]>
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> |
f7d0fe02 | 28 | #include "block/qcow2.h" |
1de7afc9 | 29 | #include "qemu/error-report.h" |
7b1b5d19 | 30 | #include "qapi/qmp/qerror.h" |
acdfb480 | 31 | #include "qapi/qmp/qbool.h" |
ffeaac9b | 32 | #include "qapi/util.h" |
85186ebd HR |
33 | #include "qapi/qmp/types.h" |
34 | #include "qapi-event.h" | |
3cce16f4 | 35 | #include "trace.h" |
1bd0e2d1 | 36 | #include "qemu/option_int.h" |
585f8587 FB |
37 | |
38 | /* | |
39 | Differences with QCOW: | |
40 | ||
41 | - Support for multiple incremental snapshots. | |
42 | - Memory management by reference counts. | |
43 | - Clusters which have a reference count of one have the bit | |
44 | QCOW_OFLAG_COPIED to optimize write performance. | |
5fafdf24 | 45 | - Size of compressed clusters is stored in sectors to reduce bit usage |
585f8587 FB |
46 | in the cluster offsets. |
47 | - Support for storing additional data (such as the VM state) in the | |
3b46e624 | 48 | snapshots. |
585f8587 FB |
49 | - If a backing store is used, the cluster size is not constrained |
50 | (could be backported to QCOW). | |
51 | - L2 tables have always a size of one cluster. | |
52 | */ | |
53 | ||
9b80ddf3 AL |
54 | |
55 | typedef struct { | |
56 | uint32_t magic; | |
57 | uint32_t len; | |
c4217f64 | 58 | } QEMU_PACKED QCowExtension; |
21d82ac9 | 59 | |
7c80ab3f JS |
60 | #define QCOW2_EXT_MAGIC_END 0 |
61 | #define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA | |
cfcc4c62 | 62 | #define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 |
9b80ddf3 | 63 | |
7c80ab3f | 64 | static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) |
585f8587 FB |
65 | { |
66 | const QCowHeader *cow_header = (const void *)buf; | |
3b46e624 | 67 | |
585f8587 FB |
68 | if (buf_size >= sizeof(QCowHeader) && |
69 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
6744cbab | 70 | be32_to_cpu(cow_header->version) >= 2) |
585f8587 FB |
71 | return 100; |
72 | else | |
73 | return 0; | |
74 | } | |
75 | ||
9b80ddf3 AL |
76 | |
77 | /* | |
78 | * read qcow2 extension and fill bs | |
79 | * start reading from start_offset | |
80 | * finish reading upon magic of value 0 or when end_offset reached | |
81 | * unknown magic is skipped (future extension this version knows nothing about) | |
82 | * return 0 upon success, non-0 otherwise | |
83 | */ | |
7c80ab3f | 84 | static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
3ef6c40a HR |
85 | uint64_t end_offset, void **p_feature_table, |
86 | Error **errp) | |
9b80ddf3 | 87 | { |
ff99129a | 88 | BDRVQcow2State *s = bs->opaque; |
9b80ddf3 AL |
89 | QCowExtension ext; |
90 | uint64_t offset; | |
75bab85c | 91 | int ret; |
9b80ddf3 AL |
92 | |
93 | #ifdef DEBUG_EXT | |
7c80ab3f | 94 | printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); |
9b80ddf3 AL |
95 | #endif |
96 | offset = start_offset; | |
97 | while (offset < end_offset) { | |
98 | ||
99 | #ifdef DEBUG_EXT | |
100 | /* Sanity check */ | |
101 | if (offset > s->cluster_size) | |
7c80ab3f | 102 | printf("qcow2_read_extension: suspicious offset %lu\n", offset); |
9b80ddf3 | 103 | |
9b2260cb | 104 | printf("attempting to read extended header in offset %lu\n", offset); |
9b80ddf3 AL |
105 | #endif |
106 | ||
3ef6c40a HR |
107 | ret = bdrv_pread(bs->file, offset, &ext, sizeof(ext)); |
108 | if (ret < 0) { | |
109 | error_setg_errno(errp, -ret, "qcow2_read_extension: ERROR: " | |
110 | "pread fail from offset %" PRIu64, offset); | |
9b80ddf3 AL |
111 | return 1; |
112 | } | |
113 | be32_to_cpus(&ext.magic); | |
114 | be32_to_cpus(&ext.len); | |
115 | offset += sizeof(ext); | |
116 | #ifdef DEBUG_EXT | |
117 | printf("ext.magic = 0x%x\n", ext.magic); | |
118 | #endif | |
2ebafc85 | 119 | if (offset > end_offset || ext.len > end_offset - offset) { |
3ef6c40a | 120 | error_setg(errp, "Header extension too large"); |
64ca6aee KW |
121 | return -EINVAL; |
122 | } | |
123 | ||
9b80ddf3 | 124 | switch (ext.magic) { |
7c80ab3f | 125 | case QCOW2_EXT_MAGIC_END: |
9b80ddf3 | 126 | return 0; |
f965509c | 127 | |
7c80ab3f | 128 | case QCOW2_EXT_MAGIC_BACKING_FORMAT: |
f965509c | 129 | if (ext.len >= sizeof(bs->backing_format)) { |
521b2b5d HR |
130 | error_setg(errp, "ERROR: ext_backing_format: len=%" PRIu32 |
131 | " too large (>=%zu)", ext.len, | |
132 | sizeof(bs->backing_format)); | |
f965509c AL |
133 | return 2; |
134 | } | |
3ef6c40a HR |
135 | ret = bdrv_pread(bs->file, offset, bs->backing_format, ext.len); |
136 | if (ret < 0) { | |
137 | error_setg_errno(errp, -ret, "ERROR: ext_backing_format: " | |
138 | "Could not read format name"); | |
f965509c | 139 | return 3; |
3ef6c40a | 140 | } |
f965509c | 141 | bs->backing_format[ext.len] = '\0'; |
e4603fe1 | 142 | s->image_backing_format = g_strdup(bs->backing_format); |
f965509c AL |
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 | { | |
ff99129a | 190 | BDRVQcow2State *s = bs->opaque; |
75bab85c KW |
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 | ||
c6bd8c70 MA |
209 | error_setg(errp, QERR_UNKNOWN_BLOCK_FORMAT_FEATURE, |
210 | bdrv_get_device_or_node_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 | 251 | { |
ff99129a | 252 | BDRVQcow2State *s = bs->opaque; |
bfe8043e SH |
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 | { | |
ff99129a | 285 | BDRVQcow2State *s = bs->opaque; |
c61d0004 SH |
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 | { | |
ff99129a | 307 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
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 | { | |
ff99129a | 319 | BDRVQcow2State *s = bs->opaque; |
69c98726 HR |
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 | { | |
ff99129a | 354 | BDRVQcow2State *s = bs->opaque; |
8c7de283 KW |
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 | }, | |
279621c0 AG |
470 | { |
471 | .name = QCOW2_OPT_CACHE_CLEAN_INTERVAL, | |
472 | .type = QEMU_OPT_NUMBER, | |
473 | .help = "Clean unused cache entries after this time (in seconds)", | |
474 | }, | |
74c4510a KW |
475 | { /* end of list */ } |
476 | }, | |
477 | }; | |
478 | ||
4092e99d HR |
479 | static const char *overlap_bool_option_names[QCOW2_OL_MAX_BITNR] = { |
480 | [QCOW2_OL_MAIN_HEADER_BITNR] = QCOW2_OPT_OVERLAP_MAIN_HEADER, | |
481 | [QCOW2_OL_ACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L1, | |
482 | [QCOW2_OL_ACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_ACTIVE_L2, | |
483 | [QCOW2_OL_REFCOUNT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_TABLE, | |
484 | [QCOW2_OL_REFCOUNT_BLOCK_BITNR] = QCOW2_OPT_OVERLAP_REFCOUNT_BLOCK, | |
485 | [QCOW2_OL_SNAPSHOT_TABLE_BITNR] = QCOW2_OPT_OVERLAP_SNAPSHOT_TABLE, | |
486 | [QCOW2_OL_INACTIVE_L1_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L1, | |
487 | [QCOW2_OL_INACTIVE_L2_BITNR] = QCOW2_OPT_OVERLAP_INACTIVE_L2, | |
488 | }; | |
489 | ||
279621c0 AG |
490 | static void cache_clean_timer_cb(void *opaque) |
491 | { | |
492 | BlockDriverState *bs = opaque; | |
ff99129a | 493 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
494 | qcow2_cache_clean_unused(bs, s->l2_table_cache); |
495 | qcow2_cache_clean_unused(bs, s->refcount_block_cache); | |
496 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
497 | (int64_t) s->cache_clean_interval * 1000); | |
498 | } | |
499 | ||
500 | static void cache_clean_timer_init(BlockDriverState *bs, AioContext *context) | |
501 | { | |
ff99129a | 502 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
503 | if (s->cache_clean_interval > 0) { |
504 | s->cache_clean_timer = aio_timer_new(context, QEMU_CLOCK_VIRTUAL, | |
505 | SCALE_MS, cache_clean_timer_cb, | |
506 | bs); | |
507 | timer_mod(s->cache_clean_timer, qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + | |
508 | (int64_t) s->cache_clean_interval * 1000); | |
509 | } | |
510 | } | |
511 | ||
512 | static void cache_clean_timer_del(BlockDriverState *bs) | |
513 | { | |
ff99129a | 514 | BDRVQcow2State *s = bs->opaque; |
279621c0 AG |
515 | if (s->cache_clean_timer) { |
516 | timer_del(s->cache_clean_timer); | |
517 | timer_free(s->cache_clean_timer); | |
518 | s->cache_clean_timer = NULL; | |
519 | } | |
520 | } | |
521 | ||
522 | static void qcow2_detach_aio_context(BlockDriverState *bs) | |
523 | { | |
524 | cache_clean_timer_del(bs); | |
525 | } | |
526 | ||
527 | static void qcow2_attach_aio_context(BlockDriverState *bs, | |
528 | AioContext *new_context) | |
529 | { | |
530 | cache_clean_timer_init(bs, new_context); | |
531 | } | |
532 | ||
bc85ef26 HR |
533 | static void read_cache_sizes(BlockDriverState *bs, QemuOpts *opts, |
534 | uint64_t *l2_cache_size, | |
6c1c8d5d HR |
535 | uint64_t *refcount_cache_size, Error **errp) |
536 | { | |
ff99129a | 537 | BDRVQcow2State *s = bs->opaque; |
6c1c8d5d HR |
538 | uint64_t combined_cache_size; |
539 | bool l2_cache_size_set, refcount_cache_size_set, combined_cache_size_set; | |
540 | ||
541 | combined_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_CACHE_SIZE); | |
542 | l2_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_L2_CACHE_SIZE); | |
543 | refcount_cache_size_set = qemu_opt_get(opts, QCOW2_OPT_REFCOUNT_CACHE_SIZE); | |
544 | ||
545 | combined_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_CACHE_SIZE, 0); | |
546 | *l2_cache_size = qemu_opt_get_size(opts, QCOW2_OPT_L2_CACHE_SIZE, 0); | |
547 | *refcount_cache_size = qemu_opt_get_size(opts, | |
548 | QCOW2_OPT_REFCOUNT_CACHE_SIZE, 0); | |
549 | ||
550 | if (combined_cache_size_set) { | |
551 | if (l2_cache_size_set && refcount_cache_size_set) { | |
552 | error_setg(errp, QCOW2_OPT_CACHE_SIZE ", " QCOW2_OPT_L2_CACHE_SIZE | |
553 | " and " QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not be set " | |
554 | "the same time"); | |
555 | return; | |
556 | } else if (*l2_cache_size > combined_cache_size) { | |
557 | error_setg(errp, QCOW2_OPT_L2_CACHE_SIZE " may not exceed " | |
558 | QCOW2_OPT_CACHE_SIZE); | |
559 | return; | |
560 | } else if (*refcount_cache_size > combined_cache_size) { | |
561 | error_setg(errp, QCOW2_OPT_REFCOUNT_CACHE_SIZE " may not exceed " | |
562 | QCOW2_OPT_CACHE_SIZE); | |
563 | return; | |
564 | } | |
565 | ||
566 | if (l2_cache_size_set) { | |
567 | *refcount_cache_size = combined_cache_size - *l2_cache_size; | |
568 | } else if (refcount_cache_size_set) { | |
569 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
570 | } else { | |
571 | *refcount_cache_size = combined_cache_size | |
572 | / (DEFAULT_L2_REFCOUNT_SIZE_RATIO + 1); | |
573 | *l2_cache_size = combined_cache_size - *refcount_cache_size; | |
574 | } | |
575 | } else { | |
576 | if (!l2_cache_size_set && !refcount_cache_size_set) { | |
bc85ef26 HR |
577 | *l2_cache_size = MAX(DEFAULT_L2_CACHE_BYTE_SIZE, |
578 | (uint64_t)DEFAULT_L2_CACHE_CLUSTERS | |
579 | * s->cluster_size); | |
6c1c8d5d HR |
580 | *refcount_cache_size = *l2_cache_size |
581 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
582 | } else if (!l2_cache_size_set) { | |
583 | *l2_cache_size = *refcount_cache_size | |
584 | * DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
585 | } else if (!refcount_cache_size_set) { | |
586 | *refcount_cache_size = *l2_cache_size | |
587 | / DEFAULT_L2_REFCOUNT_SIZE_RATIO; | |
588 | } | |
589 | } | |
590 | } | |
591 | ||
94edf3fb | 592 | static int qcow2_update_options(BlockDriverState *bs, QDict *options, |
4c75d1a1 KW |
593 | int flags, Error **errp) |
594 | { | |
595 | BDRVQcow2State *s = bs->opaque; | |
94edf3fb | 596 | QemuOpts *opts = NULL; |
4c75d1a1 KW |
597 | const char *opt_overlap_check, *opt_overlap_check_template; |
598 | int overlap_check_template = 0; | |
94edf3fb KW |
599 | uint64_t l2_cache_size, refcount_cache_size; |
600 | uint64_t cache_clean_interval; | |
4c75d1a1 | 601 | int i; |
94edf3fb | 602 | Error *local_err = NULL; |
4c75d1a1 KW |
603 | int ret; |
604 | ||
94edf3fb KW |
605 | opts = qemu_opts_create(&qcow2_runtime_opts, NULL, 0, &error_abort); |
606 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
607 | if (local_err) { | |
608 | error_propagate(errp, local_err); | |
609 | ret = -EINVAL; | |
610 | goto fail; | |
611 | } | |
612 | ||
613 | /* get L2 table/refcount block cache size from command line options */ | |
614 | read_cache_sizes(bs, opts, &l2_cache_size, &refcount_cache_size, | |
615 | &local_err); | |
616 | if (local_err) { | |
617 | error_propagate(errp, local_err); | |
618 | ret = -EINVAL; | |
619 | goto fail; | |
620 | } | |
621 | ||
622 | l2_cache_size /= s->cluster_size; | |
623 | if (l2_cache_size < MIN_L2_CACHE_SIZE) { | |
624 | l2_cache_size = MIN_L2_CACHE_SIZE; | |
625 | } | |
626 | if (l2_cache_size > INT_MAX) { | |
627 | error_setg(errp, "L2 cache size too big"); | |
628 | ret = -EINVAL; | |
629 | goto fail; | |
630 | } | |
631 | ||
632 | refcount_cache_size /= s->cluster_size; | |
633 | if (refcount_cache_size < MIN_REFCOUNT_CACHE_SIZE) { | |
634 | refcount_cache_size = MIN_REFCOUNT_CACHE_SIZE; | |
635 | } | |
636 | if (refcount_cache_size > INT_MAX) { | |
637 | error_setg(errp, "Refcount cache size too big"); | |
638 | ret = -EINVAL; | |
639 | goto fail; | |
640 | } | |
641 | ||
642 | /* alloc L2 table/refcount block cache */ | |
643 | s->l2_table_cache = qcow2_cache_create(bs, l2_cache_size); | |
644 | s->refcount_block_cache = qcow2_cache_create(bs, refcount_cache_size); | |
645 | if (s->l2_table_cache == NULL || s->refcount_block_cache == NULL) { | |
646 | error_setg(errp, "Could not allocate metadata caches"); | |
647 | ret = -ENOMEM; | |
648 | goto fail; | |
649 | } | |
650 | ||
651 | /* New interval for cache cleanup timer */ | |
652 | cache_clean_interval = | |
653 | qemu_opt_get_number(opts, QCOW2_OPT_CACHE_CLEAN_INTERVAL, 0); | |
654 | if (cache_clean_interval > UINT_MAX) { | |
655 | error_setg(errp, "Cache clean interval too big"); | |
656 | ret = -EINVAL; | |
