
2383 | finish: | |
2384 | g_free(backing_file); | |
2385 | g_free(backing_fmt); | |
2386 | g_free(buf); | |
3ef6c40a | 2387 | return ret; |
de5f3f40 KW |
2388 | } |
2389 | ||
621f0589 | 2390 | static coroutine_fn int qcow2_co_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 2391 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
621f0589 KW |
2392 | { |
2393 | int ret; | |
ff99129a | 2394 | BDRVQcow2State *s = bs->opaque; |
621f0589 KW |
2395 | |
2396 | /* Emulate misaligned zero writes */ | |
2397 | if (sector_num % s->cluster_sectors || nb_sectors % s->cluster_sectors) { | |
2398 | return -ENOTSUP; | |
2399 | } | |
2400 | ||
2401 | /* Whatever is left can use real zero clusters */ | |
2402 | qemu_co_mutex_lock(&s->lock); | |
2403 | ret = qcow2_zero_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
2404 | nb_sectors); | |
2405 | qemu_co_mutex_unlock(&s->lock); | |
2406 | ||
2407 | return ret; | |
2408 | } | |
2409 | ||
6db39ae2 PB |
2410 | static coroutine_fn int qcow2_co_discard(BlockDriverState *bs, |
2411 | int64_t sector_num, int nb_sectors) | |
5ea929e3 | 2412 | { |
6db39ae2 | 2413 | int ret; |
ff99129a | 2414 | BDRVQcow2State *s = bs->opaque; |
6db39ae2 PB |
2415 | |
2416 | qemu_co_mutex_lock(&s->lock); | |
2417 | ret = qcow2_discard_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
808c4b6f | 2418 | nb_sectors, QCOW2_DISCARD_REQUEST, false); |
6db39ae2 PB |
2419 | qemu_co_mutex_unlock(&s->lock); |
2420 | return ret; | |
5ea929e3 KW |
2421 | } |
2422 | ||
419b19d9 SH |
2423 | static int qcow2_truncate(BlockDriverState *bs, int64_t offset) |
2424 | { | |
ff99129a | 2425 | BDRVQcow2State *s = bs->opaque; |
2cf7cfa1 KW |
2426 | int64_t new_l1_size; |
2427 | int ret; | |
419b19d9 SH |
2428 | |
2429 | if (offset & 511) { | |
259b2173 | 2430 | error_report("The new size must be a multiple of 512"); |
419b19d9 SH |
2431 | return -EINVAL; |
2432 | } | |
2433 | ||
2434 | /* cannot proceed if image has snapshots */ | |
2435 | if (s->nb_snapshots) { | |
259b2173 | 2436 | error_report("Can't resize an image which has snapshots"); |
419b19d9 SH |
2437 | return -ENOTSUP; |
2438 | } | |
2439 | ||
2440 | /* shrinking is currently not supported */ | |
2441 | if (offset < bs->total_sectors * 512) { | |
259b2173 | 2442 | error_report("qcow2 doesn't support shrinking images yet"); |
419b19d9 SH |
2443 | return -ENOTSUP; |
2444 | } | |
2445 | ||
2446 | new_l1_size = size_to_l1(s, offset); | |
72893756 | 2447 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); |
419b19d9 SH |
2448 | if (ret < 0) { |
2449 | return ret; | |
2450 | } | |
2451 | ||
2452 | /* write updated header.size */ | |
2453 | offset = cpu_to_be64(offset); | |
9a4f4c31 | 2454 | ret = bdrv_pwrite_sync(bs->file->bs, offsetof(QCowHeader, size), |
8b3b7206 | 2455 | &offset, sizeof(uint64_t)); |
419b19d9 SH |
2456 | if (ret < 0) { |
2457 | return ret; | |
2458 | } | |
2459 | ||
2460 | s->l1_vm_state_index = new_l1_size; | |
2461 | return 0; | |
2462 | } | |
2463 | ||
20d97356 BS |
2464 | /* XXX: put compressed sectors first, then all the cluster aligned |
2465 | tables to avoid losing bytes in alignment */ | |
7c80ab3f JS |
2466 | static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num, |
2467 | const uint8_t *buf, int nb_sectors) | |
20d97356 | 2468 | { |
ff99129a | 2469 | BDRVQcow2State *s = bs->opaque; |
20d97356 BS |
2470 | z_stream strm; |
2471 | int ret, out_len; | |
2472 | uint8_t *out_buf; | |
2473 | uint64_t cluster_offset; | |
2474 | ||
2475 | if (nb_sectors == 0) { | |
2476 | /* align end of file to a sector boundary to ease reading with | |
2477 | sector based I/Os */ | |
9a4f4c31 KW |
2478 | cluster_offset = bdrv_getlength(bs->file->bs); |
2479 | return bdrv_truncate(bs->file->bs, cluster_offset); | |
20d97356 BS |
2480 | } |
2481 | ||
f4d38bef SH |
2482 | if (nb_sectors != s->cluster_sectors) { |
2483 | ret = -EINVAL; | |
2484 | ||
2485 | /* Zero-pad last write if image size is not cluster aligned */ | |
2486 | if (sector_num + nb_sectors == bs->total_sectors && | |
2487 | nb_sectors < s->cluster_sectors) { | |
2488 | uint8_t *pad_buf = qemu_blockalign(bs, s->cluster_size); | |
2489 | memset(pad_buf, 0, s->cluster_size); | |
2490 | memcpy(pad_buf, buf, nb_sectors * BDRV_SECTOR_SIZE); | |
2491 | ret = qcow2_write_compressed(bs, sector_num, | |
2492 | pad_buf, s->cluster_sectors); | |
2493 | qemu_vfree(pad_buf); | |
2494 | } | |
2495 | return ret; | |
2496 | } | |
20d97356 | 2497 | |
7267c094 | 2498 | out_buf = g_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); |
20d97356 BS |
2499 | |
2500 | /* best compression, small window, no zlib header */ | |
2501 | memset(&strm, 0, sizeof(strm)); | |
2502 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
2503 | Z_DEFLATED, -12, | |
2504 | 9, Z_DEFAULT_STRATEGY); | |
2505 | if (ret != 0) { | |
8f1efd00 KW |
2506 | ret = -EINVAL; |
2507 | goto fail; | |
