]>
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 | */ | |
faf07963 | 24 | #include "qemu-common.h" |
737e150e | 25 | #include "block/block_int.h" |
1de7afc9 | 26 | #include "qemu/module.h" |
585f8587 | 27 | #include <zlib.h> |
753d9b82 | 28 | #include "qemu/aes.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 | { |
75bab85c | 89 | BDRVQcowState *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 | ||
3ef6c40a HR |
108 | ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext)); |
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 | |
64ca6aee | 120 | if (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 | } | |
3ef6c40a HR |
136 | ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len); |
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 AL |
142 | bs->backing_format[ext.len] = '\0'; |
143 | #ifdef DEBUG_EXT | |
144 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
145 | #endif | |
f965509c AL |
146 | break; |
147 | ||
cfcc4c62 KW |
148 | case QCOW2_EXT_MAGIC_FEATURE_TABLE: |
149 | if (p_feature_table != NULL) { | |
150 | void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); | |
151 | ret = bdrv_pread(bs->file, offset , feature_table, ext.len); | |
152 | if (ret < 0) { | |
3ef6c40a HR |
153 | error_setg_errno(errp, -ret, "ERROR: ext_feature_table: " |
154 | "Could not read table"); | |
cfcc4c62 KW |
155 | return ret; |
156 | } | |
157 | ||
158 | *p_feature_table = feature_table; | |
159 | } | |
160 | break; | |
161 | ||
9b80ddf3 | 162 | default: |
75bab85c KW |
163 | /* unknown magic - save it in case we need to rewrite the header */ |
164 | { | |
165 | Qcow2UnknownHeaderExtension *uext; | |
166 | ||
167 | uext = g_malloc0(sizeof(*uext) + ext.len); | |
168 | uext->magic = ext.magic; | |
169 | uext->len = ext.len; | |
170 | QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); | |
171 | ||
172 | ret = bdrv_pread(bs->file, offset , uext->data, uext->len); | |
173 | if (ret < 0) { | |
3ef6c40a HR |
174 | error_setg_errno(errp, -ret, "ERROR: unknown extension: " |
175 | "Could not read data"); | |
75bab85c KW |
176 | return ret; |
177 | } | |
75bab85c | 178 | } |
9b80ddf3 AL |
179 | break; |
180 | } | |
fd29b4bb KW |
181 | |
182 | offset += ((ext.len + 7) & ~7); | |
9b80ddf3 AL |
183 | } |
184 | ||
185 | return 0; | |
186 | } | |
187 | ||
75bab85c KW |
188 | static void cleanup_unknown_header_ext(BlockDriverState *bs) |
189 | { | |
190 | BDRVQcowState *s = bs->opaque; | |
191 | Qcow2UnknownHeaderExtension *uext, *next; | |
192 | ||
193 | QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { | |
194 | QLIST_REMOVE(uext, next); | |
195 | g_free(uext); | |
196 | } | |
197 | } | |
9b80ddf3 | 198 | |
3ef6c40a HR |
199 | static void GCC_FMT_ATTR(3, 4) report_unsupported(BlockDriverState *bs, |
200 | Error **errp, const char *fmt, ...) | |
6744cbab KW |
201 | { |
202 | char msg[64]; | |
203 | va_list ap; | |
204 | ||
205 | va_start(ap, fmt); | |
206 | vsnprintf(msg, sizeof(msg), fmt, ap); | |
207 | va_end(ap); | |
208 | ||
bfb197e0 MA |
209 | error_set(errp, QERR_UNKNOWN_BLOCK_FORMAT_FEATURE, |
210 | bdrv_get_device_name(bs), "qcow2", msg); | |
6744cbab KW |
211 | } |
212 | ||
cfcc4c62 | 213 | static void report_unsupported_feature(BlockDriverState *bs, |
3ef6c40a | 214 | Error **errp, Qcow2Feature *table, uint64_t mask) |
cfcc4c62 | 215 | { |
12ac6d3d KW |
216 | char *features = g_strdup(""); |
217 | char *old; | |
218 | ||
cfcc4c62 KW |
219 | while (table && table->name[0] != '\0') { |
220 | if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) { | |
12ac6d3d KW |
221 | if (mask & (1ULL << table->bit)) { |
222 | old = features; | |
223 | features = g_strdup_printf("%s%s%.46s", old, *old ? ", " : "", | |
224 | table->name); | |
225 | g_free(old); | |
226 | mask &= ~(1ULL << table->bit); | |
cfcc4c62 KW |
227 | } |
228 | } | |
229 | table++; | |
230 | } | |
231 | ||
232 | if (mask) { | |
12ac6d3d KW |
233 | old = features; |
234 | features = g_strdup_printf("%s%sUnknown incompatible feature: %" PRIx64, | |
235 | old, *old ? ", " : "", mask); | |
236 | g_free(old); | |
cfcc4c62 | 237 | } |
12ac6d3d KW |
238 | |
239 | report_unsupported(bs, errp, "%s", features); | |
240 | g_free(features); | |
cfcc4c62 KW |
241 | } |
242 | ||
bfe8043e SH |
243 | /* |
244 | * Sets the dirty bit and flushes afterwards if necessary. | |
245 | * | |
246 | * The incompatible_features bit is only set if the image file header was | |
247 | * updated successfully. Therefore it is not required to check the return | |
248 | * value of this function. | |
249 | */ | |
280d3735 | 250 | int qcow2_mark_dirty(BlockDriverState *bs) |
bfe8043e SH |
251 | { |
252 | BDRVQcowState *s = bs->opaque; | |
253 | uint64_t val; | |
254 | int ret; | |
255 | ||
256 | assert(s->qcow_version >= 3); | |
257 | ||
258 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
259 | return 0; /* already dirty */ | |
260 | } | |
261 | ||
262 | val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); | |
263 | ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features), | |
264 | &val, sizeof(val)); | |
265 | if (ret < 0) { | |
266 | return ret; | |
267 | } | |
268 | ret = bdrv_flush(bs->file); | |
269 | if (ret < 0) { | |
270 | return ret; | |
271 | } | |
272 | ||
273 | /* Only treat image as dirty if the header was updated successfully */ | |
274 | s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; | |
275 | return 0; | |
276 | } | |
277 | ||
c61d0004 SH |
278 | /* |
279 | * Clears the dirty bit and flushes before if necessary. Only call this | |
280 | * function when there are no pending requests, it does not guard against | |
281 | * concurrent requests dirtying the image. | |
282 | */ | |
283 | static int qcow2_mark_clean(BlockDriverState *bs) | |
284 | { | |
285 | BDRVQcowState *s = bs->opaque; | |
286 | ||
287 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
4c2e5f8f KW |
288 | int ret; |
289 | ||
290 | s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; | |
291 | ||
292 | ret = bdrv_flush(bs); | |
c61d0004 SH |
293 | if (ret < 0) { |
294 | return ret; | |
295 | } | |
296 | ||
c61d0004 SH |
297 | return qcow2_update_header(bs); |
298 | } | |
299 | return 0; | |
300 | } | |
301 | ||
69c98726 HR |
302 | /* |
303 | * Marks the image as corrupt. | |
304 | */ | |
305 | int qcow2_mark_corrupt(BlockDriverState *bs) | |
306 | { | |
307 | BDRVQcowState *s = bs->opaque; | |
308 | ||
309 | s->incompatible_features |= QCOW2_INCOMPAT_CORRUPT; | |
310 | return qcow2_update_header(bs); | |
311 | } | |
312 | ||
313 | /* | |
314 | * Marks the image as consistent, i.e., unsets the corrupt bit, and flushes | |
315 | * before if necessary. | |
316 | */ | |
317 | int qcow2_mark_consistent(BlockDriverState *bs) | |
318 | { | |
319 | BDRVQcowState *s = bs->opaque; | |
320 | ||
321 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { | |
322 | int ret = bdrv_flush(bs); | |
323 | if (ret < 0) { | |
324 | return ret; | |
325 | } | |
326 | ||
327 | s->incompatible_features &= ~QCOW2_INCOMPAT_CORRUPT; | |
328 | return qcow2_update_header(bs); | |
329 | } | |
330 | return 0; | |
331 | } | |
332 | ||
acbe5982 SH |
333 | static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result, |
334 | BdrvCheckMode fix) | |
335 | { | |
336 | int ret = qcow2_check_refcounts(bs, result, fix); | |
337 | if (ret < 0) { | |
338 | return ret; | |
339 | } | |
340 | ||
341 | if (fix && result->check_errors == 0 && result->corruptions == 0) { | |
24530f3e HR |
342 | ret = qcow2_mark_clean(bs); |
343 | if (ret < 0) { | |
344 | return ret; | |
345 | } | |
346 | return qcow2_mark_consistent(bs); | |
acbe5982 SH |
347 | } |
348 | return ret; | |
349 | } | |
350 | ||
8c7de283 KW |
351 | static int validate_table_offset(BlockDriverState *bs, uint64_t offset, |
352 | uint64_t entries, size_t entry_len) | |
353 | { | |
354 | BDRVQcowState *s = bs->opaque; | |
355 | uint64_t size; | |
356 | ||
357 | /* Use signed INT64_MAX as the maximum even for uint64_t header fields, | |
358 | * because values will be passed to qemu functions taking int64_t. */ | |
359 | if (entries > INT64_MAX / entry_len) { | |
360 | return -EINVAL; | |
361 | } | |
362 | ||
363 | size = entries * entry_len; | |
364 | ||
365 | if (INT64_MAX - size < offset) { | |
366 | return -EINVAL; | |
367 | } | |
368 | ||
369 | /* Tables must be cluster aligned */ | |
370 | if (offset & (s->cluster_size - 1)) { | |
371 | return -EINVAL; | |
372 | } | |
373 | ||
374 | return 0; | |
375 | } | |
376 | ||
74c4510a KW |
377 | static QemuOptsList qcow2_runtime_opts = { |
378 | .name = "qcow2", | |
379 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_runtime_opts.head), | |
380 | .desc = { | |
381 | { | |
64aa99d3 | 382 | .name = QCOW2_OPT_LAZY_REFCOUNTS, |
74c4510a KW |
383 | .type = QEMU_OPT_BOOL, |
384 | .help = "Postpone refcount updates", | |
385 | }, | |
67af674e KW |
386 | { |
387 | .name = QCOW2_OPT_DISCARD_REQUEST, | |
388 | .type = QEMU_OPT_BOOL, | |
389 | .help = "Pass guest discard requests to the layer below", | |
390 | }, | |
391 | { | |
392 | .name = QCOW2_OPT_DISCARD_SNAPSHOT, | |
393 | .type = QEMU_OPT_BOOL, | |
394 | .help = "Generate discard requests when snapshot related space " | |
395 | "is freed", | |
396 | }, | |
397 | { | |
398 | .name = QCOW2_OPT_DISCARD_OTHER, | |
399 | .type = QEMU_OPT_BOOL, | |
400 | .help = "Generate discard requests when other clusters are freed", | |
401 | }, | |
05de7e86 HR |
402 | { |
403 | .name = QCOW2_OPT_OVERLAP, | |
404 | .type = QEMU_OPT_STRING, | |
405 | .help = "Selects which overlap checks to perform from a range of " | |
406 | "templates (none, constant, cached, all)", | |
407 | }, | |
ee42b5ce HR |
408 | { |
409 | .name = QCOW2_OPT_OVERLAP_TEMPLATE, | |
410 | .type = QEMU_OPT_STRING, | |
411 | .help = "Selects which overlap checks to perform from a range of " | |
412 | "templates (none, constant, cached, all)", | |
413 | }, | |
05de7e86 HR |
414 | { |
415 | .name = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
416 | .type = QEMU_OPT_BOOL, | |
417 | .help = "Check for unintended writes into the main qcow2 header", | |
418 | }, | |
419 | { | |
420 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
421 | .type = QEMU_OPT_BOOL, | |
422 | .help = "Check for unintended writes into the active L1 table", | |
423 | }, | |
424 | { | |
425 | .name = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
426 | .type = QEMU_OPT_BOOL, | |
427 | .help = "Check for unintended writes into an active L2 table", | |
428 | }, | |
429 | { | |
430 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
431 | .type = QEMU_OPT_BOOL, | |
432 | .help = "Check for unintended writes into the refcount table", | |
433 | }, | |
434 | { | |
435 | .name = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
436 | .type = QEMU_OPT_BOOL, | |
437 | .help = "Check for unintended writes into a refcount block", | |
438 | }, | |
439 | { | |
440 | .name = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
441 | .type = QEMU_OPT_BOOL, | |
442 | .help = "Check for unintended writes into the snapshot table", | |
443 | }, | |
444 | { | |
445 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
446 | .type = QEMU_OPT_BOOL, | |
447 | .help = "Check for unintended writes into an inactive L1 table", | |
448 | }, | |
449 | { | |
450 | .name = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
451 | .type = QEMU_OPT_BOOL, | |
452 | .help = "Check for unintended writes into an inactive L2 table", | |
453 | }, | |
6c1c8d5d HR |
454 | { |
455 | .name = QCOW2_OPT_CACHE_SIZE, | |
456 | .type = QEMU_OPT_SIZE, | |
457 | .help = "Maximum combined metadata (L2 tables and refcount blocks) " | |
458 | "cache size", | |
459 | }, | |
460 | { | |
461 | .name = QCOW2_OPT_L2_CACHE_SIZE, | |
462 | .type = QEMU_OPT_SIZE, | |
463 | .help = "Maximum L2 table cache size", | |
464 | }, | |
465 | { | |
466 | .name = QCOW2_OPT_REFCOUNT_CACHE_SIZE, | |
467 | .type = QEMU_OPT_SIZE, | |
468 | .help = "Maximum refcount block cache size", | |
469 | }, | |
74c4510a KW |
470 | { /* end of list */ } |
471 | }, | |
472 | }; | |
473 | ||
4092e99d HR |
474 | static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = { |
475 | [QCOW2_OL_MAIN_HEADER_BITNR] = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
476 | [QCOW2_OL_ACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
477 | [QCOW2_OL_ACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
478 | [QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
479 | [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
