]>
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> |
737e150e | 28 | #include "block/aes.h" |
f7d0fe02 | 29 | #include "block/qcow2.h" |
1de7afc9 | 30 | #include "qemu/error-report.h" |
7b1b5d19 | 31 | #include "qapi/qmp/qerror.h" |
3cce16f4 | 32 | #include "trace.h" |
585f8587 FB |
33 | |
34 | /* | |
35 | Differences with QCOW: | |
36 | ||
37 | - Support for multiple incremental snapshots. | |
38 | - Memory management by reference counts. | |
39 | - Clusters which have a reference count of one have the bit | |
40 | QCOW_OFLAG_COPIED to optimize write performance. | |
5fafdf24 | 41 | - Size of compressed clusters is stored in sectors to reduce bit usage |
585f8587 FB |
42 | in the cluster offsets. |
43 | - Support for storing additional data (such as the VM state) in the | |
3b46e624 | 44 | snapshots. |
585f8587 FB |
45 | - If a backing store is used, the cluster size is not constrained |
46 | (could be backported to QCOW). | |
47 | - L2 tables have always a size of one cluster. | |
48 | */ | |
49 | ||
9b80ddf3 AL |
50 | |
51 | typedef struct { | |
52 | uint32_t magic; | |
53 | uint32_t len; | |
54 | } QCowExtension; | |
21d82ac9 | 55 | |
7c80ab3f JS |
56 | #define QCOW2_EXT_MAGIC_END 0 |
57 | #define QCOW2_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA | |
cfcc4c62 | 58 | #define QCOW2_EXT_MAGIC_FEATURE_TABLE 0x6803f857 |
9b80ddf3 | 59 | |
7c80ab3f | 60 | static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) |
585f8587 FB |
61 | { |
62 | const QCowHeader *cow_header = (const void *)buf; | |
3b46e624 | 63 | |
585f8587 FB |
64 | if (buf_size >= sizeof(QCowHeader) && |
65 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
6744cbab | 66 | be32_to_cpu(cow_header->version) >= 2) |
585f8587 FB |
67 | return 100; |
68 | else | |
69 | return 0; | |
70 | } | |
71 | ||
9b80ddf3 AL |
72 | |
73 | /* | |
74 | * read qcow2 extension and fill bs | |
75 | * start reading from start_offset | |
76 | * finish reading upon magic of value 0 or when end_offset reached | |
77 | * unknown magic is skipped (future extension this version knows nothing about) | |
78 | * return 0 upon success, non-0 otherwise | |
79 | */ | |
7c80ab3f | 80 | static int qcow2_read_extensions(BlockDriverState *bs, uint64_t start_offset, |
cfcc4c62 | 81 | uint64_t end_offset, void **p_feature_table) |
9b80ddf3 | 82 | { |
75bab85c | 83 | BDRVQcowState *s = bs->opaque; |
9b80ddf3 AL |
84 | QCowExtension ext; |
85 | uint64_t offset; | |
75bab85c | 86 | int ret; |
9b80ddf3 AL |
87 | |
88 | #ifdef DEBUG_EXT | |
7c80ab3f | 89 | printf("qcow2_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); |
9b80ddf3 AL |
90 | #endif |
91 | offset = start_offset; | |
92 | while (offset < end_offset) { | |
93 | ||
94 | #ifdef DEBUG_EXT | |
95 | /* Sanity check */ | |
96 | if (offset > s->cluster_size) | |
7c80ab3f | 97 | printf("qcow2_read_extension: suspicious offset %lu\n", offset); |
9b80ddf3 | 98 | |
9b2260cb | 99 | printf("attempting to read extended header in offset %lu\n", offset); |
9b80ddf3 AL |
100 | #endif |
101 | ||
66f82cee | 102 | if (bdrv_pread(bs->file, offset, &ext, sizeof(ext)) != sizeof(ext)) { |
7c80ab3f | 103 | fprintf(stderr, "qcow2_read_extension: ERROR: " |
0bfcd599 BS |
104 | "pread fail from offset %" PRIu64 "\n", |
105 | offset); | |
9b80ddf3 AL |
106 | return 1; |
107 | } | |
108 | be32_to_cpus(&ext.magic); | |
109 | be32_to_cpus(&ext.len); | |
110 | offset += sizeof(ext); | |
111 | #ifdef DEBUG_EXT | |
112 | printf("ext.magic = 0x%x\n", ext.magic); | |
113 | #endif | |
64ca6aee KW |
114 | if (ext.len > end_offset - offset) { |
115 | error_report("Header extension too large"); | |
116 | return -EINVAL; | |
117 | } | |
118 | ||
9b80ddf3 | 119 | switch (ext.magic) { |
7c80ab3f | 120 | case QCOW2_EXT_MAGIC_END: |
9b80ddf3 | 121 | return 0; |
f965509c | 122 | |
7c80ab3f | 123 | case QCOW2_EXT_MAGIC_BACKING_FORMAT: |
f965509c AL |
124 | if (ext.len >= sizeof(bs->backing_format)) { |
125 | fprintf(stderr, "ERROR: ext_backing_format: len=%u too large" | |
4c978075 | 126 | " (>=%zu)\n", |
f965509c AL |
127 | ext.len, sizeof(bs->backing_format)); |
128 | return 2; | |
129 | } | |
66f82cee | 130 | if (bdrv_pread(bs->file, offset , bs->backing_format, |
f965509c AL |
131 | ext.len) != ext.len) |
132 | return 3; | |
133 | bs->backing_format[ext.len] = '\0'; | |
134 | #ifdef DEBUG_EXT | |
135 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
136 | #endif | |
f965509c AL |
137 | break; |
138 | ||
cfcc4c62 KW |
139 | case QCOW2_EXT_MAGIC_FEATURE_TABLE: |
140 | if (p_feature_table != NULL) { | |
141 | void* feature_table = g_malloc0(ext.len + 2 * sizeof(Qcow2Feature)); | |
142 | ret = bdrv_pread(bs->file, offset , feature_table, ext.len); | |
143 | if (ret < 0) { | |
144 | return ret; | |
145 | } | |
146 | ||
147 | *p_feature_table = feature_table; | |
148 | } | |
149 | break; | |
150 | ||
9b80ddf3 | 151 | default: |
75bab85c KW |
152 | /* unknown magic - save it in case we need to rewrite the header */ |
153 | { | |
154 | Qcow2UnknownHeaderExtension *uext; | |
155 | ||
156 | uext = g_malloc0(sizeof(*uext) + ext.len); | |
157 | uext->magic = ext.magic; | |
158 | uext->len = ext.len; | |
159 | QLIST_INSERT_HEAD(&s->unknown_header_ext, uext, next); | |
160 | ||
161 | ret = bdrv_pread(bs->file, offset , uext->data, uext->len); | |
162 | if (ret < 0) { | |
163 | return ret; | |
164 | } | |
75bab85c | 165 | } |
9b80ddf3 AL |
166 | break; |
167 | } | |
fd29b4bb KW |
168 | |
169 | offset += ((ext.len + 7) & ~7); | |
9b80ddf3 AL |
170 | } |
171 | ||
172 | return 0; | |
173 | } | |
174 | ||
75bab85c KW |
175 | static void cleanup_unknown_header_ext(BlockDriverState *bs) |
176 | { | |
177 | BDRVQcowState *s = bs->opaque; | |
178 | Qcow2UnknownHeaderExtension *uext, *next; | |
179 | ||
180 | QLIST_FOREACH_SAFE(uext, &s->unknown_header_ext, next, next) { | |
181 | QLIST_REMOVE(uext, next); | |
182 | g_free(uext); | |
183 | } | |
184 | } | |
9b80ddf3 | 185 | |
b9531b6e SW |
186 | static void GCC_FMT_ATTR(2, 3) report_unsupported(BlockDriverState *bs, |
187 | const char *fmt, ...) | |
6744cbab KW |
188 | { |
189 | char msg[64]; | |
190 | va_list ap; | |
191 | ||
192 | va_start(ap, fmt); | |
193 | vsnprintf(msg, sizeof(msg), fmt, ap); | |
194 | va_end(ap); | |
195 | ||
196 | qerror_report(QERR_UNKNOWN_BLOCK_FORMAT_FEATURE, | |
197 | bs->device_name, "qcow2", msg); | |
198 | } | |
199 | ||
cfcc4c62 KW |
200 | static void report_unsupported_feature(BlockDriverState *bs, |
201 | Qcow2Feature *table, uint64_t mask) | |
202 | { | |
203 | while (table && table->name[0] != '\0') { | |
204 | if (table->type == QCOW2_FEAT_TYPE_INCOMPATIBLE) { | |
205 | if (mask & (1 << table->bit)) { | |
206 | report_unsupported(bs, "%.46s",table->name); | |
207 | mask &= ~(1 << table->bit); | |
208 | } | |
209 | } | |
210 | table++; | |
211 | } | |
212 | ||
213 | if (mask) { | |
214 | report_unsupported(bs, "Unknown incompatible feature: %" PRIx64, mask); | |
215 | } | |
216 | } | |
217 | ||
bfe8043e SH |
218 | /* |
219 | * Sets the dirty bit and flushes afterwards if necessary. | |
220 | * | |
221 | * The incompatible_features bit is only set if the image file header was | |
222 | * updated successfully. Therefore it is not required to check the return | |
223 | * value of this function. | |
224 | */ | |
280d3735 | 225 | int qcow2_mark_dirty(BlockDriverState *bs) |
bfe8043e SH |
226 | { |
227 | BDRVQcowState *s = bs->opaque; | |
228 | uint64_t val; | |
229 | int ret; | |
230 | ||
231 | assert(s->qcow_version >= 3); | |
232 | ||
233 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
234 | return 0; /* already dirty */ | |
235 | } | |
236 | ||
237 | val = cpu_to_be64(s->incompatible_features | QCOW2_INCOMPAT_DIRTY); | |
238 | ret = bdrv_pwrite(bs->file, offsetof(QCowHeader, incompatible_features), | |
239 | &val, sizeof(val)); | |
240 | if (ret < 0) { | |
241 | return ret; | |
242 | } | |
243 | ret = bdrv_flush(bs->file); | |
244 | if (ret < 0) { | |
245 | return ret; | |
246 | } | |
247 | ||