20d97356 BS |
2508 | } |
2509 | ||
2510 | strm.avail_in = s->cluster_size; | |
2511 | strm.next_in = (uint8_t *)buf; | |
2512 | strm.avail_out = s->cluster_size; | |
2513 | strm.next_out = out_buf; | |
2514 | ||
2515 | ret = deflate(&strm, Z_FINISH); | |
2516 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
20d97356 | 2517 | deflateEnd(&strm); |
8f1efd00 KW |
2518 | ret = -EINVAL; |
2519 | goto fail; | |
20d97356 BS |
2520 | } |
2521 | out_len = strm.next_out - out_buf; | |
2522 | ||
2523 | deflateEnd(&strm); | |
2524 | ||
2525 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
2526 | /* could not compress: write normal cluster */ | |
8f1efd00 KW |
2527 | ret = bdrv_write(bs, sector_num, buf, s->cluster_sectors); |
2528 | if (ret < 0) { | |
2529 | goto fail; | |
2530 | } | |
20d97356 BS |
2531 | } else { |
2532 | cluster_offset = qcow2_alloc_compressed_cluster_offset(bs, | |
2533 | sector_num << 9, out_len); | |
8f1efd00 KW |
2534 | if (!cluster_offset) { |
2535 | ret = -EIO; | |
2536 | goto fail; | |
2537 | } | |
20d97356 | 2538 | cluster_offset &= s->cluster_offset_mask; |
cf93980e | 2539 | |
231bb267 | 2540 | ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len); |
cf93980e HR |
2541 | if (ret < 0) { |
2542 | goto fail; | |
2543 | } | |
2544 | ||
66f82cee | 2545 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); |
9a4f4c31 | 2546 | ret = bdrv_pwrite(bs->file->bs, cluster_offset, out_buf, out_len); |
8f1efd00 KW |
2547 | if (ret < 0) { |
2548 | goto fail; | |
20d97356 BS |
2549 | } |
2550 | } | |
2551 | ||
8f1efd00 KW |
2552 | ret = 0; |
2553 | fail: | |
7267c094 | 2554 | g_free(out_buf); |
8f1efd00 | 2555 | return ret; |
20d97356 BS |
2556 | } |
2557 | ||
94054183 HR |
2558 | static int make_completely_empty(BlockDriverState *bs) |
2559 | { | |
ff99129a | 2560 | BDRVQcow2State *s = bs->opaque; |
94054183 HR |
2561 | int ret, l1_clusters; |
2562 | int64_t offset; | |
2563 | uint64_t *new_reftable = NULL; | |
2564 | uint64_t rt_entry, l1_size2; | |
2565 | struct { | |
2566 | uint64_t l1_offset; | |
2567 | uint64_t reftable_offset; | |
2568 | uint32_t reftable_clusters; | |
2569 | } QEMU_PACKED l1_ofs_rt_ofs_cls; | |
2570 | ||
2571 | ret = qcow2_cache_empty(bs, s->l2_table_cache); | |
2572 | if (ret < 0) { | |
2573 | goto fail; | |
2574 | } | |
2575 | ||
2576 | ret = qcow2_cache_empty(bs, s->refcount_block_cache); | |
2577 | if (ret < 0) { | |
2578 | goto fail; | |
2579 | } | |
2580 | ||
2581 | /* Refcounts will be broken utterly */ | |
2582 | ret = qcow2_mark_dirty(bs); | |
2583 | if (ret < 0) { | |
2584 | goto fail; | |
2585 | } | |
2586 | ||
2587 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
2588 | ||
2589 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
2590 | l1_size2 = (uint64_t)s->l1_size * sizeof(uint64_t); | |
2591 | ||
2592 | /* After this call, neither the in-memory nor the on-disk refcount | |
2593 | * information accurately describe the actual references */ | |
2594 | ||
9a4f4c31 | 2595 | ret = bdrv_write_zeroes(bs->file->bs, s->l1_table_offset / BDRV_SECTOR_SIZE, |
94054183 HR |
2596 | l1_clusters * s->cluster_sectors, 0); |
2597 | if (ret < 0) { | |
2598 | goto fail_broken_refcounts; | |
2599 | } | |
2600 | memset(s->l1_table, 0, l1_size2); | |
2601 | ||
2602 | BLKDBG_EVENT(bs->file, BLKDBG_EMPTY_IMAGE_PREPARE); | |
2603 | ||
2604 | /* Overwrite enough clusters at the beginning of the sectors to place | |
2605 | * the refcount table, a refcount block and the L1 table in; this may | |
2606 | * overwrite parts of the existing refcount and L1 table, which is not | |
2607 | * an issue because the dirty flag is set, complete data loss is in fact | |
2608 | * desired and partial data loss is consequently fine as well */ | |
9a4f4c31 | 2609 | ret = bdrv_write_zeroes(bs->file->bs, s->cluster_size / BDRV_SECTOR_SIZE, |
94054183 HR |
2610 | (2 + l1_clusters) * s->cluster_size / |
2611 | BDRV_SECTOR_SIZE, 0); | |
2612 | /* This call (even if it failed overall) may have overwritten on-disk | |
2613 | * refcount structures; in that case, the in-memory refcount information | |
2614 | * will probably differ from the on-disk information which makes the BDS | |
2615 | * unusable */ | |
2616 | if (ret < 0) { | |
2617 | goto fail_broken_refcounts; | |
2618 | } | |
2619 | ||
2620 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
2621 | BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE); | |
2622 | ||
2623 | /* "Create" an empty reftable (one cluster) directly after the image | |
2624 | * header and an empty L1 table three clusters after the image header; | |
2625 | * the cluster between those two will be used as the first refblock */ | |
2626 | cpu_to_be64w(&l1_ofs_rt_ofs_cls.l1_offset, 3 * s->cluster_size); | |
2627 | cpu_to_be64w(&l1_ofs_rt_ofs_cls.reftable_offset, s->cluster_size); | |
2628 | cpu_to_be32w(&l1_ofs_rt_ofs_cls.reftable_clusters, 1); | |