480 | [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
481 | [QCOW2_OL_INACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
482 | [QCOW2_OL_INACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
483 | }; | |
484 | ||
6c1c8d5d HR |
485 | static void read_cache_sizes(QemuOpts *opts, uint64_t *l2_cache_size, |
486 | uint64_t *refcount_cache_size, Error **errp) | |
487 | { | |
488 | uint64_t combined_cache_size; | |
489 | bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set; | |
490 | ||
491 | combined_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_CACHE_SIZE); | |
492 | l2_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_SIZE); | |
493 | refcount_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
494 | ||
495 | combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0); | |
496 | *l2_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE, 0); | |
497 | *refcount_cache_size = qemu_opt_get_size(opts, | |
498 | QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0); | |
499 | ||
500 | if (combined_cache_size_set) { | |
501 | if (l2_cache_size_set && refcount_cache_size_set) { | |
502 | error_setg(errp, QCOW2_OPT_CACHE_SIZE ", " QCOW2_OPT_L2_CACHE_SIZE | |
503 | " and " QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not be set " | |
504 | "the same time"); | |
505 | return; | |
506 | } else if (*l2_cache_size > combined_cache_size) { | |
507 | error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed " | |
508 | QCOW2_OPT_CACHE_SIZE); | |
509 | return; | |
510 | } else if (*refcount_cache_size > combined_cache_size) { | |
511 | error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed " | |
512 | QCOW2_OPT_CACHE_SIZE); | |
513 | return; | |
514 | } | |
515 | ||
516 | if (l2_cache_size_set) { | |
517 | *refcount_cache_size = combined_cache_size - *l2_cache_size; | |
518 | } else if (refcount_cache_size_set) { | |
519 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
520 | } else { | |
521 | *refcount_cache_size = combined_cache_size | |
522 | / (DEFAULT_L2_REFCOUNT_SIZE_RATIO + 1); | |
523 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
524 | } | |
525 | } else { | |
526 | if (!l2_cache_size_set && !refcount_cache_size_set) { | |
527 | *l2_cache_size = DEFAULT_L2_CACHE_BYTE_SIZE; | |
528 | *refcount_cache_size = *l2_cache_size | |
529 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
530 | } else if (!l2_cache_size_set) { | |
531 | *l2_cache_size = *refcount_cache_size | |
532 | * DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
533 | } else if (!refcount_cache_size_set) { | |
534 | *refcount_cache_size = *l2_cache_size | |
535 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
536 | } | |
537 | } | |
538 | } | |
539 | ||
015a1036 HR |
540 | static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, |
541 | Error **errp) | |
585f8587 FB |
542 | { |
543 | BDRVQcowState *s = bs->opaque; | |
6d33e8e7 KW |
544 | unsigned int len, i; |
545 | int ret = 0; | |
585f8587 | 546 | QCowHeader header; |
7b17ce60 | 547 | QemuOpts *opts = NULL; |
74c4510a | 548 | Error *local_err = NULL; |
9b80ddf3 | 549 | uint64_t ext_end; |
2cf7cfa1 | 550 | uint64_t l1_vm_state_index; |
ee42b5ce | 551 | const char *opt_overlap_check, *opt_overlap_check_template; |
1fa5cc83 | 552 | int overlap_check_template = 0; |
440ba08a | 553 | uint64_t l2_cache_size, refcount_cache_size; |
585f8587 | 554 | |
6d85a57e JS |
555 | ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); |
556 | if (ret < 0) { | |
3ef6c40a | 557 | error_setg_errno(errp, -ret, "Could not read qcow2 header"); |
585f8587 | 558 | goto fail; |
6d85a57e | 559 | } |
585f8587 FB |
560 | be32_to_cpus(&header.magic); |
561 | be32_to_cpus(&header.version); | |
562 | be64_to_cpus(&header.backing_file_offset); | |
563 | be32_to_cpus(&header.backing_file_size); | |
564 | be64_to_cpus(&header.size); | |
565 | be32_to_cpus(&header.cluster_bits); | |
566 | be32_to_cpus(&header.crypt_method); | |
567 | be64_to_cpus(&header.l1_table_offset); | |
568 | be32_to_cpus(&header.l1_size); | |
569 | be64_to_cpus(&header.refcount_table_offset); | |
570 | be32_to_cpus(&header.refcount_table_clusters); | |
571 | be64_to_cpus(&header.snapshots_offset); | |
572 | be32_to_cpus(&header.nb_snapshots); | |
3b46e624 | 573 | |
e8cdcec1 | 574 | if (header.magic != QCOW_MAGIC) { |
3ef6c40a | 575 | error_setg(errp, "Image is not in qcow2 format"); |
76abe407 | 576 | ret = -EINVAL; |
585f8587 | 577 | goto fail; |
6d85a57e | 578 | } |
6744cbab | 579 | if (header.version < 2 || header.version > 3) { |
521b2b5d | 580 | report_unsupported(bs, errp, "QCOW version %" PRIu32, header.version); |
6744cbab KW |
581 | ret = -ENOTSUP; |
582 | goto fail; | |
583 | } | |
584 | ||
585 | s->qcow_version = header.version; | |
586 | ||
24342f2c KW |
587 | /* Initialise cluster size */ |
588 | if (header.cluster_bits < MIN_CLUSTER_BITS || | |
589 | header.cluster_bits > MAX_CLUSTER_BITS) { | |
521b2b5d HR |
590 | error_setg(errp, "Unsupported cluster size: 2^%" PRIu32, |
591 | header.cluster_bits); | |
24342f2c KW |
592 | ret = -EINVAL; |
593 | goto fail; | |
594 | } | |
595 | ||
596 | s->cluster_bits = header.cluster_bits; | |
597 | s->cluster_size = 1 << s->cluster_bits; | |
598 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
599 | ||
6744cbab KW |
600 | /* Initialise version 3 header fields */ |
601 | if (header.version == 2) { | |
602 | header.incompatible_features = 0; | |
603 | header.compatible_features = 0; | |
604 | header.autoclear_features = 0; | |
605 | header.refcount_order = 4; | |
606 | header.header_length = 72; | |
607 | } else { | |
608 | be64_to_cpus(&header.incompatible_features); | |
609 | be64_to_cpus(&header.compatible_features); | |
610 | be64_to_cpus(&header.autoclear_features); | |
611 | be32_to_cpus(&header.refcount_order); | |
612 | be32_to_cpus(&header.header_length); | |
24342f2c KW |
613 | |
614 | if (header.header_length < 104) { | |
615 | error_setg(errp, "qcow2 header too short"); | |
616 | ret = -EINVAL; | |
617 | goto fail; | |
618 | } | |
619 | } | |
620 | ||
621 | if (header.header_length > s->cluster_size) { | |
622 | error_setg(errp, "qcow2 header exceeds cluster size"); | |
623 | ret = -EINVAL; | |
624 | goto fail; | |
6744cbab KW |
625 | } |
626 | ||
627 | if (header.header_length > sizeof(header)) { | |
628 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
629 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
630 | ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields, | |
631 | s->unknown_header_fields_size); | |
632 | if (ret < 0) { | |
3ef6c40a HR |
633 | error_setg_errno(errp, -ret, "Could not read unknown qcow2 header " |
634 | "fields"); | |
6744cbab KW |
635 | goto fail; |
636 | } | |
637 | } | |
638 | ||
a1b3955c KW |
639 | if (header.backing_file_offset > s->cluster_size) { |
640 | error_setg(errp, "Invalid backing file offset"); | |
641 | ret = -EINVAL; | |
642 | goto fail; | |
643 | } | |
644 | ||
cfcc4c62 KW |
645 | if (header.backing_file_offset) { |
646 | ext_end = header.backing_file_offset; | |
647 | } else { | |
648 | ext_end = 1 << header.cluster_bits; | |
649 | } | |
650 | ||
6744cbab KW |
651 | /* Handle feature bits */ |
652 | s->incompatible_features = header.incompatible_features; | |
653 | s->compatible_features = header.compatible_features; | |
654 | s->autoclear_features = header.autoclear_features; | |
655 | ||
c61d0004 | 656 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { |
cfcc4c62 KW |
657 | void *feature_table = NULL; |
658 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
3ef6c40a HR |
659 | &feature_table, NULL); |
660 | report_unsupported_feature(bs, errp, feature_table, | |
c61d0004 SH |
661 | s->incompatible_features & |
662 | ~QCOW2_INCOMPAT_MASK); | |
6744cbab | 663 | ret = -ENOTSUP; |
c5a33ee9 | 664 | g_free(feature_table); |
6744cbab KW |
665 | goto fail; |
666 | } | |
667 | ||
69c98726 HR |
668 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { |
669 | /* Corrupt images may not be written to unless they are being repaired | |
670 | */ | |
671 | if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) { | |
3ef6c40a HR |
672 | error_setg(errp, "qcow2: Image is corrupt; cannot be opened " |
673 | "read/write"); | |
69c98726 HR |
674 | ret = -EACCES; |
675 | goto fail; | |
676 | } | |
677 | } | |
678 | ||
6744cbab KW |
679 | /* Check support for various header values */ |
680 | if (header.refcount_order != 4) { | |
3ef6c40a | 681 | report_unsupported(bs, errp, "%d bit reference counts", |
6744cbab | 682 | 1 << header.refcount_order); |
e8cdcec1 KW |
683 | ret = -ENOTSUP; |
684 | goto fail; | |
685 | } | |
b6481f37 | 686 | s->refcount_order = header.refcount_order; |
6744cbab | 687 | |
6d85a57e | 688 | if (header.crypt_method > QCOW_CRYPT_AES) { |
521b2b5d | 689 | error_setg(errp, "Unsupported encryption method: %" PRIu32, |
3ef6c40a | 690 | header.crypt_method); |
6d85a57e | 691 | ret = -EINVAL; |
585f8587 | 692 | goto fail; |
6d85a57e | 693 | } |
585f8587 | 694 | s->crypt_method_header = header.crypt_method; |
6d85a57e | 695 | if (s->crypt_method_header) { |
585f8587 | 696 | bs->encrypted = 1; |
6d85a57e | 697 | } |
24342f2c | 698 | |
585f8587 FB |
699 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
700 | s->l2_size = 1 << s->l2_bits; | |
1d13d654 HR |
701 | /* 2^(s->refcount_order - 3) is the refcount width in bytes */ |
702 | s->refcount_block_bits = s->cluster_bits - (s->refcount_order - 3); | |
703 | s->refcount_block_size = 1 << s->refcount_block_bits; | |
585f8587 FB |
704 | bs->total_sectors = header.size / 512; |
705 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
706 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
707 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
5dab2fad | 708 | |
585f8587 | 709 | s->refcount_table_offset = header.refcount_table_offset; |
5fafdf24 | 710 | s->refcount_table_size = |
585f8587 FB |
711 | header.refcount_table_clusters << (s->cluster_bits - 3); |
712 | ||
2b5d5953 | 713 | if (header.refcount_table_clusters > qcow2_max_refcount_clusters(s)) { |
5dab2fad KW |
714 | error_setg(errp, "Reference count table too large"); |
715 | ret = -EINVAL; | |
716 | goto fail; | |
717 | } | |
718 | ||
8c7de283 KW |
719 | ret = validate_table_offset(bs, s->refcount_table_offset, |
720 | s->refcount_table_size, sizeof(uint64_t)); | |
721 | if (ret < 0) { | |
722 | error_setg(errp, "Invalid reference count table offset"); | |
723 | goto fail; | |
724 | } | |
725 | ||
ce48f2f4 KW |
726 | /* Snapshot table offset/length */ |
727 | if (header.nb_snapshots > QCOW_MAX_SNAPSHOTS) { | |
728 | error_setg(errp, "Too many snapshots"); | |
729 | ret = -EINVAL; | |
730 | goto fail; | |
731 | } | |
732 | ||
733 | ret = validate_table_offset(bs, header.snapshots_offset, | |
734 | header.nb_snapshots, | |
735 | sizeof(QCowSnapshotHeader)); | |
736 | if (ret < 0) { | |
737 | error_setg(errp, "Invalid snapshot table offset"); | |
738 | goto fail; | |
739 | } | |
740 | ||
585f8587 | 741 | /* read the level 1 table */ |
6a83f8b5 | 742 | if (header.l1_size > QCOW_MAX_L1_SIZE) { |
2d51c32c KW |
743 | error_setg(errp, "Active L1 table too large"); |
744 | ret = -EFBIG; | |
745 | goto fail; | |
746 | } | |
585f8587 | 747 | s->l1_size = header.l1_size; |
2cf7cfa1 KW |
748 | |
749 | l1_vm_state_index = size_to_l1(s, header.size); | |
750 | if (l1_vm_state_index > INT_MAX) { | |
3ef6c40a | 751 | error_setg(errp, "Image is too big"); |
2cf7cfa1 KW |
752 | ret = -EFBIG; |
753 | goto fail; | |
754 | } | |
755 | s->l1_vm_state_index = l1_vm_state_index; | |
756 | ||
585f8587 FB |
757 | /* the L1 table must contain at least enough entries to put |
758 | header.size bytes */ | |
6d85a57e | 759 | if (s->l1_size < s->l1_vm_state_index) { |
3ef6c40a | 760 | error_setg(errp, "L1 table is too small"); |
6d85a57e | 761 | ret = -EINVAL; |
585f8587 | 762 | goto fail; |
6d85a57e | 763 | } |
2d51c32c KW |
764 | |
765 | ret = validate_table_offset(bs, header.l1_table_offset, | |
766 | header.l1_size, sizeof(uint64_t)); | |
767 | if (ret < 0) { | |
768 | error_setg(errp, "Invalid L1 table offset"); | |
769 | goto fail; | |
770 | } | |
585f8587 | 771 | s->l1_table_offset = header.l1_table_offset; |
2d51c32c KW |
772 | |
773 | ||
d191d12d | 774 | if (s->l1_size > 0) { |
de82815d | 775 | s->l1_table = qemu_try_blockalign(bs->file, |
d191d12d | 776 | align_offset(s->l1_size * sizeof(uint64_t), 512)); |
de82815d KW |
777 | if (s->l1_table == NULL) { |
778 | error_setg(errp, "Could not allocate L1 table"); | |
779 | ret = -ENOMEM; | |