657 | goto fail; | |
658 | } | |
659 | s->cache_clean_interval = cache_clean_interval; | |
660 | cache_clean_timer_init(bs, bdrv_get_aio_context(bs)); | |
661 | ||
662 | /* Enable lazy_refcounts according to image and command line options */ | |
4c75d1a1 KW |
663 | s->use_lazy_refcounts = qemu_opt_get_bool(opts, QCOW2_OPT_LAZY_REFCOUNTS, |
664 | (s->compatible_features & QCOW2_COMPAT_LAZY_REFCOUNTS)); | |
665 | ||
666 | s->discard_passthrough[QCOW2_DISCARD_NEVER] = false; | |
667 | s->discard_passthrough[QCOW2_DISCARD_ALWAYS] = true; | |
668 | s->discard_passthrough[QCOW2_DISCARD_REQUEST] = | |
669 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_REQUEST, | |
670 | flags & BDRV_O_UNMAP); | |
671 | s->discard_passthrough[QCOW2_DISCARD_SNAPSHOT] = | |
672 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_SNAPSHOT, true); | |
673 | s->discard_passthrough[QCOW2_DISCARD_OTHER] = | |
674 | qemu_opt_get_bool(opts, QCOW2_OPT_DISCARD_OTHER, false); | |
675 | ||
676 | opt_overlap_check = qemu_opt_get(opts, QCOW2_OPT_OVERLAP); | |
677 | opt_overlap_check_template = qemu_opt_get(opts, QCOW2_OPT_OVERLAP_TEMPLATE); | |
678 | if (opt_overlap_check_template && opt_overlap_check && | |
679 | strcmp(opt_overlap_check_template, opt_overlap_check)) | |
680 | { | |
681 | error_setg(errp, "Conflicting values for qcow2 options '" | |
682 | QCOW2_OPT_OVERLAP "' ('%s') and '" QCOW2_OPT_OVERLAP_TEMPLATE | |
683 | "' ('%s')", opt_overlap_check, opt_overlap_check_template); | |
684 | ret = -EINVAL; | |
685 | goto fail; | |
686 | } | |
687 | if (!opt_overlap_check) { | |
688 | opt_overlap_check = opt_overlap_check_template ?: "cached"; | |
689 | } | |
690 | ||
691 | if (!strcmp(opt_overlap_check, "none")) { | |
692 | overlap_check_template = 0; | |
693 | } else if (!strcmp(opt_overlap_check, "constant")) { | |
694 | overlap_check_template = QCOW2_OL_CONSTANT; | |
695 | } else if (!strcmp(opt_overlap_check, "cached")) { | |
696 | overlap_check_template = QCOW2_OL_CACHED; | |
697 | } else if (!strcmp(opt_overlap_check, "all")) { | |
698 | overlap_check_template = QCOW2_OL_ALL; | |
699 | } else { | |
700 | error_setg(errp, "Unsupported value '%s' for qcow2 option " | |
701 | "'overlap-check'. Allowed are any of the following: " | |
702 | "none, constant, cached, all", opt_overlap_check); | |
703 | ret = -EINVAL; | |
704 | goto fail; | |
705 | } | |
706 | ||
707 | s->overlap_check = 0; | |
708 | for (i = 0; i < QCOW2_OL_MAX_BITNR; i++) { | |
709 | /* overlap-check defines a template bitmask, but every flag may be | |
710 | * overwritten through the associated boolean option */ | |
711 | s->overlap_check |= | |
712 | qemu_opt_get_bool(opts, overlap_bool_option_names[i], | |
713 | overlap_check_template & (1 << i)) << i; | |
714 | } | |
715 | ||
716 | if (s->use_lazy_refcounts && s->qcow_version < 3) { | |
717 | error_setg(errp, "Lazy refcounts require a qcow2 image with at least " | |
718 | "qemu 1.1 compatibility level"); | |
719 | ret = -EINVAL; | |
720 | goto fail; | |
721 | } | |
722 | ||
723 | ret = 0; | |
724 | fail: | |
94edf3fb KW |
725 | qemu_opts_del(opts); |
726 | opts = NULL; | |
727 | ||
4c75d1a1 KW |
728 | return ret; |
729 | } | |
730 | ||
015a1036 HR |
731 | static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, |
732 | Error **errp) | |
585f8587 | 733 | { |
ff99129a | 734 | BDRVQcow2State *s = bs->opaque; |
6d33e8e7 KW |
735 | unsigned int len, i; |
736 | int ret = 0; | |
585f8587 | 737 | QCowHeader header; |
74c4510a | 738 | Error *local_err = NULL; |
9b80ddf3 | 739 | uint64_t ext_end; |
2cf7cfa1 | 740 | uint64_t l1_vm_state_index; |
585f8587 | 741 | |
6d85a57e JS |
742 | ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); |
743 | if (ret < 0) { | |
3ef6c40a | 744 | error_setg_errno(errp, -ret, "Could not read qcow2 header"); |
585f8587 | 745 | goto fail; |
6d85a57e | 746 | } |
585f8587 FB |
747 | be32_to_cpus(&header.magic); |
748 | be32_to_cpus(&header.version); | |
749 | be64_to_cpus(&header.backing_file_offset); | |
750 | be32_to_cpus(&header.backing_file_size); | |
751 | be64_to_cpus(&header.size); | |
752 | be32_to_cpus(&header.cluster_bits); | |
753 | be32_to_cpus(&header.crypt_method); | |
754 | be64_to_cpus(&header.l1_table_offset); | |
755 | be32_to_cpus(&header.l1_size); | |
756 | be64_to_cpus(&header.refcount_table_offset); | |
757 | be32_to_cpus(&header.refcount_table_clusters); | |
758 | be64_to_cpus(&header.snapshots_offset); | |
759 | be32_to_cpus(&header.nb_snapshots); | |
3b46e624 | 760 | |
e8cdcec1 | 761 | if (header.magic != QCOW_MAGIC) { |
3ef6c40a | 762 | error_setg(errp, "Image is not in qcow2 format"); |
76abe407 | 763 | ret = -EINVAL; |
585f8587 | 764 | goto fail; |
6d85a57e | 765 | } |
6744cbab | 766 | if (header.version < 2 || header.version > 3) { |
521b2b5d | 767 | report_unsupported(bs, errp, "QCOW version %" PRIu32, header.version); |
6744cbab KW |
768 | ret = -ENOTSUP; |
769 | goto fail; | |
770 | } | |
771 | ||
772 | s->qcow_version = header.version; | |
773 | ||
24342f2c KW |
774 | /* Initialise cluster size */ |
775 | if (header.cluster_bits < MIN_CLUSTER_BITS || | |
776 | header.cluster_bits > MAX_CLUSTER_BITS) { | |
521b2b5d HR |
777 | error_setg(errp, "Unsupported cluster size: 2^%" PRIu32, |
778 | header.cluster_bits); | |
24342f2c KW |
779 | ret = -EINVAL; |
780 | goto fail; | |
781 | } | |
782 | ||
783 | s->cluster_bits = header.cluster_bits; | |
784 | s->cluster_size = 1 << s->cluster_bits; | |
785 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
786 | ||
6744cbab KW |
787 | /* Initialise version 3 header fields */ |
788 | if (header.version == 2) { | |
789 | header.incompatible_features = 0; | |
790 | header.compatible_features = 0; | |
791 | header.autoclear_features = 0; | |
792 | header.refcount_order = 4; | |
793 | header.header_length = 72; | |
794 | } else { | |
795 | be64_to_cpus(&header.incompatible_features); | |
796 | be64_to_cpus(&header.compatible_features); | |
797 | be64_to_cpus(&header.autoclear_features); | |
798 | be32_to_cpus(&header.refcount_order); | |
799 | be32_to_cpus(&header.header_length); | |
24342f2c KW |
800 | |
801 | if (header.header_length < 104) { | |
802 | error_setg(errp, "qcow2 header too short"); | |
803 | ret = -EINVAL; | |
804 | goto fail; | |
805 | } | |
806 | } | |
807 | ||
808 | if (header.header_length > s->cluster_size) { | |
809 | error_setg(errp, "qcow2 header exceeds cluster size"); | |
810 | ret = -EINVAL; | |
811 | goto fail; | |
6744cbab KW |
812 | } |
813 | ||
814 | if (header.header_length > sizeof(header)) { | |
815 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
816 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
817 | ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields, | |
818 | s->unknown_header_fields_size); | |
819 | if (ret < 0) { | |
3ef6c40a HR |
820 | error_setg_errno(errp, -ret, "Could not read unknown qcow2 header " |
821 | "fields"); | |
6744cbab KW |
822 | goto fail; |
823 | } | |
824 | } | |
825 | ||
a1b3955c KW |
826 | if (header.backing_file_offset > s->cluster_size) { |
827 | error_setg(errp, "Invalid backing file offset"); | |
828 | ret = -EINVAL; | |
829 | goto fail; | |
830 | } | |
831 | ||
cfcc4c62 KW |
832 | if (header.backing_file_offset) { |
833 | ext_end = header.backing_file_offset; | |
834 | } else { | |
835 | ext_end = 1 << header.cluster_bits; | |
836 | } | |
837 | ||
6744cbab KW |
838 | /* Handle feature bits */ |
839 | s->incompatible_features = header.incompatible_features; | |
840 | s->compatible_features = header.compatible_features; | |
841 | s->autoclear_features = header.autoclear_features; | |
842 | ||
c61d0004 | 843 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { |
cfcc4c62 KW |
844 | void *feature_table = NULL; |
845 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
3ef6c40a HR |
846 | &feature_table, NULL); |
847 | report_unsupported_feature(bs, errp, feature_table, | |
c61d0004 SH |
848 | s->incompatible_features & |
849 | ~QCOW2_INCOMPAT_MASK); | |
6744cbab | 850 | ret = -ENOTSUP; |
c5a33ee9 | 851 | g_free(feature_table); |
6744cbab KW |
852 | goto fail; |
853 | } | |
854 | ||
69c98726 HR |
855 | if (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT) { |
856 | /* Corrupt images may not be written to unless they are being repaired | |
857 | */ | |
858 | if ((flags & BDRV_O_RDWR) && !(flags & BDRV_O_CHECK)) { | |
3ef6c40a HR |
859 | error_setg(errp, "qcow2: Image is corrupt; cannot be opened " |
860 | "read/write"); | |
69c98726 HR |
861 | ret = -EACCES; |
862 | goto fail; | |
863 | } | |
864 | } | |
865 | ||
6744cbab | 866 | /* Check support for various header values */ |
b72faf9f HR |
867 | if (header.refcount_order > 6) { |
868 | error_setg(errp, "Reference count entry width too large; may not " | |
869 | "exceed 64 bits"); | |
870 | ret = -EINVAL; | |
e8cdcec1 KW |
871 | goto fail; |
872 | } | |
b6481f37 | 873 | s->refcount_order = header.refcount_order; |
346a53df HR |
874 | s->refcount_bits = 1 << s->refcount_order; |
875 | s->refcount_max = UINT64_C(1) << (s->refcount_bits - 1); | |
876 | s->refcount_max += s->refcount_max - 1; | |
6744cbab | 877 | |
6d85a57e | 878 | if (header.crypt_method > QCOW_CRYPT_AES) { |
521b2b5d | 879 | error_setg(errp, "Unsupported encryption method: %" PRIu32, |
3ef6c40a | 880 | header.crypt_method); |
6d85a57e | 881 | ret = -EINVAL; |
585f8587 | 882 | goto fail; |
6d85a57e | 883 | } |
f6fa64f6 DB |
884 | if (!qcrypto_cipher_supports(QCRYPTO_CIPHER_ALG_AES_128)) { |
885 | error_setg(errp, "AES cipher not available"); | |
886 | ret = -EINVAL; | |
887 | goto fail; | |
888 | } | |
585f8587 | 889 | s->crypt_method_header = header.crypt_method; |
6d85a57e | 890 | if (s->crypt_method_header) { |
585f8587 | 891 | bs->encrypted = 1; |
6d85a57e | 892 | } |
24342f2c | 893 | |
585f8587 FB |
894 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ |
895 | s->l2_size = 1 << s->l2_bits; | |
1d13d654 HR |
896 | /* 2^(s->refcount_order - 3) is the refcount width in bytes */ |
897 | s->refcount_block_bits = s->cluster_bits - (s->refcount_order - 3); | |
898 | s->refcount_block_size = 1 << s->refcount_block_bits; | |
585f8587 FB |
899 | bs->total_sectors = header.size / 512; |
900 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
901 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
902 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
5dab2fad | 903 | |
585f8587 | 904 | s->refcount_table_offset = header.refcount_table_offset; |
5fafdf24 | 905 | s->refcount_table_size = |
585f8587 FB |
906 | header.refcount_table_clusters << (s->cluster_bits - 3); |
907 | ||
2b5d5953 | 908 | if (header.refcount_table_clusters > qcow2_max_refcount_clusters(s)) { |
5dab2fad KW |
909 | error_setg(errp, "Reference count table too large"); |
910 | ret = -EINVAL; | |
911 | goto fail; | |
912 | } | |
913 | ||
8c7de283 KW |
914 | ret = validate_table_offset(bs, s->refcount_table_offset, |
915 | s->refcount_table_size, sizeof(uint64_t)); | |
916 | if (ret < 0) { | |
917 | error_setg(errp, "Invalid reference count table offset"); | |
918 | goto fail; | |
919 | } | |
920 | ||
ce48f2f4 KW |
921 | /* Snapshot table offset/length */ |
922 | if (header.nb_snapshots > QCOW_MAX_SNAPSHOTS) { | |
923 | error_setg(errp, "Too many snapshots"); | |
924 | ret = -EINVAL; | |
925 | goto fail; | |
926 | } | |
927 | ||
928 | ret = validate_table_offset(bs, header.snapshots_offset, | |
929 | header.nb_snapshots, | |
930 | sizeof(QCowSnapshotHeader)); | |
931 | if (ret < 0) { | |
932 | error_setg(errp, "Invalid snapshot table offset"); | |
933 | goto fail; | |
934 | } | |
935 | ||
585f8587 | 936 | /* read the level 1 table */ |
87b86e7e | 937 | if (header.l1_size > QCOW_MAX_L1_SIZE / sizeof(uint64_t)) { |
2d51c32c KW |
938 | error_setg(errp, "Active L1 table too large"); |
939 | ret = -EFBIG; | |
940 | goto fail; | |
941 | } | |
585f8587 | 942 | s->l1_size = header.l1_size; |
2cf7cfa1 KW |
943 | |
944 | l1_vm_state_index = size_to_l1(s, header.size); | |
945 | if (l1_vm_state_index > INT_MAX) { | |
3ef6c40a | 946 | error_setg(errp, "Image is too big"); |
2cf7cfa1 KW |
947 | ret = -EFBIG; |
948 | goto fail; | |
949 | } | |
950 | s->l1_vm_state_index = l1_vm_state_index; | |
951 | ||
585f8587 FB |
952 | /* the L1 table must contain at least enough entries to put |
953 | header.size bytes */ | |
6d85a57e | 954 | if (s->l1_size < s->l1_vm_state_index) { |
3ef6c40a | 955 | error_setg(errp, "L1 table is too small"); |
6d85a57e | 956 | ret = -EINVAL; |
585f8587 | 957 | goto fail; |
6d85a57e | 958 | } |
2d51c32c KW |
959 | |
960 | ret = validate_table_offset(bs, header.l1_table_offset, | |
961 | header.l1_size, sizeof(uint64_t)); | |
962 | if (ret < 0) { | |
963 | error_setg(errp, "Invalid L1 table offset"); | |
964 | goto fail; | |
965 | } | |
585f8587 | 966 | s->l1_table_offset = header.l1_table_offset; |
2d51c32c KW |
967 | |
968 | ||
d191d12d | 969 | if (s->l1_size > 0) { |
de82815d | 970 | s->l1_table = qemu_try_blockalign(bs->file, |
d191d12d | 971 | align_offset(s->l1_size * sizeof(uint64_t), 512)); |
de82815d KW |
972 | if (s->l1_table == NULL) { |
973 | error_setg(errp, "Could not allocate L1 table"); | |
974 | ret = -ENOMEM; | |
975 | goto fail; | |
976 | } | |
6d85a57e JS |
977 | ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, |
978 | s->l1_size * sizeof(uint64_t)); | |
979 | if (ret < 0) { | |
3ef6c40a | 980 | error_setg_errno(errp, -ret, "Could not read L1 table"); |
d191d12d | 981 | goto fail; |
6d85a57e | 982 | } |
d191d12d SW |
983 | for(i = 0;i < s->l1_size; i++) { |
984 | be64_to_cpus(&s->l1_table[i]); | |
985 | } | |
585f8587 | 986 | } |
29c1a730 | 987 | |
94edf3fb KW |
988 | /* Parse driver-specific options */ |
989 | ret = qcow2_update_options(bs, options, flags, errp); | |
90efa0ea KW |
990 | if (ret < 0) { |
