]>
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" |
585f8587 | 25 | #include "block_int.h" |
5efa9d5a | 26 | #include "module.h" |
585f8587 FB |
27 | #include <zlib.h> |
28 | #include "aes.h" | |
f7d0fe02 | 29 | #include "block/qcow2.h" |
585f8587 FB |
30 | |
31 | /* | |
32 | Differences with QCOW: | |
33 | ||
34 | - Support for multiple incremental snapshots. | |
35 | - Memory management by reference counts. | |
36 | - Clusters which have a reference count of one have the bit | |
37 | QCOW_OFLAG_COPIED to optimize write performance. | |
5fafdf24 | 38 | - Size of compressed clusters is stored in sectors to reduce bit usage |
585f8587 FB |
39 | in the cluster offsets. |
40 | - Support for storing additional data (such as the VM state) in the | |
3b46e624 | 41 | snapshots. |
585f8587 FB |
42 | - If a backing store is used, the cluster size is not constrained |
43 | (could be backported to QCOW). | |
44 | - L2 tables have always a size of one cluster. | |
45 | */ | |
46 | ||
9b80ddf3 AL |
47 | |
48 | typedef struct { | |
49 | uint32_t magic; | |
50 | uint32_t len; | |
51 | } QCowExtension; | |
52 | #define QCOW_EXT_MAGIC_END 0 | |
f965509c | 53 | #define QCOW_EXT_MAGIC_BACKING_FORMAT 0xE2792ACA |
9b80ddf3 | 54 | |
20d97356 | 55 | static int qcow_create(const char *filename, QEMUOptionParameter *options); |
585f8587 FB |
56 | |
57 | static int qcow_probe(const uint8_t *buf, int buf_size, const char *filename) | |
58 | { | |
59 | const QCowHeader *cow_header = (const void *)buf; | |
3b46e624 | 60 | |
585f8587 FB |
61 | if (buf_size >= sizeof(QCowHeader) && |
62 | be32_to_cpu(cow_header->magic) == QCOW_MAGIC && | |
5fafdf24 | 63 | be32_to_cpu(cow_header->version) == QCOW_VERSION) |
585f8587 FB |
64 | return 100; |
65 | else | |
66 | return 0; | |
67 | } | |
68 | ||
9b80ddf3 AL |
69 | |
70 | /* | |
71 | * read qcow2 extension and fill bs | |
72 | * start reading from start_offset | |
73 | * finish reading upon magic of value 0 or when end_offset reached | |
74 | * unknown magic is skipped (future extension this version knows nothing about) | |
75 | * return 0 upon success, non-0 otherwise | |
76 | */ | |
77 | static int qcow_read_extensions(BlockDriverState *bs, uint64_t start_offset, | |
78 | uint64_t end_offset) | |
79 | { | |
80 | BDRVQcowState *s = bs->opaque; | |
81 | QCowExtension ext; | |
82 | uint64_t offset; | |
83 | ||
84 | #ifdef DEBUG_EXT | |
85 | printf("qcow_read_extensions: start=%ld end=%ld\n", start_offset, end_offset); | |
86 | #endif | |
87 | offset = start_offset; | |
88 | while (offset < end_offset) { | |
89 | ||
90 | #ifdef DEBUG_EXT | |
91 | /* Sanity check */ | |
92 | if (offset > s->cluster_size) | |
93 | printf("qcow_handle_extension: suspicious offset %lu\n", offset); | |
94 | ||
95 | printf("attemting to read extended header in offset %lu\n", offset); | |
96 | #endif | |
97 | ||