780 | goto fail; | |
781 | } | |
6d85a57e JS |
782 | ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, |
783 | s->l1_size * sizeof(uint64_t)); | |
784 | if (ret < 0) { | |
3ef6c40a | 785 | error_setg_errno(errp, -ret, "Could not read L1 table"); |
d191d12d | 786 | goto fail; |
6d85a57e | 787 | } |
d191d12d SW |
788 | for(i = 0;i < s->l1_size; i++) { |
789 | be64_to_cpus(&s->l1_table[i]); | |
790 | } | |
585f8587 | 791 | } |
29c1a730 | 792 | |
6c1c8d5d HR |
793 | /* get L2 table/refcount block cache size from command line options */ |
794 | opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort); | |
795 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
796 | if (local_err) { | |
797 | error_propagate(errp, local_err); | |
798 | ret = -EINVAL; | |
799 | goto fail; | |
800 | } | |
801 | ||
802 | read_cache_sizes(opts, &l2_cache_size, &refcount_cache_size, &local_err); | |
803 | if (local_err) { | |
804 | error_propagate(errp, local_err); | |
805 | ret = -EINVAL; | |
806 | goto fail; | |
807 | } | |
808 | ||
809 | l2_cache_size /= s->cluster_size; | |
440ba08a HR |
810 | if (l2_cache_size < MIN_L2_CACHE_SIZE) { |
811 | l2_cache_size = MIN_L2_CACHE_SIZE; | |
812 | } | |
6c1c8d5d HR |
813 | if (l2_cache_size > INT_MAX) { |
814 | error_setg(errp, "L2 cache size too big"); | |
815 | ret = -EINVAL; | |
816 | goto fail; | |
817 | } | |
440ba08a | 818 | |
6c1c8d5d | 819 | refcount_cache_size /= s->cluster_size; |
440ba08a HR |
820 | if (refcount_cache_size < MIN_REFCOUNT_CACHE_SIZE) { |
821 | refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE; | |
822 | } | |
6c1c8d5d HR |
823 | if (refcount_cache_size > INT_MAX) { |
824 | error_setg(errp, "Refcount cache size too big"); | |
825 | ret = -EINVAL; | |
826 | goto fail; | |
827 | } | |
440ba08a | 828 | |
6c1c8d5d | 829 | /* alloc L2 table/refcount block cache */ |
440ba08a HR |
830 | s->l2_table_cache = qcow2_cache_create(bs, l2_cache_size); |
831 | s->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size); | |
de82815d KW |
832 | if (s->l2_table_cache == NULL || s->refcount_block_cache == NULL) { |
833 | error_setg(errp, "Could not allocate metadata caches"); | |
834 | ret = -ENOMEM; | |
835 | goto fail; | |
836 | } | |
29c1a730 | 837 | |
7267c094 | 838 | s->cluster_cache = g_malloc(s->cluster_size); |
585f8587 | 839 | /* one more sector for decompressed data alignment */ |
de82815d KW |
840 | s->cluster_data = qemu_try_blockalign(bs->file, QCOW_MAX_CRYPT_CLUSTERS |
841 | * s->cluster_size + 512); | |
842 | if (s->cluster_data == NULL) { | |
843 | error_setg(errp, "Could not allocate temporary cluster buffer"); | |
844 | ret = -ENOMEM; | |
845 | goto fail; | |
846 | } | |
847 | ||
585f8587 | 848 | s->cluster_cache_offset = -1; |
06d9260f | 849 | s->flags = flags; |
3b46e624 | 850 | |
6d85a57e JS |
851 | ret = qcow2_refcount_init(bs); |
852 | if (ret != 0) { | |
3ef6c40a | 853 | error_setg_errno(errp, -ret, "Could not initialize refcount handling"); |
585f8587 | 854 | goto fail; |
6d85a57e | 855 | } |
585f8587 | 856 | |
72cf2d4f | 857 | QLIST_INIT(&s->cluster_allocs); |
0b919fae | 858 | QTAILQ_INIT(&s->discards); |
f214978a | 859 | |
9b80ddf3 | 860 | /* read qcow2 extensions */ |
3ef6c40a HR |
861 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL, |
862 | &local_err)) { | |
863 | error_propagate(errp, local_err); | |
6d85a57e | 864 | ret = -EINVAL; |
9b80ddf3 | 865 | goto fail; |
6d85a57e | 866 | } |
9b80ddf3 | 867 | |
585f8587 FB |
868 | /* read the backing file name */ |
869 | if (header.backing_file_offset != 0) { | |
870 | len = header.backing_file_size; | |
6d33e8e7 KW |
871 | if (len > MIN(1023, s->cluster_size - header.backing_file_offset)) { |
872 | error_setg(errp, "Backing file name too long"); | |
873 | ret = -EINVAL; | |
874 | goto fail; | |
6d85a57e JS |
875 | } |
876 | ret = bdrv_pread(bs->file, header.backing_file_offset, | |
877 | bs->backing_file, len); | |
878 | if (ret < 0) { | |
3ef6c40a | 879 | error_setg_errno(errp, -ret, "Could not read backing file name"); |
585f8587 | 880 | goto fail; |
6d85a57e | 881 | } |
585f8587 FB |
882 | bs->backing_file[len] = '\0'; |
883 | } | |
42deb29f | 884 | |
11b128f4 KW |
885 | /* Internal snapshots */ |
886 | s->snapshots_offset = header.snapshots_offset; | |
887 | s->nb_snapshots = header.nb_snapshots; | |
888 | ||
42deb29f KW |
889 | ret = qcow2_read_snapshots(bs); |
890 | if (ret < 0) { | |
3ef6c40a | 891 | error_setg_errno(errp, -ret, "Could not read snapshots"); |
585f8587 | 892 | goto fail; |
6d85a57e | 893 | } |
585f8587 | 894 | |
af7b708d | 895 | /* Clear unknown autoclear feature bits */ |
27eb6c09 | 896 | if (!bs->read_only && !(flags & BDRV_O_INCOMING) && s->autoclear_features) { |
af7b708d SH |
897 | s->autoclear_features = 0; |
898 | ret = qcow2_update_header(bs); | |
899 | if (ret < 0) { | |
3ef6c40a | 900 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); |
af7b708d SH |
901 | goto fail; |
902 | } | |
903 | } | |
904 | ||
68d100e9 KW |
905 | /* Initialise locks */ |
906 | qemu_co_mutex_init(&s->lock); | |
907 | ||
c61d0004 | 908 | /* Repair image if dirty */ |
27eb6c09 | 909 | if (!(flags & (BDRV_O_CHECK | BDRV_O_INCOMING)) && !bs->read_only && |
058f8f16 | 910 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { |
c61d0004 SH |
911 | BdrvCheckResult result = {0}; |
912 | ||
5b84106b | 913 | ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS); |
c61d0004 | 914 | if (ret < 0) { |
3ef6c40a | 915 | error_setg_errno(errp, -ret, "Could not repair dirty image"); |
c61d0004 SH |
916 | goto fail; |
917 | } | |
918 | } | |
919 | ||
74c4510a | 920 | /* Enable lazy_refcounts according to image and command line options */ |
acdfb480 | 921 | s->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, |
74c4510a KW |
922 | (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); |
923 | ||
67af674e KW |
924 | s->discard_passthrough[QCOW2_DISCARD_NEVER] = false; |
925 | s->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true; | |
926 | s->discard_passthrough[QCOW2_DISCARD_REQUEST] = | |
927 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST, | |
928 | flags & BDRV_O_UNMAP); | |
929 | s->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] = | |
930 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true); | |
931 | s->discard_passthrough[QCOW2_DISCARD_OTHER] = | |
932 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false); | |
933 | ||
ee42b5ce HR |
934 | opt_overlap_check = qemu_opt_get(opts, QCOW2_OPT_OVERLAP); |
935 | opt_overlap_check_template = qemu_opt_get(opts, QCOW2_OPT_OVERLAP_TEMPLATE); | |
936 | if (opt_overlap_check_template && opt_overlap_check && | |
937 | strcmp(opt_overlap_check_template, opt_overlap_check)) | |
938 | { | |
939 | error_setg(errp, "Conflicting values for qcow2 options '" | |
940 | QCOW2_OPT_OVERLAP "' ('%s') and '" QCOW2_OPT_OVERLAP_TEMPLATE | |
941 | "' ('%s')", opt_overlap_check, opt_overlap_check_template); | |
942 | ret = -EINVAL; | |
943 | goto fail; | |
944 | } | |
945 | if (!opt_overlap_check) { | |
946 | opt_overlap_check = opt_overlap_check_template ?: "cached"; | |
947 | } | |
948 | ||
1fa5cc83 HR |
949 | if (!strcmp(opt_overlap_check, "none")) { |
950 | overlap_check_template = 0; | |
951 | } else if (!strcmp(opt_overlap_check, "constant")) { | |
952 | overlap_check_template = QCOW2_OL_CONSTANT; | |
953 | } else if (!strcmp(opt_overlap_check, "cached")) { | |
954 | overlap_check_template = QCOW2_OL_CACHED; | |
955 | } else if (!strcmp(opt_overlap_check, "all")) { | |
956 | overlap_check_template = QCOW2_OL_ALL; | |
957 | } else { | |
958 | error_setg(errp, "Unsupported value '%s' for qcow2 option " | |
959 | "'overlap-check'. Allowed are either of the following: " | |
960 | "none, constant, cached, all", opt_overlap_check); | |
1fa5cc83 HR |
961 | ret = -EINVAL; |
962 | goto fail; | |
963 | } | |
964 | ||
965 | s->overlap_check = 0; | |
966 | for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) { | |
967 | /* overlap-check defines a template bitmask, but every flag may be | |
968 | * overwritten through the associated boolean option */ | |
969 | s->overlap_check |= | |
970 | qemu_opt_get_bool(opts, overlap_bool_option_names[i], | |
971 | overlap_check_template & (1 << i)) << i; | |
972 | } | |
3e355390 | 973 | |
74c4510a | 974 | qemu_opts_del(opts); |
7b17ce60 | 975 | opts = NULL; |
74c4510a KW |
976 | |
977 | if (s->use_lazy_refcounts && s->qcow_version < 3) { | |
3ef6c40a HR |
978 | error_setg(errp, "Lazy refcounts require a qcow2 image with at least " |
979 | "qemu 1.1 compatibility level"); | |
74c4510a KW |
980 | ret = -EINVAL; |
981 | goto fail; | |
982 | } | |
983 | ||
585f8587 | 984 | #ifdef DEBUG_ALLOC |
6cbc3031 PH |
985 | { |
986 | BdrvCheckResult result = {0}; | |
b35278f7 | 987 | qcow2_check_refcounts(bs, &result, 0); |
6cbc3031 | 988 | } |
585f8587 | 989 | #endif |
6d85a57e | 990 | return ret; |
585f8587 FB |
991 | |
992 | fail: | |
7b17ce60 | 993 | qemu_opts_del(opts); |
6744cbab | 994 | g_free(s->unknown_header_fields); |
75bab85c | 995 | cleanup_unknown_header_ext(bs); |
ed6ccf0f KW |
996 | qcow2_free_snapshots(bs); |
997 | qcow2_refcount_close(bs); | |
de82815d | 998 | qemu_vfree(s->l1_table); |
cf93980e HR |
999 | /* else pre-write overlap checks in cache_destroy may crash */ |
1000 | s->l1_table = NULL; | |
29c1a730 KW |
1001 | if (s->l2_table_cache) { |
1002 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
1003 | } | |
c5a33ee9 PJ |
1004 | if (s->refcount_block_cache) { |
1005 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1006 | } | |
7267c094 | 1007 | g_free(s->cluster_cache); |
dea43a65 | 1008 | qemu_vfree(s->cluster_data); |
6d85a57e | 1009 | return ret; |
585f8587 FB |
1010 | } |
1011 | ||
3baca891 | 1012 | static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd KW |
1013 | { |
1014 | BDRVQcowState *s = bs->opaque; | |
1015 | ||
1016 | bs->bl.write_zeroes_alignment = s->cluster_sectors; | |
d34682cd KW |
1017 | } |
1018 | ||
7c80ab3f | 1019 | static int qcow2_set_key(BlockDriverState *bs, const char *key) |
585f8587 FB |
1020 | { |
1021 | BDRVQcowState *s = bs->opaque; | |
1022 | uint8_t keybuf[16]; | |
1023 | int len, i; | |
3b46e624 | 1024 | |
585f8587 FB |
1025 | memset(keybuf, 0, 16); |
1026 | len = strlen(key); | |
1027 | if (len > 16) | |
1028 | len = 16; | |
1029 | /* XXX: we could compress the chars to 7 bits to increase | |
1030 | entropy */ | |
1031 | for(i = 0;i < len;i++) { | |
1032 | keybuf[i] = key[i]; | |
1033 | } | |
1034 | s->crypt_method = s->crypt_method_header; | |
1035 | ||
1036 | if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0) | |
1037 | return -1; | |
1038 | if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0) | |
1039 | return -1; | |
1040 | #if 0 | |
1041 | /* test */ | |
1042 | { | |
1043 | uint8_t in[16]; | |
1044 | uint8_t out[16]; | |
1045 | uint8_t tmp[16]; | |
1046 | for(i=0;i<16;i++) | |
1047 | in[i] = i; | |
1048 | AES_encrypt(in, tmp, &s->aes_encrypt_key); | |
1049 | AES_decrypt(tmp, out, &s->aes_decrypt_key); | |
1050 | for(i = 0; i < 16; i++) | |
1051 | printf(" %02x", tmp[i]); | |
1052 | printf("\n"); | |
1053 | for(i = 0; i < 16; i++) | |
1054 | printf(" %02x", out[i]); | |
1055 | printf("\n"); | |
1056 | } | |
1057 | #endif | |
1058 | return 0; | |
1059 | } | |
1060 | ||
4c2e5f8f KW |
1061 | /* We have no actual commit/abort logic for qcow2, but we need to write out any |
1062 | * unwritten data if we reopen read-only. */ | |
21d82ac9 JC |
1063 | static int qcow2_reopen_prepare(BDRVReopenState *state, |
1064 | BlockReopenQueue *queue, Error **errp) | |
1065 | { | |
4c2e5f8f KW |
1066 | int ret; |
1067 | ||
1068 | if ((state->flags & BDRV_O_RDWR) == 0) { | |
1069 | ret = bdrv_flush(state->bs); | |
1070 | if (ret < 0) { | |
1071 | return ret; | |
1072 | } | |
1073 | ||
1074 | ret = qcow2_mark_clean(state->bs); | |
1075 | if (ret < 0) { | |
1076 | return ret; | |
1077 | } | |
1078 | } | |
1079 | ||
21d82ac9 JC |
1080 | return 0; |
1081 | } | |
1082 | ||
b6b8a333 | 1083 | static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs, |
f8a2e5e3 | 1084 | int64_t sector_num, int nb_sectors, int *pnum) |
585f8587 | 1085 | { |
f8a2e5e3 | 1086 | BDRVQcowState *s = bs->opaque; |
585f8587 | 1087 | uint64_t cluster_offset; |
4bc74be9 PB |
1088 | int index_in_cluster, ret; |
1089 | int64_t status = 0; | |
585f8587 | 1090 | |
095a9c58 | 1091 | *pnum = nb_sectors; |
f8a2e5e3 | 1092 | qemu_co_mutex_lock(&s->lock); |
1c46efaa | 1093 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, pnum, &cluster_offset); |