991 | goto fail; | |
992 | } | |
993 | ||
7267c094 | 994 | s->cluster_cache = g_malloc(s->cluster_size); |
585f8587 | 995 | /* one more sector for decompressed data alignment */ |
de82815d KW |
996 | s->cluster_data = qemu_try_blockalign(bs->file, QCOW_MAX_CRYPT_CLUSTERS |
997 | * s->cluster_size + 512); | |
998 | if (s->cluster_data == NULL) { | |
999 | error_setg(errp, "Could not allocate temporary cluster buffer"); | |
1000 | ret = -ENOMEM; | |
1001 | goto fail; | |
1002 | } | |
1003 | ||
585f8587 | 1004 | s->cluster_cache_offset = -1; |
06d9260f | 1005 | s->flags = flags; |
3b46e624 | 1006 | |
6d85a57e JS |
1007 | ret = qcow2_refcount_init(bs); |
1008 | if (ret != 0) { | |
3ef6c40a | 1009 | error_setg_errno(errp, -ret, "Could not initialize refcount handling"); |
585f8587 | 1010 | goto fail; |
6d85a57e | 1011 | } |
585f8587 | 1012 | |
72cf2d4f | 1013 | QLIST_INIT(&s->cluster_allocs); |
0b919fae | 1014 | QTAILQ_INIT(&s->discards); |
f214978a | 1015 | |
9b80ddf3 | 1016 | /* read qcow2 extensions */ |
3ef6c40a HR |
1017 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL, |
1018 | &local_err)) { | |
1019 | error_propagate(errp, local_err); | |
6d85a57e | 1020 | ret = -EINVAL; |
9b80ddf3 | 1021 | goto fail; |
6d85a57e | 1022 | } |
9b80ddf3 | 1023 | |
585f8587 FB |
1024 | /* read the backing file name */ |
1025 | if (header.backing_file_offset != 0) { | |
1026 | len = header.backing_file_size; | |
9a29e18f | 1027 | if (len > MIN(1023, s->cluster_size - header.backing_file_offset) || |
e729fa6a | 1028 | len >= sizeof(bs->backing_file)) { |
6d33e8e7 KW |
1029 | error_setg(errp, "Backing file name too long"); |
1030 | ret = -EINVAL; | |
1031 | goto fail; | |
6d85a57e JS |
1032 | } |
1033 | ret = bdrv_pread(bs->file, header.backing_file_offset, | |
1034 | bs->backing_file, len); | |
1035 | if (ret < 0) { | |
3ef6c40a | 1036 | error_setg_errno(errp, -ret, "Could not read backing file name"); |
585f8587 | 1037 | goto fail; |
6d85a57e | 1038 | } |
585f8587 | 1039 | bs->backing_file[len] = '\0'; |
e4603fe1 | 1040 | s->image_backing_file = g_strdup(bs->backing_file); |
585f8587 | 1041 | } |
42deb29f | 1042 | |
11b128f4 KW |
1043 | /* Internal snapshots */ |
1044 | s->snapshots_offset = header.snapshots_offset; | |
1045 | s->nb_snapshots = header.nb_snapshots; | |
1046 | ||
42deb29f KW |
1047 | ret = qcow2_read_snapshots(bs); |
1048 | if (ret < 0) { | |
3ef6c40a | 1049 | error_setg_errno(errp, -ret, "Could not read snapshots"); |
585f8587 | 1050 | goto fail; |
6d85a57e | 1051 | } |
585f8587 | 1052 | |
af7b708d | 1053 | /* Clear unknown autoclear feature bits */ |
27eb6c09 | 1054 | if (!bs->read_only && !(flags & BDRV_O_INCOMING) && s->autoclear_features) { |
af7b708d SH |
1055 | s->autoclear_features = 0; |
1056 | ret = qcow2_update_header(bs); | |
1057 | if (ret < 0) { | |
3ef6c40a | 1058 | error_setg_errno(errp, -ret, "Could not update qcow2 header"); |
af7b708d SH |
1059 | goto fail; |
1060 | } | |
1061 | } | |
1062 | ||
68d100e9 KW |
1063 | /* Initialise locks */ |
1064 | qemu_co_mutex_init(&s->lock); | |
1065 | ||
c61d0004 | 1066 | /* Repair image if dirty */ |
27eb6c09 | 1067 | if (!(flags & (BDRV_O_CHECK | BDRV_O_INCOMING)) && !bs->read_only && |
058f8f16 | 1068 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { |
c61d0004 SH |
1069 | BdrvCheckResult result = {0}; |
1070 | ||
5b84106b | 1071 | ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS | BDRV_FIX_LEAKS); |
c61d0004 | 1072 | if (ret < 0) { |
3ef6c40a | 1073 | error_setg_errno(errp, -ret, "Could not repair dirty image"); |
c61d0004 SH |
1074 | goto fail; |
1075 | } | |
1076 | } | |
1077 | ||
585f8587 | 1078 | #ifdef DEBUG_ALLOC |
6cbc3031 PH |
1079 | { |
1080 | BdrvCheckResult result = {0}; | |
b35278f7 | 1081 | qcow2_check_refcounts(bs, &result, 0); |
6cbc3031 | 1082 | } |
585f8587 | 1083 | #endif |
6d85a57e | 1084 | return ret; |
585f8587 FB |
1085 | |
1086 | fail: | |
6744cbab | 1087 | g_free(s->unknown_header_fields); |
75bab85c | 1088 | cleanup_unknown_header_ext(bs); |
ed6ccf0f KW |
1089 | qcow2_free_snapshots(bs); |
1090 | qcow2_refcount_close(bs); | |
de82815d | 1091 | qemu_vfree(s->l1_table); |
cf93980e HR |
1092 | /* else pre-write overlap checks in cache_destroy may crash */ |
1093 | s->l1_table = NULL; | |
279621c0 | 1094 | cache_clean_timer_del(bs); |
29c1a730 KW |
1095 | if (s->l2_table_cache) { |
1096 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
1097 | } | |
c5a33ee9 PJ |
1098 | if (s->refcount_block_cache) { |
1099 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1100 | } | |
7267c094 | 1101 | g_free(s->cluster_cache); |
dea43a65 | 1102 | qemu_vfree(s->cluster_data); |
6d85a57e | 1103 | return ret; |
585f8587 FB |
1104 | } |
1105 | ||
3baca891 | 1106 | static void qcow2_refresh_limits(BlockDriverState *bs, Error **errp) |
d34682cd | 1107 | { |
ff99129a | 1108 | BDRVQcow2State *s = bs->opaque; |
d34682cd KW |
1109 | |
1110 | bs->bl.write_zeroes_alignment = s->cluster_sectors; | |
d34682cd KW |
1111 | } |
1112 | ||
7c80ab3f | 1113 | static int qcow2_set_key(BlockDriverState *bs, const char *key) |
585f8587 | 1114 | { |
ff99129a | 1115 | BDRVQcow2State *s = bs->opaque; |
585f8587 FB |
1116 | uint8_t keybuf[16]; |
1117 | int len, i; | |
f6fa64f6 | 1118 | Error *err = NULL; |
3b46e624 | 1119 | |
585f8587 FB |
1120 | memset(keybuf, 0, 16); |
1121 | len = strlen(key); | |
1122 | if (len > 16) | |
1123 | len = 16; | |
1124 | /* XXX: we could compress the chars to 7 bits to increase | |
1125 | entropy */ | |
1126 | for(i = 0;i < len;i++) { | |
1127 | keybuf[i] = key[i]; | |
1128 | } | |
8336aafa | 1129 | assert(bs->encrypted); |
585f8587 | 1130 | |
f6fa64f6 DB |
1131 | qcrypto_cipher_free(s->cipher); |
1132 | s->cipher = qcrypto_cipher_new( | |
1133 | QCRYPTO_CIPHER_ALG_AES_128, | |
1134 | QCRYPTO_CIPHER_MODE_CBC, | |
1135 | keybuf, G_N_ELEMENTS(keybuf), | |
1136 | &err); | |
1137 | ||
1138 | if (!s->cipher) { | |
1139 | /* XXX would be nice if errors in this method could | |
1140 | * be properly propagate to the caller. Would need | |
1141 | * the bdrv_set_key() API signature to be fixed. */ | |
1142 | error_free(err); | |
585f8587 | 1143 | return -1; |
585f8587 | 1144 | } |
585f8587 FB |
1145 | return 0; |
1146 | } | |
1147 | ||
4c2e5f8f KW |
1148 | /* We have no actual commit/abort logic for qcow2, but we need to write out any |
1149 | * unwritten data if we reopen read-only. */ | |
21d82ac9 JC |
1150 | static int qcow2_reopen_prepare(BDRVReopenState *state, |
1151 | BlockReopenQueue *queue, Error **errp) | |
1152 | { | |
4c2e5f8f KW |
1153 | int ret; |
1154 | ||
1155 | if ((state->flags & BDRV_O_RDWR) == 0) { | |
1156 | ret = bdrv_flush(state->bs); | |
1157 | if (ret < 0) { | |
1158 | return ret; | |
1159 | } | |
1160 | ||
1161 | ret = qcow2_mark_clean(state->bs); | |
1162 | if (ret < 0) { | |
1163 | return ret; | |
1164 | } | |
1165 | } | |
1166 | ||
21d82ac9 JC |
1167 | return 0; |
1168 | } | |
1169 | ||
b6b8a333 | 1170 | static int64_t coroutine_fn qcow2_co_get_block_status(BlockDriverState *bs, |
f8a2e5e3 | 1171 | int64_t sector_num, int nb_sectors, int *pnum) |
585f8587 | 1172 | { |
ff99129a | 1173 | BDRVQcow2State *s = bs->opaque; |
585f8587 | 1174 | uint64_t cluster_offset; |
4bc74be9 PB |
1175 | int index_in_cluster, ret; |
1176 | int64_t status = 0; | |
585f8587 | 1177 | |
095a9c58 | 1178 | *pnum = nb_sectors; |
f8a2e5e3 | 1179 | qemu_co_mutex_lock(&s->lock); |
1c46efaa | 1180 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, pnum, &cluster_offset); |
f8a2e5e3 | 1181 | qemu_co_mutex_unlock(&s->lock); |
1c46efaa | 1182 | if (ret < 0) { |
d663640c | 1183 | return ret; |
1c46efaa | 1184 | } |
095a9c58 | 1185 | |
4bc74be9 | 1186 | if (cluster_offset != 0 && ret != QCOW2_CLUSTER_COMPRESSED && |
f6fa64f6 | 1187 | !s->cipher) { |
4bc74be9 PB |
1188 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
1189 | cluster_offset |= (index_in_cluster << BDRV_SECTOR_BITS); | |
1190 | status |= BDRV_BLOCK_OFFSET_VALID | cluster_offset; | |
1191 | } | |
1192 | if (ret == QCOW2_CLUSTER_ZERO) { | |
1193 | status |= BDRV_BLOCK_ZERO; | |
1194 | } else if (ret != QCOW2_CLUSTER_UNALLOCATED) { | |
1195 | status |= BDRV_BLOCK_DATA; | |
1196 | } | |
1197 | return status; | |
585f8587 FB |
1198 | } |
1199 | ||
a9465922 | 1200 | /* handle reading after the end of the backing file */ |
bd28f835 KW |
1201 | int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, |
1202 | int64_t sector_num, int nb_sectors) | |
a9465922 FB |
1203 | { |
1204 | int n1; | |
1205 | if ((sector_num + nb_sectors) <= bs->total_sectors) | |
1206 | return nb_sectors; | |
1207 | if (sector_num >= bs->total_sectors) | |
1208 | n1 = 0; | |
1209 | else | |
1210 | n1 = bs->total_sectors - sector_num; | |
bd28f835 | 1211 | |
3d9b4925 | 1212 | qemu_iovec_memset(qiov, 512 * n1, 0, 512 * (nb_sectors - n1)); |
bd28f835 | 1213 | |
a9465922 FB |
1214 | return n1; |
1215 | } | |
1216 | ||
a968168c | 1217 | static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num, |
3fc48d09 | 1218 | int remaining_sectors, QEMUIOVector *qiov) |
585f8587 | 1219 | { |
ff99129a | 1220 | BDRVQcow2State *s = bs->opaque; |
a9465922 | 1221 | int index_in_cluster, n1; |
68d100e9 | 1222 | int ret; |
faf575c1 | 1223 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 1224 | uint64_t cluster_offset = 0; |
3fc48d09 FZ |
1225 | uint64_t bytes_done = 0; |
1226 | QEMUIOVector hd_qiov; | |
1227 | uint8_t *cluster_data = NULL; | |
585f8587 | 1228 | |
3fc48d09 FZ |
1229 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1230 | ||
1231 | qemu_co_mutex_lock(&s->lock); | |
1232 | ||
1233 | while (remaining_sectors != 0) { | |
bd28f835 | 1234 | |
5ebaa27e | 1235 | /* prepare next request */ |
3fc48d09 | 1236 | cur_nr_sectors = remaining_sectors; |
f6fa64f6 | 1237 | if (s->cipher) { |
5ebaa27e FZ |
1238 | cur_nr_sectors = MIN(cur_nr_sectors, |
1239 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
585f8587 | 1240 | } |
5ebaa27e | 1241 | |
3fc48d09 | 1242 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, |
5ebaa27e | 1243 | &cur_nr_sectors, &cluster_offset); |
8af36488 | 1244 | if (ret < 0) { |
3fc48d09 | 1245 | goto fail; |
8af36488 | 1246 | } |
bd28f835 | 1247 | |
3fc48d09 | 1248 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
c87c0672 | 1249 | |
3fc48d09 | 1250 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 1251 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e FZ |
1252 | cur_nr_sectors * 512); |
1253 | ||
68d000a3 KW |
1254 | switch (ret) { |
1255 | case QCOW2_CLUSTER_UNALLOCATED: | |
5ebaa27e FZ |
1256 | |
1257 | if (bs->backing_hd) { | |
1258 | /* read from the base image */ | |
3fc48d09 FZ |
1259 | n1 = qcow2_backing_read1(bs->backing_hd, &hd_qiov, |
1260 | sector_num, cur_nr_sectors); | |
5ebaa27e | 1261 | if (n1 > 0) { |
44deba5a KW |
1262 | QEMUIOVector local_qiov; |
1263 | ||
1264 | qemu_iovec_init(&local_qiov, hd_qiov.niov); | |
1265 | qemu_iovec_concat(&local_qiov, &hd_qiov, 0, | |
1266 | n1 * BDRV_SECTOR_SIZE); | |
1267 | ||
5ebaa27e FZ |
1268 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); |
1269 | qemu_co_mutex_unlock(&s->lock); | |
3fc48d09 | 1270 | ret = bdrv_co_readv(bs->backing_hd, sector_num, |
44deba5a | 1271 | n1, &local_qiov); |
5ebaa27e | 1272 | qemu_co_mutex_lock(&s->lock); |
44deba5a KW |
1273 | |
1274 | qemu_iovec_destroy(&local_qiov); | |
1275 | ||
5ebaa27e | 1276 | if (ret < 0) { |
3fc48d09 | 1277 | goto fail; |
5ebaa27e FZ |
1278 | } |
1279 | } | |
1280 | } else { | |
1281 | /* Note: in this case, no need to wait */ | |
3d9b4925 | 1282 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
5ebaa27e | 1283 | } |
68d000a3 KW |
1284 | break; |
1285 | ||
6377af48 | 1286 | case QCOW2_CLUSTER_ZERO: |
3d9b4925 | 1287 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
6377af48 KW |
1288 | break; |
1289 | ||
68d000a3 | 1290 | case QCOW2_CLUSTER_COMPRESSED: |
5ebaa27e FZ |
1291 | /* add AIO support for compressed blocks ? */ |
1292 | ret = qcow2_decompress_cluster(bs, cluster_offset); | |
1293 | if (ret < 0) { | |
3fc48d09 | 1294 | goto fail; |
bd28f835 KW |
1295 | } |
1296 | ||
03396148 | 1297 | qemu_iovec_from_buf(&hd_qiov, 0, |
5ebaa27e | 1298 | s->cluster_cache + index_in_cluster * 512, |
faf575c1 | 1299 | 512 * cur_nr_sectors); |
68d000a3 KW |
1300 | break; |
1301 | ||
1302 | case QCOW2_CLUSTER_NORMAL: | |
5ebaa27e | 1303 | if ((cluster_offset & 511) != 0) { |
3fc48d09 FZ |
1304 | ret = -EIO; |
1305 | goto fail; | |
5ebaa27e | 1306 | } |
bd28f835 | 1307 | |
8336aafa | 1308 | if (bs->encrypted) { |
f6fa64f6 | 1309 | assert(s->cipher); |
8336aafa | 1310 | |
5ebaa27e FZ |
1311 | /* |
1312 | * For encrypted images, read everything into a temporary | |
1313 | * contiguous buffer on which the AES functions can work. | |
1314 | */ | |
3fc48d09 FZ |
1315 | if (!cluster_data) { |
1316 | cluster_data = | |
de82815d KW |
1317 | qemu_try_blockalign(bs->file, QCOW_MAX_CRYPT_CLUSTERS |
1318 | * s->cluster_size); | |
1319 | if (cluster_data == NULL) { | |
1320 | ret = -ENOMEM; | |
1321 | goto fail; | |
1322 | } | |
5ebaa27e FZ |
1323 | } |
1324 | ||
1325 | assert(cur_nr_sectors <= | |
1326 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
3fc48d09 FZ |
1327 | qemu_iovec_reset(&hd_qiov); |
1328 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
1329 | 512 * cur_nr_sectors); |
1330 | } | |
1331 | ||
1332 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
1333 | qemu_co_mutex_unlock(&s->lock); | |
1334 | ret = bdrv_co_readv(bs->file, | |
1335 | (cluster_offset >> 9) + index_in_cluster, | |
3fc48d09 | 1336 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
1337 | qemu_co_mutex_lock(&s->lock); |
1338 | if (ret < 0) { | |
3fc48d09 | 1339 | goto fail; |
5ebaa27e | 1340 | } |
8336aafa | 1341 | if (bs->encrypted) { |
f6fa64f6 DB |