98 | if (bdrv_pread(s->hd, offset, &ext, sizeof(ext)) != sizeof(ext)) { | |
4c978075 AL |
99 | fprintf(stderr, "qcow_handle_extension: ERROR: pread fail from offset %llu\n", |
100 | (unsigned long long)offset); | |
9b80ddf3 AL |
101 | return 1; |
102 | } | |
103 | be32_to_cpus(&ext.magic); | |
104 | be32_to_cpus(&ext.len); | |
105 | offset += sizeof(ext); | |
106 | #ifdef DEBUG_EXT | |
107 | printf("ext.magic = 0x%x\n", ext.magic); | |
108 | #endif | |
109 | switch (ext.magic) { | |
110 | case QCOW_EXT_MAGIC_END: | |
111 | return 0; | |
f965509c AL |
112 | |
113 | case QCOW_EXT_MAGIC_BACKING_FORMAT: | |
114 | if (ext.len >= sizeof(bs->backing_format)) { | |
115 | fprintf(stderr, "ERROR: ext_backing_format: len=%u too large" | |
4c978075 | 116 | " (>=%zu)\n", |
f965509c AL |
117 | ext.len, sizeof(bs->backing_format)); |
118 | return 2; | |
119 | } | |
120 | if (bdrv_pread(s->hd, offset , bs->backing_format, | |
121 | ext.len) != ext.len) | |
122 | return 3; | |
123 | bs->backing_format[ext.len] = '\0'; | |
124 | #ifdef DEBUG_EXT | |
125 | printf("Qcow2: Got format extension %s\n", bs->backing_format); | |
126 | #endif | |
e1c7f0e3 | 127 | offset = ((offset + ext.len + 7) & ~7); |
f965509c AL |
128 | break; |
129 | ||
9b80ddf3 AL |
130 | default: |
131 | /* unknown magic -- just skip it */ | |
e1c7f0e3 | 132 | offset = ((offset + ext.len + 7) & ~7); |
9b80ddf3 AL |
133 | break; |
134 | } | |
135 | } | |
136 | ||
137 | return 0; | |
138 | } | |
139 | ||
140 | ||
585f8587 FB |
141 | static int qcow_open(BlockDriverState *bs, const char *filename, int flags) |
142 | { | |
143 | BDRVQcowState *s = bs->opaque; | |
144 | int len, i, shift, ret; | |
145 | QCowHeader header; | |
9b80ddf3 | 146 | uint64_t ext_end; |
585f8587 | 147 | |
b5eff355 | 148 | ret = bdrv_file_open(&s->hd, filename, flags); |
585f8587 FB |
149 | if (ret < 0) |
150 | return ret; | |
151 | if (bdrv_pread(s->hd, 0, &header, sizeof(header)) != sizeof(header)) | |
152 | goto fail; | |
153 | be32_to_cpus(&header.magic); | |
154 | be32_to_cpus(&header.version); | |
155 | be64_to_cpus(&header.backing_file_offset); | |
156 | be32_to_cpus(&header.backing_file_size); | |
157 | be64_to_cpus(&header.size); | |
158 | be32_to_cpus(&header.cluster_bits); | |
159 | be32_to_cpus(&header.crypt_method); | |
160 | be64_to_cpus(&header.l1_table_offset); | |
161 | be32_to_cpus(&header.l1_size); | |
162 | be64_to_cpus(&header.refcount_table_offset); | |
163 | be32_to_cpus(&header.refcount_table_clusters); | |
164 | be64_to_cpus(&header.snapshots_offset); | |
165 | be32_to_cpus(&header.nb_snapshots); | |
3b46e624 | 166 | |
585f8587 FB |
167 | if (header.magic != QCOW_MAGIC || header.version != QCOW_VERSION) |
168 | goto fail; | |
d191d12d | 169 | if (header.cluster_bits < MIN_CLUSTER_BITS || |
73c632ed | 170 | header.cluster_bits > MAX_CLUSTER_BITS) |
585f8587 FB |
171 | goto fail; |
172 | if (header.crypt_method > QCOW_CRYPT_AES) | |