f8a2e5e3 | 1094 | qemu_co_mutex_unlock(&s->lock); |
1c46efaa | 1095 | if (ret < 0) { |
d663640c | 1096 | return ret; |
1c46efaa | 1097 | } |
095a9c58 | 1098 | |
4bc74be9 PB |
1099 | if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED && |
1100 | !s->crypt_method) { | |
1101 | index_in_cluster = sector_num & (s->cluster_sectors - 1); | |
1102 | cluster_offset |= (index_in_cluster << BDRV_SECTOR_BITS); | |
1103 | status |= BDRV_BLOCK_OFFSET_VALID | cluster_offset; | |
1104 | } | |
1105 | if (ret == QCOW2_CLUSTER_ZERO) { | |
1106 | status |= BDRV_BLOCK_ZERO; | |
1107 | } else if (ret != QCOW2_CLUSTER_UNALLOCATED) { | |
1108 | status |= BDRV_BLOCK_DATA; | |
1109 | } | |
1110 | return status; | |
585f8587 FB |
1111 | } |
1112 | ||
a9465922 | 1113 | /* handle reading after the end of the backing file */ |
bd28f835 KW |
1114 | int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, |
1115 | int64_t sector_num, int nb_sectors) | |
a9465922 FB |
1116 | { |
1117 | int n1; | |
1118 | if ((sector_num + nb_sectors) <= bs->total_sectors) | |
1119 | return nb_sectors; | |
1120 | if (sector_num >= bs->total_sectors) | |
1121 | n1 = 0; | |
1122 | else | |
1123 | n1 = bs->total_sectors - sector_num; | |
bd28f835 | 1124 | |
3d9b4925 | 1125 | qemu_iovec_memset(qiov, 512 * n1, 0, 512 * (nb_sectors - n1)); |
bd28f835 | 1126 | |
a9465922 FB |
1127 | return n1; |
1128 | } | |
1129 | ||
a968168c | 1130 | static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num, |
3fc48d09 | 1131 | int remaining_sectors, QEMUIOVector *qiov) |
585f8587 | 1132 | { |
585f8587 | 1133 | BDRVQcowState *s = bs->opaque; |
a9465922 | 1134 | int index_in_cluster, n1; |
68d100e9 | 1135 | int ret; |
faf575c1 | 1136 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 1137 | uint64_t cluster_offset = 0; |
3fc48d09 FZ |
1138 | uint64_t bytes_done = 0; |
1139 | QEMUIOVector hd_qiov; | |
1140 | uint8_t *cluster_data = NULL; | |
585f8587 | 1141 | |
3fc48d09 FZ |
1142 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1143 | ||
1144 | qemu_co_mutex_lock(&s->lock); | |
1145 | ||
1146 | while (remaining_sectors != 0) { | |
bd28f835 | 1147 | |
5ebaa27e | 1148 | /* prepare next request */ |
3fc48d09 | 1149 | cur_nr_sectors = remaining_sectors; |
5ebaa27e FZ |
1150 | if (s->crypt_method) { |
1151 | cur_nr_sectors = MIN(cur_nr_sectors, | |
1152 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
585f8587 | 1153 | } |
5ebaa27e | 1154 | |
3fc48d09 | 1155 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, |
5ebaa27e | 1156 | &cur_nr_sectors, &cluster_offset); |
8af36488 | 1157 | if (ret < 0) { |
3fc48d09 | 1158 | goto fail; |
8af36488 | 1159 | } |
bd28f835 | 1160 | |
3fc48d09 | 1161 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
c87c0672 | 1162 | |
3fc48d09 | 1163 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 1164 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e FZ |
1165 | cur_nr_sectors * 512); |
1166 | ||
68d000a3 KW |
1167 | switch (ret) { |
1168 | case QCOW2_CLUSTER_UNALLOCATED: | |
5ebaa27e FZ |
1169 | |
1170 | if (bs->backing_hd) { | |
1171 | /* read from the base image */ | |
3fc48d09 FZ |
1172 | n1 = qcow2_backing_read1(bs->backing_hd, &hd_qiov, |
1173 | sector_num, cur_nr_sectors); | |
5ebaa27e | 1174 | if (n1 > 0) { |
44deba5a KW |
1175 | QEMUIOVector local_qiov; |
1176 | ||
1177 | qemu_iovec_init(&local_qiov, hd_qiov.niov); | |
1178 | qemu_iovec_concat(&local_qiov, &hd_qiov, 0, | |
1179 | n1 * BDRV_SECTOR_SIZE); | |
1180 | ||
5ebaa27e FZ |
1181 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); |
1182 | qemu_co_mutex_unlock(&s->lock); | |
3fc48d09 | 1183 | ret = bdrv_co_readv(bs->backing_hd, sector_num, |
44deba5a | 1184 | n1, &local_qiov); |
5ebaa27e | 1185 | qemu_co_mutex_lock(&s->lock); |
44deba5a KW |
1186 | |
1187 | qemu_iovec_destroy(&local_qiov); | |
1188 | ||
5ebaa27e | 1189 | if (ret < 0) { |
3fc48d09 | 1190 | goto fail; |
5ebaa27e FZ |
1191 | } |
1192 | } | |
1193 | } else { | |
1194 | /* Note: in this case, no need to wait */ | |
3d9b4925 | 1195 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
5ebaa27e | 1196 | } |
68d000a3 KW |
1197 | break; |
1198 | ||
6377af48 | 1199 | case QCOW2_CLUSTER_ZERO: |
3d9b4925 | 1200 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
6377af48 KW |
1201 | break; |
1202 | ||
68d000a3 | 1203 | case QCOW2_CLUSTER_COMPRESSED: |
5ebaa27e FZ |
1204 | /* add AIO support for compressed blocks ? */ |
1205 | ret = qcow2_decompress_cluster(bs, cluster_offset); | |
1206 | if (ret < 0) { | |
3fc48d09 | 1207 | goto fail; |
bd28f835 KW |
1208 | } |
1209 | ||
03396148 | 1210 | qemu_iovec_from_buf(&hd_qiov, 0, |
5ebaa27e | 1211 | s->cluster_cache + index_in_cluster * 512, |
faf575c1 | 1212 | 512 * cur_nr_sectors); |
68d000a3 KW |
1213 | break; |
1214 | ||
1215 | case QCOW2_CLUSTER_NORMAL: | |
5ebaa27e | 1216 | if ((cluster_offset & 511) != 0) { |
3fc48d09 FZ |
1217 | ret = -EIO; |
1218 | goto fail; | |
5ebaa27e | 1219 | } |
bd28f835 | 1220 | |
5ebaa27e FZ |
1221 | if (s->crypt_method) { |
1222 | /* | |
1223 | * For encrypted images, read everything into a temporary | |
1224 | * contiguous buffer on which the AES functions can work. | |
1225 | */ | |
3fc48d09 FZ |
1226 | if (!cluster_data) { |
1227 | cluster_data = | |
de82815d KW |
1228 | qemu_try_blockalign(bs->file, QCOW_MAX_CRYPT_CLUSTERS |
1229 | * s->cluster_size); | |
1230 | if (cluster_data == NULL) { | |
1231 | ret = -ENOMEM; | |
1232 | goto fail; | |
1233 | } | |
5ebaa27e FZ |
1234 | } |
1235 | ||
1236 | assert(cur_nr_sectors <= | |
1237 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
3fc48d09 FZ |
1238 | qemu_iovec_reset(&hd_qiov); |
1239 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
1240 | 512 * cur_nr_sectors); |
1241 | } | |
1242 | ||
1243 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
1244 | qemu_co_mutex_unlock(&s->lock); | |
1245 | ret = bdrv_co_readv(bs->file, | |
1246 | (cluster_offset >> 9) + index_in_cluster, | |
3fc48d09 | 1247 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
1248 | qemu_co_mutex_lock(&s->lock); |
1249 | if (ret < 0) { | |
3fc48d09 | 1250 | goto fail; |
5ebaa27e FZ |
1251 | } |
1252 | if (s->crypt_method) { | |
3fc48d09 FZ |
1253 | qcow2_encrypt_sectors(s, sector_num, cluster_data, |
1254 | cluster_data, cur_nr_sectors, 0, &s->aes_decrypt_key); | |
03396148 MT |
1255 | qemu_iovec_from_buf(qiov, bytes_done, |
1256 | cluster_data, 512 * cur_nr_sectors); | |
5ebaa27e | 1257 | } |
68d000a3 KW |
1258 | break; |
1259 | ||
1260 | default: | |
1261 | g_assert_not_reached(); | |
1262 | ret = -EIO; | |
1263 | goto fail; | |
faf575c1 | 1264 | } |
f141eafe | 1265 | |
3fc48d09 FZ |
1266 | remaining_sectors -= cur_nr_sectors; |
1267 | sector_num += cur_nr_sectors; | |
1268 | bytes_done += cur_nr_sectors * 512; | |
5ebaa27e | 1269 | } |
3fc48d09 | 1270 | ret = 0; |
faf575c1 | 1271 | |
3fc48d09 | 1272 | fail: |
68d100e9 | 1273 | qemu_co_mutex_unlock(&s->lock); |
42496d62 | 1274 | |
3fc48d09 | 1275 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1276 | qemu_vfree(cluster_data); |
68d100e9 KW |
1277 | |
1278 | return ret; | |
585f8587 FB |
1279 | } |
1280 | ||
a968168c | 1281 | static coroutine_fn int qcow2_co_writev(BlockDriverState *bs, |
3fc48d09 FZ |
1282 | int64_t sector_num, |
1283 | int remaining_sectors, | |
1284 | QEMUIOVector *qiov) | |
585f8587 | 1285 | { |
585f8587 | 1286 | BDRVQcowState *s = bs->opaque; |
585f8587 | 1287 | int index_in_cluster; |
68d100e9 | 1288 | int ret; |
faf575c1 | 1289 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 1290 | uint64_t cluster_offset; |
3fc48d09 FZ |
1291 | QEMUIOVector hd_qiov; |
1292 | uint64_t bytes_done = 0; | |
1293 | uint8_t *cluster_data = NULL; | |
8d2497c3 | 1294 | QCowL2Meta *l2meta = NULL; |
c2271403 | 1295 | |
3cce16f4 KW |
1296 | trace_qcow2_writev_start_req(qemu_coroutine_self(), sector_num, |
1297 | remaining_sectors); | |
1298 | ||
3fc48d09 FZ |
1299 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1300 | ||
1301 | s->cluster_cache_offset = -1; /* disable compressed cache */ | |
3b46e624 | 1302 | |
3fc48d09 FZ |
1303 | qemu_co_mutex_lock(&s->lock); |
1304 | ||
1305 | while (remaining_sectors != 0) { | |
1306 | ||
f50f88b9 | 1307 | l2meta = NULL; |
cf5c1a23 | 1308 | |
3cce16f4 | 1309 | trace_qcow2_writev_start_part(qemu_coroutine_self()); |
3fc48d09 | 1310 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
16f0587e | 1311 | cur_nr_sectors = remaining_sectors; |
5ebaa27e | 1312 | if (s->crypt_method && |
16f0587e HT |
1313 | cur_nr_sectors > |
1314 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster) { | |
1315 | cur_nr_sectors = | |
1316 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster; | |
5ebaa27e | 1317 | } |
095a9c58 | 1318 | |
3fc48d09 | 1319 | ret = qcow2_alloc_cluster_offset(bs, sector_num << 9, |
16f0587e | 1320 | &cur_nr_sectors, &cluster_offset, &l2meta); |
5ebaa27e | 1321 | if (ret < 0) { |
3fc48d09 | 1322 | goto fail; |
5ebaa27e | 1323 | } |
148da7ea | 1324 | |
5ebaa27e | 1325 | assert((cluster_offset & 511) == 0); |
148da7ea | 1326 | |
3fc48d09 | 1327 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 1328 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e | 1329 | cur_nr_sectors * 512); |
6f5f060b | 1330 | |
5ebaa27e | 1331 | if (s->crypt_method) { |
3fc48d09 | 1332 | if (!cluster_data) { |
de82815d KW |
1333 | cluster_data = qemu_try_blockalign(bs->file, |
1334 | QCOW_MAX_CRYPT_CLUSTERS | |
1335 | * s->cluster_size); | |
1336 | if (cluster_data == NULL) { | |
1337 | ret = -ENOMEM; | |
1338 | goto fail; | |
1339 | } | |
5ebaa27e | 1340 | } |
6f5f060b | 1341 | |
3fc48d09 | 1342 | assert(hd_qiov.size <= |
5ebaa27e | 1343 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
d5e6b161 | 1344 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); |
6f5f060b | 1345 | |
3fc48d09 FZ |
1346 | qcow2_encrypt_sectors(s, sector_num, cluster_data, |
1347 | cluster_data, cur_nr_sectors, 1, &s->aes_encrypt_key); | |
6f5f060b | 1348 | |
3fc48d09 FZ |
1349 | qemu_iovec_reset(&hd_qiov); |
1350 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
1351 | cur_nr_sectors * 512); |
1352 | } | |
6f5f060b | 1353 | |
231bb267 | 1354 | ret = qcow2_pre_write_overlap_check(bs, 0, |
cf93980e HR |
1355 | cluster_offset + index_in_cluster * BDRV_SECTOR_SIZE, |
1356 | cur_nr_sectors * BDRV_SECTOR_SIZE); | |
1357 | if (ret < 0) { | |
1358 | goto fail; | |
1359 | } | |
1360 | ||
5ebaa27e | 1361 | qemu_co_mutex_unlock(&s->lock); |
67a7a0eb | 1362 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); |
3cce16f4 KW |
1363 | trace_qcow2_writev_data(qemu_coroutine_self(), |
1364 | (cluster_offset >> 9) + index_in_cluster); | |
5ebaa27e FZ |
1365 | ret = bdrv_co_writev(bs->file, |
1366 | (cluster_offset >> 9) + index_in_cluster, | |
3fc48d09 | 1367 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
1368 | qemu_co_mutex_lock(&s->lock); |
1369 | if (ret < 0) { | |
3fc48d09 | 1370 | goto fail; |
5ebaa27e | 1371 | } |
f141eafe | 1372 | |
88c6588c KW |
1373 | while (l2meta != NULL) { |
1374 | QCowL2Meta *next; | |
1375 | ||
f50f88b9 KW |
1376 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); |
1377 | if (ret < 0) { | |
1378 | goto fail; | |
1379 | } | |
faf575c1 | 1380 | |
4e95314e KW |
1381 | /* Take the request off the list of running requests */ |
1382 | if (l2meta->nb_clusters != 0) { | |
1383 | QLIST_REMOVE(l2meta, next_in_flight); | |
1384 | } | |
1385 | ||
4e95314e | 1386 | qemu_co_queue_restart_all(&l2meta->dependent_requests); |
4e95314e | 1387 | |
88c6588c | 1388 | next = l2meta->next; |
f50f88b9 | 1389 | g_free(l2meta); |
88c6588c | 1390 | l2meta = next; |
f50f88b9 | 1391 | } |
0fa9131a | 1392 | |
3fc48d09 FZ |
1393 | remaining_sectors -= cur_nr_sectors; |
1394 | sector_num += cur_nr_sectors; | |
1395 | bytes_done += cur_nr_sectors * 512; | |
3cce16f4 | 1396 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_nr_sectors); |
5ebaa27e | 1397 | } |
3fc48d09 | 1398 | ret = 0; |
faf575c1 | 1399 | |
3fc48d09 | 1400 | fail: |
4e95314e KW |
1401 | qemu_co_mutex_unlock(&s->lock); |
1402 | ||
88c6588c KW |
1403 | while (l2meta != NULL) { |
1404 | QCowL2Meta *next; | |
1405 | ||
4e95314e KW |
1406 | if (l2meta->nb_clusters != 0) { |
1407 | QLIST_REMOVE(l2meta, next_in_flight); | |
1408 | } | |
1409 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
88c6588c KW |
1410 | |
1411 | next = l2meta->next; | |
cf5c1a23 | 1412 | g_free(l2meta); |
88c6588c | 1413 | l2meta = next; |
cf5c1a23 | 1414 | } |
0fa9131a | 1415 | |