9a4f4c31 | 2629 | ret = bdrv_pwrite_sync(bs->file->bs, offsetof(QCowHeader, l1_table_offset), |
94054183 HR |
2630 | &l1_ofs_rt_ofs_cls, sizeof(l1_ofs_rt_ofs_cls)); |
2631 | if (ret < 0) { | |
2632 | goto fail_broken_refcounts; | |
2633 | } | |
2634 | ||
2635 | s->l1_table_offset = 3 * s->cluster_size; | |
2636 | ||
2637 | new_reftable = g_try_new0(uint64_t, s->cluster_size / sizeof(uint64_t)); | |
2638 | if (!new_reftable) { | |
2639 | ret = -ENOMEM; | |
2640 | goto fail_broken_refcounts; | |
2641 | } | |
2642 | ||
2643 | s->refcount_table_offset = s->cluster_size; | |
2644 | s->refcount_table_size = s->cluster_size / sizeof(uint64_t); | |
2645 | ||
2646 | g_free(s->refcount_table); | |
2647 | s->refcount_table = new_reftable; | |
2648 | new_reftable = NULL; | |
2649 | ||
2650 | /* Now the in-memory refcount information again corresponds to the on-disk | |
2651 | * information (reftable is empty and no refblocks (the refblock cache is | |
2652 | * empty)); however, this means some clusters (e.g. the image header) are | |
2653 | * referenced, but not refcounted, but the normal qcow2 code assumes that | |
2654 | * the in-memory information is always correct */ | |
2655 | ||
2656 | BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC); | |
2657 | ||
2658 | /* Enter the first refblock into the reftable */ | |
2659 | rt_entry = cpu_to_be64(2 * s->cluster_size); | |
9a4f4c31 | 2660 | ret = bdrv_pwrite_sync(bs->file->bs, s->cluster_size, |
94054183 HR |
2661 | &rt_entry, sizeof(rt_entry)); |
2662 | if (ret < 0) { | |
2663 | goto fail_broken_refcounts; | |
2664 | } | |
2665 | s->refcount_table[0] = 2 * s->cluster_size; | |
2666 | ||
2667 | s->free_cluster_index = 0; | |
2668 | assert(3 + l1_clusters <= s->refcount_block_size); | |
2669 | offset = qcow2_alloc_clusters(bs, 3 * s->cluster_size + l1_size2); | |
2670 | if (offset < 0) { | |
2671 | ret = offset; | |
2672 | goto fail_broken_refcounts; | |
2673 | } else if (offset > 0) { | |
2674 | error_report("First cluster in emptied image is in use"); | |
2675 | abort(); | |
2676 | } | |
2677 | ||
2678 | /* Now finally the in-memory information corresponds to the on-disk | |
2679 | * structures and is correct */ | |
2680 | ret = qcow2_mark_clean(bs); | |
2681 | if (ret < 0) { | |
2682 | goto fail; | |
2683 | } | |
2684 | ||
9a4f4c31 | 2685 | ret = bdrv_truncate(bs->file->bs, (3 + l1_clusters) * s->cluster_size); |
94054183 HR |
2686 | if (ret < 0) { |
2687 | goto fail; | |
2688 | } | |
2689 | ||
2690 | return 0; | |
2691 | ||
2692 | fail_broken_refcounts: | |
2693 | /* The BDS is unusable at this point. If we wanted to make it usable, we | |
2694 | * would have to call qcow2_refcount_close(), qcow2_refcount_init(), | |
2695 | * qcow2_check_refcounts(), qcow2_refcount_close() and qcow2_refcount_init() | |
2696 | * again. However, because the functions which could have caused this error | |
2697 | * path to be taken are used by those functions as well, it's very likely | |
2698 | * that that sequence will fail as well. Therefore, just eject the BDS. */ | |
2699 | bs->drv = NULL; | |
2700 | ||
2701 | fail: | |
2702 | g_free(new_reftable); | |
2703 | return ret; | |
2704 | } | |
2705 | ||
491d27e2 HR |
2706 | static int qcow2_make_empty(BlockDriverState *bs) |
2707 | { | |
ff99129a | 2708 | BDRVQcow2State *s = bs->opaque; |
491d27e2 HR |
2709 | uint64_t start_sector; |
2710 | int sector_step = INT_MAX / BDRV_SECTOR_SIZE; | |
94054183 HR |
2711 | int l1_clusters, ret = 0; |
2712 | ||
2713 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
2714 | ||
2715 | if (s->qcow_version >= 3 && !s->snapshots && | |
2716 | 3 + l1_clusters <= s->refcount_block_size) { | |
2717 | /* The following function only works for qcow2 v3 images (it requires | |
2718 | * the dirty flag) and only as long as there are no snapshots (because | |
2719 | * it completely empties the image). Furthermore, the L1 table and three | |
2720 | * additional clusters (image header, refcount table, one refcount | |
2721 | * block) have to fit inside one refcount block. */ | |
2722 | return make_completely_empty(bs); | |
2723 | } | |
491d27e2 | 2724 | |
94054183 HR |
2725 | /* This fallback code simply discards every active cluster; this is slow, |
2726 | * but works in all cases */ | |
491d27e2 HR |
2727 | for (start_sector = 0; start_sector < bs->total_sectors; |
2728 | start_sector += sector_step) | |
2729 | { | |
2730 | /* As this function is generally used after committing an external | |
2731 | * snapshot, QCOW2_DISCARD_SNAPSHOT seems appropriate. Also, the | |
2732 | * default action for this kind of discard is to pass the discard, | |
2733 | * which will ideally result in an actually smaller image file, as | |
2734 | * is probably desired. */ | |
2735 | ret = qcow2_discard_clusters(bs, start_sector * BDRV_SECTOR_SIZE, | |
2736 | MIN(sector_step, | |
2737 | bs->total_sectors - start_sector), | |