173 | goto fail; | |
174 | s->crypt_method_header = header.crypt_method; | |
175 | if (s->crypt_method_header) | |
176 | bs->encrypted = 1; | |
177 | s->cluster_bits = header.cluster_bits; | |
178 | s->cluster_size = 1 << s->cluster_bits; | |
179 | s->cluster_sectors = 1 << (s->cluster_bits - 9); | |
180 | s->l2_bits = s->cluster_bits - 3; /* L2 is always one cluster */ | |
181 | s->l2_size = 1 << s->l2_bits; | |
182 | bs->total_sectors = header.size / 512; | |
183 | s->csize_shift = (62 - (s->cluster_bits - 8)); | |
184 | s->csize_mask = (1 << (s->cluster_bits - 8)) - 1; | |
185 | s->cluster_offset_mask = (1LL << s->csize_shift) - 1; | |
186 | s->refcount_table_offset = header.refcount_table_offset; | |
5fafdf24 | 187 | s->refcount_table_size = |
585f8587 FB |
188 | header.refcount_table_clusters << (s->cluster_bits - 3); |
189 | ||
190 | s->snapshots_offset = header.snapshots_offset; | |
191 | s->nb_snapshots = header.nb_snapshots; | |
192 | ||
193 | /* read the level 1 table */ | |
194 | s->l1_size = header.l1_size; | |
195 | shift = s->cluster_bits + s->l2_bits; | |
196 | s->l1_vm_state_index = (header.size + (1LL << shift) - 1) >> shift; | |
197 | /* the L1 table must contain at least enough entries to put | |
198 | header.size bytes */ | |
199 | if (s->l1_size < s->l1_vm_state_index) | |
200 | goto fail; | |
201 | s->l1_table_offset = header.l1_table_offset; | |
d191d12d SW |
202 | if (s->l1_size > 0) { |
203 | s->l1_table = qemu_mallocz( | |
204 | align_offset(s->l1_size * sizeof(uint64_t), 512)); | |
205 | if (bdrv_pread(s->hd, s->l1_table_offset, s->l1_table, s->l1_size * sizeof(uint64_t)) != | |
206 | s->l1_size * sizeof(uint64_t)) | |
207 | goto fail; | |
208 | for(i = 0;i < s->l1_size; i++) { | |
209 | be64_to_cpus(&s->l1_table[i]); | |
210 | } | |
585f8587 FB |
211 | } |
212 | /* alloc L2 cache */ | |
213 | s->l2_cache = qemu_malloc(s->l2_size * L2_CACHE_SIZE * sizeof(uint64_t)); | |
585f8587 | 214 | s->cluster_cache = qemu_malloc(s->cluster_size); |
585f8587 | 215 | /* one more sector for decompressed data alignment */ |
095a9c58 AL |
216 | s->cluster_data = qemu_malloc(QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size |
217 | + 512); | |
585f8587 | 218 | s->cluster_cache_offset = -1; |
3b46e624 | 219 | |
ed6ccf0f | 220 | if (qcow2_refcount_init(bs) < 0) |
585f8587 FB |
221 | goto fail; |
222 | ||
72cf2d4f | 223 | QLIST_INIT(&s->cluster_allocs); |
f214978a | 224 | |
9b80ddf3 AL |
225 | /* read qcow2 extensions */ |
226 | if (header.backing_file_offset) | |
227 | ext_end = header.backing_file_offset; | |
228 | else | |
229 | ext_end = s->cluster_size; | |
230 | if (qcow_read_extensions(bs, sizeof(header), ext_end)) | |
231 | goto fail; | |
232 | ||
585f8587 FB |
233 | /* read the backing file name */ |
234 | if (header.backing_file_offset != 0) { | |
235 | len = header.backing_file_size; | |
236 | if (len > 1023) | |