3fc48d09 | 1416 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1417 | qemu_vfree(cluster_data); |
3cce16f4 | 1418 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
42496d62 | 1419 | |
68d100e9 | 1420 | return ret; |
585f8587 FB |
1421 | } |
1422 | ||
7c80ab3f | 1423 | static void qcow2_close(BlockDriverState *bs) |
585f8587 FB |
1424 | { |
1425 | BDRVQcowState *s = bs->opaque; | |
de82815d | 1426 | qemu_vfree(s->l1_table); |
cf93980e HR |
1427 | /* else pre-write overlap checks in cache_destroy may crash */ |
1428 | s->l1_table = NULL; | |
29c1a730 | 1429 | |
27eb6c09 KW |
1430 | if (!(bs->open_flags & BDRV_O_INCOMING)) { |
1431 | qcow2_cache_flush(bs, s->l2_table_cache); | |
1432 | qcow2_cache_flush(bs, s->refcount_block_cache); | |
29c1a730 | 1433 | |
27eb6c09 KW |
1434 | qcow2_mark_clean(bs); |
1435 | } | |
c61d0004 | 1436 | |
29c1a730 KW |
1437 | qcow2_cache_destroy(bs, s->l2_table_cache); |
1438 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1439 | ||
6744cbab | 1440 | g_free(s->unknown_header_fields); |
75bab85c | 1441 | cleanup_unknown_header_ext(bs); |
6744cbab | 1442 | |
7267c094 | 1443 | g_free(s->cluster_cache); |
dea43a65 | 1444 | qemu_vfree(s->cluster_data); |
ed6ccf0f | 1445 | qcow2_refcount_close(bs); |
28c1202b | 1446 | qcow2_free_snapshots(bs); |
585f8587 FB |
1447 | } |
1448 | ||
5a8a30db | 1449 | static void qcow2_invalidate_cache(BlockDriverState *bs, Error **errp) |
06d9260f AL |
1450 | { |
1451 | BDRVQcowState *s = bs->opaque; | |
1452 | int flags = s->flags; | |
1453 | AES_KEY aes_encrypt_key; | |
1454 | AES_KEY aes_decrypt_key; | |
1455 | uint32_t crypt_method = 0; | |
acdfb480 | 1456 | QDict *options; |
5a8a30db KW |
1457 | Error *local_err = NULL; |
1458 | int ret; | |
06d9260f AL |
1459 | |
1460 | /* | |
1461 | * Backing files are read-only which makes all of their metadata immutable, | |
1462 | * that means we don't have to worry about reopening them here. | |
1463 | */ | |
1464 | ||
1465 | if (s->crypt_method) { | |
1466 | crypt_method = s->crypt_method; | |
1467 | memcpy(&aes_encrypt_key, &s->aes_encrypt_key, sizeof(aes_encrypt_key)); | |
1468 | memcpy(&aes_decrypt_key, &s->aes_decrypt_key, sizeof(aes_decrypt_key)); | |
1469 | } | |
1470 | ||
1471 | qcow2_close(bs); | |
1472 | ||
5a8a30db KW |
1473 | bdrv_invalidate_cache(bs->file, &local_err); |
1474 | if (local_err) { | |
1475 | error_propagate(errp, local_err); | |
1476 | return; | |
1477 | } | |
3456a8d1 | 1478 | |
06d9260f | 1479 | memset(s, 0, sizeof(BDRVQcowState)); |
d475e5ac | 1480 | options = qdict_clone_shallow(bs->options); |
5a8a30db KW |
1481 | |
1482 | ret = qcow2_open(bs, options, flags, &local_err); | |
a1904e48 | 1483 | QDECREF(options); |
5a8a30db KW |
1484 | if (local_err) { |
1485 | error_setg(errp, "Could not reopen qcow2 layer: %s", | |
1486 | error_get_pretty(local_err)); | |
1487 | error_free(local_err); | |
1488 | return; | |
1489 | } else if (ret < 0) { | |
1490 | error_setg_errno(errp, -ret, "Could not reopen qcow2 layer"); | |
1491 | return; | |
1492 | } | |
acdfb480 | 1493 | |
06d9260f AL |
1494 | if (crypt_method) { |
1495 | s->crypt_method = crypt_method; | |
1496 | memcpy(&s->aes_encrypt_key, &aes_encrypt_key, sizeof(aes_encrypt_key)); | |
1497 | memcpy(&s->aes_decrypt_key, &aes_decrypt_key, sizeof(aes_decrypt_key)); | |
1498 | } | |
1499 | } | |
1500 | ||
e24e49e6 KW |
1501 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
1502 | size_t len, size_t buflen) | |
1503 | { | |
1504 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
1505 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
1506 | ||
1507 | if (buflen < ext_len) { | |
1508 | return -ENOSPC; | |
1509 | } | |
1510 | ||
1511 | *ext_backing_fmt = (QCowExtension) { | |
1512 | .magic = cpu_to_be32(magic), | |
1513 | .len = cpu_to_be32(len), | |
1514 | }; | |
1515 | memcpy(buf + sizeof(QCowExtension), s, len); | |
1516 | ||
1517 | return ext_len; | |
1518 | } | |
1519 | ||
756e6736 | 1520 | /* |
e24e49e6 KW |
1521 | * Updates the qcow2 header, including the variable length parts of it, i.e. |
1522 | * the backing file name and all extensions. qcow2 was not designed to allow | |
1523 | * such changes, so if we run out of space (we can only use the first cluster) | |
1524 | * this function may fail. | |
756e6736 KW |
1525 | * |
1526 | * Returns 0 on success, -errno in error cases. | |
1527 | */ | |
e24e49e6 | 1528 | int qcow2_update_header(BlockDriverState *bs) |
756e6736 | 1529 | { |
756e6736 | 1530 | BDRVQcowState *s = bs->opaque; |
e24e49e6 KW |
1531 | QCowHeader *header; |
1532 | char *buf; | |
1533 | size_t buflen = s->cluster_size; | |
756e6736 | 1534 | int ret; |
e24e49e6 KW |
1535 | uint64_t total_size; |
1536 | uint32_t refcount_table_clusters; | |
6744cbab | 1537 | size_t header_length; |
75bab85c | 1538 | Qcow2UnknownHeaderExtension *uext; |
756e6736 | 1539 | |
e24e49e6 | 1540 | buf = qemu_blockalign(bs, buflen); |
756e6736 | 1541 | |
e24e49e6 KW |
1542 | /* Header structure */ |
1543 | header = (QCowHeader*) buf; | |
756e6736 | 1544 | |
e24e49e6 KW |
1545 | if (buflen < sizeof(*header)) { |
1546 | ret = -ENOSPC; | |
1547 | goto fail; | |
756e6736 KW |
1548 | } |
1549 | ||
6744cbab | 1550 | header_length = sizeof(*header) + s->unknown_header_fields_size; |
e24e49e6 KW |
1551 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
1552 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
1553 | ||
1554 | *header = (QCowHeader) { | |
6744cbab | 1555 | /* Version 2 fields */ |
e24e49e6 | 1556 | .magic = cpu_to_be32(QCOW_MAGIC), |
6744cbab | 1557 | .version = cpu_to_be32(s->qcow_version), |
e24e49e6 KW |
1558 | .backing_file_offset = 0, |
1559 | .backing_file_size = 0, | |
1560 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
1561 | .size = cpu_to_be64(total_size), | |
1562 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
1563 | .l1_size = cpu_to_be32(s->l1_size), | |
1564 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
1565 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
1566 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
1567 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
1568 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
6744cbab KW |
1569 | |
1570 | /* Version 3 fields */ | |
1571 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
1572 | .compatible_features = cpu_to_be64(s->compatible_features), | |
1573 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
b6481f37 | 1574 | .refcount_order = cpu_to_be32(s->refcount_order), |
6744cbab | 1575 | .header_length = cpu_to_be32(header_length), |
e24e49e6 | 1576 | }; |
756e6736 | 1577 | |
6744cbab KW |
1578 | /* For older versions, write a shorter header */ |
1579 | switch (s->qcow_version) { | |
1580 | case 2: | |
1581 | ret = offsetof(QCowHeader, incompatible_features); | |
1582 | break; | |
1583 | case 3: | |
1584 | ret = sizeof(*header); | |
1585 | break; | |
1586 | default: | |
b6c14762 JM |
1587 | ret = -EINVAL; |
1588 | goto fail; | |
6744cbab KW |
1589 | } |
1590 | ||
1591 | buf += ret; | |
1592 | buflen -= ret; | |
1593 | memset(buf, 0, buflen); | |
1594 | ||
1595 | /* Preserve any unknown field in the header */ | |
1596 | if (s->unknown_header_fields_size) { | |
1597 | if (buflen < s->unknown_header_fields_size) { | |
1598 | ret = -ENOSPC; | |
1599 | goto fail; | |
1600 | } | |
1601 | ||
1602 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
1603 | buf += s->unknown_header_fields_size; | |
1604 | buflen -= s->unknown_header_fields_size; | |
1605 | } | |
756e6736 | 1606 | |
e24e49e6 KW |
1607 | /* Backing file format header extension */ |
1608 | if (*bs->backing_format) { | |
1609 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, | |
1610 | bs->backing_format, strlen(bs->backing_format), | |
1611 | buflen); | |
1612 | if (ret < 0) { | |
1613 | goto fail; | |
756e6736 KW |
1614 | } |
1615 | ||
e24e49e6 KW |
1616 | buf += ret; |
1617 | buflen -= ret; | |
756e6736 KW |
1618 | } |
1619 | ||
cfcc4c62 KW |
1620 | /* Feature table */ |
1621 | Qcow2Feature features[] = { | |
c61d0004 SH |
1622 | { |
1623 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1624 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
1625 | .name = "dirty bit", | |
1626 | }, | |
69c98726 HR |
1627 | { |
1628 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1629 | .bit = QCOW2_INCOMPAT_CORRUPT_BITNR, | |
1630 | .name = "corrupt bit", | |
1631 | }, | |
bfe8043e SH |
1632 | { |
1633 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
1634 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
1635 | .name = "lazy refcounts", | |
1636 | }, | |
cfcc4c62 KW |
1637 | }; |
1638 | ||
1639 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
1640 | features, sizeof(features), buflen); | |
1641 | if (ret < 0) { | |
1642 | goto fail; | |
1643 | } | |
1644 | buf += ret; | |
1645 | buflen -= ret; | |
1646 | ||
75bab85c KW |
1647 | /* Keep unknown header extensions */ |
1648 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
1649 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
1650 | if (ret < 0) { | |
1651 | goto fail; | |
1652 | } | |
1653 | ||
1654 | buf += ret; | |
1655 | buflen -= ret; | |
1656 | } | |
1657 | ||
e24e49e6 KW |
1658 | /* End of header extensions */ |
1659 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
756e6736 KW |
1660 | if (ret < 0) { |
1661 | goto fail; | |
1662 | } | |
1663 | ||
e24e49e6 KW |
1664 | buf += ret; |
1665 | buflen -= ret; | |
756e6736 | 1666 | |
e24e49e6 KW |
1667 | /* Backing file name */ |
1668 | if (*bs->backing_file) { | |
1669 | size_t backing_file_len = strlen(bs->backing_file); | |
1670 | ||
1671 | if (buflen < backing_file_len) { | |
1672 | ret = -ENOSPC; | |
1673 | goto fail; | |
1674 | } | |
1675 | ||
00ea1881 | 1676 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ |
e24e49e6 KW |
1677 | strncpy(buf, bs->backing_file, buflen); |
1678 | ||
1679 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
1680 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
756e6736 KW |
1681 | } |
1682 | ||
e24e49e6 KW |
1683 | /* Write the new header */ |
1684 | ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size); | |
756e6736 KW |
1685 | if (ret < 0) { |
1686 | goto fail; | |
1687 | } | |
1688 | ||
1689 | ret = 0; | |
1690 | fail: | |
e24e49e6 | 1691 | qemu_vfree(header); |
756e6736 KW |
1692 | return ret; |
1693 | } | |
1694 | ||
1695 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
1696 | const char *backing_file, const char *backing_fmt) | |
1697 | { | |
e24e49e6 KW |
1698 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
1699 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1700 | ||
1701 | return qcow2_update_header(bs); | |
756e6736 KW |
1702 | } |
1703 | ||
a35e1c17 KW |
1704 | static int preallocate(BlockDriverState *bs) |
1705 | { | |
a35e1c17 KW |
1706 | uint64_t nb_sectors; |
1707 | uint64_t offset; | |
060bee89 | 1708 | uint64_t host_offset = 0; |
a35e1c17 | 1709 | int num; |
148da7ea | 1710 | int ret; |
f50f88b9 | 1711 | QCowL2Meta *meta; |
a35e1c17 | 1712 | |
57322b78 | 1713 | nb_sectors = bdrv_nb_sectors(bs); |
a35e1c17 KW |
1714 | offset = 0; |
1715 | ||
1716 | while (nb_sectors) { | |
7c2bbf4a | 1717 | num = MIN(nb_sectors, INT_MAX >> BDRV_SECTOR_BITS); |
16f0587e | 1718 | ret = qcow2_alloc_cluster_offset(bs, offset, &num, |
060bee89 | 1719 | &host_offset, &meta); |
148da7ea | 1720 | if (ret < 0) { |
19dbcbf7 | 1721 | return ret; |
a35e1c17 KW |
1722 | } |
1723 | ||
c792707f SH |
1724 | while (meta) { |
1725 | QCowL2Meta *next = meta->next; | |
1726 | ||
7c2bbf4a HT |
1727 | ret = qcow2_alloc_cluster_link_l2(bs, meta); |
1728 | if (ret < 0) { | |
1729 | qcow2_free_any_clusters(bs, meta->alloc_offset, | |
1730 | meta->nb_clusters, QCOW2_DISCARD_NEVER); | |
1731 | return ret; | |
1732 | } | |
1733 | ||
1734 | /* There are no dependent requests, but we need to remove our | |
1735 | * request from the list of in-flight requests */ | |
4e95314e | 1736 | QLIST_REMOVE(meta, next_in_flight); |
c792707f SH |
1737 | |
1738 | g_free(meta); | |
1739 | meta = next; | |
f50f88b9 | 1740 | } |
f214978a | 1741 | |
a35e1c17 KW |
1742 | /* TODO Preallocate data if requested */ |
1743 | ||
1744 | nb_sectors -= num; | |
7c2bbf4a | 1745 | offset += num << BDRV_SECTOR_BITS; |
a35e1c17 KW |
1746 | } |
1747 | ||
1748 | /* | |
1749 | * It is expected that the image file is large enough to actually contain | |
1750 | * all of the allocated clusters (otherwise we get failing reads after | |
1751 | * EOF). Extend the image to the last allocated sector. | |
1752 | */ | |