2738 | QCOW2_DISCARD_SNAPSHOT, true); | |
2739 | if (ret < 0) { | |
2740 | break; | |
2741 | } | |
2742 | } | |
2743 | ||
2744 | return ret; | |
2745 | } | |
2746 | ||
a968168c | 2747 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
20d97356 | 2748 | { |
ff99129a | 2749 | BDRVQcow2State *s = bs->opaque; |
29c1a730 KW |
2750 | int ret; |
2751 | ||
8b94ff85 | 2752 | qemu_co_mutex_lock(&s->lock); |
29c1a730 KW |
2753 | ret = qcow2_cache_flush(bs, s->l2_table_cache); |
2754 | if (ret < 0) { | |
c95de7e2 | 2755 | qemu_co_mutex_unlock(&s->lock); |
8b94ff85 | 2756 | return ret; |
29c1a730 KW |
2757 | } |
2758 | ||
bfe8043e SH |
2759 | if (qcow2_need_accurate_refcounts(s)) { |
2760 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
2761 | if (ret < 0) { | |
2762 | qemu_co_mutex_unlock(&s->lock); | |
2763 | return ret; | |
2764 | } | |
29c1a730 | 2765 | } |
8b94ff85 | 2766 | qemu_co_mutex_unlock(&s->lock); |
29c1a730 | 2767 | |
eb489bb1 KW |
2768 | return 0; |
2769 | } | |
2770 | ||
7c80ab3f | 2771 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
20d97356 | 2772 | { |
ff99129a | 2773 | BDRVQcow2State *s = bs->opaque; |
95de6d70 PB |
2774 | bdi->unallocated_blocks_are_zero = true; |
2775 | bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3); | |
20d97356 | 2776 | bdi->cluster_size = s->cluster_size; |
7c80ab3f | 2777 | bdi->vm_state_offset = qcow2_vm_state_offset(s); |
20d97356 BS |
2778 | return 0; |
2779 | } | |
2780 | ||
37764dfb HR |
2781 | static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs) |
2782 | { | |
ff99129a | 2783 | BDRVQcow2State *s = bs->opaque; |
37764dfb HR |
2784 | ImageInfoSpecific *spec_info = g_new(ImageInfoSpecific, 1); |
2785 | ||
2786 | *spec_info = (ImageInfoSpecific){ | |
6a8f9661 EB |
2787 | .type = IMAGE_INFO_SPECIFIC_KIND_QCOW2, |
2788 | .u.qcow2 = g_new(ImageInfoSpecificQCow2, 1), | |
37764dfb HR |
2789 | }; |
2790 | if (s->qcow_version == 2) { | |
6a8f9661 | 2791 | *spec_info->u.qcow2 = (ImageInfoSpecificQCow2){ |
0709c5a1 HR |
2792 | .compat = g_strdup("0.10"), |
2793 | .refcount_bits = s->refcount_bits, | |
37764dfb HR |
2794 | }; |
2795 | } else if (s->qcow_version == 3) { | |
6a8f9661 | 2796 | *spec_info->u.qcow2 = (ImageInfoSpecificQCow2){ |
37764dfb HR |
2797 | .compat = g_strdup("1.1"), |
2798 | .lazy_refcounts = s->compatible_features & | |
2799 | QCOW2_COMPAT_LAZY_REFCOUNTS, | |
2800 | .has_lazy_refcounts = true, | |
9009b196 HR |
2801 | .corrupt = s->incompatible_features & |
2802 | QCOW2_INCOMPAT_CORRUPT, | |
2803 | .has_corrupt = true, | |
0709c5a1 | 2804 | .refcount_bits = s->refcount_bits, |
37764dfb | 2805 | }; |
b1fc8f93 DL |
2806 | } else { |
2807 | /* if this assertion fails, this probably means a new version was | |
2808 | * added without having it covered here */ | |
2809 | assert(false); | |
37764dfb HR |
2810 | } |
2811 | ||
2812 | return spec_info; | |
2813 | } | |
2814 | ||
20d97356 BS |
2815 | #if 0 |
2816 | static void dump_refcounts(BlockDriverState *bs) | |
2817 | { | |
ff99129a | 2818 | BDRVQcow2State *s = bs->opaque; |
20d97356 BS |
2819 | int64_t nb_clusters, k, k1, size; |
2820 | int refcount; | |
2821 | ||
9a4f4c31 | 2822 | size = bdrv_getlength(bs->file->bs); |
20d97356 BS |
2823 | nb_clusters = size_to_clusters(s, size); |
2824 | for(k = 0; k < nb_clusters;) { | |
2825 | k1 = k; | |
2826 | refcount = get_refcount(bs, k); | |
2827 | k++; | |
2828 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
2829 | k++; | |
0bfcd599 BS |
2830 | printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount, |
2831 | k - k1); | |
20d97356 BS |
2832 | } |
2833 | } | |
2834 | #endif | |
2835 | ||
cf8074b3 KW |
2836 | static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
2837 | int64_t pos) | |
20d97356 | 2838 | { |
ff99129a | 2839 | BDRVQcow2State *s = bs->opaque; |
eedff66f | 2840 | int64_t total_sectors = bs->total_sectors; |
6e13610a | 2841 | bool zero_beyond_eof = bs->zero_beyond_eof; |
20d97356 BS |
2842 | int ret; |
2843 | ||
66f82cee | 2844 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
6e13610a | 2845 | bs->zero_beyond_eof = false; |
8d3b1a2d | 2846 | ret = bdrv_pwritev(bs, qcow2_vm_state_offset(s) + pos, qiov); |
6e13610a | 2847 | bs->zero_beyond_eof = zero_beyond_eof; |
20d97356 | 2848 | |
eedff66f HR |
2849 | /* bdrv_co_do_writev will have increased the total_sectors value to include |
2850 | * the VM state - the VM state is however not an actual part of the block | |
2851 | * device, therefore, we need to restore the old value. */ | |
2852 | bs->total_sectors = total_sectors; | |
2853 | ||
20d97356 BS |
2854 | return ret; |
2855 | } | |
2856 | ||
7c80ab3f JS |
2857 | static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
2858 | int64_t pos, int size) | |
20d97356 | 2859 | { |
ff99129a | 2860 | BDRVQcow2State *s = bs->opaque; |
0d51b4de | 2861 | bool zero_beyond_eof = bs->zero_beyond_eof; |