1342 | assert(s->cipher); |
1343 | Error *err = NULL; | |
1344 | if (qcow2_encrypt_sectors(s, sector_num, cluster_data, | |
1345 | cluster_data, cur_nr_sectors, false, | |
1346 | &err) < 0) { | |
1347 | error_free(err); | |
1348 | ret = -EIO; | |
1349 | goto fail; | |
1350 | } | |
03396148 MT |
1351 | qemu_iovec_from_buf(qiov, bytes_done, |
1352 | cluster_data, 512 * cur_nr_sectors); | |
5ebaa27e | 1353 | } |
68d000a3 KW |
1354 | break; |
1355 | ||
1356 | default: | |
1357 | g_assert_not_reached(); | |
1358 | ret = -EIO; | |
1359 | goto fail; | |
faf575c1 | 1360 | } |
f141eafe | 1361 | |
3fc48d09 FZ |
1362 | remaining_sectors -= cur_nr_sectors; |
1363 | sector_num += cur_nr_sectors; | |
1364 | bytes_done += cur_nr_sectors * 512; | |
5ebaa27e | 1365 | } |
3fc48d09 | 1366 | ret = 0; |
faf575c1 | 1367 | |
3fc48d09 | 1368 | fail: |
68d100e9 | 1369 | qemu_co_mutex_unlock(&s->lock); |
42496d62 | 1370 | |
3fc48d09 | 1371 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1372 | qemu_vfree(cluster_data); |
68d100e9 KW |
1373 | |
1374 | return ret; | |
585f8587 FB |
1375 | } |
1376 | ||
a968168c | 1377 | static coroutine_fn int qcow2_co_writev(BlockDriverState *bs, |
3fc48d09 FZ |
1378 | int64_t sector_num, |
1379 | int remaining_sectors, | |
1380 | QEMUIOVector *qiov) | |
585f8587 | 1381 | { |
ff99129a | 1382 | BDRVQcow2State *s = bs->opaque; |
585f8587 | 1383 | int index_in_cluster; |
68d100e9 | 1384 | int ret; |
faf575c1 | 1385 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 1386 | uint64_t cluster_offset; |
3fc48d09 FZ |
1387 | QEMUIOVector hd_qiov; |
1388 | uint64_t bytes_done = 0; | |
1389 | uint8_t *cluster_data = NULL; | |
8d2497c3 | 1390 | QCowL2Meta *l2meta = NULL; |
c2271403 | 1391 | |
3cce16f4 KW |
1392 | trace_qcow2_writev_start_req(qemu_coroutine_self(), sector_num, |
1393 | remaining_sectors); | |
1394 | ||
3fc48d09 FZ |
1395 | qemu_iovec_init(&hd_qiov, qiov->niov); |
1396 | ||
1397 | s->cluster_cache_offset = -1; /* disable compressed cache */ | |
3b46e624 | 1398 | |
3fc48d09 FZ |
1399 | qemu_co_mutex_lock(&s->lock); |
1400 | ||
1401 | while (remaining_sectors != 0) { | |
1402 | ||
f50f88b9 | 1403 | l2meta = NULL; |
cf5c1a23 | 1404 | |
3cce16f4 | 1405 | trace_qcow2_writev_start_part(qemu_coroutine_self()); |
3fc48d09 | 1406 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
16f0587e | 1407 | cur_nr_sectors = remaining_sectors; |
8336aafa | 1408 | if (bs->encrypted && |
16f0587e HT |
1409 | cur_nr_sectors > |
1410 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster) { | |
1411 | cur_nr_sectors = | |
1412 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors - index_in_cluster; | |
5ebaa27e | 1413 | } |
095a9c58 | 1414 | |
3fc48d09 | 1415 | ret = qcow2_alloc_cluster_offset(bs, sector_num << 9, |
16f0587e | 1416 | &cur_nr_sectors, &cluster_offset, &l2meta); |
5ebaa27e | 1417 | if (ret < 0) { |
3fc48d09 | 1418 | goto fail; |
5ebaa27e | 1419 | } |
148da7ea | 1420 | |
5ebaa27e | 1421 | assert((cluster_offset & 511) == 0); |
148da7ea | 1422 | |
3fc48d09 | 1423 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 1424 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e | 1425 | cur_nr_sectors * 512); |
6f5f060b | 1426 | |
8336aafa | 1427 | if (bs->encrypted) { |
f6fa64f6 DB |
1428 | Error *err = NULL; |
1429 | assert(s->cipher); | |
3fc48d09 | 1430 | if (!cluster_data) { |
de82815d KW |
1431 | cluster_data = qemu_try_blockalign(bs->file, |
1432 | QCOW_MAX_CRYPT_CLUSTERS | |
1433 | * s->cluster_size); | |
1434 | if (cluster_data == NULL) { | |
1435 | ret = -ENOMEM; | |
1436 | goto fail; | |
1437 | } | |
5ebaa27e | 1438 | } |
6f5f060b | 1439 | |
3fc48d09 | 1440 | assert(hd_qiov.size <= |
5ebaa27e | 1441 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
d5e6b161 | 1442 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); |
6f5f060b | 1443 | |
f6fa64f6 DB |
1444 | if (qcow2_encrypt_sectors(s, sector_num, cluster_data, |
1445 | cluster_data, cur_nr_sectors, | |
1446 | true, &err) < 0) { | |
1447 | error_free(err); | |
1448 | ret = -EIO; | |
1449 | goto fail; | |
1450 | } | |
6f5f060b | 1451 | |
3fc48d09 FZ |
1452 | qemu_iovec_reset(&hd_qiov); |
1453 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
1454 | cur_nr_sectors * 512); |
1455 | } | |
6f5f060b | 1456 | |
231bb267 | 1457 | ret = qcow2_pre_write_overlap_check(bs, 0, |
cf93980e HR |
1458 | cluster_offset + index_in_cluster * BDRV_SECTOR_SIZE, |
1459 | cur_nr_sectors * BDRV_SECTOR_SIZE); | |
1460 | if (ret < 0) { | |
1461 | goto fail; | |
1462 | } | |
1463 | ||
5ebaa27e | 1464 | qemu_co_mutex_unlock(&s->lock); |
67a7a0eb | 1465 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); |
3cce16f4 KW |
1466 | trace_qcow2_writev_data(qemu_coroutine_self(), |
1467 | (cluster_offset >> 9) + index_in_cluster); | |
5ebaa27e FZ |
1468 | ret = bdrv_co_writev(bs->file, |
1469 | (cluster_offset >> 9) + index_in_cluster, | |
3fc48d09 | 1470 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
1471 | qemu_co_mutex_lock(&s->lock); |
1472 | if (ret < 0) { | |
3fc48d09 | 1473 | goto fail; |
5ebaa27e | 1474 | } |
f141eafe | 1475 | |
88c6588c KW |
1476 | while (l2meta != NULL) { |
1477 | QCowL2Meta *next; | |
1478 | ||
f50f88b9 KW |
1479 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); |
1480 | if (ret < 0) { | |
1481 | goto fail; | |
1482 | } | |
faf575c1 | 1483 | |
4e95314e KW |
1484 | /* Take the request off the list of running requests */ |
1485 | if (l2meta->nb_clusters != 0) { | |
1486 | QLIST_REMOVE(l2meta, next_in_flight); | |
1487 | } | |
1488 | ||
4e95314e | 1489 | qemu_co_queue_restart_all(&l2meta->dependent_requests); |
4e95314e | 1490 | |
88c6588c | 1491 | next = l2meta->next; |
f50f88b9 | 1492 | g_free(l2meta); |
88c6588c | 1493 | l2meta = next; |
f50f88b9 | 1494 | } |
0fa9131a | 1495 | |
3fc48d09 FZ |
1496 | remaining_sectors -= cur_nr_sectors; |
1497 | sector_num += cur_nr_sectors; | |
1498 | bytes_done += cur_nr_sectors * 512; | |
3cce16f4 | 1499 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_nr_sectors); |
5ebaa27e | 1500 | } |
3fc48d09 | 1501 | ret = 0; |
faf575c1 | 1502 | |
3fc48d09 | 1503 | fail: |
4e95314e KW |
1504 | qemu_co_mutex_unlock(&s->lock); |
1505 | ||
88c6588c KW |
1506 | while (l2meta != NULL) { |
1507 | QCowL2Meta *next; | |
1508 | ||
4e95314e KW |
1509 | if (l2meta->nb_clusters != 0) { |
1510 | QLIST_REMOVE(l2meta, next_in_flight); | |
1511 | } | |
1512 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
88c6588c KW |
1513 | |
1514 | next = l2meta->next; | |
cf5c1a23 | 1515 | g_free(l2meta); |
88c6588c | 1516 | l2meta = next; |
cf5c1a23 | 1517 | } |
0fa9131a | 1518 | |
3fc48d09 | 1519 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 1520 | qemu_vfree(cluster_data); |
3cce16f4 | 1521 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
42496d62 | 1522 | |
68d100e9 | 1523 | return ret; |
585f8587 FB |
1524 | } |
1525 | ||
7c80ab3f | 1526 | static void qcow2_close(BlockDriverState *bs) |
585f8587 | 1527 | { |
ff99129a | 1528 | BDRVQcow2State *s = bs->opaque; |
de82815d | 1529 | qemu_vfree(s->l1_table); |
cf93980e HR |
1530 | /* else pre-write overlap checks in cache_destroy may crash */ |
1531 | s->l1_table = NULL; | |
29c1a730 | 1532 | |
27eb6c09 | 1533 | if (!(bs->open_flags & BDRV_O_INCOMING)) { |
3b5e14c7 | 1534 | int ret1, ret2; |
29c1a730 | 1535 | |
3b5e14c7 HR |
1536 | ret1 = qcow2_cache_flush(bs, s->l2_table_cache); |
1537 | ret2 = qcow2_cache_flush(bs, s->refcount_block_cache); | |
1538 | ||
1539 | if (ret1) { | |
1540 | error_report("Failed to flush the L2 table cache: %s", | |
1541 | strerror(-ret1)); | |
1542 | } | |
1543 | if (ret2) { | |
1544 | error_report("Failed to flush the refcount block cache: %s", | |
1545 | strerror(-ret2)); | |
1546 | } | |
1547 | ||
1548 | if (!ret1 && !ret2) { | |
1549 | qcow2_mark_clean(bs); | |
1550 | } | |
27eb6c09 | 1551 | } |
c61d0004 | 1552 | |
279621c0 | 1553 | cache_clean_timer_del(bs); |
29c1a730 KW |
1554 | qcow2_cache_destroy(bs, s->l2_table_cache); |
1555 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
1556 | ||
f6fa64f6 DB |
1557 | qcrypto_cipher_free(s->cipher); |
1558 | s->cipher = NULL; | |
1559 | ||
6744cbab | 1560 | g_free(s->unknown_header_fields); |
75bab85c | 1561 | cleanup_unknown_header_ext(bs); |
6744cbab | 1562 | |
e4603fe1 KW |
1563 | g_free(s->image_backing_file); |
1564 | g_free(s->image_backing_format); | |
1565 | ||
7267c094 | 1566 | g_free(s->cluster_cache); |
dea43a65 | 1567 | qemu_vfree(s->cluster_data); |
ed6ccf0f | 1568 | qcow2_refcount_close(bs); |
28c1202b | 1569 | qcow2_free_snapshots(bs); |
585f8587 FB |
1570 | } |
1571 | ||
5a8a30db | 1572 | static void qcow2_invalidate_cache(BlockDriverState *bs, Error **errp) |
06d9260f | 1573 | { |
ff99129a | 1574 | BDRVQcow2State *s = bs->opaque; |
06d9260f | 1575 | int flags = s->flags; |
f6fa64f6 | 1576 | QCryptoCipher *cipher = NULL; |
acdfb480 | 1577 | QDict *options; |
5a8a30db KW |
1578 | Error *local_err = NULL; |
1579 | int ret; | |
06d9260f AL |
1580 | |
1581 | /* | |
1582 | * Backing files are read-only which makes all of their metadata immutable, | |
1583 | * that means we don't have to worry about reopening them here. | |
1584 | */ | |
1585 | ||
f6fa64f6 DB |
1586 | cipher = s->cipher; |
1587 | s->cipher = NULL; | |
06d9260f AL |
1588 | |
1589 | qcow2_close(bs); | |
1590 | ||
5a8a30db KW |
1591 | bdrv_invalidate_cache(bs->file, &local_err); |
1592 | if (local_err) { | |
1593 | error_propagate(errp, local_err); | |
1594 | return; | |
1595 | } | |
3456a8d1 | 1596 | |
ff99129a | 1597 | memset(s, 0, sizeof(BDRVQcow2State)); |
d475e5ac | 1598 | options = qdict_clone_shallow(bs->options); |
5a8a30db KW |
1599 | |
1600 | ret = qcow2_open(bs, options, flags, &local_err); | |
a1904e48 | 1601 | QDECREF(options); |
5a8a30db KW |
1602 | if (local_err) { |
1603 | error_setg(errp, "Could not reopen qcow2 layer: %s", | |
1604 | error_get_pretty(local_err)); | |
1605 | error_free(local_err); | |
1606 | return; | |
1607 | } else if (ret < 0) { | |
1608 | error_setg_errno(errp, -ret, "Could not reopen qcow2 layer"); | |
1609 | return; | |
1610 | } | |
acdfb480 | 1611 | |
f6fa64f6 | 1612 | s->cipher = cipher; |
06d9260f AL |
1613 | } |
1614 | ||
e24e49e6 KW |
1615 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
1616 | size_t len, size_t buflen) | |
1617 | { | |
1618 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
1619 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
1620 | ||
1621 | if (buflen < ext_len) { | |
1622 | return -ENOSPC; | |
1623 | } | |
1624 | ||
1625 | *ext_backing_fmt = (QCowExtension) { | |
1626 | .magic = cpu_to_be32(magic), | |
1627 | .len = cpu_to_be32(len), | |
1628 | }; | |
1629 | memcpy(buf + sizeof(QCowExtension), s, len); | |
1630 | ||
1631 | return ext_len; | |
1632 | } | |
1633 | ||
756e6736 | 1634 | /* |
e24e49e6 KW |
1635 | * Updates the qcow2 header, including the variable length parts of it, i.e. |
1636 | * the backing file name and all extensions. qcow2 was not designed to allow | |
1637 | * such changes, so if we run out of space (we can only use the first cluster) | |
1638 | * this function may fail. | |
756e6736 KW |
1639 | * |
1640 | * Returns 0 on success, -errno in error cases. | |
1641 | */ | |
e24e49e6 | 1642 | int qcow2_update_header(BlockDriverState *bs) |
756e6736 | 1643 | { |
ff99129a | 1644 | BDRVQcow2State *s = bs->opaque; |
e24e49e6 KW |
1645 | QCowHeader *header; |
1646 | char *buf; | |
1647 | size_t buflen = s->cluster_size; | |
756e6736 | 1648 | int ret; |
e24e49e6 KW |
1649 | uint64_t total_size; |
1650 | uint32_t refcount_table_clusters; | |
6744cbab | 1651 | size_t header_length; |
75bab85c | 1652 | Qcow2UnknownHeaderExtension *uext; |
756e6736 | 1653 | |
e24e49e6 | 1654 | buf = qemu_blockalign(bs, buflen); |
756e6736 | 1655 | |
e24e49e6 KW |
1656 | /* Header structure */ |
1657 | header = (QCowHeader*) buf; | |
756e6736 | 1658 | |
e24e49e6 KW |
1659 | if (buflen < sizeof(*header)) { |
1660 | ret = -ENOSPC; | |
1661 | goto fail; | |
756e6736 KW |
1662 | } |
1663 | ||
6744cbab | 1664 | header_length = sizeof(*header) + s->unknown_header_fields_size; |
e24e49e6 KW |
1665 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
1666 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
1667 | ||
1668 | *header = (QCowHeader) { | |
6744cbab | 1669 | /* Version 2 fields */ |
e24e49e6 | 1670 | .magic = cpu_to_be32(QCOW_MAGIC), |
6744cbab | 1671 | .version = cpu_to_be32(s->qcow_version), |
e24e49e6 KW |
1672 | .backing_file_offset = 0, |
1673 | .backing_file_size = 0, | |
1674 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
1675 | .size = cpu_to_be64(total_size), | |
1676 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
1677 | .l1_size = cpu_to_be32(s->l1_size), | |
1678 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
1679 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
1680 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
1681 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
1682 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
6744cbab KW |
1683 | |
1684 | /* Version 3 fields */ | |
1685 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
1686 | .compatible_features = cpu_to_be64(s->compatible_features), | |
1687 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
b6481f37 | 1688 | .refcount_order = cpu_to_be32(s->refcount_order), |
6744cbab | 1689 | .header_length = cpu_to_be32(header_length), |
e24e49e6 | 1690 | }; |
756e6736 | 1691 | |
6744cbab KW |
1692 | /* For older versions, write a shorter header */ |
1693 | switch (s->qcow_version) { | |
1694 | case 2: | |
1695 | ret = offsetof(QCowHeader, incompatible_features); | |
1696 | break; | |
1697 | case 3: | |
1698 | ret = sizeof(*header); | |
1699 | break; | |
1700 | default: | |
b6c14762 JM |
1701 | ret = -EINVAL; |