060bee89 | 1753 | if (host_offset != 0) { |
7c2bbf4a HT |
1754 | uint8_t buf[BDRV_SECTOR_SIZE]; |
1755 | memset(buf, 0, BDRV_SECTOR_SIZE); | |
1756 | ret = bdrv_write(bs->file, (host_offset >> BDRV_SECTOR_BITS) + num - 1, | |
1757 | buf, 1); | |
19dbcbf7 KW |
1758 | if (ret < 0) { |
1759 | return ret; | |
1760 | } | |
a35e1c17 KW |
1761 | } |
1762 | ||
1763 | return 0; | |
1764 | } | |
1765 | ||
7c80ab3f JS |
1766 | static int qcow2_create2(const char *filename, int64_t total_size, |
1767 | const char *backing_file, const char *backing_format, | |
ffeaac9b | 1768 | int flags, size_t cluster_size, PreallocMode prealloc, |
1bd0e2d1 | 1769 | QemuOpts *opts, int version, |
3ef6c40a | 1770 | Error **errp) |
a9420734 | 1771 | { |
9b2260cb | 1772 | /* Calculate cluster_bits */ |
a9420734 KW |
1773 | int cluster_bits; |
1774 | cluster_bits = ffs(cluster_size) - 1; | |
1775 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || | |
1776 | (1 << cluster_bits) != cluster_size) | |
1777 | { | |
3ef6c40a HR |
1778 | error_setg(errp, "Cluster size must be a power of two between %d and " |
1779 | "%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); | |
a9420734 KW |
1780 | return -EINVAL; |
1781 | } | |
1782 | ||
1783 | /* | |
1784 | * Open the image file and write a minimal qcow2 header. | |
1785 | * | |
1786 | * We keep things simple and start with a zero-sized image. We also | |
1787 | * do without refcount blocks or a L1 table for now. We'll fix the | |
1788 | * inconsistency later. | |
1789 | * | |
1790 | * We do need a refcount table because growing the refcount table means | |
1791 | * allocating two new refcount blocks - the seconds of which would be at | |
1792 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
1793 | * size for any qcow2 image. | |
1794 | */ | |
1795 | BlockDriverState* bs; | |
f8413b3c | 1796 | QCowHeader *header; |
b106ad91 | 1797 | uint64_t* refcount_table; |
3ef6c40a | 1798 | Error *local_err = NULL; |
a9420734 KW |
1799 | int ret; |
1800 | ||
0e4271b7 HT |
1801 | if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) { |
1802 | int64_t meta_size = 0; | |
1803 | uint64_t nreftablee, nrefblocke, nl1e, nl2e; | |
1804 | int64_t aligned_total_size = align_offset(total_size, cluster_size); | |
1805 | ||
1806 | /* header: 1 cluster */ | |
1807 | meta_size += cluster_size; | |
1808 | ||
1809 | /* total size of L2 tables */ | |
1810 | nl2e = aligned_total_size / cluster_size; | |
1811 | nl2e = align_offset(nl2e, cluster_size / sizeof(uint64_t)); | |
1812 | meta_size += nl2e * sizeof(uint64_t); | |
1813 | ||
1814 | /* total size of L1 tables */ | |
1815 | nl1e = nl2e * sizeof(uint64_t) / cluster_size; | |
1816 | nl1e = align_offset(nl1e, cluster_size / sizeof(uint64_t)); | |
1817 | meta_size += nl1e * sizeof(uint64_t); | |
1818 | ||
1819 | /* total size of refcount blocks | |
1820 | * | |
1821 | * note: every host cluster is reference-counted, including metadata | |
1822 | * (even refcount blocks are recursively included). | |
1823 | * Let: | |
1824 | * a = total_size (this is the guest disk size) | |
1825 | * m = meta size not including refcount blocks and refcount tables | |
1826 | * c = cluster size | |
1827 | * y1 = number of refcount blocks entries | |
1828 | * y2 = meta size including everything | |
1829 | * then, | |
1830 | * y1 = (y2 + a)/c | |
1831 | * y2 = y1 * sizeof(u16) + y1 * sizeof(u16) * sizeof(u64) / c + m | |
1832 | * we can get y1: | |
1833 | * y1 = (a + m) / (c - sizeof(u16) - sizeof(u16) * sizeof(u64) / c) | |
1834 | */ | |
1835 | nrefblocke = (aligned_total_size + meta_size + cluster_size) / | |
1836 | (cluster_size - sizeof(uint16_t) - | |
1837 | 1.0 * sizeof(uint16_t) * sizeof(uint64_t) / cluster_size); | |
1838 | nrefblocke = align_offset(nrefblocke, cluster_size / sizeof(uint16_t)); | |
1839 | meta_size += nrefblocke * sizeof(uint16_t); | |
1840 | ||
1841 | /* total size of refcount tables */ | |
1842 | nreftablee = nrefblocke * sizeof(uint16_t) / cluster_size; | |
1843 | nreftablee = align_offset(nreftablee, cluster_size / sizeof(uint64_t)); | |
1844 | meta_size += nreftablee * sizeof(uint64_t); | |
1845 | ||
1846 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, | |
1847 | aligned_total_size + meta_size); | |
1848 | qemu_opt_set(opts, BLOCK_OPT_PREALLOC, PreallocMode_lookup[prealloc]); | |
1849 | } | |
1850 | ||
c282e1fd | 1851 | ret = bdrv_create_file(filename, opts, &local_err); |
a9420734 | 1852 | if (ret < 0) { |
3ef6c40a | 1853 | error_propagate(errp, local_err); |
a9420734 KW |
1854 | return ret; |
1855 | } | |
1856 | ||
2e40134b HR |
1857 | bs = NULL; |
1858 | ret = bdrv_open(&bs, filename, NULL, NULL, BDRV_O_RDWR | BDRV_O_PROTOCOL, | |
1859 | NULL, &local_err); | |
a9420734 | 1860 | if (ret < 0) { |
3ef6c40a | 1861 | error_propagate(errp, local_err); |
a9420734 KW |
1862 | return ret; |
1863 | } | |
1864 | ||
1865 | /* Write the header */ | |
f8413b3c KW |
1866 | QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header)); |
1867 | header = g_malloc0(cluster_size); | |
1868 | *header = (QCowHeader) { | |
1869 | .magic = cpu_to_be32(QCOW_MAGIC), | |
1870 | .version = cpu_to_be32(version), | |
1871 | .cluster_bits = cpu_to_be32(cluster_bits), | |
1872 | .size = cpu_to_be64(0), | |
1873 | .l1_table_offset = cpu_to_be64(0), | |
1874 | .l1_size = cpu_to_be32(0), | |
1875 | .refcount_table_offset = cpu_to_be64(cluster_size), | |
1876 | .refcount_table_clusters = cpu_to_be32(1), | |
1877 | .refcount_order = cpu_to_be32(3 + REFCOUNT_SHIFT), | |
1878 | .header_length = cpu_to_be32(sizeof(*header)), | |
1879 | }; | |
a9420734 KW |
1880 | |
1881 | if (flags & BLOCK_FLAG_ENCRYPT) { | |
f8413b3c | 1882 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_AES); |
a9420734 | 1883 | } else { |
f8413b3c | 1884 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); |
a9420734 KW |
1885 | } |
1886 | ||
bfe8043e | 1887 | if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) { |
f8413b3c | 1888 | header->compatible_features |= |
bfe8043e SH |
1889 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); |
1890 | } | |
1891 | ||
f8413b3c KW |
1892 | ret = bdrv_pwrite(bs, 0, header, cluster_size); |
1893 | g_free(header); | |
a9420734 | 1894 | if (ret < 0) { |
3ef6c40a | 1895 | error_setg_errno(errp, -ret, "Could not write qcow2 header"); |
a9420734 KW |
1896 | goto out; |
1897 | } | |
1898 | ||
b106ad91 KW |
1899 | /* Write a refcount table with one refcount block */ |
1900 | refcount_table = g_malloc0(2 * cluster_size); | |
1901 | refcount_table[0] = cpu_to_be64(2 * cluster_size); | |
1902 | ret = bdrv_pwrite(bs, cluster_size, refcount_table, 2 * cluster_size); | |
7267c094 | 1903 | g_free(refcount_table); |
a9420734 KW |
1904 | |
1905 | if (ret < 0) { | |
3ef6c40a | 1906 | error_setg_errno(errp, -ret, "Could not write refcount table"); |
a9420734 KW |
1907 | goto out; |
1908 | } | |
1909 | ||
f67503e5 HR |
1910 | bdrv_unref(bs); |
1911 | bs = NULL; | |
a9420734 KW |
1912 | |
1913 | /* | |
1914 | * And now open the image and make it consistent first (i.e. increase the | |
1915 | * refcount of the cluster that is occupied by the header and the refcount | |
1916 | * table) | |
1917 | */ | |
1918 | BlockDriver* drv = bdrv_find_format("qcow2"); | |
1919 | assert(drv != NULL); | |
ddf5636d | 1920 | ret = bdrv_open(&bs, filename, NULL, NULL, |
3ef6c40a | 1921 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_FLUSH, drv, &local_err); |
a9420734 | 1922 | if (ret < 0) { |
3ef6c40a | 1923 | error_propagate(errp, local_err); |
a9420734 KW |
1924 | goto out; |
1925 | } | |
1926 | ||
b106ad91 | 1927 | ret = qcow2_alloc_clusters(bs, 3 * cluster_size); |
a9420734 | 1928 | if (ret < 0) { |
3ef6c40a HR |
1929 | error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 " |
1930 | "header and refcount table"); | |
a9420734 KW |
1931 | goto out; |
1932 | ||
1933 | } else if (ret != 0) { | |
1934 | error_report("Huh, first cluster in empty image is already in use?"); | |
1935 | abort(); | |
1936 | } | |
1937 | ||
1938 | /* Okay, now that we have a valid image, let's give it the right size */ | |
180e9526 | 1939 | ret = bdrv_truncate(bs, total_size); |
a9420734 | 1940 | if (ret < 0) { |
3ef6c40a | 1941 | error_setg_errno(errp, -ret, "Could not resize image"); |
a9420734 KW |
1942 | goto out; |
1943 | } | |
1944 | ||
1945 | /* Want a backing file? There you go.*/ | |
1946 | if (backing_file) { | |
1947 | ret = bdrv_change_backing_file(bs, backing_file, backing_format); | |
1948 | if (ret < 0) { | |
3ef6c40a HR |
1949 | error_setg_errno(errp, -ret, "Could not assign backing file '%s' " |
1950 | "with format '%s'", backing_file, backing_format); | |
a9420734 KW |
1951 | goto out; |
1952 | } | |
1953 | } | |
1954 | ||
1955 | /* And if we're supposed to preallocate metadata, do that now */ | |
0e4271b7 | 1956 | if (prealloc != PREALLOC_MODE_OFF) { |
15552c4a ZYW |
1957 | BDRVQcowState *s = bs->opaque; |
1958 | qemu_co_mutex_lock(&s->lock); | |
a9420734 | 1959 | ret = preallocate(bs); |
15552c4a | 1960 | qemu_co_mutex_unlock(&s->lock); |
a9420734 | 1961 | if (ret < 0) { |
3ef6c40a | 1962 | error_setg_errno(errp, -ret, "Could not preallocate metadata"); |
a9420734 KW |
1963 | goto out; |
1964 | } | |
1965 | } | |
1966 | ||
f67503e5 HR |
1967 | bdrv_unref(bs); |
1968 | bs = NULL; | |
ba2ab2f2 HR |
1969 | |
1970 | /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning */ | |
ddf5636d | 1971 | ret = bdrv_open(&bs, filename, NULL, NULL, |
c9fbb99d KW |
1972 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_BACKING, |
1973 | drv, &local_err); | |
84d18f06 | 1974 | if (local_err) { |
ba2ab2f2 HR |
1975 | error_propagate(errp, local_err); |
1976 | goto out; | |
1977 | } | |
1978 | ||
a9420734 KW |
1979 | ret = 0; |
1980 | out: | |
f67503e5 HR |
1981 | if (bs) { |
1982 | bdrv_unref(bs); | |
1983 | } | |
a9420734 KW |
1984 | return ret; |
1985 | } | |
de5f3f40 | 1986 | |
1bd0e2d1 | 1987 | static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) |
de5f3f40 | 1988 | { |
1bd0e2d1 CL |
1989 | char *backing_file = NULL; |
1990 | char *backing_fmt = NULL; | |
1991 | char *buf = NULL; | |
180e9526 | 1992 | uint64_t size = 0; |
de5f3f40 | 1993 | int flags = 0; |
99cce9fa | 1994 | size_t cluster_size = DEFAULT_CLUSTER_SIZE; |
ffeaac9b | 1995 | PreallocMode prealloc; |
8ad1898c | 1996 | int version = 3; |
3ef6c40a HR |
1997 | Error *local_err = NULL; |
1998 | int ret; | |
de5f3f40 KW |
1999 | |
2000 | /* Read out options */ | |
180e9526 HT |
2001 | size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
2002 | BDRV_SECTOR_SIZE); | |
1bd0e2d1 CL |
2003 | backing_file = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE); |
2004 | backing_fmt = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FMT); | |
2005 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_ENCRYPT, false)) { | |
2006 | flags |= BLOCK_FLAG_ENCRYPT; | |
2007 | } | |
2008 | cluster_size = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, | |
2009 | DEFAULT_CLUSTER_SIZE); | |
2010 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
ffeaac9b HT |
2011 | prealloc = qapi_enum_parse(PreallocMode_lookup, buf, |
2012 | PREALLOC_MODE_MAX, PREALLOC_MODE_OFF, | |
2013 | &local_err); | |
2014 | if (local_err) { | |
2015 | error_propagate(errp, local_err); | |
1bd0e2d1 CL |
2016 | ret = -EINVAL; |
2017 | goto finish; | |
2018 | } | |
2019 | g_free(buf); | |
2020 | buf = qemu_opt_get_del(opts, BLOCK_OPT_COMPAT_LEVEL); | |
2021 | if (!buf) { | |
2022 | /* keep the default */ | |
2023 | } else if (!strcmp(buf, "0.10")) { | |
2024 | version = 2; | |
2025 | } else if (!strcmp(buf, "1.1")) { | |
2026 | version = 3; | |
2027 | } else { | |
2028 | error_setg(errp, "Invalid compatibility level: '%s'", buf); | |
2029 | ret = -EINVAL; | |
2030 | goto finish; | |
2031 | } | |
2032 | ||
2033 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_LAZY_REFCOUNTS, false)) { | |
2034 | flags |= BLOCK_FLAG_LAZY_REFCOUNTS; | |
de5f3f40 KW |
2035 | } |
2036 | ||
ffeaac9b | 2037 | if (backing_file && prealloc != PREALLOC_MODE_OFF) { |
3ef6c40a HR |
2038 | error_setg(errp, "Backing file and preallocation cannot be used at " |
2039 | "the same time"); | |
1bd0e2d1 CL |
2040 | ret = -EINVAL; |
2041 | goto finish; | |
de5f3f40 KW |
2042 | } |
2043 | ||
bfe8043e | 2044 | if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { |
3ef6c40a HR |
2045 | error_setg(errp, "Lazy refcounts only supported with compatibility " |
2046 | "level 1.1 and above (use compat=1.1 or greater)"); | |
1bd0e2d1 CL |
2047 | ret = -EINVAL; |