248 | /* Only treat image as dirty if the header was updated successfully */ | |
249 | s->incompatible_features |= QCOW2_INCOMPAT_DIRTY; | |
250 | return 0; | |
251 | } | |
252 | ||
c61d0004 SH |
253 | /* |
254 | * Clears the dirty bit and flushes before if necessary. Only call this | |
255 | * function when there are no pending requests, it does not guard against | |
256 | * concurrent requests dirtying the image. | |
257 | */ | |
258 | static int qcow2_mark_clean(BlockDriverState *bs) | |
259 | { | |
260 | BDRVQcowState *s = bs->opaque; | |
261 | ||
262 | if (s->incompatible_features & QCOW2_INCOMPAT_DIRTY) { | |
263 | int ret = bdrv_flush(bs); | |
264 | if (ret < 0) { | |
265 | return ret; | |
266 | } | |
267 | ||
268 | s->incompatible_features &= ~QCOW2_INCOMPAT_DIRTY; | |
269 | return qcow2_update_header(bs); | |
270 | } | |
271 | return 0; | |
272 | } | |
273 | ||
acbe5982 SH |
274 | static int qcow2_check(BlockDriverState *bs, BdrvCheckResult *result, |
275 | BdrvCheckMode fix) | |
276 | { | |
277 | int ret = qcow2_check_refcounts(bs, result, fix); | |
278 | if (ret < 0) { | |
279 | return ret; | |
280 | } | |
281 | ||
282 | if (fix && result->check_errors == 0 && result->corruptions == 0) { | |
283 | return qcow2_mark_clean(bs); | |
284 | } | |
285 | return ret; | |
286 | } | |
287 | ||
7c80ab3f | 288 | static int qcow2_open(BlockDriverState *bs, int flags) |
585f8587 FB |
289 | { |
290 | BDRVQcowState *s = bs->opaque; | |
6d85a57e | 291 | int len, i, ret = 0; |
585f8587 | 292 | QCowHeader header; |
9b80ddf3 | 293 | uint64_t ext_end; |
585f8587 | 294 | |
6d85a57e JS |
295 | ret = bdrv_pread(bs->file, 0, &header, sizeof(header)); |
296 | if (ret < 0) { | |
585f8587 | 297 | goto fail; |
6d85a57e | 298 | } |
585f8587 FB |
299 | be32_to_cpus(&header.magic); |
300 | be32_to_cpus(&header.version); | |
301 | be64_to_cpus(&header.backing_file_offset); | |
302 | be32_to_cpus(&header.backing_file_size); | |
303 | be64_to_cpus(&header.size); | |
304 | be32_to_cpus(&header.cluster_bits); | |
305 | be32_to_cpus(&header.crypt_method); | |
306 | be64_to_cpus(&header.l1_table_offset); | |
307 | be32_to_cpus(&header.l1_size); | |
308 | be64_to_cpus(&header.refcount_table_offset); | |
309 | be32_to_cpus(&header.refcount_table_clusters); | |
310 | be64_to_cpus(&header.snapshots_offset); | |
311 | be32_to_cpus(&header.nb_snapshots); | |
3b46e624 | 312 | |
e8cdcec1 | 313 | if (header.magic != QCOW_MAGIC) { |
6d85a57e | 314 | ret = -EINVAL; |
585f8587 | 315 | goto fail; |
6d85a57e | 316 | } |
6744cbab KW |
317 | if (header.version < 2 || header.version > 3) { |
318 | report_unsupported(bs, "QCOW version %d", header.version); | |
319 | ret = -ENOTSUP; | |
320 | goto fail; | |
321 | } | |
322 | ||
323 | s->qcow_version = header.version; | |
324 | ||
325 | /* Initialise version 3 header fields */ | |
326 | if (header.version == 2) { | |
327 | header.incompatible_features = 0; | |
328 | header.compatible_features = 0; | |
329 | header.autoclear_features = 0; | |
330 | header.refcount_order = 4; | |
331 | header.header_length = 72; | |
332 | } else { | |
333 | be64_to_cpus(&header.incompatible_features); | |
334 | be64_to_cpus(&header.compatible_features); | |
335 | be64_to_cpus(&header.autoclear_features); | |
336 | be32_to_cpus(&header.refcount_order); | |
337 | be32_to_cpus(&header.header_length); | |
338 | } | |
339 | ||
340 | if (header.header_length > sizeof(header)) { | |
341 | s->unknown_header_fields_size = header.header_length - sizeof(header); | |
342 | s->unknown_header_fields = g_malloc(s->unknown_header_fields_size); | |
343 | ret = bdrv_pread(bs->file, sizeof(header), s->unknown_header_fields, | |
344 | s->unknown_header_fields_size); | |
345 | if (ret < 0) { | |
346 | goto fail; | |
347 | } | |
348 | } | |
349 | ||
cfcc4c62 KW |
350 | if (header.backing_file_offset) { |
351 | ext_end = header.backing_file_offset; | |
352 | } else { | |
353 | ext_end = 1 << header.cluster_bits; | |
354 | } | |
355 | ||
6744cbab KW |
356 | /* Handle feature bits */ |
357 | s->incompatible_features = header.incompatible_features; | |
358 | s->compatible_features = header.compatible_features; | |
359 | s->autoclear_features = header.autoclear_features; | |
360 | ||
c61d0004 | 361 | if (s->incompatible_features & ~QCOW2_INCOMPAT_MASK) { |
cfcc4c62 KW |
362 | void *feature_table = NULL; |
363 | qcow2_read_extensions(bs, header.header_length, ext_end, | |
364 | &feature_table); | |
365 | report_unsupported_feature(bs, feature_table, | |
c61d0004 SH |
366 | s->incompatible_features & |
367 | ~QCOW2_INCOMPAT_MASK); | |
6744cbab KW |
368 | ret = -ENOTSUP; |
369 | goto fail; | |
370 | } | |
371 | ||
6744cbab KW |
372 | /* Check support for various header values */ |
373 | if (header.refcount_order != 4) { | |
374 | report_unsupported(bs, "%d bit reference counts", | |
375 | 1 << header.refcount_order); | |
e8cdcec1 KW |
376 | ret = -ENOTSUP; |
377 | goto fail; | |
378 | } | |
6744cbab | 379 | |
d191d12d | 380 | if (header.cluster_bits < MIN_CLUSTER_BITS || |
6d85a57e JS |
381 | header.cluster_bits > MAX_CLUSTER_BITS) { |
382 | ret = -EINVAL; | |
585f8587 | 383 | goto fail; |
6d85a57e JS |
384 | } |
385 | if (header.crypt_method > QCOW_CRYPT_AES) { | |
386 | ret = -EINVAL; | |
585f8587 | 387 | goto fail; |
6d85a57e | 388 | } |
585f8587 | 389 | s->crypt_method_header = header.crypt_method; |
6d85a57e | 390 | if (s->crypt_method_header) { |
585f8587 | 391 | bs->encrypted = 1; |
6d85a57e | 392 | } |
585f8587 FB |
393 | s->cluster_bits = header.cluster_bits; |
394 | s->cluster_size = 1 << s->cluster_bits; | |
395 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
396 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ | |
397 | s->l2_size = 1 << s->l2_bits; | |
398 | bs->total_sectors = header.size / 512; | |
399 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
400 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
401 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
402 | s->refcount_table_offset = header.refcount_table_offset; | |
5fafdf24 | 403 | s->refcount_table_size = |
585f8587 FB |
404 | header.refcount_table_clusters << (s->cluster_bits - 3); |
405 | ||
406 | s->snapshots_offset = header.snapshots_offset; | |
407 | s->nb_snapshots = header.nb_snapshots; | |
408 | ||
409 | /* read the level 1 table */ | |
410 | s->l1_size = header.l1_size; | |
419b19d9 | 411 | s->l1_vm_state_index = size_to_l1(s, header.size); |
585f8587 FB |
412 | /* the L1 table must contain at least enough entries to put |
413 | header.size bytes */ | |
6d85a57e JS |
414 | if (s->l1_size < s->l1_vm_state_index) { |
415 | ret = -EINVAL; | |
585f8587 | 416 | goto fail; |
6d85a57e | 417 | } |
585f8587 | 418 | s->l1_table_offset = header.l1_table_offset; |
d191d12d | 419 | if (s->l1_size > 0) { |
7267c094 | 420 | s->l1_table = g_malloc0( |
d191d12d | 421 | align_offset(s->l1_size * sizeof(uint64_t), 512)); |
6d85a57e JS |
422 | ret = bdrv_pread(bs->file, s->l1_table_offset, s->l1_table, |
423 | s->l1_size * sizeof(uint64_t)); | |
424 | if (ret < 0) { | |
d191d12d | 425 | goto fail; |
6d85a57e | 426 | } |
d191d12d SW |
427 | for(i = 0;i < s->l1_size; i++) { |
428 | be64_to_cpus(&s->l1_table[i]); | |
429 | } | |
585f8587 | 430 | } |
29c1a730 KW |
431 | |
432 | /* alloc L2 table/refcount block cache */ | |
6af4e9ea PB |
433 | s->l2_table_cache = qcow2_cache_create(bs, L2_CACHE_SIZE); |
434 | s->refcount_block_cache = qcow2_cache_create(bs, REFCOUNT_CACHE_SIZE); | |
29c1a730 | 435 | |
7267c094 | 436 | s->cluster_cache = g_malloc(s->cluster_size); |
585f8587 | 437 | /* one more sector for decompressed data alignment */ |
dea43a65 | 438 | s->cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size |
095a9c58 | 439 | + 512); |
585f8587 | 440 | s->cluster_cache_offset = -1; |
06d9260f | 441 | s->flags = flags; |
3b46e624 | 442 | |
6d85a57e JS |
443 | ret = qcow2_refcount_init(bs); |
444 | if (ret != 0) { | |
585f8587 | 445 | goto fail; |
6d85a57e | 446 | } |
585f8587 | 447 | |