237 | len = 1023; | |
238 | if (bdrv_pread(s->hd, header.backing_file_offset, bs->backing_file, len) != len) | |
239 | goto fail; | |
240 | bs->backing_file[len] = '\0'; | |
241 | } | |
ed6ccf0f | 242 | if (qcow2_read_snapshots(bs) < 0) |
585f8587 FB |
243 | goto fail; |
244 | ||
245 | #ifdef DEBUG_ALLOC | |
14899cdf | 246 | qcow2_check_refcounts(bs); |
585f8587 FB |
247 | #endif |
248 | return 0; | |
249 | ||
250 | fail: | |
ed6ccf0f KW |
251 | qcow2_free_snapshots(bs); |
252 | qcow2_refcount_close(bs); | |
585f8587 FB |
253 | qemu_free(s->l1_table); |
254 | qemu_free(s->l2_cache); | |
255 | qemu_free(s->cluster_cache); | |
256 | qemu_free(s->cluster_data); | |
257 | bdrv_delete(s->hd); | |
258 | return -1; | |
259 | } | |
260 | ||
261 | static int qcow_set_key(BlockDriverState *bs, const char *key) | |
262 | { | |
263 | BDRVQcowState *s = bs->opaque; | |
264 | uint8_t keybuf[16]; | |
265 | int len, i; | |
3b46e624 | 266 | |
585f8587 FB |
267 | memset(keybuf, 0, 16); |
268 | len = strlen(key); | |
269 | if (len > 16) | |
270 | len = 16; | |
271 | /* XXX: we could compress the chars to 7 bits to increase | |
272 | entropy */ | |
273 | for(i = 0;i < len;i++) { | |
274 | keybuf[i] = key[i]; | |
275 | } | |
276 | s->crypt_method = s->crypt_method_header; | |
277 | ||
278 | if (AES_set_encrypt_key(keybuf, 128, &s->aes_encrypt_key) != 0) | |
279 | return -1; | |
280 | if (AES_set_decrypt_key(keybuf, 128, &s->aes_decrypt_key) != 0) | |
281 | return -1; | |
282 | #if 0 | |
283 | /* test */ | |
284 | { | |
285 | uint8_t in[16]; | |
286 | uint8_t out[16]; | |
287 | uint8_t tmp[16]; | |
288 | for(i=0;i<16;i++) | |
289 | in[i] = i; | |
290 | AES_encrypt(in, tmp, &s->aes_encrypt_key); | |
291 | AES_decrypt(tmp, out, &s->aes_decrypt_key); | |
292 | for(i = 0; i < 16; i++) | |
293 | printf(" %02x", tmp[i]); | |
294 | printf("\n"); | |
295 | for(i = 0; i < 16; i++) | |
296 | printf(" %02x", out[i]); | |
297 | printf("\n"); | |
298 | } | |
299 | #endif | |
300 | return 0; | |
301 | } | |
302 | ||
5fafdf24 | 303 | static int qcow_is_allocated(BlockDriverState *bs, int64_t sector_num, |
585f8587 FB |
304 | int nb_sectors, int *pnum) |
305 | { | |
585f8587 FB |
306 | uint64_t cluster_offset; |
307 | ||
095a9c58 | 308 | *pnum = nb_sectors; |
ed6ccf0f | 309 | cluster_offset = qcow2_get_cluster_offset(bs, sector_num << 9, pnum); |
095a9c58 | 310 | |
585f8587 FB |
311 | return (cluster_offset != 0); |
312 | } | |
313 | ||
a9465922 | 314 | /* handle reading after the end of the backing file */ |
ed6ccf0f | 315 | int qcow2_backing_read1(BlockDriverState *bs, |
45aba42f | 316 | int64_t sector_num, uint8_t *buf, int nb_sectors) |
a9465922 FB |
317 | { |
318 | int n1; | |
319 | if ((sector_num + nb_sectors) <= bs->total_sectors) | |
320 | return nb_sectors; | |
321 | if (sector_num >= bs->total_sectors) | |
322 | n1 = 0; | |
323 | else | |