2048 | goto finish; | |
bfe8043e SH |
2049 | } |
2050 | ||
180e9526 | 2051 | ret = qcow2_create2(filename, size, backing_file, backing_fmt, flags, |
1bd0e2d1 | 2052 | cluster_size, prealloc, opts, version, &local_err); |
84d18f06 | 2053 | if (local_err) { |
3ef6c40a HR |
2054 | error_propagate(errp, local_err); |
2055 | } | |
1bd0e2d1 CL |
2056 | |
2057 | finish: | |
2058 | g_free(backing_file); | |
2059 | g_free(backing_fmt); | |
2060 | g_free(buf); | |
3ef6c40a | 2061 | return ret; |
de5f3f40 KW |
2062 | } |
2063 | ||
621f0589 | 2064 | static coroutine_fn int qcow2_co_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 2065 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
621f0589 KW |
2066 | { |
2067 | int ret; | |
2068 | BDRVQcowState *s = bs->opaque; | |
2069 | ||
2070 | /* Emulate misaligned zero writes */ | |
2071 | if (sector_num % s->cluster_sectors || nb_sectors % s->cluster_sectors) { | |
2072 | return -ENOTSUP; | |
2073 | } | |
2074 | ||
2075 | /* Whatever is left can use real zero clusters */ | |
2076 | qemu_co_mutex_lock(&s->lock); | |
2077 | ret = qcow2_zero_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
2078 | nb_sectors); | |
2079 | qemu_co_mutex_unlock(&s->lock); | |
2080 | ||
2081 | return ret; | |
2082 | } | |
2083 | ||
6db39ae2 PB |
2084 | static coroutine_fn int qcow2_co_discard(BlockDriverState *bs, |
2085 | int64_t sector_num, int nb_sectors) | |
5ea929e3 | 2086 | { |
6db39ae2 PB |
2087 | int ret; |
2088 | BDRVQcowState *s = bs->opaque; | |
2089 | ||
2090 | qemu_co_mutex_lock(&s->lock); | |
2091 | ret = qcow2_discard_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
670df5e3 | 2092 | nb_sectors, QCOW2_DISCARD_REQUEST); |
6db39ae2 PB |
2093 | qemu_co_mutex_unlock(&s->lock); |
2094 | return ret; | |
5ea929e3 KW |
2095 | } |
2096 | ||
419b19d9 SH |
2097 | static int qcow2_truncate(BlockDriverState *bs, int64_t offset) |
2098 | { | |
2099 | BDRVQcowState *s = bs->opaque; | |
2cf7cfa1 KW |
2100 | int64_t new_l1_size; |
2101 | int ret; | |
419b19d9 SH |
2102 | |
2103 | if (offset & 511) { | |
259b2173 | 2104 | error_report("The new size must be a multiple of 512"); |
419b19d9 SH |
2105 | return -EINVAL; |
2106 | } | |
2107 | ||
2108 | /* cannot proceed if image has snapshots */ | |
2109 | if (s->nb_snapshots) { | |
259b2173 | 2110 | error_report("Can't resize an image which has snapshots"); |
419b19d9 SH |
2111 | return -ENOTSUP; |
2112 | } | |
2113 | ||
2114 | /* shrinking is currently not supported */ | |
2115 | if (offset < bs->total_sectors * 512) { | |
259b2173 | 2116 | error_report("qcow2 doesn't support shrinking images yet"); |
419b19d9 SH |
2117 | return -ENOTSUP; |
2118 | } | |
2119 | ||
2120 | new_l1_size = size_to_l1(s, offset); | |
72893756 | 2121 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); |
419b19d9 SH |
2122 | if (ret < 0) { |
2123 | return ret; | |
2124 | } | |
2125 | ||
2126 | /* write updated header.size */ | |
2127 | offset = cpu_to_be64(offset); | |
8b3b7206 KW |
2128 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), |
2129 | &offset, sizeof(uint64_t)); | |
419b19d9 SH |
2130 | if (ret < 0) { |
2131 | return ret; | |
2132 | } | |
2133 | ||
2134 | s->l1_vm_state_index = new_l1_size; | |
2135 | return 0; | |
2136 | } | |
2137 | ||
20d97356 BS |
2138 | /* XXX: put compressed sectors first, then all the cluster aligned |
2139 | tables to avoid losing bytes in alignment */ | |
7c80ab3f JS |
2140 | static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num, |
2141 | const uint8_t *buf, int nb_sectors) | |
20d97356 BS |
2142 | { |
2143 | BDRVQcowState *s = bs->opaque; | |
2144 | z_stream strm; | |
2145 | int ret, out_len; | |
2146 | uint8_t *out_buf; | |
2147 | uint64_t cluster_offset; | |
2148 | ||
2149 | if (nb_sectors == 0) { | |
2150 | /* align end of file to a sector boundary to ease reading with | |
2151 | sector based I/Os */ | |
66f82cee | 2152 | cluster_offset = bdrv_getlength(bs->file); |
66f82cee | 2153 | bdrv_truncate(bs->file, cluster_offset); |
20d97356 BS |
2154 | return 0; |
2155 | } | |
2156 | ||
f4d38bef SH |
2157 | if (nb_sectors != s->cluster_sectors) { |
2158 | ret = -EINVAL; | |
2159 | ||
2160 | /* Zero-pad last write if image size is not cluster aligned */ | |
2161 | if (sector_num + nb_sectors == bs->total_sectors && | |
2162 | nb_sectors < s->cluster_sectors) { | |
2163 | uint8_t *pad_buf = qemu_blockalign(bs, s->cluster_size); | |
2164 | memset(pad_buf, 0, s->cluster_size); | |
2165 | memcpy(pad_buf, buf, nb_sectors * BDRV_SECTOR_SIZE); | |
2166 | ret = qcow2_write_compressed(bs, sector_num, | |
2167 | pad_buf, s->cluster_sectors); | |
2168 | qemu_vfree(pad_buf); | |
2169 | } | |
2170 | return ret; | |
2171 | } | |
20d97356 | 2172 | |
7267c094 | 2173 | out_buf = g_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); |
20d97356 BS |
2174 | |
2175 | /* best compression, small window, no zlib header */ | |
2176 | memset(&strm, 0, sizeof(strm)); | |
2177 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
2178 | Z_DEFLATED, -12, | |
2179 | 9, Z_DEFAULT_STRATEGY); | |
2180 | if (ret != 0) { | |
8f1efd00 KW |
2181 | ret = -EINVAL; |
2182 | goto fail; | |
20d97356 BS |
2183 | } |
2184 | ||
2185 | strm.avail_in = s->cluster_size; | |
2186 | strm.next_in = (uint8_t *)buf; | |
2187 | strm.avail_out = s->cluster_size; | |
2188 | strm.next_out = out_buf; | |
2189 | ||
2190 | ret = deflate(&strm, Z_FINISH); | |
2191 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
20d97356 | 2192 | deflateEnd(&strm); |
8f1efd00 KW |
2193 | ret = -EINVAL; |
2194 | goto fail; | |
20d97356 BS |
2195 | } |
2196 | out_len = strm.next_out - out_buf; | |
2197 | ||
2198 | deflateEnd(&strm); | |
2199 | ||
2200 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
2201 | /* could not compress: write normal cluster */ | |
8f1efd00 KW |
2202 | ret = bdrv_write(bs, sector_num, buf, s->cluster_sectors); |
2203 | if (ret < 0) { | |
2204 | goto fail; | |
2205 | } | |
20d97356 BS |
2206 | } else { |
2207 | cluster_offset = qcow2_alloc_compressed_cluster_offset(bs, | |
2208 | sector_num << 9, out_len); | |
8f1efd00 KW |
2209 | if (!cluster_offset) { |
2210 | ret = -EIO; | |
2211 | goto fail; | |
2212 | } | |
20d97356 | 2213 | cluster_offset &= s->cluster_offset_mask; |
cf93980e | 2214 | |
231bb267 | 2215 | ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len); |
cf93980e HR |
2216 | if (ret < 0) { |
2217 | goto fail; | |
2218 | } | |
2219 | ||
66f82cee | 2220 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); |
8f1efd00 KW |
2221 | ret = bdrv_pwrite(bs->file, cluster_offset, out_buf, out_len); |
2222 | if (ret < 0) { | |
2223 | goto fail; | |
20d97356 BS |
2224 | } |
2225 | } | |
2226 | ||
8f1efd00 KW |
2227 | ret = 0; |
2228 | fail: | |
7267c094 | 2229 | g_free(out_buf); |
8f1efd00 | 2230 | return ret; |
20d97356 BS |
2231 | } |
2232 | ||
a968168c | 2233 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
20d97356 | 2234 | { |
29c1a730 KW |
2235 | BDRVQcowState *s = bs->opaque; |
2236 | int ret; | |
2237 | ||
8b94ff85 | 2238 | qemu_co_mutex_lock(&s->lock); |
29c1a730 KW |
2239 | ret = qcow2_cache_flush(bs, s->l2_table_cache); |
2240 | if (ret < 0) { | |
c95de7e2 | 2241 | qemu_co_mutex_unlock(&s->lock); |
8b94ff85 | 2242 | return ret; |
29c1a730 KW |
2243 | } |
2244 | ||
bfe8043e SH |
2245 | if (qcow2_need_accurate_refcounts(s)) { |
2246 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
2247 | if (ret < 0) { | |
2248 | qemu_co_mutex_unlock(&s->lock); | |
2249 | return ret; | |
2250 | } | |
29c1a730 | 2251 | } |
8b94ff85 | 2252 | qemu_co_mutex_unlock(&s->lock); |
29c1a730 | 2253 | |
eb489bb1 KW |
2254 | return 0; |
2255 | } | |
2256 | ||
7c80ab3f | 2257 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
20d97356 BS |
2258 | { |
2259 | BDRVQcowState *s = bs->opaque; | |
95de6d70 PB |
2260 | bdi->unallocated_blocks_are_zero = true; |
2261 | bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3); | |
20d97356 | 2262 | bdi->cluster_size = s->cluster_size; |
7c80ab3f | 2263 | bdi->vm_state_offset = qcow2_vm_state_offset(s); |
20d97356 BS |
2264 | return 0; |
2265 | } | |
2266 | ||
37764dfb HR |
2267 | static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs) |
2268 | { | |
2269 | BDRVQcowState *s = bs->opaque; | |
2270 | ImageInfoSpecific *spec_info = g_new(ImageInfoSpecific, 1); | |
2271 | ||
2272 | *spec_info = (ImageInfoSpecific){ | |
2273 | .kind = IMAGE_INFO_SPECIFIC_KIND_QCOW2, | |
2274 | { | |
2275 | .qcow2 = g_new(ImageInfoSpecificQCow2, 1), | |
2276 | }, | |
2277 | }; | |
2278 | if (s->qcow_version == 2) { | |
2279 | *spec_info->qcow2 = (ImageInfoSpecificQCow2){ | |
2280 | .compat = g_strdup("0.10"), | |
2281 | }; | |
2282 | } else if (s->qcow_version == 3) { | |
2283 | *spec_info->qcow2 = (ImageInfoSpecificQCow2){ | |
2284 | .compat = g_strdup("1.1"), | |
2285 | .lazy_refcounts = s->compatible_features & | |
2286 | QCOW2_COMPAT_LAZY_REFCOUNTS, | |
2287 | .has_lazy_refcounts = true, | |
9009b196 HR |
2288 | .corrupt = s->incompatible_features & |
2289 | QCOW2_INCOMPAT_CORRUPT, | |
2290 | .has_corrupt = true, | |
37764dfb HR |
2291 | }; |
2292 | } | |
2293 | ||
2294 | return spec_info; | |
2295 | } | |
2296 | ||
20d97356 BS |
2297 | #if 0 |
2298 | static void dump_refcounts(BlockDriverState *bs) | |
2299 | { | |
2300 | BDRVQcowState *s = bs->opaque; | |
2301 | int64_t nb_clusters, k, k1, size; | |
2302 | int refcount; | |
2303 | ||
66f82cee | 2304 | size = bdrv_getlength(bs->file); |
20d97356 BS |
2305 | nb_clusters = size_to_clusters(s, size); |
2306 | for(k = 0; k < nb_clusters;) { | |
2307 | k1 = k; | |
2308 | refcount = get_refcount(bs, k); | |
2309 | k++; | |
2310 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
2311 | k++; | |
0bfcd599 BS |
2312 | printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount, |
2313 | k - k1); | |
20d97356 BS |
2314 | } |
2315 | } | |
2316 | #endif | |
2317 | ||
cf8074b3 KW |
2318 | static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
2319 | int64_t pos) | |
20d97356 BS |
2320 | { |
2321 | BDRVQcowState *s = bs->opaque; | |
eedff66f | 2322 | int64_t total_sectors = bs->total_sectors; |
20d97356 | 2323 | int growable = bs->growable; |
6e13610a | 2324 | bool zero_beyond_eof = bs->zero_beyond_eof; |
20d97356 BS |
2325 | int ret; |
2326 | ||
66f82cee | 2327 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
20d97356 | 2328 | bs->growable = 1; |
6e13610a | 2329 | bs->zero_beyond_eof = false; |
8d3b1a2d | 2330 | ret = bdrv_pwritev(bs, qcow2_vm_state_offset(s) + pos, qiov); |
20d97356 | 2331 | bs->growable = growable; |
6e13610a | 2332 | bs->zero_beyond_eof = zero_beyond_eof; |
20d97356 | 2333 | |
eedff66f HR |
2334 | /* bdrv_co_do_writev will have increased the total_sectors value to include |
2335 | * the VM state - the VM state is however not an actual part of the block | |
2336 | * device, therefore, we need to restore the old value. */ | |
2337 | bs->total_sectors = total_sectors; | |
2338 | ||
20d97356 BS |
2339 | return ret; |
2340 | } | |
2341 | ||
7c80ab3f JS |
2342 | static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
2343 | int64_t pos, int size) | |
20d97356 BS |
2344 | { |
2345 | BDRVQcowState *s = bs->opaque; | |
2346 | int growable = bs->growable; | |
0d51b4de | 2347 | bool zero_beyond_eof = bs->zero_beyond_eof; |
20d97356 BS |
2348 | int ret; |
2349 | ||
66f82cee | 2350 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
20d97356 | 2351 | bs->growable = 1; |
0d51b4de | 2352 | bs->zero_beyond_eof = false; |
7c80ab3f | 2353 | ret = bdrv_pread(bs, qcow2_vm_state_offset(s) + pos, buf, size); |
20d97356 | 2354 | bs->growable = growable; |
0d51b4de | 2355 | bs->zero_beyond_eof = zero_beyond_eof; |
20d97356 BS |
2356 | |
2357 | return ret; | |
2358 | } | |
2359 | ||
9296b3ed HR |
2360 | /* |
2361 | * Downgrades an image's version. To achieve this, any incompatible features | |
2362 | * have to be removed. | |
2363 | */ | |
2364 | static int qcow2_downgrade(BlockDriverState *bs, int target_version) | |
2365 | { | |
2366 | BDRVQcowState *s = bs->opaque; | |
2367 | int current_version = s->qcow_version; | |
2368 | int ret; | |
2369 | ||
2370 | if (target_version == current_version) { | |
2371 | return 0; | |
2372 | } else if (target_version > current_version) { | |
2373 | return -EINVAL; | |
2374 | } else if (target_version != 2) { | |
2375 | return -EINVAL; | |
2376 | } | |
2377 | ||
2378 | if (s->refcount_order != 4) { | |
2379 | /* we would have to convert the image to a refcount_order == 4 image | |