72cf2d4f | 448 | QLIST_INIT(&s->cluster_allocs); |
f214978a | 449 | |
9b80ddf3 | 450 | /* read qcow2 extensions */ |
cfcc4c62 | 451 | if (qcow2_read_extensions(bs, header.header_length, ext_end, NULL)) { |
6d85a57e | 452 | ret = -EINVAL; |
9b80ddf3 | 453 | goto fail; |
6d85a57e | 454 | } |
9b80ddf3 | 455 | |
585f8587 FB |
456 | /* read the backing file name */ |
457 | if (header.backing_file_offset != 0) { | |
458 | len = header.backing_file_size; | |
6d85a57e | 459 | if (len > 1023) { |
585f8587 | 460 | len = 1023; |
6d85a57e JS |
461 | } |
462 | ret = bdrv_pread(bs->file, header.backing_file_offset, | |
463 | bs->backing_file, len); | |
464 | if (ret < 0) { | |
585f8587 | 465 | goto fail; |
6d85a57e | 466 | } |
585f8587 FB |
467 | bs->backing_file[len] = '\0'; |
468 | } | |
42deb29f KW |
469 | |
470 | ret = qcow2_read_snapshots(bs); | |
471 | if (ret < 0) { | |
585f8587 | 472 | goto fail; |
6d85a57e | 473 | } |
585f8587 | 474 | |
af7b708d SH |
475 | /* Clear unknown autoclear feature bits */ |
476 | if (!bs->read_only && s->autoclear_features != 0) { | |
477 | s->autoclear_features = 0; | |
478 | ret = qcow2_update_header(bs); | |
479 | if (ret < 0) { | |
480 | goto fail; | |
481 | } | |
482 | } | |
483 | ||
68d100e9 KW |
484 | /* Initialise locks */ |
485 | qemu_co_mutex_init(&s->lock); | |
486 | ||
c61d0004 | 487 | /* Repair image if dirty */ |
058f8f16 SH |
488 | if (!(flags & BDRV_O_CHECK) && !bs->read_only && |
489 | (s->incompatible_features & QCOW2_INCOMPAT_DIRTY)) { | |
c61d0004 SH |
490 | BdrvCheckResult result = {0}; |
491 | ||
acbe5982 | 492 | ret = qcow2_check(bs, &result, BDRV_FIX_ERRORS); |
c61d0004 SH |
493 | if (ret < 0) { |
494 | goto fail; | |
495 | } | |
496 | } | |
497 | ||
585f8587 | 498 | #ifdef DEBUG_ALLOC |
6cbc3031 PH |
499 | { |
500 | BdrvCheckResult result = {0}; | |
b35278f7 | 501 | qcow2_check_refcounts(bs, &result, 0); |
6cbc3031 | 502 | } |
585f8587 | 503 | #endif |
6d85a57e | 504 | return ret; |
585f8587 FB |
505 | |
506 | fail: | |
6744cbab | 507 | g_free(s->unknown_header_fields); |
75bab85c | 508 | cleanup_unknown_header_ext(bs); |
ed6ccf0f KW |
509 | qcow2_free_snapshots(bs); |
510 | qcow2_refcount_close(bs); | |
7267c094 | 511 | g_free(s->l1_table); |
29c1a730 KW |
512 | if (s->l2_table_cache) { |
513 | qcow2_cache_destroy(bs, s->l2_table_cache); | |
514 | } | |
7267c094 | 515 | g_free(s->cluster_cache); |
dea43a65 | 516 | qemu_vfree(s->cluster_data); |
6d85a57e | 517 | return ret; |
585f8587 FB |
518 | } |
519 | ||
7c80ab3f | 520 | static int qcow2_set_key(BlockDriverState *bs, const char *key) |
585f8587 FB |
521 | { |
522 | BDRVQcowState *s = bs->opaque; | |
523 | uint8_t keybuf[16]; | |
524 | int len, i; | |
3b46e624 | 525 | |
585f8587 FB |
526 | memset(keybuf, 0, 16); |
527 | len = strlen(key); | |
528 | if (len > 16) | |
529 | len = 16; | |
530 | /* XXX: we could compress the chars to 7 bits to increase | |
531 | entropy */ | |
532 | for(i = 0;i < len;i++) { | |
533 | keybuf[i] = key[i]; | |
534 | } | |
535 | s->crypt_method = s->crypt_method_header; | |
536 | ||
537 | if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0) | |
538 | return -1; | |
539 | if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0) | |
540 | return -1; | |
541 | #if 0 | |
542 | /* test */ | |
543 | { | |
544 | uint8_t in[16]; | |
545 | uint8_t out[16]; | |
546 | uint8_t tmp[16]; | |
547 | for(i=0;i<16;i++) | |
548 | in[i] = i; | |
549 | AES_encrypt(in, tmp, &s->aes_encrypt_key); | |
550 | AES_decrypt(tmp, out, &s->aes_decrypt_key); | |
551 | for(i = 0; i < 16; i++) | |
552 | printf(" %02x", tmp[i]); | |
553 | printf("\n"); | |
554 | for(i = 0; i < 16; i++) | |
555 | printf(" %02x", out[i]); | |
556 | printf("\n"); | |
557 | } | |
558 | #endif | |
559 | return 0; | |
560 | } | |
561 | ||
21d82ac9 JC |
562 | /* We have nothing to do for QCOW2 reopen, stubs just return |
563 | * success */ | |
564 | static int qcow2_reopen_prepare(BDRVReopenState *state, | |
565 | BlockReopenQueue *queue, Error **errp) | |
566 | { | |
567 | return 0; | |
568 | } | |
569 | ||
f8a2e5e3 SH |
570 | static int coroutine_fn qcow2_co_is_allocated(BlockDriverState *bs, |
571 | int64_t sector_num, int nb_sectors, int *pnum) | |
585f8587 | 572 | { |
f8a2e5e3 | 573 | BDRVQcowState *s = bs->opaque; |
585f8587 | 574 | uint64_t cluster_offset; |
1c46efaa | 575 | int ret; |
585f8587 | 576 | |
095a9c58 | 577 | *pnum = nb_sectors; |
f8a2e5e3 SH |
578 | /* FIXME We can get errors here, but the bdrv_co_is_allocated interface |
579 | * can't pass them on today */ | |
580 | qemu_co_mutex_lock(&s->lock); | |
1c46efaa | 581 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, pnum, &cluster_offset); |
f8a2e5e3 | 582 | qemu_co_mutex_unlock(&s->lock); |
1c46efaa KW |
583 | if (ret < 0) { |
584 | *pnum = 0; | |
585 | } | |
095a9c58 | 586 | |
585f8587 FB |
587 | return (cluster_offset != 0); |
588 | } | |
589 | ||
a9465922 | 590 | /* handle reading after the end of the backing file */ |
bd28f835 KW |
591 | int qcow2_backing_read1(BlockDriverState *bs, QEMUIOVector *qiov, |
592 | int64_t sector_num, int nb_sectors) | |
a9465922 FB |
593 | { |
594 | int n1; | |
595 | if ((sector_num + nb_sectors) <= bs->total_sectors) | |
596 | return nb_sectors; | |
597 | if (sector_num >= bs->total_sectors) | |
598 | n1 = 0; | |
599 | else | |
600 | n1 = bs->total_sectors - sector_num; | |
bd28f835 | 601 | |
3d9b4925 | 602 | qemu_iovec_memset(qiov, 512 * n1, 0, 512 * (nb_sectors - n1)); |
bd28f835 | 603 | |
a9465922 FB |
604 | return n1; |
605 | } | |
606 | ||
a968168c | 607 | static coroutine_fn int qcow2_co_readv(BlockDriverState *bs, int64_t sector_num, |
3fc48d09 | 608 | int remaining_sectors, QEMUIOVector *qiov) |
585f8587 | 609 | { |
585f8587 | 610 | BDRVQcowState *s = bs->opaque; |
a9465922 | 611 | int index_in_cluster, n1; |
68d100e9 | 612 | int ret; |
faf575c1 | 613 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 614 | uint64_t cluster_offset = 0; |
3fc48d09 FZ |
615 | uint64_t bytes_done = 0; |
616 | QEMUIOVector hd_qiov; | |
617 | uint8_t *cluster_data = NULL; | |
585f8587 | 618 | |
3fc48d09 FZ |
619 | qemu_iovec_init(&hd_qiov, qiov->niov); |
620 | ||
621 | qemu_co_mutex_lock(&s->lock); | |
622 | ||
623 | while (remaining_sectors != 0) { | |
bd28f835 | 624 | |
5ebaa27e | 625 | /* prepare next request */ |
3fc48d09 | 626 | cur_nr_sectors = remaining_sectors; |
5ebaa27e FZ |
627 | if (s->crypt_method) { |
628 | cur_nr_sectors = MIN(cur_nr_sectors, | |
629 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
585f8587 | 630 | } |
5ebaa27e | 631 | |
3fc48d09 | 632 | ret = qcow2_get_cluster_offset(bs, sector_num << 9, |
5ebaa27e | 633 | &cur_nr_sectors, &cluster_offset); |
8af36488 | 634 | if (ret < 0) { |
3fc48d09 | 635 | goto fail; |
8af36488 | 636 | } |
bd28f835 | 637 | |
3fc48d09 | 638 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
c87c0672 | 639 | |
3fc48d09 | 640 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 641 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e FZ |
642 | cur_nr_sectors * 512); |
643 | ||
68d000a3 KW |
644 | switch (ret) { |
645 | case QCOW2_CLUSTER_UNALLOCATED: | |
5ebaa27e FZ |
646 | |
647 | if (bs->backing_hd) { | |
648 | /* read from the base image */ | |
3fc48d09 FZ |
649 | n1 = qcow2_backing_read1(bs->backing_hd, &hd_qiov, |
650 | sector_num, cur_nr_sectors); | |
5ebaa27e FZ |
651 | if (n1 > 0) { |
652 | BLKDBG_EVENT(bs->file, BLKDBG_READ_BACKING_AIO); | |
653 | qemu_co_mutex_unlock(&s->lock); | |
3fc48d09 FZ |
654 | ret = bdrv_co_readv(bs->backing_hd, sector_num, |
655 | n1, &hd_qiov); | |
5ebaa27e FZ |
656 | qemu_co_mutex_lock(&s->lock); |
657 | if (ret < 0) { | |
3fc48d09 | 658 | goto fail; |
5ebaa27e FZ |
659 | } |
660 | } | |
661 | } else { | |
662 | /* Note: in this case, no need to wait */ | |
3d9b4925 | 663 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
5ebaa27e | 664 | } |
68d000a3 KW |
665 | break; |
666 | ||
6377af48 KW |
667 | case QCOW2_CLUSTER_ZERO: |
668 | if (s->qcow_version < 3) { | |
669 | ret = -EIO; | |