324 | n1 = bs->total_sectors - sector_num; | |
325 | memset(buf + n1 * 512, 0, 512 * (nb_sectors - n1)); | |
326 | return n1; | |
327 | } | |
328 | ||
ce1a14dc PB |
329 | typedef struct QCowAIOCB { |
330 | BlockDriverAIOCB common; | |
585f8587 | 331 | int64_t sector_num; |
f141eafe | 332 | QEMUIOVector *qiov; |
585f8587 | 333 | uint8_t *buf; |
f141eafe | 334 | void *orig_buf; |
7b88e48b CH |
335 | int remaining_sectors; |
336 | int cur_nr_sectors; /* number of sectors in current iteration */ | |
585f8587 | 337 | uint64_t cluster_offset; |
5fafdf24 | 338 | uint8_t *cluster_data; |
585f8587 | 339 | BlockDriverAIOCB *hd_aiocb; |
c87c0672 AL |
340 | struct iovec hd_iov; |
341 | QEMUIOVector hd_qiov; | |
1490791f | 342 | QEMUBH *bh; |
e976c6a1 | 343 | QCowL2Meta l2meta; |
72cf2d4f | 344 | QLIST_ENTRY(QCowAIOCB) next_depend; |
585f8587 FB |
345 | } QCowAIOCB; |
346 | ||
c16b5a2c CH |
347 | static void qcow_aio_cancel(BlockDriverAIOCB *blockacb) |
348 | { | |
349 | QCowAIOCB *acb = (QCowAIOCB *)blockacb; | |
350 | if (acb->hd_aiocb) | |
351 | bdrv_aio_cancel(acb->hd_aiocb); | |
352 | qemu_aio_release(acb); | |
353 | } | |
354 | ||
355 | static AIOPool qcow_aio_pool = { | |
356 | .aiocb_size = sizeof(QCowAIOCB), | |
357 | .cancel = qcow_aio_cancel, | |
358 | }; | |
359 | ||
1490791f AL |
360 | static void qcow_aio_read_cb(void *opaque, int ret); |
361 | static void qcow_aio_read_bh(void *opaque) | |
362 | { | |
363 | QCowAIOCB *acb = opaque; | |
364 | qemu_bh_delete(acb->bh); | |
365 | acb->bh = NULL; | |
366 | qcow_aio_read_cb(opaque, 0); | |
367 | } | |
368 | ||
a32ef786 AL |
369 | static int qcow_schedule_bh(QEMUBHFunc *cb, QCowAIOCB *acb) |
370 | { | |
371 | if (acb->bh) | |
372 | return -EIO; | |
373 | ||
374 | acb->bh = qemu_bh_new(cb, acb); | |
375 | if (!acb->bh) | |
376 | return -EIO; | |
377 | ||
378 | qemu_bh_schedule(acb->bh); | |
379 | ||
380 | return 0; | |
381 | } | |
382 | ||
585f8587 FB |
383 | static void qcow_aio_read_cb(void *opaque, int ret) |
384 | { | |
ce1a14dc PB |
385 | QCowAIOCB *acb = opaque; |
386 | BlockDriverState *bs = acb->common.bs; | |
585f8587 | 387 | BDRVQcowState *s = bs->opaque; |
a9465922 | 388 | int index_in_cluster, n1; |
585f8587 | 389 | |
ce1a14dc | 390 | acb->hd_aiocb = NULL; |
f141eafe AL |
391 | if (ret < 0) |
392 | goto done; | |
585f8587 | 393 | |
585f8587 | 394 | /* post process the read buffer */ |
ce1a14dc | 395 | if (!acb->cluster_offset) { |
585f8587 | 396 | /* nothing to do */ |
ce1a14dc | 397 | } else if (acb->cluster_offset & QCOW_OFLAG_COMPRESSED) { |
585f8587 FB |
398 | /* nothing to do */ |
399 | } else { | |
400 | if (s->crypt_method) { | |
ed6ccf0f | 401 | qcow2_encrypt_sectors(s, acb->sector_num, acb->buf, acb->buf, |
7b88e48b | 402 | acb->cur_nr_sectors, 0, |
585f8587 FB |
403 | &s->aes_decrypt_key); |
404 | } | |
405 | } | |
406 | ||
7b88e48b CH |
407 | acb->remaining_sectors -= acb->cur_nr_sectors; |