1702 | goto fail; | |
6744cbab KW |
1703 | } |
1704 | ||
1705 | buf += ret; | |
1706 | buflen -= ret; | |
1707 | memset(buf, 0, buflen); | |
1708 | ||
1709 | /* Preserve any unknown field in the header */ | |
1710 | if (s->unknown_header_fields_size) { | |
1711 | if (buflen < s->unknown_header_fields_size) { | |
1712 | ret = -ENOSPC; | |
1713 | goto fail; | |
1714 | } | |
1715 | ||
1716 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
1717 | buf += s->unknown_header_fields_size; | |
1718 | buflen -= s->unknown_header_fields_size; | |
1719 | } | |
756e6736 | 1720 | |
e24e49e6 | 1721 | /* Backing file format header extension */ |
e4603fe1 | 1722 | if (s->image_backing_format) { |
e24e49e6 | 1723 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, |
e4603fe1 KW |
1724 | s->image_backing_format, |
1725 | strlen(s->image_backing_format), | |
e24e49e6 KW |
1726 | buflen); |
1727 | if (ret < 0) { | |
1728 | goto fail; | |
756e6736 KW |
1729 | } |
1730 | ||
e24e49e6 KW |
1731 | buf += ret; |
1732 | buflen -= ret; | |
756e6736 KW |
1733 | } |
1734 | ||
cfcc4c62 KW |
1735 | /* Feature table */ |
1736 | Qcow2Feature features[] = { | |
c61d0004 SH |
1737 | { |
1738 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1739 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
1740 | .name = "dirty bit", | |
1741 | }, | |
69c98726 HR |
1742 | { |
1743 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1744 | .bit = QCOW2_INCOMPAT_CORRUPT_BITNR, | |
1745 | .name = "corrupt bit", | |
1746 | }, | |
bfe8043e SH |
1747 | { |
1748 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
1749 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
1750 | .name = "lazy refcounts", | |
1751 | }, | |
cfcc4c62 KW |
1752 | }; |
1753 | ||
1754 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
1755 | features, sizeof(features), buflen); | |
1756 | if (ret < 0) { | |
1757 | goto fail; | |
1758 | } | |
1759 | buf += ret; | |
1760 | buflen -= ret; | |
1761 | ||
75bab85c KW |
1762 | /* Keep unknown header extensions */ |
1763 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
1764 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
1765 | if (ret < 0) { | |
1766 | goto fail; | |
1767 | } | |
1768 | ||
1769 | buf += ret; | |
1770 | buflen -= ret; | |
1771 | } | |
1772 | ||
e24e49e6 KW |
1773 | /* End of header extensions */ |
1774 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
756e6736 KW |
1775 | if (ret < 0) { |
1776 | goto fail; | |
1777 | } | |
1778 | ||
e24e49e6 KW |
1779 | buf += ret; |
1780 | buflen -= ret; | |
756e6736 | 1781 | |
e24e49e6 | 1782 | /* Backing file name */ |
e4603fe1 KW |
1783 | if (s->image_backing_file) { |
1784 | size_t backing_file_len = strlen(s->image_backing_file); | |
e24e49e6 KW |
1785 | |
1786 | if (buflen < backing_file_len) { | |
1787 | ret = -ENOSPC; | |
1788 | goto fail; | |
1789 | } | |
1790 | ||
00ea1881 | 1791 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ |
e4603fe1 | 1792 | strncpy(buf, s->image_backing_file, buflen); |
e24e49e6 KW |
1793 | |
1794 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
1795 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
756e6736 KW |
1796 | } |
1797 | ||
e24e49e6 KW |
1798 | /* Write the new header */ |
1799 | ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size); | |
756e6736 KW |
1800 | if (ret < 0) { |
1801 | goto fail; | |
1802 | } | |
1803 | ||
1804 | ret = 0; | |
1805 | fail: | |
e24e49e6 | 1806 | qemu_vfree(header); |
756e6736 KW |
1807 | return ret; |
1808 | } | |
1809 | ||
1810 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
1811 | const char *backing_file, const char *backing_fmt) | |
1812 | { | |
ff99129a | 1813 | BDRVQcow2State *s = bs->opaque; |
e4603fe1 | 1814 | |
e24e49e6 KW |
1815 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
1816 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1817 | ||
e4603fe1 KW |
1818 | g_free(s->image_backing_file); |
1819 | g_free(s->image_backing_format); | |
1820 | ||
1821 | s->image_backing_file = backing_file ? g_strdup(bs->backing_file) : NULL; | |
1822 | s->image_backing_format = backing_fmt ? g_strdup(bs->backing_format) : NULL; | |
1823 | ||
e24e49e6 | 1824 | return qcow2_update_header(bs); |
756e6736 KW |
1825 | } |
1826 | ||
a35e1c17 KW |
1827 | static int preallocate(BlockDriverState *bs) |
1828 | { | |
a35e1c17 KW |
1829 | uint64_t nb_sectors; |
1830 | uint64_t offset; | |
060bee89 | 1831 | uint64_t host_offset = 0; |
a35e1c17 | 1832 | int num; |
148da7ea | 1833 | int ret; |
f50f88b9 | 1834 | QCowL2Meta *meta; |
a35e1c17 | 1835 | |
57322b78 | 1836 | nb_sectors = bdrv_nb_sectors(bs); |
a35e1c17 KW |
1837 | offset = 0; |
1838 | ||
1839 | while (nb_sectors) { | |
7c2bbf4a | 1840 | num = MIN(nb_sectors, INT_MAX >> BDRV_SECTOR_BITS); |
16f0587e | 1841 | ret = qcow2_alloc_cluster_offset(bs, offset, &num, |
060bee89 | 1842 | &host_offset, &meta); |
148da7ea | 1843 | if (ret < 0) { |
19dbcbf7 | 1844 | return ret; |
a35e1c17 KW |
1845 | } |
1846 | ||
c792707f SH |
1847 | while (meta) { |
1848 | QCowL2Meta *next = meta->next; | |
1849 | ||
7c2bbf4a HT |
1850 | ret = qcow2_alloc_cluster_link_l2(bs, meta); |
1851 | if (ret < 0) { | |
1852 | qcow2_free_any_clusters(bs, meta->alloc_offset, | |
1853 | meta->nb_clusters, QCOW2_DISCARD_NEVER); | |
1854 | return ret; | |
1855 | } | |
1856 | ||
1857 | /* There are no dependent requests, but we need to remove our | |
1858 | * request from the list of in-flight requests */ | |
4e95314e | 1859 | QLIST_REMOVE(meta, next_in_flight); |
c792707f SH |
1860 | |
1861 | g_free(meta); | |
1862 | meta = next; | |
f50f88b9 | 1863 | } |
f214978a | 1864 | |
a35e1c17 KW |
1865 | /* TODO Preallocate data if requested */ |
1866 | ||
1867 | nb_sectors -= num; | |
7c2bbf4a | 1868 | offset += num << BDRV_SECTOR_BITS; |
a35e1c17 KW |
1869 | } |
1870 | ||
1871 | /* | |
1872 | * It is expected that the image file is large enough to actually contain | |
1873 | * all of the allocated clusters (otherwise we get failing reads after | |
1874 | * EOF). Extend the image to the last allocated sector. | |
1875 | */ | |
060bee89 | 1876 | if (host_offset != 0) { |
7c2bbf4a HT |
1877 | uint8_t buf[BDRV_SECTOR_SIZE]; |
1878 | memset(buf, 0, BDRV_SECTOR_SIZE); | |
1879 | ret = bdrv_write(bs->file, (host_offset >> BDRV_SECTOR_BITS) + num - 1, | |
1880 | buf, 1); | |
19dbcbf7 KW |
1881 | if (ret < 0) { |
1882 | return ret; | |
1883 | } | |
a35e1c17 KW |
1884 | } |
1885 | ||
1886 | return 0; | |
1887 | } | |
1888 | ||
7c80ab3f JS |
1889 | static int qcow2_create2(const char *filename, int64_t total_size, |
1890 | const char *backing_file, const char *backing_format, | |
ffeaac9b | 1891 | int flags, size_t cluster_size, PreallocMode prealloc, |
bd4b167f | 1892 | QemuOpts *opts, int version, int refcount_order, |
3ef6c40a | 1893 | Error **errp) |
a9420734 | 1894 | { |
a9420734 | 1895 | int cluster_bits; |
e6641719 HR |
1896 | QDict *options; |
1897 | ||
1898 | /* Calculate cluster_bits */ | |
786a4ea8 | 1899 | cluster_bits = ctz32(cluster_size); |
a9420734 KW |
1900 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || |
1901 | (1 << cluster_bits) != cluster_size) | |
1902 | { | |
3ef6c40a HR |
1903 | error_setg(errp, "Cluster size must be a power of two between %d and " |
1904 | "%dk", 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); | |
a9420734 KW |
1905 | return -EINVAL; |
1906 | } | |
1907 | ||
1908 | /* | |
1909 | * Open the image file and write a minimal qcow2 header. | |
1910 | * | |
1911 | * We keep things simple and start with a zero-sized image. We also | |
1912 | * do without refcount blocks or a L1 table for now. We'll fix the | |
1913 | * inconsistency later. | |
1914 | * | |
1915 | * We do need a refcount table because growing the refcount table means | |
1916 | * allocating two new refcount blocks - the seconds of which would be at | |
1917 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
1918 | * size for any qcow2 image. | |
1919 | */ | |
1920 | BlockDriverState* bs; | |
f8413b3c | 1921 | QCowHeader *header; |
b106ad91 | 1922 | uint64_t* refcount_table; |
3ef6c40a | 1923 | Error *local_err = NULL; |
a9420734 KW |
1924 | int ret; |
1925 | ||
0e4271b7 | 1926 | if (prealloc == PREALLOC_MODE_FULL || prealloc == PREALLOC_MODE_FALLOC) { |
bd4b167f HR |
1927 | /* Note: The following calculation does not need to be exact; if it is a |
1928 | * bit off, either some bytes will be "leaked" (which is fine) or we | |
1929 | * will need to increase the file size by some bytes (which is fine, | |
1930 | * too, as long as the bulk is allocated here). Therefore, using | |
1931 | * floating point arithmetic is fine. */ | |
0e4271b7 HT |
1932 | int64_t meta_size = 0; |
1933 | uint64_t nreftablee, nrefblocke, nl1e, nl2e; | |
1934 | int64_t aligned_total_size = align_offset(total_size, cluster_size); | |
bd4b167f HR |
1935 | int refblock_bits, refblock_size; |
1936 | /* refcount entry size in bytes */ | |
1937 | double rces = (1 << refcount_order) / 8.; | |
1938 | ||
1939 | /* see qcow2_open() */ | |
1940 | refblock_bits = cluster_bits - (refcount_order - 3); | |
1941 | refblock_size = 1 << refblock_bits; | |
0e4271b7 HT |
1942 | |
1943 | /* header: 1 cluster */ | |
1944 | meta_size += cluster_size; | |
1945 | ||
1946 | /* total size of L2 tables */ | |
1947 | nl2e = aligned_total_size / cluster_size; | |
1948 | nl2e = align_offset(nl2e, cluster_size / sizeof(uint64_t)); | |
1949 | meta_size += nl2e * sizeof(uint64_t); | |
1950 | ||
1951 | /* total size of L1 tables */ | |
1952 | nl1e = nl2e * sizeof(uint64_t) / cluster_size; | |
1953 | nl1e = align_offset(nl1e, cluster_size / sizeof(uint64_t)); | |
1954 | meta_size += nl1e * sizeof(uint64_t); | |
1955 | ||
1956 | /* total size of refcount blocks | |
1957 | * | |
1958 | * note: every host cluster is reference-counted, including metadata | |
1959 | * (even refcount blocks are recursively included). | |
1960 | * Let: | |
1961 | * a = total_size (this is the guest disk size) | |
1962 | * m = meta size not including refcount blocks and refcount tables | |
1963 | * c = cluster size | |
1964 | * y1 = number of refcount blocks entries | |
1965 | * y2 = meta size including everything | |
bd4b167f | 1966 | * rces = refcount entry size in bytes |
0e4271b7 HT |
1967 | * then, |
1968 | * y1 = (y2 + a)/c | |
bd4b167f | 1969 | * y2 = y1 * rces + y1 * rces * sizeof(u64) / c + m |
0e4271b7 | 1970 | * we can get y1: |
bd4b167f | 1971 | * y1 = (a + m) / (c - rces - rces * sizeof(u64) / c) |
0e4271b7 | 1972 | */ |
bd4b167f HR |
1973 | nrefblocke = (aligned_total_size + meta_size + cluster_size) |
1974 | / (cluster_size - rces - rces * sizeof(uint64_t) | |
1975 | / cluster_size); | |
1976 | meta_size += DIV_ROUND_UP(nrefblocke, refblock_size) * cluster_size; | |
0e4271b7 HT |
1977 | |
1978 | /* total size of refcount tables */ | |
bd4b167f | 1979 | nreftablee = nrefblocke / refblock_size; |
0e4271b7 HT |
1980 | nreftablee = align_offset(nreftablee, cluster_size / sizeof(uint64_t)); |
1981 | meta_size += nreftablee * sizeof(uint64_t); | |
1982 | ||
1983 | qemu_opt_set_number(opts, BLOCK_OPT_SIZE, | |
39101f25 | 1984 | aligned_total_size + meta_size, &error_abort); |
f43e47db MA |
1985 | qemu_opt_set(opts, BLOCK_OPT_PREALLOC, PreallocMode_lookup[prealloc], |
1986 | &error_abort); | |
0e4271b7 HT |
1987 | } |
1988 | ||
c282e1fd | 1989 | ret = bdrv_create_file(filename, opts, &local_err); |
a9420734 | 1990 | if (ret < 0) { |
3ef6c40a | 1991 | error_propagate(errp, local_err); |
a9420734 KW |
1992 | return ret; |
1993 | } | |
1994 | ||
2e40134b HR |
1995 | bs = NULL; |
1996 | ret = bdrv_open(&bs, filename, NULL, NULL, BDRV_O_RDWR | BDRV_O_PROTOCOL, | |
6ebf9aa2 | 1997 | &local_err); |
a9420734 | 1998 | if (ret < 0) { |
3ef6c40a | 1999 | error_propagate(errp, local_err); |
a9420734 KW |
2000 | return ret; |
2001 | } | |
2002 | ||
2003 | /* Write the header */ | |
f8413b3c KW |
2004 | QEMU_BUILD_BUG_ON((1 << MIN_CLUSTER_BITS) < sizeof(*header)); |
2005 | header = g_malloc0(cluster_size); | |
2006 | *header = (QCowHeader) { | |
2007 | .magic = cpu_to_be32(QCOW_MAGIC), | |
2008 | .version = cpu_to_be32(version), | |
2009 | .cluster_bits = cpu_to_be32(cluster_bits), | |
2010 | .size = cpu_to_be64(0), | |
2011 | .l1_table_offset = cpu_to_be64(0), | |
2012 | .l1_size = cpu_to_be32(0), | |
2013 | .refcount_table_offset = cpu_to_be64(cluster_size), | |
2014 | .refcount_table_clusters = cpu_to_be32(1), | |
bd4b167f | 2015 | .refcount_order = cpu_to_be32(refcount_order), |
f8413b3c KW |
2016 | .header_length = cpu_to_be32(sizeof(*header)), |
2017 | }; | |
a9420734 KW |
2018 | |
2019 | if (flags & BLOCK_FLAG_ENCRYPT) { | |
f8413b3c | 2020 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_AES); |
a9420734 | 2021 | } else { |
f8413b3c | 2022 | header->crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); |
a9420734 KW |
2023 | } |
2024 | ||
bfe8043e | 2025 | if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) { |
f8413b3c | 2026 | header->compatible_features |= |
bfe8043e SH |
2027 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); |
2028 | } | |
2029 | ||
f8413b3c KW |
2030 | ret = bdrv_pwrite(bs, 0, header, cluster_size); |
2031 | g_free(header); | |
a9420734 | 2032 | if (ret < 0) { |
3ef6c40a | 2033 | error_setg_errno(errp, -ret, "Could not write qcow2 header"); |
a9420734 KW |
2034 | goto out; |
2035 | } | |
2036 | ||
b106ad91 KW |
2037 | /* Write a refcount table with one refcount block */ |
2038 | refcount_table = g_malloc0(2 * cluster_size); | |
2039 | refcount_table[0] = cpu_to_be64(2 * cluster_size); | |
2040 | ret = bdrv_pwrite(bs, cluster_size, refcount_table, 2 * cluster_size); | |
7267c094 | 2041 | g_free(refcount_table); |
a9420734 KW |
2042 | |
2043 | if (ret < 0) { | |