20d97356 BS |
2862 | int ret; |
2863 | ||
66f82cee | 2864 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
0d51b4de | 2865 | bs->zero_beyond_eof = false; |
7c80ab3f | 2866 | ret = bdrv_pread(bs, qcow2_vm_state_offset(s) + pos, buf, size); |
0d51b4de | 2867 | bs->zero_beyond_eof = zero_beyond_eof; |
20d97356 BS |
2868 | |
2869 | return ret; | |
2870 | } | |
2871 | ||
9296b3ed HR |
2872 | /* |
2873 | * Downgrades an image's version. To achieve this, any incompatible features | |
2874 | * have to be removed. | |
2875 | */ | |
4057a2b2 | 2876 | static int qcow2_downgrade(BlockDriverState *bs, int target_version, |
8b13976d | 2877 | BlockDriverAmendStatusCB *status_cb, void *cb_opaque) |
9296b3ed | 2878 | { |
ff99129a | 2879 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
2880 | int current_version = s->qcow_version; |
2881 | int ret; | |
2882 | ||
2883 | if (target_version == current_version) { | |
2884 | return 0; | |
2885 | } else if (target_version > current_version) { | |
2886 | return -EINVAL; | |
2887 | } else if (target_version != 2) { | |
2888 | return -EINVAL; | |
2889 | } | |
2890 | ||
2891 | if (s->refcount_order != 4) { | |
61ce55fc | 2892 | error_report("compat=0.10 requires refcount_bits=16"); |
9296b3ed HR |
2893 | return -ENOTSUP; |
2894 | } | |
2895 | ||
2896 | /* clear incompatible features */ | |
2897 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
2898 | ret = qcow2_mark_clean(bs); | |
2899 | if (ret < 0) { | |
2900 | return ret; | |
2901 | } | |
2902 | } | |
2903 | ||
2904 | /* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in | |
2905 | * the first place; if that happens nonetheless, returning -ENOTSUP is the | |
2906 | * best thing to do anyway */ | |
2907 | ||
2908 | if (s->incompatible_features) { | |
2909 | return -ENOTSUP; | |
2910 | } | |
2911 | ||
2912 | /* since we can ignore compatible features, we can set them to 0 as well */ | |
2913 | s->compatible_features = 0; | |
2914 | /* if lazy refcounts have been used, they have already been fixed through | |
2915 | * clearing the dirty flag */ | |
2916 | ||
2917 | /* clearing autoclear features is trivial */ | |
2918 | s->autoclear_features = 0; | |
2919 | ||
8b13976d | 2920 | ret = qcow2_expand_zero_clusters(bs, status_cb, cb_opaque); |
9296b3ed HR |
2921 | if (ret < 0) { |
2922 | return ret; | |
2923 | } | |
2924 | ||
2925 | s->qcow_version = target_version; | |
2926 | ret = qcow2_update_header(bs); | |
2927 | if (ret < 0) { | |
2928 | s->qcow_version = current_version; | |
2929 | return ret; | |
2930 | } | |
2931 | return 0; | |
2932 | } | |
2933 | ||
c293a809 HR |
2934 | typedef enum Qcow2AmendOperation { |
2935 | /* This is the value Qcow2AmendHelperCBInfo::last_operation will be | |
2936 | * statically initialized to so that the helper CB can discern the first | |
2937 | * invocation from an operation change */ | |
2938 | QCOW2_NO_OPERATION = 0, | |
2939 | ||
61ce55fc | 2940 | QCOW2_CHANGING_REFCOUNT_ORDER, |
c293a809 HR |
2941 | QCOW2_DOWNGRADING, |
2942 | } Qcow2AmendOperation; | |
2943 | ||
2944 | typedef struct Qcow2AmendHelperCBInfo { | |
2945 | /* The code coordinating the amend operations should only modify | |
2946 | * these four fields; the rest will be managed by the CB */ | |
2947 | BlockDriverAmendStatusCB *original_status_cb; | |
2948 | void *original_cb_opaque; | |
2949 | ||
2950 | Qcow2AmendOperation current_operation; | |
2951 | ||
2952 | /* Total number of operations to perform (only set once) */ | |
2953 | int total_operations; | |
2954 | ||
2955 | /* The following fields are managed by the CB */ | |
2956 | ||
2957 | /* Number of operations completed */ | |
2958 | int operations_completed; | |
2959 | ||
2960 | /* Cumulative offset of all completed operations */ | |
2961 | int64_t offset_completed; | |
2962 | ||
2963 | Qcow2AmendOperation last_operation; | |
2964 | int64_t last_work_size; | |
2965 | } Qcow2AmendHelperCBInfo; | |
2966 | ||
2967 | static void qcow2_amend_helper_cb(BlockDriverState *bs, | |
2968 | int64_t operation_offset, | |
2969 | int64_t operation_work_size, void *opaque) | |
2970 | { | |
2971 | Qcow2AmendHelperCBInfo *info = opaque; | |
2972 | int64_t current_work_size; | |
2973 | int64_t projected_work_size; | |
2974 | ||
2975 | if (info->current_operation != info->last_operation) { | |
2976 | if (info->last_operation != QCOW2_NO_OPERATION) { | |
2977 | info->offset_completed += info->last_work_size; | |
2978 | info->operations_completed++; | |
2979 | } | |
2980 | ||
2981 | info->last_operation = info->current_operation; | |
2982 | } | |
2983 | ||
2984 | assert(info->total_operations > 0); | |
2985 | assert(info->operations_completed < info->total_operations); | |
2986 | ||
2987 | info->last_work_size = operation_work_size; | |
2988 | ||
2989 | current_work_size = info->offset_completed + operation_work_size; | |
2990 | ||