408 | acb->sector_num += acb->cur_nr_sectors; | |
409 | acb->buf += acb->cur_nr_sectors * 512; | |
585f8587 | 410 | |
7b88e48b | 411 | if (acb->remaining_sectors == 0) { |
585f8587 | 412 | /* request completed */ |
f141eafe AL |
413 | ret = 0; |
414 | goto done; | |
585f8587 | 415 | } |
3b46e624 | 416 | |
585f8587 | 417 | /* prepare next AIO request */ |
7b88e48b CH |
418 | acb->cur_nr_sectors = acb->remaining_sectors; |
419 | acb->cluster_offset = qcow2_get_cluster_offset(bs, acb->sector_num << 9, | |
420 | &acb->cur_nr_sectors); | |
ce1a14dc | 421 | index_in_cluster = acb->sector_num & (s->cluster_sectors - 1); |
ce1a14dc PB |
422 | |
423 | if (!acb->cluster_offset) { | |
585f8587 FB |
424 | if (bs->backing_hd) { |
425 | /* read from the base image */ | |
ed6ccf0f | 426 | n1 = qcow2_backing_read1(bs->backing_hd, acb->sector_num, |
7b88e48b | 427 | acb->buf, acb->cur_nr_sectors); |
a9465922 | 428 | if (n1 > 0) { |
3f4cb3d3 | 429 | acb->hd_iov.iov_base = (void *)acb->buf; |
7b88e48b | 430 | acb->hd_iov.iov_len = acb->cur_nr_sectors * 512; |
c87c0672 | 431 | qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); |
8252278a | 432 | BLKDBG_EVENT(s->hd, BLKDBG_READ_BACKING_AIO); |
c87c0672 | 433 | acb->hd_aiocb = bdrv_aio_readv(bs->backing_hd, acb->sector_num, |
7b88e48b | 434 | &acb->hd_qiov, acb->cur_nr_sectors, |
c87c0672 | 435 | qcow_aio_read_cb, acb); |
ce1a14dc | 436 | if (acb->hd_aiocb == NULL) |
f141eafe | 437 | goto done; |
a9465922 | 438 | } else { |
a32ef786 AL |
439 | ret = qcow_schedule_bh(qcow_aio_read_bh, acb); |
440 | if (ret < 0) | |
f141eafe | 441 | goto done; |
a9465922 | 442 | } |
585f8587 FB |
443 | } else { |
444 | /* Note: in this case, no need to wait */ | |
7b88e48b | 445 | memset(acb->buf, 0, 512 * acb->cur_nr_sectors); |
a32ef786 AL |
446 | ret = qcow_schedule_bh(qcow_aio_read_bh, acb); |
447 | if (ret < 0) | |
f141eafe | 448 | goto done; |
585f8587 | 449 | } |
ce1a14dc | 450 | } else if (acb->cluster_offset & QCOW_OFLAG_COMPRESSED) { |
585f8587 | 451 | /* add AIO support for compressed blocks ? */ |
ed6ccf0f | 452 | if (qcow2_decompress_cluster(s, acb->cluster_offset) < 0) |
f141eafe | 453 | goto done; |
7b88e48b CH |
454 | memcpy(acb->buf, s->cluster_cache + index_in_cluster * 512, |
455 | 512 * acb->cur_nr_sectors); | |
a32ef786 AL |
456 | ret = qcow_schedule_bh(qcow_aio_read_bh, acb); |
457 | if (ret < 0) | |
f141eafe | 458 | goto done; |
585f8587 | 459 | } else { |
ce1a14dc | 460 | if ((acb->cluster_offset & 511) != 0) { |
585f8587 | 461 | ret = -EIO; |
f141eafe | 462 | goto done; |
585f8587 | 463 | } |
c87c0672 | 464 | |
3f4cb3d3 | 465 | acb->hd_iov.iov_base = (void *)acb->buf; |
7b88e48b | 466 | acb->hd_iov.iov_len = acb->cur_nr_sectors * 512; |
c87c0672 | 467 | qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); |
8252278a | 468 | BLKDBG_EVENT(s->hd, BLKDBG_READ_AIO); |
c87c0672 | 469 | acb->hd_aiocb = bdrv_aio_readv(s->hd, |