3ef6c40a | 2044 | error_setg_errno(errp, -ret, "Could not write refcount table"); |
a9420734 KW |
2045 | goto out; |
2046 | } | |
2047 | ||
f67503e5 HR |
2048 | bdrv_unref(bs); |
2049 | bs = NULL; | |
a9420734 KW |
2050 | |
2051 | /* | |
2052 | * And now open the image and make it consistent first (i.e. increase the | |
2053 | * refcount of the cluster that is occupied by the header and the refcount | |
2054 | * table) | |
2055 | */ | |
e6641719 HR |
2056 | options = qdict_new(); |
2057 | qdict_put(options, "driver", qstring_from_str("qcow2")); | |
2058 | ret = bdrv_open(&bs, filename, NULL, options, | |
ef810437 | 2059 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_FLUSH, |
6ebf9aa2 | 2060 | &local_err); |
a9420734 | 2061 | if (ret < 0) { |
3ef6c40a | 2062 | error_propagate(errp, local_err); |
a9420734 KW |
2063 | goto out; |
2064 | } | |
2065 | ||
b106ad91 | 2066 | ret = qcow2_alloc_clusters(bs, 3 * cluster_size); |
a9420734 | 2067 | if (ret < 0) { |
3ef6c40a HR |
2068 | error_setg_errno(errp, -ret, "Could not allocate clusters for qcow2 " |
2069 | "header and refcount table"); | |
a9420734 KW |
2070 | goto out; |
2071 | ||
2072 | } else if (ret != 0) { | |
2073 | error_report("Huh, first cluster in empty image is already in use?"); | |
2074 | abort(); | |
2075 | } | |
2076 | ||
2077 | /* Okay, now that we have a valid image, let's give it the right size */ | |
180e9526 | 2078 | ret = bdrv_truncate(bs, total_size); |
a9420734 | 2079 | if (ret < 0) { |
3ef6c40a | 2080 | error_setg_errno(errp, -ret, "Could not resize image"); |
a9420734 KW |
2081 | goto out; |
2082 | } | |
2083 | ||
2084 | /* Want a backing file? There you go.*/ | |
2085 | if (backing_file) { | |
2086 | ret = bdrv_change_backing_file(bs, backing_file, backing_format); | |
2087 | if (ret < 0) { | |
3ef6c40a HR |
2088 | error_setg_errno(errp, -ret, "Could not assign backing file '%s' " |
2089 | "with format '%s'", backing_file, backing_format); | |
a9420734 KW |
2090 | goto out; |
2091 | } | |
2092 | } | |
2093 | ||
2094 | /* And if we're supposed to preallocate metadata, do that now */ | |
0e4271b7 | 2095 | if (prealloc != PREALLOC_MODE_OFF) { |
ff99129a | 2096 | BDRVQcow2State *s = bs->opaque; |
15552c4a | 2097 | qemu_co_mutex_lock(&s->lock); |
a9420734 | 2098 | ret = preallocate(bs); |
15552c4a | 2099 | qemu_co_mutex_unlock(&s->lock); |
a9420734 | 2100 | if (ret < 0) { |
3ef6c40a | 2101 | error_setg_errno(errp, -ret, "Could not preallocate metadata"); |
a9420734 KW |
2102 | goto out; |
2103 | } | |
2104 | } | |
2105 | ||
f67503e5 HR |
2106 | bdrv_unref(bs); |
2107 | bs = NULL; | |
ba2ab2f2 HR |
2108 | |
2109 | /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning */ | |
e6641719 HR |
2110 | options = qdict_new(); |
2111 | qdict_put(options, "driver", qstring_from_str("qcow2")); | |
2112 | ret = bdrv_open(&bs, filename, NULL, options, | |
c9fbb99d | 2113 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_BACKING, |
6ebf9aa2 | 2114 | &local_err); |
84d18f06 | 2115 | if (local_err) { |
ba2ab2f2 HR |
2116 | error_propagate(errp, local_err); |
2117 | goto out; | |
2118 | } | |
2119 | ||
a9420734 KW |
2120 | ret = 0; |
2121 | out: | |
f67503e5 HR |
2122 | if (bs) { |
2123 | bdrv_unref(bs); | |
2124 | } | |
a9420734 KW |
2125 | return ret; |
2126 | } | |
de5f3f40 | 2127 | |
1bd0e2d1 | 2128 | static int qcow2_create(const char *filename, QemuOpts *opts, Error **errp) |
de5f3f40 | 2129 | { |
1bd0e2d1 CL |
2130 | char *backing_file = NULL; |
2131 | char *backing_fmt = NULL; | |
2132 | char *buf = NULL; | |
180e9526 | 2133 | uint64_t size = 0; |
de5f3f40 | 2134 | int flags = 0; |
99cce9fa | 2135 | size_t cluster_size = DEFAULT_CLUSTER_SIZE; |
ffeaac9b | 2136 | PreallocMode prealloc; |
8ad1898c | 2137 | int version = 3; |
bd4b167f HR |
2138 | uint64_t refcount_bits = 16; |
2139 | int refcount_order; | |
3ef6c40a HR |
2140 | Error *local_err = NULL; |
2141 | int ret; | |
de5f3f40 KW |
2142 | |
2143 | /* Read out options */ | |
180e9526 HT |
2144 | size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
2145 | BDRV_SECTOR_SIZE); | |
1bd0e2d1 CL |
2146 | backing_file = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE); |
2147 | backing_fmt = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FMT); | |
2148 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_ENCRYPT, false)) { | |
2149 | flags |= BLOCK_FLAG_ENCRYPT; | |
2150 | } | |
2151 | cluster_size = qemu_opt_get_size_del(opts, BLOCK_OPT_CLUSTER_SIZE, | |
2152 | DEFAULT_CLUSTER_SIZE); | |
2153 | buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
ffeaac9b HT |
2154 | prealloc = qapi_enum_parse(PreallocMode_lookup, buf, |
2155 | PREALLOC_MODE_MAX, PREALLOC_MODE_OFF, | |
2156 | &local_err); | |
2157 | if (local_err) { | |
2158 | error_propagate(errp, local_err); | |
1bd0e2d1 CL |
2159 | ret = -EINVAL; |
2160 | goto finish; | |
2161 | } | |
2162 | g_free(buf); | |
2163 | buf = qemu_opt_get_del(opts, BLOCK_OPT_COMPAT_LEVEL); | |
2164 | if (!buf) { | |
2165 | /* keep the default */ | |
2166 | } else if (!strcmp(buf, "0.10")) { | |
2167 | version = 2; | |
2168 | } else if (!strcmp(buf, "1.1")) { | |
2169 | version = 3; | |
2170 | } else { | |
2171 | error_setg(errp, "Invalid compatibility level: '%s'", buf); | |
2172 | ret = -EINVAL; | |
2173 | goto finish; | |
2174 | } | |
2175 | ||
2176 | if (qemu_opt_get_bool_del(opts, BLOCK_OPT_LAZY_REFCOUNTS, false)) { | |
2177 | flags |= BLOCK_FLAG_LAZY_REFCOUNTS; | |
de5f3f40 KW |
2178 | } |
2179 | ||
ffeaac9b | 2180 | if (backing_file && prealloc != PREALLOC_MODE_OFF) { |
3ef6c40a HR |
2181 | error_setg(errp, "Backing file and preallocation cannot be used at " |
2182 | "the same time"); | |
1bd0e2d1 CL |
2183 | ret = -EINVAL; |
2184 | goto finish; | |
de5f3f40 KW |
2185 | } |
2186 | ||
bfe8043e | 2187 | if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { |
3ef6c40a HR |
2188 | error_setg(errp, "Lazy refcounts only supported with compatibility " |
2189 | "level 1.1 and above (use compat=1.1 or greater)"); | |
1bd0e2d1 CL |
2190 | ret = -EINVAL; |
2191 | goto finish; | |
bfe8043e SH |
2192 | } |
2193 | ||
06d05fa7 HR |
2194 | refcount_bits = qemu_opt_get_number_del(opts, BLOCK_OPT_REFCOUNT_BITS, |
2195 | refcount_bits); | |
2196 | if (refcount_bits > 64 || !is_power_of_2(refcount_bits)) { | |
2197 | error_setg(errp, "Refcount width must be a power of two and may not " | |
2198 | "exceed 64 bits"); | |
2199 | ret = -EINVAL; | |
2200 | goto finish; | |
2201 | } | |
2202 | ||
bd4b167f HR |
2203 | if (version < 3 && refcount_bits != 16) { |
2204 | error_setg(errp, "Different refcount widths than 16 bits require " | |
2205 | "compatibility level 1.1 or above (use compat=1.1 or " | |
2206 | "greater)"); | |
2207 | ret = -EINVAL; | |
2208 | goto finish; | |
2209 | } | |
2210 | ||
786a4ea8 | 2211 | refcount_order = ctz32(refcount_bits); |
bd4b167f | 2212 | |
180e9526 | 2213 | ret = qcow2_create2(filename, size, backing_file, backing_fmt, flags, |
bd4b167f HR |
2214 | cluster_size, prealloc, opts, version, refcount_order, |
2215 | &local_err); | |
84d18f06 | 2216 | if (local_err) { |
3ef6c40a HR |
2217 | error_propagate(errp, local_err); |
2218 | } | |
1bd0e2d1 CL |
2219 | |
2220 | finish: | |
2221 | g_free(backing_file); | |
2222 | g_free(backing_fmt); | |
2223 | g_free(buf); | |
3ef6c40a | 2224 | return ret; |
de5f3f40 KW |
2225 | } |
2226 | ||
621f0589 | 2227 | static coroutine_fn int qcow2_co_write_zeroes(BlockDriverState *bs, |
aa7bfbff | 2228 | int64_t sector_num, int nb_sectors, BdrvRequestFlags flags) |
621f0589 KW |
2229 | { |
2230 | int ret; | |
ff99129a | 2231 | BDRVQcow2State *s = bs->opaque; |
621f0589 KW |
2232 | |
2233 | /* Emulate misaligned zero writes */ | |
2234 | if (sector_num % s->cluster_sectors || nb_sectors % s->cluster_sectors) { | |
2235 | return -ENOTSUP; | |
2236 | } | |
2237 | ||
2238 | /* Whatever is left can use real zero clusters */ | |
2239 | qemu_co_mutex_lock(&s->lock); | |
2240 | ret = qcow2_zero_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
2241 | nb_sectors); | |
2242 | qemu_co_mutex_unlock(&s->lock); | |
2243 | ||
2244 | return ret; | |
2245 | } | |
2246 | ||
6db39ae2 PB |
2247 | static coroutine_fn int qcow2_co_discard(BlockDriverState *bs, |
2248 | int64_t sector_num, int nb_sectors) | |
5ea929e3 | 2249 | { |
6db39ae2 | 2250 | int ret; |
ff99129a | 2251 | BDRVQcow2State *s = bs->opaque; |
6db39ae2 PB |
2252 | |
2253 | qemu_co_mutex_lock(&s->lock); | |
2254 | ret = qcow2_discard_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
808c4b6f | 2255 | nb_sectors, QCOW2_DISCARD_REQUEST, false); |
6db39ae2 PB |
2256 | qemu_co_mutex_unlock(&s->lock); |
2257 | return ret; | |
5ea929e3 KW |
2258 | } |
2259 | ||
419b19d9 SH |
2260 | static int qcow2_truncate(BlockDriverState *bs, int64_t offset) |
2261 | { | |
ff99129a | 2262 | BDRVQcow2State *s = bs->opaque; |
2cf7cfa1 KW |
2263 | int64_t new_l1_size; |
2264 | int ret; | |
419b19d9 SH |
2265 | |
2266 | if (offset & 511) { | |
259b2173 | 2267 | error_report("The new size must be a multiple of 512"); |
419b19d9 SH |
2268 | return -EINVAL; |
2269 | } | |
2270 | ||
2271 | /* cannot proceed if image has snapshots */ | |
2272 | if (s->nb_snapshots) { | |
259b2173 | 2273 | error_report("Can't resize an image which has snapshots"); |
419b19d9 SH |
2274 | return -ENOTSUP; |
2275 | } | |
2276 | ||
2277 | /* shrinking is currently not supported */ | |
2278 | if (offset < bs->total_sectors * 512) { | |
259b2173 | 2279 | error_report("qcow2 doesn't support shrinking images yet"); |
419b19d9 SH |
2280 | return -ENOTSUP; |
2281 | } | |
2282 | ||
2283 | new_l1_size = size_to_l1(s, offset); | |
72893756 | 2284 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); |
419b19d9 SH |
2285 | if (ret < 0) { |
2286 | return ret; | |
2287 | } | |
2288 | ||
2289 | /* write updated header.size */ | |
2290 | offset = cpu_to_be64(offset); | |
8b3b7206 KW |
2291 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), |
2292 | &offset, sizeof(uint64_t)); | |
419b19d9 SH |
2293 | if (ret < 0) { |
2294 | return ret; | |
2295 | } | |
2296 | ||
2297 | s->l1_vm_state_index = new_l1_size; | |
2298 | return 0; | |
2299 | } | |
2300 | ||
20d97356 BS |
2301 | /* XXX: put compressed sectors first, then all the cluster aligned |
2302 | tables to avoid losing bytes in alignment */ | |
7c80ab3f JS |
2303 | static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num, |
2304 | const uint8_t *buf, int nb_sectors) | |
20d97356 | 2305 | { |
ff99129a | 2306 | BDRVQcow2State *s = bs->opaque; |
20d97356 BS |
2307 | z_stream strm; |
2308 | int ret, out_len; | |
2309 | uint8_t *out_buf; | |
2310 | uint64_t cluster_offset; | |
2311 | ||
2312 | if (nb_sectors == 0) { | |
2313 | /* align end of file to a sector boundary to ease reading with | |
2314 | sector based I/Os */ | |
66f82cee | 2315 | cluster_offset = bdrv_getlength(bs->file); |
6a69b962 | 2316 | return bdrv_truncate(bs->file, cluster_offset); |
20d97356 BS |
2317 | } |
2318 | ||
f4d38bef SH |
2319 | if (nb_sectors != s->cluster_sectors) { |
2320 | ret = -EINVAL; | |
2321 | ||
2322 | /* Zero-pad last write if image size is not cluster aligned */ | |
2323 | if (sector_num + nb_sectors == bs->total_sectors && | |
2324 | nb_sectors < s->cluster_sectors) { | |
2325 | uint8_t *pad_buf = qemu_blockalign(bs, s->cluster_size); | |
2326 | memset(pad_buf, 0, s->cluster_size); | |
2327 | memcpy(pad_buf, buf, nb_sectors * BDRV_SECTOR_SIZE); | |
2328 | ret = qcow2_write_compressed(bs, sector_num, | |
2329 | pad_buf, s->cluster_sectors); | |
2330 | qemu_vfree(pad_buf); | |
2331 | } | |
2332 | return ret; | |
2333 | } | |
20d97356 | 2334 | |
7267c094 | 2335 | out_buf = g_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); |
20d97356 BS |
2336 | |
2337 | /* best compression, small window, no zlib header */ | |
2338 | memset(&strm, 0, sizeof(strm)); | |
2339 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
2340 | Z_DEFLATED, -12, | |
2341 | 9, Z_DEFAULT_STRATEGY); | |
2342 | if (ret != 0) { | |
8f1efd00 KW |
2343 | ret = -EINVAL; |
2344 | goto fail; | |
20d97356 BS |
2345 | } |
2346 | ||
2347 | strm.avail_in = s->cluster_size; | |
2348 | strm.next_in = (uint8_t *)buf; | |
2349 | strm.avail_out = s->cluster_size; | |
2350 | strm.next_out = out_buf; | |
2351 | ||
2352 | ret = deflate(&strm, Z_FINISH); | |
2353 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
20d97356 | 2354 | deflateEnd(&strm); |
8f1efd00 KW |
2355 | ret = -EINVAL; |
2356 | goto fail; | |
20d97356 BS |
2357 | } |
2358 | out_len = strm.next_out - out_buf; | |
2359 | ||
2360 | deflateEnd(&strm); | |
2361 | ||
2362 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
2363 | /* could not compress: write normal cluster */ | |
8f1efd00 KW |
2364 | ret = bdrv_write(bs, sector_num, buf, s->cluster_sectors); |
2365 | if (ret < 0) { | |
2366 | goto fail; | |
2367 | } | |
20d97356 BS |
2368 | } else { |
2369 | cluster_offset = qcow2_alloc_compressed_cluster_offset(bs, | |
2370 | sector_num << 9, out_len); | |
8f1efd00 KW |
2371 | if (!cluster_offset) { |
2372 | ret = -EIO; | |
2373 | goto fail; | |
2374 | } | |
20d97356 | 2375 | cluster_offset &= s->cluster_offset_mask; |
cf93980e | 2376 | |
231bb267 | 2377 | ret = qcow2_pre_write_overlap_check(bs, 0, cluster_offset, out_len); |
cf93980e HR |
2378 | if (ret < 0) { |
2379 | goto fail; | |
2380 | } | |
2381 | ||
66f82cee | 2382 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); |
8f1efd00 KW |
2383 | ret = bdrv_pwrite(bs->file, cluster_offset, out_buf, out_len); |
2384 | if (ret < 0) { | |
2385 | goto fail; | |
20d97356 BS |
2386 | } |
2387 | } | |
2388 | ||
8f1efd00 KW |
2389 | ret = 0; |
2390 | fail: | |
7267c094 | 2391 | g_free(out_buf); |
8f1efd00 | 2392 | return ret; |
20d97356 BS |
2393 | } |
2394 | ||
94054183 HR |
2395 | static int make_completely_empty(BlockDriverState *bs) |