2991 | /* current_work_size is the total work size for (operations_completed + 1) | |
2992 | * operations (which includes this one), so multiply it by the number of | |
2993 | * operations not covered and divide it by the number of operations | |
2994 | * covered to get a projection for the operations not covered */ | |
2995 | projected_work_size = current_work_size * (info->total_operations - | |
2996 | info->operations_completed - 1) | |
2997 | / (info->operations_completed + 1); | |
2998 | ||
2999 | info->original_status_cb(bs, info->offset_completed + operation_offset, | |
3000 | current_work_size + projected_work_size, | |
3001 | info->original_cb_opaque); | |
3002 | } | |
3003 | ||
77485434 | 3004 | static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts, |
8b13976d HR |
3005 | BlockDriverAmendStatusCB *status_cb, |
3006 | void *cb_opaque) | |
9296b3ed | 3007 | { |
ff99129a | 3008 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
3009 | int old_version = s->qcow_version, new_version = old_version; |
3010 | uint64_t new_size = 0; | |
3011 | const char *backing_file = NULL, *backing_format = NULL; | |
3012 | bool lazy_refcounts = s->use_lazy_refcounts; | |
1bd0e2d1 CL |
3013 | const char *compat = NULL; |
3014 | uint64_t cluster_size = s->cluster_size; | |
3015 | bool encrypt; | |
61ce55fc | 3016 | int refcount_bits = s->refcount_bits; |
9296b3ed | 3017 | int ret; |
1bd0e2d1 | 3018 | QemuOptDesc *desc = opts->list->desc; |
c293a809 | 3019 | Qcow2AmendHelperCBInfo helper_cb_info; |
9296b3ed | 3020 | |
1bd0e2d1 CL |
3021 | while (desc && desc->name) { |
3022 | if (!qemu_opt_find(opts, desc->name)) { | |
9296b3ed | 3023 | /* only change explicitly defined options */ |
1bd0e2d1 | 3024 | desc++; |
9296b3ed HR |
3025 | continue; |
3026 | } | |
3027 | ||
8a17b83c HR |
3028 | if (!strcmp(desc->name, BLOCK_OPT_COMPAT_LEVEL)) { |
3029 | compat = qemu_opt_get(opts, BLOCK_OPT_COMPAT_LEVEL); | |
1bd0e2d1 | 3030 | if (!compat) { |
9296b3ed | 3031 | /* preserve default */ |
1bd0e2d1 | 3032 | } else if (!strcmp(compat, "0.10")) { |
9296b3ed | 3033 | new_version = 2; |
1bd0e2d1 | 3034 | } else if (!strcmp(compat, "1.1")) { |
9296b3ed HR |
3035 | new_version = 3; |
3036 | } else { | |
29d72431 | 3037 | error_report("Unknown compatibility level %s", compat); |
9296b3ed HR |
3038 | return -EINVAL; |
3039 | } | |
8a17b83c | 3040 | } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) { |
29d72431 | 3041 | error_report("Cannot change preallocation mode"); |
9296b3ed | 3042 | return -ENOTSUP; |
8a17b83c HR |
3043 | } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) { |
3044 | new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); | |
3045 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FILE)) { | |
3046 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); | |
3047 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FMT)) { | |
3048 | backing_format = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); | |
3049 | } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT)) { | |
3050 | encrypt = qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, | |
f6fa64f6 DB |
3051 | !!s->cipher); |
3052 | ||
3053 | if (encrypt != !!s->cipher) { | |
29d72431 | 3054 | error_report("Changing the encryption flag is not supported"); |
9296b3ed HR |
3055 | return -ENOTSUP; |
3056 | } | |
8a17b83c HR |
3057 | } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) { |
3058 | cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, | |
1bd0e2d1 CL |
3059 | cluster_size); |
3060 | if (cluster_size != s->cluster_size) { | |
29d72431 | 3061 | error_report("Changing the cluster size is not supported"); |
9296b3ed HR |
3062 | return -ENOTSUP; |
3063 | } | |
8a17b83c HR |
3064 | } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) { |
3065 | lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS, | |
1bd0e2d1 | 3066 | lazy_refcounts); |
06d05fa7 | 3067 | } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) { |
61ce55fc HR |
3068 | refcount_bits = qemu_opt_get_number(opts, BLOCK_OPT_REFCOUNT_BITS, |
3069 | refcount_bits); | |
3070 | ||
3071 | if (refcount_bits <= 0 || refcount_bits > 64 || | |
3072 | !is_power_of_2(refcount_bits)) | |
3073 | { | |
3074 | error_report("Refcount width must be a power of two and may " | |
3075 | "not exceed 64 bits"); | |
3076 | return -EINVAL; | |
3077 | } | |
9296b3ed | 3078 | } else { |
164e0f89 | 3079 | /* if this point is reached, this probably means a new option was |
9296b3ed | 3080 | * added without having it covered here */ |
164e0f89 | 3081 | abort(); |
9296b3ed | 3082 | } |
1bd0e2d1 CL |
3083 | |
3084 | desc++; | |
9296b3ed HR |
3085 | } |
3086 | ||
c293a809 HR |
3087 | helper_cb_info = (Qcow2AmendHelperCBInfo){ |
3088 | .original_status_cb = status_cb, | |
3089 | .original_cb_opaque = cb_opaque, | |
3090 | .total_operations = (new_version < old_version) | |