5fafdf24 | 470 | (acb->cluster_offset >> 9) + index_in_cluster, |
7b88e48b CH |
471 | &acb->hd_qiov, acb->cur_nr_sectors, |
472 | qcow_aio_read_cb, acb); | |
171e3d6b KW |
473 | if (acb->hd_aiocb == NULL) { |
474 | ret = -EIO; | |
f141eafe | 475 | goto done; |
171e3d6b | 476 | } |
f141eafe AL |
477 | } |
478 | ||
479 | return; | |
480 | done: | |
481 | if (acb->qiov->niov > 1) { | |
482 | qemu_iovec_from_buffer(acb->qiov, acb->orig_buf, acb->qiov->size); | |
483 | qemu_vfree(acb->orig_buf); | |
585f8587 | 484 | } |
f141eafe AL |
485 | acb->common.cb(acb->common.opaque, ret); |
486 | qemu_aio_release(acb); | |
585f8587 FB |
487 | } |
488 | ||
ce1a14dc | 489 | static QCowAIOCB *qcow_aio_setup(BlockDriverState *bs, |
f141eafe AL |
490 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, |
491 | BlockDriverCompletionFunc *cb, void *opaque, int is_write) | |
585f8587 | 492 | { |
ce1a14dc PB |
493 | QCowAIOCB *acb; |
494 | ||
c16b5a2c | 495 | acb = qemu_aio_get(&qcow_aio_pool, bs, cb, opaque); |
ce1a14dc PB |
496 | if (!acb) |
497 | return NULL; | |
498 | acb->hd_aiocb = NULL; | |
499 | acb->sector_num = sector_num; | |
f141eafe AL |
500 | acb->qiov = qiov; |
501 | if (qiov->niov > 1) { | |
e268ca52 | 502 | acb->buf = acb->orig_buf = qemu_blockalign(bs, qiov->size); |
f141eafe AL |
503 | if (is_write) |
504 | qemu_iovec_to_buffer(qiov, acb->buf); | |
3f4cb3d3 BS |
505 | } else { |
506 | acb->buf = (uint8_t *)qiov->iov->iov_base; | |
507 | } | |
7b88e48b CH |
508 | acb->remaining_sectors = nb_sectors; |
509 | acb->cur_nr_sectors = 0; | |
ce1a14dc | 510 | acb->cluster_offset = 0; |
e976c6a1 | 511 | acb->l2meta.nb_clusters = 0; |
72cf2d4f | 512 | QLIST_INIT(&acb->l2meta.dependent_requests); |
ce1a14dc PB |
513 | return acb; |
514 | } | |
515 | ||
f141eafe AL |
516 | static BlockDriverAIOCB *qcow_aio_readv(BlockDriverState *bs, |
517 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc PB |
518 | BlockDriverCompletionFunc *cb, void *opaque) |
519 | { | |
520 | QCowAIOCB *acb; | |
521 | ||
f141eafe | 522 | acb = qcow_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 0); |
ce1a14dc PB |
523 | if (!acb) |
524 | return NULL; | |
585f8587 FB |
525 | |
526 | qcow_aio_read_cb(acb, 0); | |
ce1a14dc | 527 | return &acb->common; |
585f8587 FB |
528 | } |
529 | ||
f214978a KW |
530 | static void qcow_aio_write_cb(void *opaque, int ret); |
531 | ||
532 | static void run_dependent_requests(QCowL2Meta *m) | |
533 | { | |
534 | QCowAIOCB *req; | |
535 | QCowAIOCB *next; | |
536 | ||
537 | /* Take the request off the list of running requests */ | |
538 | if (m->nb_clusters != 0) { | |
72cf2d4f | 539 | QLIST_REMOVE(m, next_in_flight); |
f214978a KW |
540 | } |
541 | ||
d4c146f0 SH |
542 | /* Restart all dependent requests */ |
543 | QLIST_FOREACH_SAFE(req, &m->dependent_requests, next_depend, next) { | |
f214978a KW |