2396 | { | |
ff99129a | 2397 | BDRVQcow2State *s = bs->opaque; |
94054183 HR |
2398 | int ret, l1_clusters; |
2399 | int64_t offset; | |
2400 | uint64_t *new_reftable = NULL; | |
2401 | uint64_t rt_entry, l1_size2; | |
2402 | struct { | |
2403 | uint64_t l1_offset; | |
2404 | uint64_t reftable_offset; | |
2405 | uint32_t reftable_clusters; | |
2406 | } QEMU_PACKED l1_ofs_rt_ofs_cls; | |
2407 | ||
2408 | ret = qcow2_cache_empty(bs, s->l2_table_cache); | |
2409 | if (ret < 0) { | |
2410 | goto fail; | |
2411 | } | |
2412 | ||
2413 | ret = qcow2_cache_empty(bs, s->refcount_block_cache); | |
2414 | if (ret < 0) { | |
2415 | goto fail; | |
2416 | } | |
2417 | ||
2418 | /* Refcounts will be broken utterly */ | |
2419 | ret = qcow2_mark_dirty(bs); | |
2420 | if (ret < 0) { | |
2421 | goto fail; | |
2422 | } | |
2423 | ||
2424 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
2425 | ||
2426 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
2427 | l1_size2 = (uint64_t)s->l1_size * sizeof(uint64_t); | |
2428 | ||
2429 | /* After this call, neither the in-memory nor the on-disk refcount | |
2430 | * information accurately describe the actual references */ | |
2431 | ||
2432 | ret = bdrv_write_zeroes(bs->file, s->l1_table_offset / BDRV_SECTOR_SIZE, | |
2433 | l1_clusters * s->cluster_sectors, 0); | |
2434 | if (ret < 0) { | |
2435 | goto fail_broken_refcounts; | |
2436 | } | |
2437 | memset(s->l1_table, 0, l1_size2); | |
2438 | ||
2439 | BLKDBG_EVENT(bs->file, BLKDBG_EMPTY_IMAGE_PREPARE); | |
2440 | ||
2441 | /* Overwrite enough clusters at the beginning of the sectors to place | |
2442 | * the refcount table, a refcount block and the L1 table in; this may | |
2443 | * overwrite parts of the existing refcount and L1 table, which is not | |
2444 | * an issue because the dirty flag is set, complete data loss is in fact | |
2445 | * desired and partial data loss is consequently fine as well */ | |
2446 | ret = bdrv_write_zeroes(bs->file, s->cluster_size / BDRV_SECTOR_SIZE, | |
2447 | (2 + l1_clusters) * s->cluster_size / | |
2448 | BDRV_SECTOR_SIZE, 0); | |
2449 | /* This call (even if it failed overall) may have overwritten on-disk | |
2450 | * refcount structures; in that case, the in-memory refcount information | |
2451 | * will probably differ from the on-disk information which makes the BDS | |
2452 | * unusable */ | |
2453 | if (ret < 0) { | |
2454 | goto fail_broken_refcounts; | |
2455 | } | |
2456 | ||
2457 | BLKDBG_EVENT(bs->file, BLKDBG_L1_UPDATE); | |
2458 | BLKDBG_EVENT(bs->file, BLKDBG_REFTABLE_UPDATE); | |
2459 | ||
2460 | /* "Create" an empty reftable (one cluster) directly after the image | |
2461 | * header and an empty L1 table three clusters after the image header; | |
2462 | * the cluster between those two will be used as the first refblock */ | |
2463 | cpu_to_be64w(&l1_ofs_rt_ofs_cls.l1_offset, 3 * s->cluster_size); | |
2464 | cpu_to_be64w(&l1_ofs_rt_ofs_cls.reftable_offset, s->cluster_size); | |
2465 | cpu_to_be32w(&l1_ofs_rt_ofs_cls.reftable_clusters, 1); | |
2466 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, l1_table_offset), | |
2467 | &l1_ofs_rt_ofs_cls, sizeof(l1_ofs_rt_ofs_cls)); | |
2468 | if (ret < 0) { | |
2469 | goto fail_broken_refcounts; | |
2470 | } | |
2471 | ||
2472 | s->l1_table_offset = 3 * s->cluster_size; | |
2473 | ||
2474 | new_reftable = g_try_new0(uint64_t, s->cluster_size / sizeof(uint64_t)); | |
2475 | if (!new_reftable) { | |
2476 | ret = -ENOMEM; | |
2477 | goto fail_broken_refcounts; | |
2478 | } | |
2479 | ||
2480 | s->refcount_table_offset = s->cluster_size; | |
2481 | s->refcount_table_size = s->cluster_size / sizeof(uint64_t); | |
2482 | ||
2483 | g_free(s->refcount_table); | |
2484 | s->refcount_table = new_reftable; | |
2485 | new_reftable = NULL; | |
2486 | ||
2487 | /* Now the in-memory refcount information again corresponds to the on-disk | |
2488 | * information (reftable is empty and no refblocks (the refblock cache is | |
2489 | * empty)); however, this means some clusters (e.g. the image header) are | |
2490 | * referenced, but not refcounted, but the normal qcow2 code assumes that | |
2491 | * the in-memory information is always correct */ | |
2492 | ||
2493 | BLKDBG_EVENT(bs->file, BLKDBG_REFBLOCK_ALLOC); | |
2494 | ||
2495 | /* Enter the first refblock into the reftable */ | |
2496 | rt_entry = cpu_to_be64(2 * s->cluster_size); | |
2497 | ret = bdrv_pwrite_sync(bs->file, s->cluster_size, | |
2498 | &rt_entry, sizeof(rt_entry)); | |
2499 | if (ret < 0) { | |
2500 | goto fail_broken_refcounts; | |
2501 | } | |
2502 | s->refcount_table[0] = 2 * s->cluster_size; | |
2503 | ||
2504 | s->free_cluster_index = 0; | |
2505 | assert(3 + l1_clusters <= s->refcount_block_size); | |
2506 | offset = qcow2_alloc_clusters(bs, 3 * s->cluster_size + l1_size2); | |
2507 | if (offset < 0) { | |
2508 | ret = offset; | |
2509 | goto fail_broken_refcounts; | |
2510 | } else if (offset > 0) { | |
2511 | error_report("First cluster in emptied image is in use"); | |
2512 | abort(); | |
2513 | } | |
2514 | ||
2515 | /* Now finally the in-memory information corresponds to the on-disk | |
2516 | * structures and is correct */ | |
2517 | ret = qcow2_mark_clean(bs); | |
2518 | if (ret < 0) { | |
2519 | goto fail; | |
2520 | } | |
2521 | ||
2522 | ret = bdrv_truncate(bs->file, (3 + l1_clusters) * s->cluster_size); | |
2523 | if (ret < 0) { | |
2524 | goto fail; | |
2525 | } | |
2526 | ||
2527 | return 0; | |
2528 | ||
2529 | fail_broken_refcounts: | |
2530 | /* The BDS is unusable at this point. If we wanted to make it usable, we | |
2531 | * would have to call qcow2_refcount_close(), qcow2_refcount_init(), | |
2532 | * qcow2_check_refcounts(), qcow2_refcount_close() and qcow2_refcount_init() | |
2533 | * again. However, because the functions which could have caused this error | |
2534 | * path to be taken are used by those functions as well, it's very likely | |
2535 | * that that sequence will fail as well. Therefore, just eject the BDS. */ | |
2536 | bs->drv = NULL; | |
2537 | ||
2538 | fail: | |
2539 | g_free(new_reftable); | |
2540 | return ret; | |
2541 | } | |
2542 | ||
491d27e2 HR |
2543 | static int qcow2_make_empty(BlockDriverState *bs) |
2544 | { | |
ff99129a | 2545 | BDRVQcow2State *s = bs->opaque; |
491d27e2 HR |
2546 | uint64_t start_sector; |
2547 | int sector_step = INT_MAX / BDRV_SECTOR_SIZE; | |
94054183 HR |
2548 | int l1_clusters, ret = 0; |
2549 | ||
2550 | l1_clusters = DIV_ROUND_UP(s->l1_size, s->cluster_size / sizeof(uint64_t)); | |
2551 | ||
2552 | if (s->qcow_version >= 3 && !s->snapshots && | |
2553 | 3 + l1_clusters <= s->refcount_block_size) { | |
2554 | /* The following function only works for qcow2 v3 images (it requires | |
2555 | * the dirty flag) and only as long as there are no snapshots (because | |
2556 | * it completely empties the image). Furthermore, the L1 table and three | |
2557 | * additional clusters (image header, refcount table, one refcount | |
2558 | * block) have to fit inside one refcount block. */ | |
2559 | return make_completely_empty(bs); | |
2560 | } | |
491d27e2 | 2561 | |
94054183 HR |
2562 | /* This fallback code simply discards every active cluster; this is slow, |
2563 | * but works in all cases */ | |
491d27e2 HR |
2564 | for (start_sector = 0; start_sector < bs->total_sectors; |
2565 | start_sector += sector_step) | |
2566 | { | |
2567 | /* As this function is generally used after committing an external | |
2568 | * snapshot, QCOW2_DISCARD_SNAPSHOT seems appropriate. Also, the | |
2569 | * default action for this kind of discard is to pass the discard, | |
2570 | * which will ideally result in an actually smaller image file, as | |
2571 | * is probably desired. */ | |
2572 | ret = qcow2_discard_clusters(bs, start_sector * BDRV_SECTOR_SIZE, | |
2573 | MIN(sector_step, | |
2574 | bs->total_sectors - start_sector), | |
2575 | QCOW2_DISCARD_SNAPSHOT, true); | |
2576 | if (ret < 0) { | |
2577 | break; | |
2578 | } | |
2579 | } | |
2580 | ||
2581 | return ret; | |
2582 | } | |
2583 | ||
a968168c | 2584 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
20d97356 | 2585 | { |
ff99129a | 2586 | BDRVQcow2State *s = bs->opaque; |
29c1a730 KW |
2587 | int ret; |
2588 | ||
8b94ff85 | 2589 | qemu_co_mutex_lock(&s->lock); |
29c1a730 KW |
2590 | ret = qcow2_cache_flush(bs, s->l2_table_cache); |
2591 | if (ret < 0) { | |
c95de7e2 | 2592 | qemu_co_mutex_unlock(&s->lock); |
8b94ff85 | 2593 | return ret; |
29c1a730 KW |
2594 | } |
2595 | ||
bfe8043e SH |
2596 | if (qcow2_need_accurate_refcounts(s)) { |
2597 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
2598 | if (ret < 0) { | |
2599 | qemu_co_mutex_unlock(&s->lock); | |
2600 | return ret; | |
2601 | } | |
29c1a730 | 2602 | } |
8b94ff85 | 2603 | qemu_co_mutex_unlock(&s->lock); |
29c1a730 | 2604 | |
eb489bb1 KW |
2605 | return 0; |
2606 | } | |
2607 | ||
7c80ab3f | 2608 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
20d97356 | 2609 | { |
ff99129a | 2610 | BDRVQcow2State *s = bs->opaque; |
95de6d70 PB |
2611 | bdi->unallocated_blocks_are_zero = true; |
2612 | bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3); | |
20d97356 | 2613 | bdi->cluster_size = s->cluster_size; |
7c80ab3f | 2614 | bdi->vm_state_offset = qcow2_vm_state_offset(s); |
20d97356 BS |
2615 | return 0; |
2616 | } | |
2617 | ||
37764dfb HR |
2618 | static ImageInfoSpecific *qcow2_get_specific_info(BlockDriverState *bs) |
2619 | { | |
ff99129a | 2620 | BDRVQcow2State *s = bs->opaque; |
37764dfb HR |
2621 | ImageInfoSpecific *spec_info = g_new(ImageInfoSpecific, 1); |
2622 | ||
2623 | *spec_info = (ImageInfoSpecific){ | |
2624 | .kind = IMAGE_INFO_SPECIFIC_KIND_QCOW2, | |
2625 | { | |
2626 | .qcow2 = g_new(ImageInfoSpecificQCow2, 1), | |
2627 | }, | |
2628 | }; | |
2629 | if (s->qcow_version == 2) { | |
2630 | *spec_info->qcow2 = (ImageInfoSpecificQCow2){ | |
0709c5a1 HR |
2631 | .compat = g_strdup("0.10"), |
2632 | .refcount_bits = s->refcount_bits, | |
37764dfb HR |
2633 | }; |
2634 | } else if (s->qcow_version == 3) { | |
2635 | *spec_info->qcow2 = (ImageInfoSpecificQCow2){ | |
2636 | .compat = g_strdup("1.1"), | |
2637 | .lazy_refcounts = s->compatible_features & | |
2638 | QCOW2_COMPAT_LAZY_REFCOUNTS, | |
2639 | .has_lazy_refcounts = true, | |
9009b196 HR |
2640 | .corrupt = s->incompatible_features & |
2641 | QCOW2_INCOMPAT_CORRUPT, | |
2642 | .has_corrupt = true, | |
0709c5a1 | 2643 | .refcount_bits = s->refcount_bits, |
37764dfb HR |
2644 | }; |
2645 | } | |
2646 | ||
2647 | return spec_info; | |
2648 | } | |
2649 | ||
20d97356 BS |
2650 | #if 0 |
2651 | static void dump_refcounts(BlockDriverState *bs) | |
2652 | { | |
ff99129a | 2653 | BDRVQcow2State *s = bs->opaque; |
20d97356 BS |
2654 | int64_t nb_clusters, k, k1, size; |
2655 | int refcount; | |
2656 | ||
66f82cee | 2657 | size = bdrv_getlength(bs->file); |
20d97356 BS |
2658 | nb_clusters = size_to_clusters(s, size); |
2659 | for(k = 0; k < nb_clusters;) { | |
2660 | k1 = k; | |
2661 | refcount = get_refcount(bs, k); | |
2662 | k++; | |
2663 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
2664 | k++; | |
0bfcd599 BS |
2665 | printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount, |
2666 | k - k1); | |
20d97356 BS |
2667 | } |
2668 | } | |
2669 | #endif | |
2670 | ||
cf8074b3 KW |
2671 | static int qcow2_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, |
2672 | int64_t pos) | |
20d97356 | 2673 | { |
ff99129a | 2674 | BDRVQcow2State *s = bs->opaque; |
eedff66f | 2675 | int64_t total_sectors = bs->total_sectors; |
6e13610a | 2676 | bool zero_beyond_eof = bs->zero_beyond_eof; |
20d97356 BS |
2677 | int ret; |
2678 | ||
66f82cee | 2679 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
6e13610a | 2680 | bs->zero_beyond_eof = false; |
8d3b1a2d | 2681 | ret = bdrv_pwritev(bs, qcow2_vm_state_offset(s) + pos, qiov); |
6e13610a | 2682 | bs->zero_beyond_eof = zero_beyond_eof; |
20d97356 | 2683 | |
eedff66f HR |
2684 | /* bdrv_co_do_writev will have increased the total_sectors value to include |
2685 | * the VM state - the VM state is however not an actual part of the block | |
2686 | * device, therefore, we need to restore the old value. */ | |
2687 | bs->total_sectors = total_sectors; | |
2688 | ||
20d97356 BS |
2689 | return ret; |
2690 | } | |
2691 | ||
7c80ab3f JS |
2692 | static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
2693 | int64_t pos, int size) | |
20d97356 | 2694 | { |
ff99129a | 2695 | BDRVQcow2State *s = bs->opaque; |
0d51b4de | 2696 | bool zero_beyond_eof = bs->zero_beyond_eof; |
20d97356 BS |
2697 | int ret; |
2698 | ||
66f82cee | 2699 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
0d51b4de | 2700 | bs->zero_beyond_eof = false; |
7c80ab3f | 2701 | ret = bdrv_pread(bs, qcow2_vm_state_offset(s) + pos, buf, size); |
0d51b4de | 2702 | bs->zero_beyond_eof = zero_beyond_eof; |
20d97356 BS |
2703 | |
2704 | return ret; | |
2705 | } | |
2706 | ||
9296b3ed HR |
2707 | /* |
2708 | * Downgrades an image's version. To achieve this, any incompatible features | |
2709 | * have to be removed. | |
2710 | */ | |
4057a2b2 HR |
2711 | static int qcow2_downgrade(BlockDriverState *bs, int target_version, |
2712 | BlockDriverAmendStatusCB *status_cb) | |
9296b3ed | 2713 | { |
ff99129a | 2714 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
2715 | int current_version = s->qcow_version; |
2716 | int ret; | |
2717 | ||
2718 | if (target_version == current_version) { | |
2719 | return 0; | |
2720 | } else if (target_version > current_version) { | |
2721 | return -EINVAL; | |
2722 | } else if (target_version != 2) { | |
2723 | return -EINVAL; | |
2724 | } | |
2725 | ||
2726 | if (s->refcount_order != 4) { | |
2727 | /* we would have to convert the image to a refcount_order == 4 image | |