2380 | * here; however, since qemu (at the time of writing this) does not | |
2381 | * support anything different than 4 anyway, there is no point in doing | |
2382 | * so right now; however, we should error out (if qemu supports this in | |
2383 | * the future and this code has not been adapted) */ | |
9e3f0892 | 2384 | error_report("qcow2_downgrade: Image refcount orders other than 4 are " |
9296b3ed HR |
2385 | "currently not supported."); |
2386 | return -ENOTSUP; | |
2387 | } | |
2388 | ||
2389 | /* clear incompatible features */ | |
2390 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
2391 | ret = qcow2_mark_clean(bs); | |
2392 | if (ret < 0) { | |
2393 | return ret; | |
2394 | } | |
2395 | } | |
2396 | ||
2397 | /* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in | |
2398 | * the first place; if that happens nonetheless, returning -ENOTSUP is the | |
2399 | * best thing to do anyway */ | |
2400 | ||
2401 | if (s->incompatible_features) { | |
2402 | return -ENOTSUP; | |
2403 | } | |
2404 | ||
2405 | /* since we can ignore compatible features, we can set them to 0 as well */ | |
2406 | s->compatible_features = 0; | |
2407 | /* if lazy refcounts have been used, they have already been fixed through | |
2408 | * clearing the dirty flag */ | |
2409 | ||
2410 | /* clearing autoclear features is trivial */ | |
2411 | s->autoclear_features = 0; | |
2412 | ||
2413 | ret = qcow2_expand_zero_clusters(bs); | |
2414 | if (ret < 0) { | |
2415 | return ret; | |
2416 | } | |
2417 | ||
2418 | s->qcow_version = target_version; | |
2419 | ret = qcow2_update_header(bs); | |
2420 | if (ret < 0) { | |
2421 | s->qcow_version = current_version; | |
2422 | return ret; | |
2423 | } | |
2424 | return 0; | |
2425 | } | |
2426 | ||
1bd0e2d1 | 2427 | static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts) |
9296b3ed HR |
2428 | { |
2429 | BDRVQcowState *s = bs->opaque; | |
2430 | int old_version = s->qcow_version, new_version = old_version; | |
2431 | uint64_t new_size = 0; | |
2432 | const char *backing_file = NULL, *backing_format = NULL; | |
2433 | bool lazy_refcounts = s->use_lazy_refcounts; | |
1bd0e2d1 CL |
2434 | const char *compat = NULL; |
2435 | uint64_t cluster_size = s->cluster_size; | |
2436 | bool encrypt; | |
9296b3ed | 2437 | int ret; |
1bd0e2d1 | 2438 | QemuOptDesc *desc = opts->list->desc; |
9296b3ed | 2439 | |
1bd0e2d1 CL |
2440 | while (desc && desc->name) { |
2441 | if (!qemu_opt_find(opts, desc->name)) { | |
9296b3ed | 2442 | /* only change explicitly defined options */ |
1bd0e2d1 | 2443 | desc++; |
9296b3ed HR |
2444 | continue; |
2445 | } | |
2446 | ||
1bd0e2d1 CL |
2447 | if (!strcmp(desc->name, "compat")) { |
2448 | compat = qemu_opt_get(opts, "compat"); | |
2449 | if (!compat) { | |
9296b3ed | 2450 | /* preserve default */ |
1bd0e2d1 | 2451 | } else if (!strcmp(compat, "0.10")) { |
9296b3ed | 2452 | new_version = 2; |
1bd0e2d1 | 2453 | } else if (!strcmp(compat, "1.1")) { |
9296b3ed HR |
2454 | new_version = 3; |
2455 | } else { | |
1bd0e2d1 | 2456 | fprintf(stderr, "Unknown compatibility level %s.\n", compat); |
9296b3ed HR |
2457 | return -EINVAL; |
2458 | } | |
1bd0e2d1 | 2459 | } else if (!strcmp(desc->name, "preallocation")) { |
9296b3ed HR |
2460 | fprintf(stderr, "Cannot change preallocation mode.\n"); |
2461 | return -ENOTSUP; | |
1bd0e2d1 CL |
2462 | } else if (!strcmp(desc->name, "size")) { |
2463 | new_size = qemu_opt_get_size(opts, "size", 0); | |
2464 | } else if (!strcmp(desc->name, "backing_file")) { | |
2465 | backing_file = qemu_opt_get(opts, "backing_file"); | |
2466 | } else if (!strcmp(desc->name, "backing_fmt")) { | |
2467 | backing_format = qemu_opt_get(opts, "backing_fmt"); | |
2468 | } else if (!strcmp(desc->name, "encryption")) { | |
2469 | encrypt = qemu_opt_get_bool(opts, "encryption", s->crypt_method); | |
2470 | if (encrypt != !!s->crypt_method) { | |
9296b3ed HR |
2471 | fprintf(stderr, "Changing the encryption flag is not " |
2472 | "supported.\n"); | |
2473 | return -ENOTSUP; | |
2474 | } | |
1bd0e2d1 CL |
2475 | } else if (!strcmp(desc->name, "cluster_size")) { |
2476 | cluster_size = qemu_opt_get_size(opts, "cluster_size", | |
2477 | cluster_size); | |
2478 | if (cluster_size != s->cluster_size) { | |
9296b3ed HR |
2479 | fprintf(stderr, "Changing the cluster size is not " |
2480 | "supported.\n"); | |
2481 | return -ENOTSUP; | |
2482 | } | |
1bd0e2d1 CL |
2483 | } else if (!strcmp(desc->name, "lazy_refcounts")) { |
2484 | lazy_refcounts = qemu_opt_get_bool(opts, "lazy_refcounts", | |
2485 | lazy_refcounts); | |
9296b3ed HR |
2486 | } else { |
2487 | /* if this assertion fails, this probably means a new option was | |
2488 | * added without having it covered here */ | |
2489 | assert(false); | |
2490 | } | |
1bd0e2d1 CL |
2491 | |
2492 | desc++; | |
9296b3ed HR |
2493 | } |
2494 | ||
2495 | if (new_version != old_version) { | |
2496 | if (new_version > old_version) { | |
2497 | /* Upgrade */ | |
2498 | s->qcow_version = new_version; | |
2499 | ret = qcow2_update_header(bs); | |
2500 | if (ret < 0) { | |
2501 | s->qcow_version = old_version; | |
2502 | return ret; | |
2503 | } | |
2504 | } else { | |
2505 | ret = qcow2_downgrade(bs, new_version); | |
2506 | if (ret < 0) { | |
2507 | return ret; | |
2508 | } | |
2509 | } | |
2510 | } | |
2511 | ||
2512 | if (backing_file || backing_format) { | |
2513 | ret = qcow2_change_backing_file(bs, backing_file ?: bs->backing_file, | |
2514 | backing_format ?: bs->backing_format); | |
2515 | if (ret < 0) { | |
2516 | return ret; | |
2517 | } | |
2518 | } | |
2519 | ||
2520 | if (s->use_lazy_refcounts != lazy_refcounts) { | |
2521 | if (lazy_refcounts) { | |
2522 | if (s->qcow_version < 3) { | |
2523 | fprintf(stderr, "Lazy refcounts only supported with compatibility " | |
2524 | "level 1.1 and above (use compat=1.1 or greater)\n"); | |
2525 | return -EINVAL; | |
2526 | } | |
2527 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2528 | ret = qcow2_update_header(bs); | |
2529 | if (ret < 0) { | |
2530 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2531 | return ret; | |
2532 | } | |
2533 | s->use_lazy_refcounts = true; | |
2534 | } else { | |
2535 | /* make image clean first */ | |
2536 | ret = qcow2_mark_clean(bs); | |
2537 | if (ret < 0) { | |
2538 | return ret; | |
2539 | } | |
2540 | /* now disallow lazy refcounts */ | |
2541 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2542 | ret = qcow2_update_header(bs); | |
2543 | if (ret < 0) { | |
2544 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2545 | return ret; | |
2546 | } | |
2547 | s->use_lazy_refcounts = false; | |
2548 | } | |
2549 | } | |
2550 | ||
2551 | if (new_size) { | |
2552 | ret = bdrv_truncate(bs, new_size); | |
2553 | if (ret < 0) { | |
2554 | return ret; | |
2555 | } | |
2556 | } | |
2557 | ||
2558 | return 0; | |
2559 | } | |
2560 | ||
85186ebd HR |
2561 | /* |
2562 | * If offset or size are negative, respectively, they will not be included in | |
2563 | * the BLOCK_IMAGE_CORRUPTED event emitted. | |
2564 | * fatal will be ignored for read-only BDS; corruptions found there will always | |
2565 | * be considered non-fatal. | |
2566 | */ | |
2567 | void qcow2_signal_corruption(BlockDriverState *bs, bool fatal, int64_t offset, | |
2568 | int64_t size, const char *message_format, ...) | |
2569 | { | |
2570 | BDRVQcowState *s = bs->opaque; | |
2571 | char *message; | |
2572 | va_list ap; | |
2573 | ||
2574 | fatal = fatal && !bs->read_only; | |
2575 | ||
2576 | if (s->signaled_corruption && | |
2577 | (!fatal || (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT))) | |
2578 | { | |
2579 | return; | |
2580 | } | |
2581 | ||
2582 | va_start(ap, message_format); | |
2583 | message = g_strdup_vprintf(message_format, ap); | |
2584 | va_end(ap); | |
2585 | ||
2586 | if (fatal) { | |
2587 | fprintf(stderr, "qcow2: Marking image as corrupt: %s; further " | |
2588 | "corruption events will be suppressed\n", message); | |
2589 | } else { | |
2590 | fprintf(stderr, "qcow2: Image is corrupt: %s; further non-fatal " | |
2591 | "corruption events will be suppressed\n", message); | |
2592 | } | |
2593 | ||
2594 | qapi_event_send_block_image_corrupted(bdrv_get_device_name(bs), message, | |
2595 | offset >= 0, offset, size >= 0, size, | |
2596 | fatal, &error_abort); | |
2597 | g_free(message); | |
2598 | ||
2599 | if (fatal) { | |
2600 | qcow2_mark_corrupt(bs); | |
2601 | bs->drv = NULL; /* make BDS unusable */ | |
2602 | } | |
2603 | ||
2604 | s->signaled_corruption = true; | |
2605 | } | |
2606 | ||
1bd0e2d1 CL |
2607 | static QemuOptsList qcow2_create_opts = { |
2608 | .name = "qcow2-create-opts", | |
2609 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_create_opts.head), | |
2610 | .desc = { | |
2611 | { | |
2612 | .name = BLOCK_OPT_SIZE, | |
2613 | .type = QEMU_OPT_SIZE, | |
2614 | .help = "Virtual disk size" | |
2615 | }, | |
2616 | { | |
2617 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
2618 | .type = QEMU_OPT_STRING, | |
2619 | .help = "Compatibility level (0.10 or 1.1)" | |
2620 | }, | |
2621 | { | |
2622 | .name = BLOCK_OPT_BACKING_FILE, | |
2623 | .type = QEMU_OPT_STRING, | |
2624 | .help = "File name of a base image" | |
2625 | }, | |
2626 | { | |
2627 | .name = BLOCK_OPT_BACKING_FMT, | |
2628 | .type = QEMU_OPT_STRING, | |
2629 | .help = "Image format of the base image" | |
2630 | }, | |
2631 | { | |
2632 | .name = BLOCK_OPT_ENCRYPT, | |
2633 | .type = QEMU_OPT_BOOL, | |
2634 | .help = "Encrypt the image", | |
2635 | .def_value_str = "off" | |
2636 | }, | |
2637 | { | |
2638 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
2639 | .type = QEMU_OPT_SIZE, | |
2640 | .help = "qcow2 cluster size", | |
2641 | .def_value_str = stringify(DEFAULT_CLUSTER_SIZE) | |
2642 | }, | |
2643 | { | |
2644 | .name = BLOCK_OPT_PREALLOC, | |
2645 | .type = QEMU_OPT_STRING, | |
0e4271b7 HT |
2646 | .help = "Preallocation mode (allowed values: off, metadata, " |
2647 | "falloc, full)" | |
1bd0e2d1 CL |
2648 | }, |
2649 | { | |
2650 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
2651 | .type = QEMU_OPT_BOOL, | |
2652 | .help = "Postpone refcount updates", | |
2653 | .def_value_str = "off" | |
2654 | }, | |
2655 | { /* end of list */ } | |
2656 | } | |
20d97356 BS |
2657 | }; |
2658 | ||
2659 | static BlockDriver bdrv_qcow2 = { | |
7c80ab3f JS |
2660 | .format_name = "qcow2", |
2661 | .instance_size = sizeof(BDRVQcowState), | |
2662 | .bdrv_probe = qcow2_probe, | |
2663 | .bdrv_open = qcow2_open, | |
2664 | .bdrv_close = qcow2_close, | |
21d82ac9 | 2665 | .bdrv_reopen_prepare = qcow2_reopen_prepare, |
c282e1fd | 2666 | .bdrv_create = qcow2_create, |
3ac21627 | 2667 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 2668 | .bdrv_co_get_block_status = qcow2_co_get_block_status, |
7c80ab3f | 2669 | .bdrv_set_key = qcow2_set_key, |
7c80ab3f | 2670 | |
c68b89ac KW |
2671 | .bdrv_co_readv = qcow2_co_readv, |
2672 | .bdrv_co_writev = qcow2_co_writev, | |
eb489bb1 | 2673 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
419b19d9 | 2674 | |
621f0589 | 2675 | .bdrv_co_write_zeroes = qcow2_co_write_zeroes, |
6db39ae2 | 2676 | .bdrv_co_discard = qcow2_co_discard, |
419b19d9 | 2677 | .bdrv_truncate = qcow2_truncate, |
7c80ab3f | 2678 | .bdrv_write_compressed = qcow2_write_compressed, |
20d97356 BS |
2679 | |
2680 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
2681 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
2682 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
2683 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1bd0e2d1 CL |
2684 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
2685 | .bdrv_get_info = qcow2_get_info, | |
37764dfb | 2686 | .bdrv_get_specific_info = qcow2_get_specific_info, |
20d97356 | 2687 | |
7c80ab3f JS |
2688 | .bdrv_save_vmstate = qcow2_save_vmstate, |
2689 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
20d97356 | 2690 | |
8ee79e70 | 2691 | .supports_backing = true, |
20d97356 BS |
2692 | .bdrv_change_backing_file = qcow2_change_backing_file, |
2693 | ||
d34682cd | 2694 | .bdrv_refresh_limits = qcow2_refresh_limits, |
06d9260f AL |
2695 | .bdrv_invalidate_cache = qcow2_invalidate_cache, |
2696 | ||
1bd0e2d1 CL |
2697 | .create_opts = &qcow2_create_opts, |
2698 | .bdrv_check = qcow2_check, | |
c282e1fd | 2699 | .bdrv_amend_options = qcow2_amend_options, |
20d97356 BS |
2700 | }; |
2701 | ||
5efa9d5a AL |
2702 | static void bdrv_qcow2_init(void) |
2703 | { | |
2704 | bdrv_register(&bdrv_qcow2); | |
2705 | } | |
2706 | ||
2707 | block_init(bdrv_qcow2_init); |