670 | goto fail; | |
671 | } | |
3d9b4925 | 672 | qemu_iovec_memset(&hd_qiov, 0, 0, 512 * cur_nr_sectors); |
6377af48 KW |
673 | break; |
674 | ||
68d000a3 | 675 | case QCOW2_CLUSTER_COMPRESSED: |
5ebaa27e FZ |
676 | /* add AIO support for compressed blocks ? */ |
677 | ret = qcow2_decompress_cluster(bs, cluster_offset); | |
678 | if (ret < 0) { | |
3fc48d09 | 679 | goto fail; |
bd28f835 KW |
680 | } |
681 | ||
03396148 | 682 | qemu_iovec_from_buf(&hd_qiov, 0, |
5ebaa27e | 683 | s->cluster_cache + index_in_cluster * 512, |
faf575c1 | 684 | 512 * cur_nr_sectors); |
68d000a3 KW |
685 | break; |
686 | ||
687 | case QCOW2_CLUSTER_NORMAL: | |
5ebaa27e | 688 | if ((cluster_offset & 511) != 0) { |
3fc48d09 FZ |
689 | ret = -EIO; |
690 | goto fail; | |
5ebaa27e | 691 | } |
bd28f835 | 692 | |
5ebaa27e FZ |
693 | if (s->crypt_method) { |
694 | /* | |
695 | * For encrypted images, read everything into a temporary | |
696 | * contiguous buffer on which the AES functions can work. | |
697 | */ | |
3fc48d09 FZ |
698 | if (!cluster_data) { |
699 | cluster_data = | |
dea43a65 | 700 | qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
5ebaa27e FZ |
701 | } |
702 | ||
703 | assert(cur_nr_sectors <= | |
704 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors); | |
3fc48d09 FZ |
705 | qemu_iovec_reset(&hd_qiov); |
706 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
707 | 512 * cur_nr_sectors); |
708 | } | |
709 | ||
710 | BLKDBG_EVENT(bs->file, BLKDBG_READ_AIO); | |
711 | qemu_co_mutex_unlock(&s->lock); | |
712 | ret = bdrv_co_readv(bs->file, | |
713 | (cluster_offset >> 9) + index_in_cluster, | |
3fc48d09 | 714 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
715 | qemu_co_mutex_lock(&s->lock); |
716 | if (ret < 0) { | |
3fc48d09 | 717 | goto fail; |
5ebaa27e FZ |
718 | } |
719 | if (s->crypt_method) { | |
3fc48d09 FZ |
720 | qcow2_encrypt_sectors(s, sector_num, cluster_data, |
721 | cluster_data, cur_nr_sectors, 0, &s->aes_decrypt_key); | |
03396148 MT |
722 | qemu_iovec_from_buf(qiov, bytes_done, |
723 | cluster_data, 512 * cur_nr_sectors); | |
5ebaa27e | 724 | } |
68d000a3 KW |
725 | break; |
726 | ||
727 | default: | |
728 | g_assert_not_reached(); | |
729 | ret = -EIO; | |
730 | goto fail; | |
faf575c1 | 731 | } |
f141eafe | 732 | |
3fc48d09 FZ |
733 | remaining_sectors -= cur_nr_sectors; |
734 | sector_num += cur_nr_sectors; | |
735 | bytes_done += cur_nr_sectors * 512; | |
5ebaa27e | 736 | } |
3fc48d09 | 737 | ret = 0; |
faf575c1 | 738 | |
3fc48d09 | 739 | fail: |
68d100e9 | 740 | qemu_co_mutex_unlock(&s->lock); |
42496d62 | 741 | |
3fc48d09 | 742 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 743 | qemu_vfree(cluster_data); |
68d100e9 KW |
744 | |
745 | return ret; | |
585f8587 FB |
746 | } |
747 | ||
a968168c | 748 | static coroutine_fn int qcow2_co_writev(BlockDriverState *bs, |
3fc48d09 FZ |
749 | int64_t sector_num, |
750 | int remaining_sectors, | |
751 | QEMUIOVector *qiov) | |
585f8587 | 752 | { |
585f8587 | 753 | BDRVQcowState *s = bs->opaque; |
585f8587 | 754 | int index_in_cluster; |
095a9c58 | 755 | int n_end; |
68d100e9 | 756 | int ret; |
faf575c1 | 757 | int cur_nr_sectors; /* number of sectors in current iteration */ |
c2bdd990 | 758 | uint64_t cluster_offset; |
3fc48d09 FZ |
759 | QEMUIOVector hd_qiov; |
760 | uint64_t bytes_done = 0; | |
761 | uint8_t *cluster_data = NULL; | |
8d2497c3 | 762 | QCowL2Meta *l2meta = NULL; |
c2271403 | 763 | |
3cce16f4 KW |
764 | trace_qcow2_writev_start_req(qemu_coroutine_self(), sector_num, |
765 | remaining_sectors); | |
766 | ||
3fc48d09 FZ |
767 | qemu_iovec_init(&hd_qiov, qiov->niov); |
768 | ||
769 | s->cluster_cache_offset = -1; /* disable compressed cache */ | |
3b46e624 | 770 | |
3fc48d09 FZ |
771 | qemu_co_mutex_lock(&s->lock); |
772 | ||
773 | while (remaining_sectors != 0) { | |
774 | ||
f50f88b9 | 775 | l2meta = NULL; |
cf5c1a23 | 776 | |
3cce16f4 | 777 | trace_qcow2_writev_start_part(qemu_coroutine_self()); |
3fc48d09 FZ |
778 | index_in_cluster = sector_num & (s->cluster_sectors - 1); |
779 | n_end = index_in_cluster + remaining_sectors; | |
5ebaa27e FZ |
780 | if (s->crypt_method && |
781 | n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors) { | |
782 | n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors; | |
783 | } | |
095a9c58 | 784 | |
3fc48d09 | 785 | ret = qcow2_alloc_cluster_offset(bs, sector_num << 9, |
f50f88b9 | 786 | index_in_cluster, n_end, &cur_nr_sectors, &cluster_offset, &l2meta); |
5ebaa27e | 787 | if (ret < 0) { |
3fc48d09 | 788 | goto fail; |
5ebaa27e | 789 | } |
148da7ea | 790 | |
5ebaa27e | 791 | assert((cluster_offset & 511) == 0); |
148da7ea | 792 | |
3fc48d09 | 793 | qemu_iovec_reset(&hd_qiov); |
1b093c48 | 794 | qemu_iovec_concat(&hd_qiov, qiov, bytes_done, |
5ebaa27e | 795 | cur_nr_sectors * 512); |
6f5f060b | 796 | |
5ebaa27e | 797 | if (s->crypt_method) { |
3fc48d09 | 798 | if (!cluster_data) { |
dea43a65 | 799 | cluster_data = qemu_blockalign(bs, QCOW_MAX_CRYPT_CLUSTERS * |
5ebaa27e FZ |
800 | s->cluster_size); |
801 | } | |
6f5f060b | 802 | |
3fc48d09 | 803 | assert(hd_qiov.size <= |
5ebaa27e | 804 | QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size); |
d5e6b161 | 805 | qemu_iovec_to_buf(&hd_qiov, 0, cluster_data, hd_qiov.size); |
6f5f060b | 806 | |
3fc48d09 FZ |
807 | qcow2_encrypt_sectors(s, sector_num, cluster_data, |
808 | cluster_data, cur_nr_sectors, 1, &s->aes_encrypt_key); | |
6f5f060b | 809 | |
3fc48d09 FZ |
810 | qemu_iovec_reset(&hd_qiov); |
811 | qemu_iovec_add(&hd_qiov, cluster_data, | |
5ebaa27e FZ |
812 | cur_nr_sectors * 512); |
813 | } | |
6f5f060b | 814 | |
5ebaa27e | 815 | qemu_co_mutex_unlock(&s->lock); |
67a7a0eb | 816 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_AIO); |
3cce16f4 KW |
817 | trace_qcow2_writev_data(qemu_coroutine_self(), |
818 | (cluster_offset >> 9) + index_in_cluster); | |
5ebaa27e FZ |
819 | ret = bdrv_co_writev(bs->file, |
820 | (cluster_offset >> 9) + index_in_cluster, | |
3fc48d09 | 821 | cur_nr_sectors, &hd_qiov); |
5ebaa27e FZ |
822 | qemu_co_mutex_lock(&s->lock); |
823 | if (ret < 0) { | |
3fc48d09 | 824 | goto fail; |
5ebaa27e | 825 | } |
f141eafe | 826 | |
f50f88b9 KW |
827 | if (l2meta != NULL) { |
828 | ret = qcow2_alloc_cluster_link_l2(bs, l2meta); | |
829 | if (ret < 0) { | |
830 | goto fail; | |
831 | } | |
faf575c1 | 832 | |
4e95314e KW |
833 | /* Take the request off the list of running requests */ |
834 | if (l2meta->nb_clusters != 0) { | |
835 | QLIST_REMOVE(l2meta, next_in_flight); | |
836 | } | |
837 | ||
838 | qemu_co_mutex_unlock(&s->lock); | |
839 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
840 | qemu_co_mutex_lock(&s->lock); | |
841 | ||
f50f88b9 KW |
842 | g_free(l2meta); |
843 | l2meta = NULL; | |
844 | } | |
0fa9131a | 845 | |
3fc48d09 FZ |
846 | remaining_sectors -= cur_nr_sectors; |
847 | sector_num += cur_nr_sectors; | |
848 | bytes_done += cur_nr_sectors * 512; | |
3cce16f4 | 849 | trace_qcow2_writev_done_part(qemu_coroutine_self(), cur_nr_sectors); |
5ebaa27e | 850 | } |
3fc48d09 | 851 | ret = 0; |
faf575c1 | 852 | |
3fc48d09 | 853 | fail: |
4e95314e KW |
854 | qemu_co_mutex_unlock(&s->lock); |
855 | ||
cf5c1a23 | 856 | if (l2meta != NULL) { |
4e95314e KW |
857 | if (l2meta->nb_clusters != 0) { |
858 | QLIST_REMOVE(l2meta, next_in_flight); | |
859 | } | |
860 | qemu_co_queue_restart_all(&l2meta->dependent_requests); | |
cf5c1a23 KW |
861 | g_free(l2meta); |
862 | } | |
0fa9131a | 863 | |
3fc48d09 | 864 | qemu_iovec_destroy(&hd_qiov); |
dea43a65 | 865 | qemu_vfree(cluster_data); |
3cce16f4 | 866 | trace_qcow2_writev_done_req(qemu_coroutine_self(), ret); |
42496d62 | 867 | |
68d100e9 | 868 | return ret; |
585f8587 FB |
869 | } |
870 | ||
7c80ab3f | 871 | static void qcow2_close(BlockDriverState *bs) |
585f8587 FB |
872 | { |
873 | BDRVQcowState *s = bs->opaque; | |
7267c094 | 874 | g_free(s->l1_table); |
29c1a730 KW |
875 | |
876 | qcow2_cache_flush(bs, s->l2_table_cache); | |
877 | qcow2_cache_flush(bs, s->refcount_block_cache); | |