2728 | * here; however, since qemu (at the time of writing this) does not | |
2729 | * support anything different than 4 anyway, there is no point in doing | |
2730 | * so right now; however, we should error out (if qemu supports this in | |
2731 | * the future and this code has not been adapted) */ | |
9e3f0892 | 2732 | error_report("qcow2_downgrade: Image refcount orders other than 4 are " |
9296b3ed HR |
2733 | "currently not supported."); |
2734 | return -ENOTSUP; | |
2735 | } | |
2736 | ||
2737 | /* clear incompatible features */ | |
2738 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
2739 | ret = qcow2_mark_clean(bs); | |
2740 | if (ret < 0) { | |
2741 | return ret; | |
2742 | } | |
2743 | } | |
2744 | ||
2745 | /* with QCOW2_INCOMPAT_CORRUPT, it is pretty much impossible to get here in | |
2746 | * the first place; if that happens nonetheless, returning -ENOTSUP is the | |
2747 | * best thing to do anyway */ | |
2748 | ||
2749 | if (s->incompatible_features) { | |
2750 | return -ENOTSUP; | |
2751 | } | |
2752 | ||
2753 | /* since we can ignore compatible features, we can set them to 0 as well */ | |
2754 | s->compatible_features = 0; | |
2755 | /* if lazy refcounts have been used, they have already been fixed through | |
2756 | * clearing the dirty flag */ | |
2757 | ||
2758 | /* clearing autoclear features is trivial */ | |
2759 | s->autoclear_features = 0; | |
2760 | ||
4057a2b2 | 2761 | ret = qcow2_expand_zero_clusters(bs, status_cb); |
9296b3ed HR |
2762 | if (ret < 0) { |
2763 | return ret; | |
2764 | } | |
2765 | ||
2766 | s->qcow_version = target_version; | |
2767 | ret = qcow2_update_header(bs); | |
2768 | if (ret < 0) { | |
2769 | s->qcow_version = current_version; | |
2770 | return ret; | |
2771 | } | |
2772 | return 0; | |
2773 | } | |
2774 | ||
77485434 HR |
2775 | static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts, |
2776 | BlockDriverAmendStatusCB *status_cb) | |
9296b3ed | 2777 | { |
ff99129a | 2778 | BDRVQcow2State *s = bs->opaque; |
9296b3ed HR |
2779 | int old_version = s->qcow_version, new_version = old_version; |
2780 | uint64_t new_size = 0; | |
2781 | const char *backing_file = NULL, *backing_format = NULL; | |
2782 | bool lazy_refcounts = s->use_lazy_refcounts; | |
1bd0e2d1 CL |
2783 | const char *compat = NULL; |
2784 | uint64_t cluster_size = s->cluster_size; | |
2785 | bool encrypt; | |
9296b3ed | 2786 | int ret; |
1bd0e2d1 | 2787 | QemuOptDesc *desc = opts->list->desc; |
9296b3ed | 2788 | |
1bd0e2d1 CL |
2789 | while (desc && desc->name) { |
2790 | if (!qemu_opt_find(opts, desc->name)) { | |
9296b3ed | 2791 | /* only change explicitly defined options */ |
1bd0e2d1 | 2792 | desc++; |
9296b3ed HR |
2793 | continue; |
2794 | } | |
2795 | ||
8a17b83c HR |
2796 | if (!strcmp(desc->name, BLOCK_OPT_COMPAT_LEVEL)) { |
2797 | compat = qemu_opt_get(opts, BLOCK_OPT_COMPAT_LEVEL); | |
1bd0e2d1 | 2798 | if (!compat) { |
9296b3ed | 2799 | /* preserve default */ |
1bd0e2d1 | 2800 | } else if (!strcmp(compat, "0.10")) { |
9296b3ed | 2801 | new_version = 2; |
1bd0e2d1 | 2802 | } else if (!strcmp(compat, "1.1")) { |
9296b3ed HR |
2803 | new_version = 3; |
2804 | } else { | |
1bd0e2d1 | 2805 | fprintf(stderr, "Unknown compatibility level %s.\n", compat); |
9296b3ed HR |
2806 | return -EINVAL; |
2807 | } | |
8a17b83c | 2808 | } else if (!strcmp(desc->name, BLOCK_OPT_PREALLOC)) { |
9296b3ed HR |
2809 | fprintf(stderr, "Cannot change preallocation mode.\n"); |
2810 | return -ENOTSUP; | |
8a17b83c HR |
2811 | } else if (!strcmp(desc->name, BLOCK_OPT_SIZE)) { |
2812 | new_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0); | |
2813 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FILE)) { | |
2814 | backing_file = qemu_opt_get(opts, BLOCK_OPT_BACKING_FILE); | |
2815 | } else if (!strcmp(desc->name, BLOCK_OPT_BACKING_FMT)) { | |
2816 | backing_format = qemu_opt_get(opts, BLOCK_OPT_BACKING_FMT); | |
2817 | } else if (!strcmp(desc->name, BLOCK_OPT_ENCRYPT)) { | |
2818 | encrypt = qemu_opt_get_bool(opts, BLOCK_OPT_ENCRYPT, | |
f6fa64f6 DB |
2819 | !!s->cipher); |
2820 | ||
2821 | if (encrypt != !!s->cipher) { | |
9296b3ed HR |
2822 | fprintf(stderr, "Changing the encryption flag is not " |
2823 | "supported.\n"); | |
2824 | return -ENOTSUP; | |
2825 | } | |
8a17b83c HR |
2826 | } else if (!strcmp(desc->name, BLOCK_OPT_CLUSTER_SIZE)) { |
2827 | cluster_size = qemu_opt_get_size(opts, BLOCK_OPT_CLUSTER_SIZE, | |
1bd0e2d1 CL |
2828 | cluster_size); |
2829 | if (cluster_size != s->cluster_size) { | |
9296b3ed HR |
2830 | fprintf(stderr, "Changing the cluster size is not " |
2831 | "supported.\n"); | |
2832 | return -ENOTSUP; | |
2833 | } | |
8a17b83c HR |
2834 | } else if (!strcmp(desc->name, BLOCK_OPT_LAZY_REFCOUNTS)) { |
2835 | lazy_refcounts = qemu_opt_get_bool(opts, BLOCK_OPT_LAZY_REFCOUNTS, | |
1bd0e2d1 | 2836 | lazy_refcounts); |
06d05fa7 HR |
2837 | } else if (!strcmp(desc->name, BLOCK_OPT_REFCOUNT_BITS)) { |
2838 | error_report("Cannot change refcount entry width"); | |
2839 | return -ENOTSUP; | |
9296b3ed HR |
2840 | } else { |
2841 | /* if this assertion fails, this probably means a new option was | |
2842 | * added without having it covered here */ | |
2843 | assert(false); | |
2844 | } | |
1bd0e2d1 CL |
2845 | |
2846 | desc++; | |
9296b3ed HR |
2847 | } |
2848 | ||
2849 | if (new_version != old_version) { | |
2850 | if (new_version > old_version) { | |
2851 | /* Upgrade */ | |
2852 | s->qcow_version = new_version; | |
2853 | ret = qcow2_update_header(bs); | |
2854 | if (ret < 0) { | |
2855 | s->qcow_version = old_version; | |
2856 | return ret; | |
2857 | } | |
2858 | } else { | |
4057a2b2 | 2859 | ret = qcow2_downgrade(bs, new_version, status_cb); |
9296b3ed HR |
2860 | if (ret < 0) { |
2861 | return ret; | |
2862 | } | |
2863 | } | |
2864 | } | |
2865 | ||
2866 | if (backing_file || backing_format) { | |
e4603fe1 KW |
2867 | ret = qcow2_change_backing_file(bs, |
2868 | backing_file ?: s->image_backing_file, | |
2869 | backing_format ?: s->image_backing_format); | |
9296b3ed HR |
2870 | if (ret < 0) { |
2871 | return ret; | |
2872 | } | |
2873 | } | |
2874 | ||
2875 | if (s->use_lazy_refcounts != lazy_refcounts) { | |
2876 | if (lazy_refcounts) { | |
2877 | if (s->qcow_version < 3) { | |
2878 | fprintf(stderr, "Lazy refcounts only supported with compatibility " | |
2879 | "level 1.1 and above (use compat=1.1 or greater)\n"); | |
2880 | return -EINVAL; | |
2881 | } | |
2882 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2883 | ret = qcow2_update_header(bs); | |
2884 | if (ret < 0) { | |
2885 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2886 | return ret; | |
2887 | } | |
2888 | s->use_lazy_refcounts = true; | |
2889 | } else { | |
2890 | /* make image clean first */ | |
2891 | ret = qcow2_mark_clean(bs); | |
2892 | if (ret < 0) { | |
2893 | return ret; | |
2894 | } | |
2895 | /* now disallow lazy refcounts */ | |
2896 | s->compatible_features &= ~QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2897 | ret = qcow2_update_header(bs); | |
2898 | if (ret < 0) { | |
2899 | s->compatible_features |= QCOW2_COMPAT_LAZY_REFCOUNTS; | |
2900 | return ret; | |
2901 | } | |
2902 | s->use_lazy_refcounts = false; | |
2903 | } | |
2904 | } | |
2905 | ||
2906 | if (new_size) { | |
2907 | ret = bdrv_truncate(bs, new_size); | |
2908 | if (ret < 0) { | |
2909 | return ret; | |
2910 | } | |
2911 | } | |
2912 | ||
2913 | return 0; | |
2914 | } | |
2915 | ||
85186ebd HR |
2916 | /* |
2917 | * If offset or size are negative, respectively, they will not be included in | |
2918 | * the BLOCK_IMAGE_CORRUPTED event emitted. | |
2919 | * fatal will be ignored for read-only BDS; corruptions found there will always | |
2920 | * be considered non-fatal. | |
2921 | */ | |
2922 | void qcow2_signal_corruption(BlockDriverState *bs, bool fatal, int64_t offset, | |
2923 | int64_t size, const char *message_format, ...) | |
2924 | { | |
ff99129a | 2925 | BDRVQcow2State *s = bs->opaque; |
dc881b44 | 2926 | const char *node_name; |
85186ebd HR |
2927 | char *message; |
2928 | va_list ap; | |
2929 | ||
2930 | fatal = fatal && !bs->read_only; | |
2931 | ||
2932 | if (s->signaled_corruption && | |
2933 | (!fatal || (s->incompatible_features & QCOW2_INCOMPAT_CORRUPT))) | |
2934 | { | |
2935 | return; | |
2936 | } | |
2937 | ||
2938 | va_start(ap, message_format); | |
2939 | message = g_strdup_vprintf(message_format, ap); | |
2940 | va_end(ap); | |
2941 | ||
2942 | if (fatal) { | |
2943 | fprintf(stderr, "qcow2: Marking image as corrupt: %s; further " | |
2944 | "corruption events will be suppressed\n", message); | |
2945 | } else { | |
2946 | fprintf(stderr, "qcow2: Image is corrupt: %s; further non-fatal " | |
2947 | "corruption events will be suppressed\n", message); | |
2948 | } | |
2949 | ||
dc881b44 AG |
2950 | node_name = bdrv_get_node_name(bs); |
2951 | qapi_event_send_block_image_corrupted(bdrv_get_device_name(bs), | |
2952 | *node_name != '\0', node_name, | |
2953 | message, offset >= 0, offset, | |
2954 | size >= 0, size, | |
85186ebd HR |
2955 | fatal, &error_abort); |
2956 | g_free(message); | |
2957 | ||
2958 | if (fatal) { | |
2959 | qcow2_mark_corrupt(bs); | |
2960 | bs->drv = NULL; /* make BDS unusable */ | |
2961 | } | |
2962 | ||
2963 | s->signaled_corruption = true; | |
2964 | } | |
2965 | ||
1bd0e2d1 CL |
2966 | static QemuOptsList qcow2_create_opts = { |
2967 | .name = "qcow2-create-opts", | |
2968 | .head = QTAILQ_HEAD_INITIALIZER(qcow2_create_opts.head), | |
2969 | .desc = { | |
2970 | { | |
2971 | .name = BLOCK_OPT_SIZE, | |
2972 | .type = QEMU_OPT_SIZE, | |
2973 | .help = "Virtual disk size" | |
2974 | }, | |
2975 | { | |
2976 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
2977 | .type = QEMU_OPT_STRING, | |
2978 | .help = "Compatibility level (0.10 or 1.1)" | |
2979 | }, | |
2980 | { | |
2981 | .name = BLOCK_OPT_BACKING_FILE, | |
2982 | .type = QEMU_OPT_STRING, | |
2983 | .help = "File name of a base image" | |
2984 | }, | |
2985 | { | |
2986 | .name = BLOCK_OPT_BACKING_FMT, | |
2987 | .type = QEMU_OPT_STRING, | |
2988 | .help = "Image format of the base image" | |
2989 | }, | |
2990 | { | |
2991 | .name = BLOCK_OPT_ENCRYPT, | |
2992 | .type = QEMU_OPT_BOOL, | |
2993 | .help = "Encrypt the image", | |
2994 | .def_value_str = "off" | |
2995 | }, | |
2996 | { | |
2997 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
2998 | .type = QEMU_OPT_SIZE, | |
2999 | .help = "qcow2 cluster size", | |
3000 | .def_value_str = stringify(DEFAULT_CLUSTER_SIZE) | |
3001 | }, | |
3002 | { | |
3003 | .name = BLOCK_OPT_PREALLOC, | |
3004 | .type = QEMU_OPT_STRING, | |
0e4271b7 HT |
3005 | .help = "Preallocation mode (allowed values: off, metadata, " |
3006 | "falloc, full)" | |
1bd0e2d1 CL |
3007 | }, |
3008 | { | |
3009 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
3010 | .type = QEMU_OPT_BOOL, | |
3011 | .help = "Postpone refcount updates", | |
3012 | .def_value_str = "off" | |
3013 | }, | |
06d05fa7 HR |
3014 | { |
3015 | .name = BLOCK_OPT_REFCOUNT_BITS, | |
3016 | .type = QEMU_OPT_NUMBER, | |
3017 | .help = "Width of a reference count entry in bits", | |
3018 | .def_value_str = "16" | |
3019 | }, | |
1bd0e2d1 CL |
3020 | { /* end of list */ } |
3021 | } | |
20d97356 BS |
3022 | }; |
3023 | ||
5f535a94 | 3024 | BlockDriver bdrv_qcow2 = { |
7c80ab3f | 3025 | .format_name = "qcow2", |
ff99129a | 3026 | .instance_size = sizeof(BDRVQcow2State), |
7c80ab3f JS |
3027 | .bdrv_probe = qcow2_probe, |
3028 | .bdrv_open = qcow2_open, | |
3029 | .bdrv_close = qcow2_close, | |
21d82ac9 | 3030 | .bdrv_reopen_prepare = qcow2_reopen_prepare, |
c282e1fd | 3031 | .bdrv_create = qcow2_create, |
3ac21627 | 3032 | .bdrv_has_zero_init = bdrv_has_zero_init_1, |
b6b8a333 | 3033 | .bdrv_co_get_block_status = qcow2_co_get_block_status, |
7c80ab3f | 3034 | .bdrv_set_key = qcow2_set_key, |
7c80ab3f | 3035 | |
c68b89ac KW |
3036 | .bdrv_co_readv = qcow2_co_readv, |
3037 | .bdrv_co_writev = qcow2_co_writev, | |
eb489bb1 | 3038 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
419b19d9 | 3039 | |
621f0589 | 3040 | .bdrv_co_write_zeroes = qcow2_co_write_zeroes, |
6db39ae2 | 3041 | .bdrv_co_discard = qcow2_co_discard, |
419b19d9 | 3042 | .bdrv_truncate = qcow2_truncate, |
7c80ab3f | 3043 | .bdrv_write_compressed = qcow2_write_compressed, |
491d27e2 | 3044 | .bdrv_make_empty = qcow2_make_empty, |
20d97356 BS |
3045 | |
3046 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
3047 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
3048 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
3049 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1bd0e2d1 CL |
3050 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
3051 | .bdrv_get_info = qcow2_get_info, | |
37764dfb | 3052 | .bdrv_get_specific_info = qcow2_get_specific_info, |
20d97356 | 3053 | |
7c80ab3f JS |
3054 | .bdrv_save_vmstate = qcow2_save_vmstate, |
3055 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
20d97356 | 3056 | |
8ee79e70 | 3057 | .supports_backing = true, |
20d97356 BS |
3058 | .bdrv_change_backing_file = qcow2_change_backing_file, |
3059 | ||
d34682cd | 3060 | .bdrv_refresh_limits = qcow2_refresh_limits, |
06d9260f AL |
3061 | .bdrv_invalidate_cache = qcow2_invalidate_cache, |
3062 | ||
1bd0e2d1 CL |
3063 | .create_opts = &qcow2_create_opts, |
3064 | .bdrv_check = qcow2_check, | |
c282e1fd | 3065 | .bdrv_amend_options = qcow2_amend_options, |
279621c0 AG |
3066 | |
3067 | .bdrv_detach_aio_context = qcow2_detach_aio_context, | |
3068 | .bdrv_attach_aio_context = qcow2_attach_aio_context, | |
20d97356 BS |
3069 | }; |
3070 | ||
5efa9d5a AL |
3071 | static void bdrv_qcow2_init(void) |
3072 | { | |
3073 | bdrv_register(&bdrv_qcow2); | |
3074 | } | |
3075 | ||
3076 | block_init(bdrv_qcow2_init); |