61ce55fc | 3091 | + (s->refcount_bits != refcount_bits) |
c293a809 HR |
3092 | }; |
3093 | ||
1038bbb8 HR |
3094 | /* Upgrade first (some features may require compat=1.1) */ |
3095 | if (new_version > old_version) { | |
3096 | s->qcow_version = new_version; | |
3097 | ret = qcow2_update_header(bs); | |
3098 | if (ret < 0) { | |
3099 | s->qcow_version = old_version; | |
3100 | return ret; | |
9296b3ed HR |
3101 | } |
3102 | } | |
3103 | ||
61ce55fc HR |
3104 | if (s->refcount_bits != refcount_bits) { |
3105 | int refcount_order = ctz32(refcount_bits); | |
3106 | Error *local_error = NULL; | |
3107 | ||
3108 | if (new_version < 3 && refcount_bits != 16) { | |
3109 | error_report("Different refcount widths than 16 bits require " | |
3110 | "compatibility level 1.1 or above (use compat=1.1 or " | |
3111 | "greater)"); | |
3112 | return -EINVAL; | |
3113 | } | |
3114 | ||
3115 | helper_cb_info.current_operation = QCOW2_CHANGING_REFCOUNT_ORDER; | |
3116 | ret = qcow2_change_refcount_order(bs, refcount_order, | |
3117 | &qcow2_amend_helper_cb, | |
3118 | &helper_cb_info, &local_error); | |
3119 | if (ret < 0) { | |
3120 | error_report_err(local_error); | |
3121 | return ret; | |
3122 | } | |
3123 | } | |
3124 | ||
9296b3ed | 3125 | if (backing_file || backing_format) { |
e4603fe1 KW |
3126 | ret = qcow2_change_backing_file(bs, |
3127 | backing_file ?: s->image_backing_file, | |
3128 | backing_format ?: s->image_backing_format); | |
9296b3ed HR |
3129 | if (ret < 0) { |
3130 | return ret; | |
3131 | } | |
3132 | } | |
3133 | ||
3134 | if (s->use_lazy_refcounts != lazy_refcounts) { | |
3135 | if (lazy_refcounts) { | |
1038bbb8 | 3136 | if (new_version < 3) { |
29d72431 HR |
3137 | error_report("Lazy refcounts only supported with compatibility " |
3138 | "level 1.1 and above (use compat=1.1 or greater)"); | |
9296b3ed HR |
3139 | return -EINVAL; |
3140 | } | |
3141 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3142 | ret = qcow2_update_header(bs); | |
3143 | if (ret < 0) { | |
3144 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3145 | return ret; | |
3146 | } | |
3147 | s->use_lazy_refcounts = true; | |
3148 | } else { | |
3149 | /* make image clean first */ | |
3150 | ret = qcow2_mark_clean(bs); | |
3151 | if (ret < 0) { | |
3152 | return ret; | |
3153 | } | |
3154 | /* now disallow lazy refcounts */ | |
3155 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3156 | ret = qcow2_update_header(bs); | |
3157 | if (ret < 0) { | |
3158 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
3159 | return ret; | |
3160 | } | |
3161 | s->use_lazy_refcounts = false; | |
3162 | } | |
3163 | } | |
3164 | ||
3165 | if (new_size) { | |
3166 | ret = bdrv_truncate(bs, new_size); | |
3167 | if (ret < 0) { | |
3168 | return ret; | |
3169 | } | |
3170 | } | |
3171 | ||
1038bbb8 HR |
3172 | /* Downgrade last (so unsupported features can be removed before) */ |
3173 | if (new_version < old_version) { | |
c293a809 HR |
3174 | helper_cb_info.current_operation = QCOW2_DOWNGRADING; |
3175 | ret = qcow2_downgrade(bs, new_version, &qcow2_amend_helper_cb, | |
3176 | &helper_cb_info); | |
1038bbb8 HR |
3177 | if (ret < 0) { |
3178 | return ret; | |
3179 | } | |
3180 | } | |
3181 | ||
9296b3ed HR |
3182 | return 0; |
3183 | } | |
3184 | ||
85186ebd HR |
3185 | /* |
3186 | * If offset or size are negative, respectively, they will not be included in | |
3187 | * the BLOCK_IMAGE_CORRUPTED event emitted. | |
3188 | * fatal will be ignored for read-only BDS; corruptions found there will always | |
3189 | * be considered non-fatal. | |
3190 | */ | |
3191 | void qcow2_signal_corruption(BlockDriverState *bs, bool fatal, int64_t offset, | |
3192 | int64_t size, const char *message_format, ...) | |
3193 | { | |
ff99129a | 3194 | BDRVQcow2State *s = bs->opaque; |
dc881b44 | 3195 | const char *node_name; |
85186ebd HR |
3196 | char *message; |
3197 | va_list ap; | |
3198 | ||
3199 | fatal = fatal && !bs->read_only; | |
3200 | ||
3201 | if (s->signaled_corruption && | |
3202 | (!fatal || (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT))) | |
3203 | { | |
3204 | return; | |
3205 | } | |
3206 | ||
3207 | va_start(ap, message_format); | |
3208 | message = g_strdup_vprintf(message_format, ap); | |
3209 | va_end(ap); | |
3210 | ||
3211 | if (fatal) { | |
3212 | fprintf(stderr, "qcow2: Marking image as corrupt: %s; further " | |
3213 | "corruption events will be suppressed\n", message); | |
3214 | } else { | |
3215 | fprintf(stderr, "qcow2: Image is corrupt: %s; further non-fatal " | |
3216 | "corruption events will be suppressed\n", message); | |
3217 | } | |
3218 | ||
dc881b44 AG |
3219 | node_name = bdrv_get_node_name(bs); |
3220 | qapi_event_send_block_image_corrupted(bdrv_get_device_name(bs), | |
3221 | *node_name != '\0', node_name, | |
3222 | message, offset >= 0, offset, | |
3223 | size >= 0, size, | |
85186ebd HR |
3224 | fatal, &error_abort); |