878 | ||
c61d0004 SH |
879 | qcow2_mark_clean(bs); |
880 | ||
29c1a730 KW |
881 | qcow2_cache_destroy(bs, s->l2_table_cache); |
882 | qcow2_cache_destroy(bs, s->refcount_block_cache); | |
883 | ||
6744cbab | 884 | g_free(s->unknown_header_fields); |
75bab85c | 885 | cleanup_unknown_header_ext(bs); |
6744cbab | 886 | |
7267c094 | 887 | g_free(s->cluster_cache); |
dea43a65 | 888 | qemu_vfree(s->cluster_data); |
ed6ccf0f | 889 | qcow2_refcount_close(bs); |
28c1202b | 890 | qcow2_free_snapshots(bs); |
585f8587 FB |
891 | } |
892 | ||
06d9260f AL |
893 | static void qcow2_invalidate_cache(BlockDriverState *bs) |
894 | { | |
895 | BDRVQcowState *s = bs->opaque; | |
896 | int flags = s->flags; | |
897 | AES_KEY aes_encrypt_key; | |
898 | AES_KEY aes_decrypt_key; | |
899 | uint32_t crypt_method = 0; | |
900 | ||
901 | /* | |
902 | * Backing files are read-only which makes all of their metadata immutable, | |
903 | * that means we don't have to worry about reopening them here. | |
904 | */ | |
905 | ||
906 | if (s->crypt_method) { | |
907 | crypt_method = s->crypt_method; | |
908 | memcpy(&aes_encrypt_key, &s->aes_encrypt_key, sizeof(aes_encrypt_key)); | |
909 | memcpy(&aes_decrypt_key, &s->aes_decrypt_key, sizeof(aes_decrypt_key)); | |
910 | } | |
911 | ||
912 | qcow2_close(bs); | |
913 | ||
914 | memset(s, 0, sizeof(BDRVQcowState)); | |
915 | qcow2_open(bs, flags); | |
916 | ||
917 | if (crypt_method) { | |
918 | s->crypt_method = crypt_method; | |
919 | memcpy(&s->aes_encrypt_key, &aes_encrypt_key, sizeof(aes_encrypt_key)); | |
920 | memcpy(&s->aes_decrypt_key, &aes_decrypt_key, sizeof(aes_decrypt_key)); | |
921 | } | |
922 | } | |
923 | ||
e24e49e6 KW |
924 | static size_t header_ext_add(char *buf, uint32_t magic, const void *s, |
925 | size_t len, size_t buflen) | |
926 | { | |
927 | QCowExtension *ext_backing_fmt = (QCowExtension*) buf; | |
928 | size_t ext_len = sizeof(QCowExtension) + ((len + 7) & ~7); | |
929 | ||
930 | if (buflen < ext_len) { | |
931 | return -ENOSPC; | |
932 | } | |
933 | ||
934 | *ext_backing_fmt = (QCowExtension) { | |
935 | .magic = cpu_to_be32(magic), | |
936 | .len = cpu_to_be32(len), | |
937 | }; | |
938 | memcpy(buf + sizeof(QCowExtension), s, len); | |
939 | ||
940 | return ext_len; | |
941 | } | |
942 | ||
756e6736 | 943 | /* |
e24e49e6 KW |
944 | * Updates the qcow2 header, including the variable length parts of it, i.e. |
945 | * the backing file name and all extensions. qcow2 was not designed to allow | |
946 | * such changes, so if we run out of space (we can only use the first cluster) | |
947 | * this function may fail. | |
756e6736 KW |
948 | * |
949 | * Returns 0 on success, -errno in error cases. | |
950 | */ | |
e24e49e6 | 951 | int qcow2_update_header(BlockDriverState *bs) |
756e6736 | 952 | { |
756e6736 | 953 | BDRVQcowState *s = bs->opaque; |
e24e49e6 KW |
954 | QCowHeader *header; |
955 | char *buf; | |
956 | size_t buflen = s->cluster_size; | |
756e6736 | 957 | int ret; |
e24e49e6 KW |
958 | uint64_t total_size; |
959 | uint32_t refcount_table_clusters; | |
6744cbab | 960 | size_t header_length; |
75bab85c | 961 | Qcow2UnknownHeaderExtension *uext; |
756e6736 | 962 | |
e24e49e6 | 963 | buf = qemu_blockalign(bs, buflen); |
756e6736 | 964 | |
e24e49e6 KW |
965 | /* Header structure */ |
966 | header = (QCowHeader*) buf; | |
756e6736 | 967 | |
e24e49e6 KW |
968 | if (buflen < sizeof(*header)) { |
969 | ret = -ENOSPC; | |
970 | goto fail; | |
756e6736 KW |
971 | } |
972 | ||
6744cbab | 973 | header_length = sizeof(*header) + s->unknown_header_fields_size; |
e24e49e6 KW |
974 | total_size = bs->total_sectors * BDRV_SECTOR_SIZE; |
975 | refcount_table_clusters = s->refcount_table_size >> (s->cluster_bits - 3); | |
976 | ||
977 | *header = (QCowHeader) { | |
6744cbab | 978 | /* Version 2 fields */ |
e24e49e6 | 979 | .magic = cpu_to_be32(QCOW_MAGIC), |
6744cbab | 980 | .version = cpu_to_be32(s->qcow_version), |
e24e49e6 KW |
981 | .backing_file_offset = 0, |
982 | .backing_file_size = 0, | |
983 | .cluster_bits = cpu_to_be32(s->cluster_bits), | |
984 | .size = cpu_to_be64(total_size), | |
985 | .crypt_method = cpu_to_be32(s->crypt_method_header), | |
986 | .l1_size = cpu_to_be32(s->l1_size), | |
987 | .l1_table_offset = cpu_to_be64(s->l1_table_offset), | |
988 | .refcount_table_offset = cpu_to_be64(s->refcount_table_offset), | |
989 | .refcount_table_clusters = cpu_to_be32(refcount_table_clusters), | |
990 | .nb_snapshots = cpu_to_be32(s->nb_snapshots), | |
991 | .snapshots_offset = cpu_to_be64(s->snapshots_offset), | |
6744cbab KW |
992 | |
993 | /* Version 3 fields */ | |
994 | .incompatible_features = cpu_to_be64(s->incompatible_features), | |
995 | .compatible_features = cpu_to_be64(s->compatible_features), | |
996 | .autoclear_features = cpu_to_be64(s->autoclear_features), | |
997 | .refcount_order = cpu_to_be32(3 + REFCOUNT_SHIFT), | |
998 | .header_length = cpu_to_be32(header_length), | |
e24e49e6 | 999 | }; |
756e6736 | 1000 | |
6744cbab KW |
1001 | /* For older versions, write a shorter header */ |
1002 | switch (s->qcow_version) { | |
1003 | case 2: | |
1004 | ret = offsetof(QCowHeader, incompatible_features); | |
1005 | break; | |
1006 | case 3: | |
1007 | ret = sizeof(*header); | |
1008 | break; | |
1009 | default: | |
b6c14762 JM |
1010 | ret = -EINVAL; |
1011 | goto fail; | |
6744cbab KW |
1012 | } |
1013 | ||
1014 | buf += ret; | |
1015 | buflen -= ret; | |
1016 | memset(buf, 0, buflen); | |
1017 | ||
1018 | /* Preserve any unknown field in the header */ | |
1019 | if (s->unknown_header_fields_size) { | |
1020 | if (buflen < s->unknown_header_fields_size) { | |
1021 | ret = -ENOSPC; | |
1022 | goto fail; | |
1023 | } | |
1024 | ||
1025 | memcpy(buf, s->unknown_header_fields, s->unknown_header_fields_size); | |
1026 | buf += s->unknown_header_fields_size; | |
1027 | buflen -= s->unknown_header_fields_size; | |
1028 | } | |
756e6736 | 1029 | |
e24e49e6 KW |
1030 | /* Backing file format header extension */ |
1031 | if (*bs->backing_format) { | |
1032 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_BACKING_FORMAT, | |
1033 | bs->backing_format, strlen(bs->backing_format), | |
1034 | buflen); | |
1035 | if (ret < 0) { | |
1036 | goto fail; | |
756e6736 KW |
1037 | } |
1038 | ||
e24e49e6 KW |
1039 | buf += ret; |
1040 | buflen -= ret; | |
756e6736 KW |
1041 | } |
1042 | ||
cfcc4c62 KW |
1043 | /* Feature table */ |
1044 | Qcow2Feature features[] = { | |
c61d0004 SH |
1045 | { |
1046 | .type = QCOW2_FEAT_TYPE_INCOMPATIBLE, | |
1047 | .bit = QCOW2_INCOMPAT_DIRTY_BITNR, | |
1048 | .name = "dirty bit", | |
1049 | }, | |
bfe8043e SH |
1050 | { |
1051 | .type = QCOW2_FEAT_TYPE_COMPATIBLE, | |
1052 | .bit = QCOW2_COMPAT_LAZY_REFCOUNTS_BITNR, | |
1053 | .name = "lazy refcounts", | |
1054 | }, | |
cfcc4c62 KW |
1055 | }; |
1056 | ||
1057 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_FEATURE_TABLE, | |
1058 | features, sizeof(features), buflen); | |
1059 | if (ret < 0) { | |
1060 | goto fail; | |
1061 | } | |
1062 | buf += ret; | |
1063 | buflen -= ret; | |
1064 | ||
75bab85c KW |
1065 | /* Keep unknown header extensions */ |
1066 | QLIST_FOREACH(uext, &s->unknown_header_ext, next) { | |
1067 | ret = header_ext_add(buf, uext->magic, uext->data, uext->len, buflen); | |
1068 | if (ret < 0) { | |
1069 | goto fail; | |
1070 | } | |
1071 | ||
1072 | buf += ret; | |
1073 | buflen -= ret; | |
1074 | } | |
1075 | ||
e24e49e6 KW |
1076 | /* End of header extensions */ |
1077 | ret = header_ext_add(buf, QCOW2_EXT_MAGIC_END, NULL, 0, buflen); | |
756e6736 KW |
1078 | if (ret < 0) { |
1079 | goto fail; | |
1080 | } | |
1081 | ||
e24e49e6 KW |
1082 | buf += ret; |
1083 | buflen -= ret; | |
756e6736 | 1084 | |
e24e49e6 KW |
1085 | /* Backing file name */ |
1086 | if (*bs->backing_file) { | |
1087 | size_t backing_file_len = strlen(bs->backing_file); | |
1088 | ||
1089 | if (buflen < backing_file_len) { | |
1090 | ret = -ENOSPC; | |
1091 | goto fail; | |
1092 | } | |
1093 | ||
00ea1881 | 1094 | /* Using strncpy is ok here, since buf is not NUL-terminated. */ |
e24e49e6 KW |
1095 | strncpy(buf, bs->backing_file, buflen); |
1096 | ||
1097 | header->backing_file_offset = cpu_to_be64(buf - ((char*) header)); | |
1098 | header->backing_file_size = cpu_to_be32(backing_file_len); | |
756e6736 KW |
1099 | } |
1100 | ||
e24e49e6 KW |
1101 | /* Write the new header */ |
1102 | ret = bdrv_pwrite(bs->file, 0, header, s->cluster_size); | |
756e6736 KW |
1103 | if (ret < 0) { |
1104 | goto fail; | |
1105 | } | |
1106 | ||
1107 | ret = 0; | |
1108 | fail: | |
e24e49e6 | 1109 | qemu_vfree(header); |
756e6736 KW |
1110 | return ret; |
1111 | } | |
1112 | ||
1113 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
1114 | const char *backing_file, const char *backing_fmt) | |
1115 | { | |
e24e49e6 KW |
1116 | pstrcpy(bs->backing_file, sizeof(bs->backing_file), backing_file ?: ""); |
1117 | pstrcpy(bs->backing_format, sizeof(bs->backing_format), backing_fmt ?: ""); | |
1118 | ||
1119 | return qcow2_update_header(bs); | |
756e6736 KW |
1120 | } |
1121 | ||
a35e1c17 KW |
1122 | static int preallocate(BlockDriverState *bs) |
1123 | { | |
a35e1c17 KW |
1124 | uint64_t nb_sectors; |
1125 | uint64_t offset; | |
060bee89 | 1126 | uint64_t host_offset = 0; |
a35e1c17 | 1127 | int num; |
148da7ea | 1128 | int ret; |
f50f88b9 | 1129 | QCowL2Meta *meta; |
a35e1c17 KW |
1130 | |
1131 | nb_sectors = bdrv_getlength(bs) >> 9; | |
1132 | offset = 0; | |
1133 | ||
1134 | while (nb_sectors) { | |
1135 | num = MIN(nb_sectors, INT_MAX >> 9); | |
060bee89 KW |
1136 | ret = qcow2_alloc_cluster_offset(bs, offset, 0, num, &num, |
1137 | &host_offset, &meta); | |
148da7ea | 1138 | if (ret < 0) { |
19dbcbf7 | 1139 | return ret; |
a35e1c17 KW |
1140 | } |
1141 | ||
f50f88b9 | 1142 | ret = qcow2_alloc_cluster_link_l2(bs, meta); |
19dbcbf7 | 1143 | if (ret < 0) { |
f50f88b9 | 1144 | qcow2_free_any_clusters(bs, meta->alloc_offset, meta->nb_clusters); |
19dbcbf7 | 1145 | return ret; |
a35e1c17 KW |
1146 | } |
1147 | ||
f214978a KW |
1148 | /* There are no dependent requests, but we need to remove our request |
1149 | * from the list of in-flight requests */ | |
f50f88b9 | 1150 | if (meta != NULL) { |
4e95314e | 1151 | QLIST_REMOVE(meta, next_in_flight); |
f50f88b9 | 1152 | } |
f214978a | 1153 | |
a35e1c17 KW |
1154 | /* TODO Preallocate data if requested */ |
1155 | ||
1156 | nb_sectors -= num; | |
1157 | offset += num << 9; | |
1158 | } | |
1159 | ||
1160 | /* | |
1161 | * It is expected that the image file is large enough to actually contain | |
1162 | * all of the allocated clusters (otherwise we get failing reads after | |
1163 | * EOF). Extend the image to the last allocated sector. | |
1164 | */ | |
060bee89 | 1165 | if (host_offset != 0) { |
ea80b906 KW |
1166 | uint8_t buf[512]; |
1167 | memset(buf, 0, 512); | |
060bee89 | 1168 | ret = bdrv_write(bs->file, (host_offset >> 9) + num - 1, buf, 1); |
19dbcbf7 KW |
1169 | if (ret < 0) { |
1170 | return ret; | |
1171 | } | |
a35e1c17 KW |
1172 | } |
1173 | ||
1174 | return 0; | |
1175 | } | |
1176 | ||
7c80ab3f JS |
1177 | static int qcow2_create2(const char *filename, int64_t total_size, |
1178 | const char *backing_file, const char *backing_format, | |
1179 | int flags, size_t cluster_size, int prealloc, | |
6744cbab | 1180 | QEMUOptionParameter *options, int version) |
a9420734 | 1181 | { |
9b2260cb | 1182 | /* Calculate cluster_bits */ |
a9420734 KW |
1183 | int cluster_bits; |
1184 | cluster_bits = ffs(cluster_size) - 1; | |
1185 | if (cluster_bits < MIN_CLUSTER_BITS || cluster_bits > MAX_CLUSTER_BITS || | |
1186 | (1 << cluster_bits) != cluster_size) | |
1187 | { | |
1188 | error_report( | |
6daf194d | 1189 | "Cluster size must be a power of two between %d and %dk", |
a9420734 KW |
1190 | 1 << MIN_CLUSTER_BITS, 1 << (MAX_CLUSTER_BITS - 10)); |
1191 | return -EINVAL; | |
1192 | } | |
1193 | ||
1194 | /* | |
1195 | * Open the image file and write a minimal qcow2 header. | |
1196 | * | |
1197 | * We keep things simple and start with a zero-sized image. We also | |
1198 | * do without refcount blocks or a L1 table for now. We'll fix the | |
1199 | * inconsistency later. | |
1200 | * | |
1201 | * We do need a refcount table because growing the refcount table means | |
1202 | * allocating two new refcount blocks - the seconds of which would be at | |
1203 | * 2 GB for 64k clusters, and we don't want to have a 2 GB initial file | |
1204 | * size for any qcow2 image. | |
1205 | */ | |
1206 | BlockDriverState* bs; | |
1207 | QCowHeader header; | |
1208 | uint8_t* refcount_table; | |
1209 | int ret; | |
1210 | ||
1211 | ret = bdrv_create_file(filename, options); | |
1212 | if (ret < 0) { | |
1213 | return ret; | |
1214 | } | |
1215 | ||
1216 | ret = bdrv_file_open(&bs, filename, BDRV_O_RDWR); | |
1217 | if (ret < 0) { | |
1218 | return ret; | |
1219 | } | |
1220 | ||
1221 | /* Write the header */ | |
1222 | memset(&header, 0, sizeof(header)); | |
1223 | header.magic = cpu_to_be32(QCOW_MAGIC); | |
6744cbab | 1224 | header.version = cpu_to_be32(version); |
a9420734 KW |
1225 | header.cluster_bits = cpu_to_be32(cluster_bits); |
1226 | header.size = cpu_to_be64(0); | |
1227 | header.l1_table_offset = cpu_to_be64(0); | |
1228 | header.l1_size = cpu_to_be32(0); | |
1229 | header.refcount_table_offset = cpu_to_be64(cluster_size); | |
1230 | header.refcount_table_clusters = cpu_to_be32(1); | |
6744cbab KW |
1231 | header.refcount_order = cpu_to_be32(3 + REFCOUNT_SHIFT); |
1232 | header.header_length = cpu_to_be32(sizeof(header)); | |
a9420734 KW |
1233 | |
1234 | if (flags & BLOCK_FLAG_ENCRYPT) { | |
1235 | header.crypt_method = cpu_to_be32(QCOW_CRYPT_AES); | |
1236 | } else { | |
1237 | header.crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); | |
1238 | } | |
1239 | ||
bfe8043e SH |
1240 | if (flags & BLOCK_FLAG_LAZY_REFCOUNTS) { |
1241 | header.compatible_features |= | |
1242 | cpu_to_be64(QCOW2_COMPAT_LAZY_REFCOUNTS); | |
1243 | } | |
1244 | ||
a9420734 KW |
1245 | ret = bdrv_pwrite(bs, 0, &header, sizeof(header)); |
1246 | if (ret < 0) { | |
1247 | goto out; | |
1248 | } | |
1249 | ||
1250 | /* Write an empty refcount table */ | |
7267c094 | 1251 | refcount_table = g_malloc0(cluster_size); |
a9420734 | 1252 | ret = bdrv_pwrite(bs, cluster_size, refcount_table, cluster_size); |
7267c094 | 1253 | g_free(refcount_table); |
a9420734 KW |
1254 | |
1255 | if (ret < 0) { | |
1256 | goto out; | |
1257 | } | |
1258 | ||
1259 | bdrv_close(bs); | |
1260 | ||
1261 | /* | |
1262 | * And now open the image and make it consistent first (i.e. increase the | |
1263 | * refcount of the cluster that is occupied by the header and the refcount | |
1264 | * table) | |
1265 | */ | |
1266 | BlockDriver* drv = bdrv_find_format("qcow2"); | |
1267 | assert(drv != NULL); | |
e1a7107f KW |
1268 | ret = bdrv_open(bs, filename, |
1269 | BDRV_O_RDWR | BDRV_O_CACHE_WB | BDRV_O_NO_FLUSH, drv); | |
a9420734 KW |
1270 | if (ret < 0) { |
1271 | goto out; | |
1272 | } | |
1273 | ||
1274 | ret = qcow2_alloc_clusters(bs, 2 * cluster_size); | |
1275 | if (ret < 0) { | |
1276 | goto out; | |
1277 | ||
1278 | } else if (ret != 0) { | |
1279 | error_report("Huh, first cluster in empty image is already in use?"); | |
1280 | abort(); | |
1281 | } | |
1282 | ||
1283 | /* Okay, now that we have a valid image, let's give it the right size */ | |
1284 | ret = bdrv_truncate(bs, total_size * BDRV_SECTOR_SIZE); | |
1285 | if (ret < 0) { | |
1286 | goto out; | |
1287 | } | |
1288 | ||
1289 | /* Want a backing file? There you go.*/ | |
1290 | if (backing_file) { | |
1291 | ret = bdrv_change_backing_file(bs, backing_file, backing_format); | |
1292 | if (ret < 0) { | |
1293 | goto out; | |
1294 | } | |
1295 | } | |
1296 | ||
1297 | /* And if we're supposed to preallocate metadata, do that now */ | |
1298 | if (prealloc) { | |
15552c4a ZYW |
1299 | BDRVQcowState *s = bs->opaque; |
1300 | qemu_co_mutex_lock(&s->lock); | |
a9420734 | 1301 | ret = preallocate(bs); |