3225 | g_free(message); | |
3226 | ||
3227 | if (fatal) { | |
3228 | qcow2_mark_corrupt(bs); | |
3229 | bs->drv = NULL; /* make BDS unusable */ | |
3230 | } | |
3231 | ||
3232 | s->signaled_corruption = true; | |
3233 | } | |
3234 | ||
1bd0e2d1 CL |
3235 | static QemuOptsList qcow2_create_opts = { |
3236 | .name = "qcow2-create-opts", | |
3237 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_create_opts.head), | |
3238 | .desc = { | |
3239 | { | |
3240 | .name = BLOCK_OPT_SIZE, | |
3241 | .type = QEMU_OPT_SIZE, | |
3242 | .help = "Virtual disk size" | |
3243 | }, | |
3244 | { | |
3245 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
3246 | .type = QEMU_OPT_STRING, | |
3247 | .help = "Compatibility level (0.10 or 1.1)" | |
3248 | }, | |
3249 | { | |
3250 | .name = BLOCK_OPT_BACKING_FILE, | |
3251 | .type = QEMU_OPT_STRING, | |
3252 | .help = "File name of a base image" | |
3253 | }, | |
3254 | { | |
3255 | .name = BLOCK_OPT_BACKING_FMT, | |
3256 | .type = QEMU_OPT_STRING, | |
3257 | .help = "Image format of the base image" | |
3258 | }, | |
3259 | { | |
3260 | .name = BLOCK_OPT_ENCRYPT, | |
3261 | .type = QEMU_OPT_BOOL, | |
3262 | .help = "Encrypt the image", | |
3263 | .def_value_str = "off" | |
3264 | }, | |
3265 | { | |
3266 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
3267 | .type = QEMU_OPT_SIZE, | |
3268 | .help = "qcow2 cluster size", | |
3269 | .def_value_str = stringify(DEFAULT_CLUSTER_SIZE) | |
3270 | }, | |
3271 | { | |
3272 | .name = BLOCK_OPT_PREALLOC, | |
3273 | .type = QEMU_OPT_STRING, | |
0e4271b7 HT |
3274 | .help = "Preallocation mode (allowed values: off, metadata, " |
3275 | "falloc, full)" | |
1bd0e2d1 CL |
3276 | }, |
3277 | { | |
3278 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
3279 | .type = QEMU_OPT_BOOL, | |
3280 | .help = "Postpone refcount updates", | |
3281 | .def_value_str = "off" | |
3282 | }, | |
06d05fa7 HR |
3283 | { |
3284 | .name = BLOCK_OPT_REFCOUNT_BITS, | |
3285 | .type = QEMU_OPT_NUMBER, | |
3286 | .help = "Width of a reference count entry in bits", | |
3287 | .def_value_str = "16" | |
3288 | }, | |
1bd0e2d1 CL |
3289 | { /* end of list */ } |
3290 | } | |
20d97356 BS |
3291 | }; |
3292 | ||
5f535a94 | 3293 | BlockDriver bdrv_qcow2 = { |
7c80ab3f | 3294 | .format_name = "qcow2", |
ff99129a | 3295 | .instance_size = sizeof(BDRVQcow2State), |
7c80ab3f JS |
3296 | .bdrv_probe = qcow2_probe, |
3297 | .bdrv_open = qcow2_open, | |
3298 | .bdrv_close = qcow2_close, | |
21d82ac9 | 3299 | .bdrv_reopen_prepare = qcow2_reopen_prepare, |
5b0959a7 KW |
3300 | .bdrv_reopen_commit = qcow2_reopen_commit, |
3301 | .bdrv_reopen_abort = qcow2_reopen_abort, | |
5365f44d | 3302 | .bdrv_join_options = qcow2_join_options, |
c282e1fd | 3303 | .bdrv_create = qcow2_create, |
3ac21627 | 3304 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 3305 | .bdrv_co_get_block_status = qcow2_co_get_block_status, |
7c80ab3f | 3306 | .bdrv_set_key = qcow2_set_key, |
7c80ab3f | 3307 | |
c68b89ac KW |
3308 | .bdrv_co_readv = qcow2_co_readv, |
3309 | .bdrv_co_writev = qcow2_co_writev, | |
eb489bb1 | 3310 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
419b19d9 | 3311 | |
621f0589 | 3312 | .bdrv_co_write_zeroes = qcow2_co_write_zeroes, |
6db39ae2 | 3313 | .bdrv_co_discard = qcow2_co_discard, |
419b19d9 | 3314 | .bdrv_truncate = qcow2_truncate, |
7c80ab3f | 3315 | .bdrv_write_compressed = qcow2_write_compressed, |
491d27e2 | 3316 | .bdrv_make_empty = qcow2_make_empty, |
20d97356 BS |
3317 | |
3318 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
3319 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
3320 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
3321 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1bd0e2d1 CL |
3322 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
3323 | .bdrv_get_info = qcow2_get_info, | |
37764dfb | 3324 | .bdrv_get_specific_info = qcow2_get_specific_info, |
20d97356 | 3325 | |
7c80ab3f JS |
3326 | .bdrv_save_vmstate = qcow2_save_vmstate, |
3327 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
20d97356 | 3328 | |
8ee79e70 | 3329 | .supports_backing = true, |
20d97356 BS |
3330 | .bdrv_change_backing_file = qcow2_change_backing_file, |
3331 | ||
d34682cd | 3332 | .bdrv_refresh_limits = qcow2_refresh_limits, |
06d9260f AL |
3333 | .bdrv_invalidate_cache = qcow2_invalidate_cache, |
3334 | ||
1bd0e2d1 CL |
3335 | .create_opts = &qcow2_create_opts, |
3336 | .bdrv_check = qcow2_check, | |
c282e1fd | 3337 | .bdrv_amend_options = qcow2_amend_options, |
279621c0 AG |
3338 | |
3339 | .bdrv_detach_aio_context = qcow2_detach_aio_context, | |
3340 | .bdrv_attach_aio_context = qcow2_attach_aio_context, | |
20d97356 BS |
3341 | }; |
3342 | ||
5efa9d5a AL |
3343 | static void bdrv_qcow2_init(void) |
3344 | { | |
3345 | bdrv_register(&bdrv_qcow2); | |
3346 | } | |
3347 | ||
3348 | block_init(bdrv_qcow2_init); |