15552c4a | 1302 | qemu_co_mutex_unlock(&s->lock); |
a9420734 KW |
1303 | if (ret < 0) { |
1304 | goto out; | |
1305 | } | |
1306 | } | |
1307 | ||
1308 | ret = 0; | |
1309 | out: | |
1310 | bdrv_delete(bs); | |
1311 | return ret; | |
1312 | } | |
de5f3f40 | 1313 | |
7c80ab3f | 1314 | static int qcow2_create(const char *filename, QEMUOptionParameter *options) |
de5f3f40 KW |
1315 | { |
1316 | const char *backing_file = NULL; | |
1317 | const char *backing_fmt = NULL; | |
1318 | uint64_t sectors = 0; | |
1319 | int flags = 0; | |
99cce9fa | 1320 | size_t cluster_size = DEFAULT_CLUSTER_SIZE; |
de5f3f40 | 1321 | int prealloc = 0; |
6744cbab | 1322 | int version = 2; |
de5f3f40 KW |
1323 | |
1324 | /* Read out options */ | |
1325 | while (options && options->name) { | |
1326 | if (!strcmp(options->name, BLOCK_OPT_SIZE)) { | |
1327 | sectors = options->value.n / 512; | |
1328 | } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) { | |
1329 | backing_file = options->value.s; | |
1330 | } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FMT)) { | |
1331 | backing_fmt = options->value.s; | |
1332 | } else if (!strcmp(options->name, BLOCK_OPT_ENCRYPT)) { | |
1333 | flags |= options->value.n ? BLOCK_FLAG_ENCRYPT : 0; | |
1334 | } else if (!strcmp(options->name, BLOCK_OPT_CLUSTER_SIZE)) { | |
1335 | if (options->value.n) { | |
1336 | cluster_size = options->value.n; | |
1337 | } | |
1338 | } else if (!strcmp(options->name, BLOCK_OPT_PREALLOC)) { | |
1339 | if (!options->value.s || !strcmp(options->value.s, "off")) { | |
1340 | prealloc = 0; | |
1341 | } else if (!strcmp(options->value.s, "metadata")) { | |
1342 | prealloc = 1; | |
1343 | } else { | |
1344 | fprintf(stderr, "Invalid preallocation mode: '%s'\n", | |
1345 | options->value.s); | |
1346 | return -EINVAL; | |
1347 | } | |
6744cbab KW |
1348 | } else if (!strcmp(options->name, BLOCK_OPT_COMPAT_LEVEL)) { |
1349 | if (!options->value.s || !strcmp(options->value.s, "0.10")) { | |
1350 | version = 2; | |
1351 | } else if (!strcmp(options->value.s, "1.1")) { | |
1352 | version = 3; | |
1353 | } else { | |
1354 | fprintf(stderr, "Invalid compatibility level: '%s'\n", | |
1355 | options->value.s); | |
1356 | return -EINVAL; | |
1357 | } | |
bfe8043e SH |
1358 | } else if (!strcmp(options->name, BLOCK_OPT_LAZY_REFCOUNTS)) { |
1359 | flags |= options->value.n ? BLOCK_FLAG_LAZY_REFCOUNTS : 0; | |
de5f3f40 KW |
1360 | } |
1361 | options++; | |
1362 | } | |
1363 | ||
1364 | if (backing_file && prealloc) { | |
1365 | fprintf(stderr, "Backing file and preallocation cannot be used at " | |
1366 | "the same time\n"); | |
1367 | return -EINVAL; | |
1368 | } | |
1369 | ||
bfe8043e SH |
1370 | if (version < 3 && (flags & BLOCK_FLAG_LAZY_REFCOUNTS)) { |
1371 | fprintf(stderr, "Lazy refcounts only supported with compatibility " | |
1372 | "level 1.1 and above (use compat=1.1 or greater)\n"); | |
1373 | return -EINVAL; | |
1374 | } | |
1375 | ||
7c80ab3f | 1376 | return qcow2_create2(filename, sectors, backing_file, backing_fmt, flags, |
6744cbab | 1377 | cluster_size, prealloc, options, version); |
de5f3f40 KW |
1378 | } |
1379 | ||
7c80ab3f | 1380 | static int qcow2_make_empty(BlockDriverState *bs) |
20d97356 BS |
1381 | { |
1382 | #if 0 | |
1383 | /* XXX: not correct */ | |
1384 | BDRVQcowState *s = bs->opaque; | |
1385 | uint32_t l1_length = s->l1_size * sizeof(uint64_t); | |
1386 | int ret; | |
1387 | ||
1388 | memset(s->l1_table, 0, l1_length); | |
66f82cee | 1389 | if (bdrv_pwrite(bs->file, s->l1_table_offset, s->l1_table, l1_length) < 0) |
20d97356 | 1390 | return -1; |
66f82cee | 1391 | ret = bdrv_truncate(bs->file, s->l1_table_offset + l1_length); |
20d97356 BS |
1392 | if (ret < 0) |
1393 | return ret; | |
1394 | ||
1395 | l2_cache_reset(bs); | |
1396 | #endif | |
1397 | return 0; | |
1398 | } | |
1399 | ||
621f0589 KW |
1400 | static coroutine_fn int qcow2_co_write_zeroes(BlockDriverState *bs, |
1401 | int64_t sector_num, int nb_sectors) | |
1402 | { | |
1403 | int ret; | |
1404 | BDRVQcowState *s = bs->opaque; | |
1405 | ||
1406 | /* Emulate misaligned zero writes */ | |
1407 | if (sector_num % s->cluster_sectors || nb_sectors % s->cluster_sectors) { | |
1408 | return -ENOTSUP; | |
1409 | } | |
1410 | ||
1411 | /* Whatever is left can use real zero clusters */ | |
1412 | qemu_co_mutex_lock(&s->lock); | |
1413 | ret = qcow2_zero_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
1414 | nb_sectors); | |
1415 | qemu_co_mutex_unlock(&s->lock); | |
1416 | ||
1417 | return ret; | |
1418 | } | |
1419 | ||
6db39ae2 PB |
1420 | static coroutine_fn int qcow2_co_discard(BlockDriverState *bs, |
1421 | int64_t sector_num, int nb_sectors) | |
5ea929e3 | 1422 | { |
6db39ae2 PB |
1423 | int ret; |
1424 | BDRVQcowState *s = bs->opaque; | |
1425 | ||
1426 | qemu_co_mutex_lock(&s->lock); | |
1427 | ret = qcow2_discard_clusters(bs, sector_num << BDRV_SECTOR_BITS, | |
5ea929e3 | 1428 | nb_sectors); |
6db39ae2 PB |
1429 | qemu_co_mutex_unlock(&s->lock); |
1430 | return ret; | |
5ea929e3 KW |
1431 | } |
1432 | ||
419b19d9 SH |
1433 | static int qcow2_truncate(BlockDriverState *bs, int64_t offset) |
1434 | { | |
1435 | BDRVQcowState *s = bs->opaque; | |
1436 | int ret, new_l1_size; | |
1437 | ||
1438 | if (offset & 511) { | |
259b2173 | 1439 | error_report("The new size must be a multiple of 512"); |
419b19d9 SH |
1440 | return -EINVAL; |
1441 | } | |
1442 | ||
1443 | /* cannot proceed if image has snapshots */ | |
1444 | if (s->nb_snapshots) { | |
259b2173 | 1445 | error_report("Can't resize an image which has snapshots"); |
419b19d9 SH |
1446 | return -ENOTSUP; |
1447 | } | |
1448 | ||
1449 | /* shrinking is currently not supported */ | |
1450 | if (offset < bs->total_sectors * 512) { | |
259b2173 | 1451 | error_report("qcow2 doesn't support shrinking images yet"); |
419b19d9 SH |
1452 | return -ENOTSUP; |
1453 | } | |
1454 | ||
1455 | new_l1_size = size_to_l1(s, offset); | |
72893756 | 1456 | ret = qcow2_grow_l1_table(bs, new_l1_size, true); |
419b19d9 SH |
1457 | if (ret < 0) { |
1458 | return ret; | |
1459 | } | |
1460 | ||
1461 | /* write updated header.size */ | |
1462 | offset = cpu_to_be64(offset); | |
8b3b7206 KW |
1463 | ret = bdrv_pwrite_sync(bs->file, offsetof(QCowHeader, size), |
1464 | &offset, sizeof(uint64_t)); | |
419b19d9 SH |
1465 | if (ret < 0) { |
1466 | return ret; | |
1467 | } | |
1468 | ||
1469 | s->l1_vm_state_index = new_l1_size; | |
1470 | return 0; | |
1471 | } | |
1472 | ||
20d97356 BS |
1473 | /* XXX: put compressed sectors first, then all the cluster aligned |
1474 | tables to avoid losing bytes in alignment */ | |
7c80ab3f JS |
1475 | static int qcow2_write_compressed(BlockDriverState *bs, int64_t sector_num, |
1476 | const uint8_t *buf, int nb_sectors) | |
20d97356 BS |
1477 | { |
1478 | BDRVQcowState *s = bs->opaque; | |
1479 | z_stream strm; | |
1480 | int ret, out_len; | |
1481 | uint8_t *out_buf; | |
1482 | uint64_t cluster_offset; | |
1483 | ||
1484 | if (nb_sectors == 0) { | |
1485 | /* align end of file to a sector boundary to ease reading with | |
1486 | sector based I/Os */ | |
66f82cee | 1487 | cluster_offset = bdrv_getlength(bs->file); |
20d97356 | 1488 | cluster_offset = (cluster_offset + 511) & ~511; |
66f82cee | 1489 | bdrv_truncate(bs->file, cluster_offset); |
20d97356 BS |
1490 | return 0; |
1491 | } | |
1492 | ||
1493 | if (nb_sectors != s->cluster_sectors) | |
1494 | return -EINVAL; | |
1495 | ||
7267c094 | 1496 | out_buf = g_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); |
20d97356 BS |
1497 | |
1498 | /* best compression, small window, no zlib header */ | |
1499 | memset(&strm, 0, sizeof(strm)); | |
1500 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
1501 | Z_DEFLATED, -12, | |
1502 | 9, Z_DEFAULT_STRATEGY); | |
1503 | if (ret != 0) { | |
8f1efd00 KW |
1504 | ret = -EINVAL; |
1505 | goto fail; | |
20d97356 BS |
1506 | } |
1507 | ||
1508 | strm.avail_in = s->cluster_size; | |
1509 | strm.next_in = (uint8_t *)buf; | |
1510 | strm.avail_out = s->cluster_size; | |
1511 | strm.next_out = out_buf; | |
1512 | ||
1513 | ret = deflate(&strm, Z_FINISH); | |
1514 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
20d97356 | 1515 | deflateEnd(&strm); |
8f1efd00 KW |
1516 | ret = -EINVAL; |
1517 | goto fail; | |
20d97356 BS |
1518 | } |
1519 | out_len = strm.next_out - out_buf; | |
1520 | ||
1521 | deflateEnd(&strm); | |
1522 | ||
1523 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
1524 | /* could not compress: write normal cluster */ | |
8f1efd00 KW |
1525 | ret = bdrv_write(bs, sector_num, buf, s->cluster_sectors); |
1526 | if (ret < 0) { | |
1527 | goto fail; | |
1528 | } | |
20d97356 BS |
1529 | } else { |
1530 | cluster_offset = qcow2_alloc_compressed_cluster_offset(bs, | |
1531 | sector_num << 9, out_len); | |
8f1efd00 KW |
1532 | if (!cluster_offset) { |
1533 | ret = -EIO; | |
1534 | goto fail; | |
1535 | } | |
20d97356 | 1536 | cluster_offset &= s->cluster_offset_mask; |
66f82cee | 1537 | BLKDBG_EVENT(bs->file, BLKDBG_WRITE_COMPRESSED); |
8f1efd00 KW |
1538 | ret = bdrv_pwrite(bs->file, cluster_offset, out_buf, out_len); |
1539 | if (ret < 0) { | |
1540 | goto fail; | |
20d97356 BS |
1541 | } |
1542 | } | |
1543 | ||
8f1efd00 KW |
1544 | ret = 0; |
1545 | fail: | |
7267c094 | 1546 | g_free(out_buf); |
8f1efd00 | 1547 | return ret; |
20d97356 BS |
1548 | } |
1549 | ||
a968168c | 1550 | static coroutine_fn int qcow2_co_flush_to_os(BlockDriverState *bs) |
20d97356 | 1551 | { |
29c1a730 KW |
1552 | BDRVQcowState *s = bs->opaque; |
1553 | int ret; | |
1554 | ||
8b94ff85 | 1555 | qemu_co_mutex_lock(&s->lock); |
29c1a730 KW |
1556 | ret = qcow2_cache_flush(bs, s->l2_table_cache); |
1557 | if (ret < 0) { | |
c95de7e2 | 1558 | qemu_co_mutex_unlock(&s->lock); |
8b94ff85 | 1559 | return ret; |
29c1a730 KW |
1560 | } |
1561 | ||
bfe8043e SH |
1562 | if (qcow2_need_accurate_refcounts(s)) { |
1563 | ret = qcow2_cache_flush(bs, s->refcount_block_cache); | |
1564 | if (ret < 0) { | |
1565 | qemu_co_mutex_unlock(&s->lock); | |
1566 | return ret; | |
1567 | } | |
29c1a730 | 1568 | } |
8b94ff85 | 1569 | qemu_co_mutex_unlock(&s->lock); |
29c1a730 | 1570 | |
eb489bb1 KW |
1571 | return 0; |
1572 | } | |
1573 | ||
7c80ab3f | 1574 | static int64_t qcow2_vm_state_offset(BDRVQcowState *s) |
20d97356 BS |
1575 | { |
1576 | return (int64_t)s->l1_vm_state_index << (s->cluster_bits + s->l2_bits); | |
1577 | } | |
1578 | ||
7c80ab3f | 1579 | static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) |
20d97356 BS |
1580 | { |
1581 | BDRVQcowState *s = bs->opaque; | |
1582 | bdi->cluster_size = s->cluster_size; | |
7c80ab3f | 1583 | bdi->vm_state_offset = qcow2_vm_state_offset(s); |
20d97356 BS |
1584 | return 0; |
1585 | } | |
1586 | ||
20d97356 BS |
1587 | #if 0 |
1588 | static void dump_refcounts(BlockDriverState *bs) | |
1589 | { | |
1590 | BDRVQcowState *s = bs->opaque; | |
1591 | int64_t nb_clusters, k, k1, size; | |
1592 | int refcount; | |
1593 | ||
66f82cee | 1594 | size = bdrv_getlength(bs->file); |
20d97356 BS |
1595 | nb_clusters = size_to_clusters(s, size); |
1596 | for(k = 0; k < nb_clusters;) { | |
1597 | k1 = k; | |
1598 | refcount = get_refcount(bs, k); | |
1599 | k++; | |
1600 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
1601 | k++; | |
0bfcd599 BS |
1602 | printf("%" PRId64 ": refcount=%d nb=%" PRId64 "\n", k, refcount, |
1603 | k - k1); | |
20d97356 BS |
1604 | } |
1605 | } | |
1606 | #endif | |
1607 | ||
7c80ab3f JS |
1608 | static int qcow2_save_vmstate(BlockDriverState *bs, const uint8_t *buf, |
1609 | int64_t pos, int size) | |
20d97356 BS |
1610 | { |
1611 | BDRVQcowState *s = bs->opaque; | |
1612 | int growable = bs->growable; | |
1613 | int ret; | |
1614 | ||
66f82cee | 1615 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_SAVE); |
20d97356 | 1616 | bs->growable = 1; |
7c80ab3f | 1617 | ret = bdrv_pwrite(bs, qcow2_vm_state_offset(s) + pos, buf, size); |
20d97356 BS |
1618 | bs->growable = growable; |
1619 | ||
1620 | return ret; | |
1621 | } | |
1622 | ||
7c80ab3f JS |
1623 | static int qcow2_load_vmstate(BlockDriverState *bs, uint8_t *buf, |
1624 | int64_t pos, int size) | |
20d97356 BS |
1625 | { |
1626 | BDRVQcowState *s = bs->opaque; | |
1627 | int growable = bs->growable; | |
1628 | int ret; | |
1629 | ||
66f82cee | 1630 | BLKDBG_EVENT(bs->file, BLKDBG_VMSTATE_LOAD); |
20d97356 | 1631 | bs->growable = 1; |
7c80ab3f | 1632 | ret = bdrv_pread(bs, qcow2_vm_state_offset(s) + pos, buf, size); |
20d97356 BS |
1633 | bs->growable = growable; |
1634 | ||
1635 | return ret; | |
1636 | } | |
1637 | ||
7c80ab3f | 1638 | static QEMUOptionParameter qcow2_create_options[] = { |
20d97356 BS |
1639 | { |
1640 | .name = BLOCK_OPT_SIZE, | |
1641 | .type = OPT_SIZE, | |
1642 | .help = "Virtual disk size" | |
1643 | }, | |
6744cbab KW |
1644 | { |
1645 | .name = BLOCK_OPT_COMPAT_LEVEL, | |
1646 | .type = OPT_STRING, | |
1647 | .help = "Compatibility level (0.10 or 1.1)" | |
1648 | }, | |
20d97356 BS |
1649 | { |
1650 | .name = BLOCK_OPT_BACKING_FILE, | |
1651 | .type = OPT_STRING, | |
1652 | .help = "File name of a base image" | |
1653 | }, | |
1654 | { | |
1655 | .name = BLOCK_OPT_BACKING_FMT, | |
1656 | .type = OPT_STRING, | |
1657 | .help = "Image format of the base image" | |
1658 | }, | |
1659 | { | |
1660 | .name = BLOCK_OPT_ENCRYPT, | |
1661 | .type = OPT_FLAG, | |
1662 | .help = "Encrypt the image" | |
1663 | }, | |
1664 | { | |
1665 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
1666 | .type = OPT_SIZE, | |
99cce9fa KW |
1667 | .help = "qcow2 cluster size", |
1668 | .value = { .n = DEFAULT_CLUSTER_SIZE }, | |
20d97356 BS |
1669 | }, |
1670 | { | |
1671 | .name = BLOCK_OPT_PREALLOC, | |
1672 | .type = OPT_STRING, | |
1673 | .help = "Preallocation mode (allowed values: off, metadata)" | |
1674 | }, | |
bfe8043e SH |
1675 | { |
1676 | .name = BLOCK_OPT_LAZY_REFCOUNTS, | |
1677 | .type = OPT_FLAG, | |
1678 | .help = "Postpone refcount updates", | |
1679 | }, | |
20d97356 BS |
1680 | { NULL } |
1681 | }; | |
1682 | ||
1683 | static BlockDriver bdrv_qcow2 = { | |
7c80ab3f JS |
1684 | .format_name = "qcow2", |
1685 | .instance_size = sizeof(BDRVQcowState), | |
1686 | .bdrv_probe = qcow2_probe, | |
1687 | .bdrv_open = qcow2_open, | |
1688 | .bdrv_close = qcow2_close, | |
21d82ac9 | 1689 | .bdrv_reopen_prepare = qcow2_reopen_prepare, |
7c80ab3f | 1690 | .bdrv_create = qcow2_create, |
f8a2e5e3 | 1691 | .bdrv_co_is_allocated = qcow2_co_is_allocated, |
7c80ab3f JS |
1692 | .bdrv_set_key = qcow2_set_key, |
1693 | .bdrv_make_empty = qcow2_make_empty, | |
1694 | ||
c68b89ac KW |
1695 | .bdrv_co_readv = qcow2_co_readv, |
1696 | .bdrv_co_writev = qcow2_co_writev, | |
eb489bb1 | 1697 | .bdrv_co_flush_to_os = qcow2_co_flush_to_os, |
419b19d9 | 1698 | |
621f0589 | 1699 | .bdrv_co_write_zeroes = qcow2_co_write_zeroes, |
6db39ae2 | 1700 | .bdrv_co_discard = qcow2_co_discard, |
419b19d9 | 1701 | .bdrv_truncate = qcow2_truncate, |
7c80ab3f | 1702 | .bdrv_write_compressed = qcow2_write_compressed, |
20d97356 BS |
1703 | |
1704 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
1705 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
1706 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
1707 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
51ef6727 | 1708 | .bdrv_snapshot_load_tmp = qcow2_snapshot_load_tmp, |
7c80ab3f | 1709 | .bdrv_get_info = qcow2_get_info, |
20d97356 | 1710 | |
7c80ab3f JS |
1711 | .bdrv_save_vmstate = qcow2_save_vmstate, |
1712 | .bdrv_load_vmstate = qcow2_load_vmstate, | |
20d97356 BS |
1713 | |
1714 | .bdrv_change_backing_file = qcow2_change_backing_file, | |
1715 | ||
06d9260f AL |
1716 | .bdrv_invalidate_cache = qcow2_invalidate_cache, |
1717 | ||
7c80ab3f JS |
1718 | .create_options = qcow2_create_options, |
1719 | .bdrv_check = qcow2_check, | |
20d97356 BS |
1720 | }; |
1721 | ||
5efa9d5a AL |
1722 | static void bdrv_qcow2_init(void) |
1723 | { | |
1724 | bdrv_register(&bdrv_qcow2); | |
1725 | } | |
1726 | ||
1727 | block_init(bdrv_qcow2_init); |