544 | qcow_aio_write_cb(req, 0); |
545 | } | |
546 | ||
547 | /* Empty the list for the next part of the request */ | |
72cf2d4f | 548 | QLIST_INIT(&m->dependent_requests); |
f214978a KW |
549 | } |
550 | ||
585f8587 FB |
551 | static void qcow_aio_write_cb(void *opaque, int ret) |
552 | { | |
ce1a14dc PB |
553 | QCowAIOCB *acb = opaque; |
554 | BlockDriverState *bs = acb->common.bs; | |
585f8587 | 555 | BDRVQcowState *s = bs->opaque; |
585f8587 | 556 | int index_in_cluster; |
585f8587 | 557 | const uint8_t *src_buf; |
095a9c58 | 558 | int n_end; |
ce1a14dc PB |
559 | |
560 | acb->hd_aiocb = NULL; | |
561 | ||
f214978a | 562 | if (ret >= 0) { |
148da7ea | 563 | ret = qcow2_alloc_cluster_link_l2(bs, &acb->l2meta); |
f214978a KW |
564 | } |
565 | ||
566 | run_dependent_requests(&acb->l2meta); | |
567 | ||
f141eafe AL |
568 | if (ret < 0) |
569 | goto done; | |
585f8587 | 570 | |
7b88e48b CH |
571 | acb->remaining_sectors -= acb->cur_nr_sectors; |
572 | acb->sector_num += acb->cur_nr_sectors; | |
573 | acb->buf += acb->cur_nr_sectors * 512; | |
585f8587 | 574 | |
7b88e48b | 575 | if (acb->remaining_sectors == 0) { |
585f8587 | 576 | /* request completed */ |
f141eafe AL |
577 | ret = 0; |
578 | goto done; | |
585f8587 | 579 | } |
3b46e624 | 580 | |
ce1a14dc | 581 | index_in_cluster = acb->sector_num & (s->cluster_sectors - 1); |
7b88e48b | 582 | n_end = index_in_cluster + acb->remaining_sectors; |
095a9c58 AL |
583 | if (s->crypt_method && |
584 | n_end > QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors) | |
585 | n_end = QCOW_MAX_CRYPT_CLUSTERS * s->cluster_sectors; | |
586 | ||
148da7ea | 587 | ret = qcow2_alloc_cluster_offset(bs, acb->sector_num << 9, |
7b88e48b | 588 | index_in_cluster, n_end, &acb->cur_nr_sectors, &acb->l2meta); |
148da7ea KW |
589 | if (ret < 0) { |
590 | goto done; | |
591 | } | |
592 | ||
593 | acb->cluster_offset = acb->l2meta.cluster_offset; | |
f214978a KW |
594 | |
595 | /* Need to wait for another request? If so, we are done for now. */ | |
148da7ea | 596 | if (acb->l2meta.nb_clusters == 0 && acb->l2meta.depends_on != NULL) { |
72cf2d4f | 597 | QLIST_INSERT_HEAD(&acb->l2meta.depends_on->dependent_requests, |
f214978a KW |
598 | acb, next_depend); |
599 | return; | |
600 | } | |
601 | ||
148da7ea KW |
602 | assert((acb->cluster_offset & 511) == 0); |
603 | ||
585f8587 | 604 | if (s->crypt_method) { |
ce1a14dc | 605 | if (!acb->cluster_data) { |
095a9c58 AL |
606 | acb->cluster_data = qemu_mallocz(QCOW_MAX_CRYPT_CLUSTERS * |
607 | s->cluster_size); | |
585f8587 | 608 | } |
ed6ccf0f | 609 | qcow2_encrypt_sectors(s, acb->sector_num, acb->cluster_data, acb->buf, |
7b88e48b | 610 | acb->cur_nr_sectors, 1, &s->aes_encrypt_key); |
ce1a14dc | 611 | src_buf = acb->cluster_data; |
585f8587 | 612 | } else { |
ce1a14dc | 613 | src_buf = acb->buf; |
585f8587 | 614 | } |
c87c0672 | 615 | acb->hd_iov.iov_base = (void *)src_buf; |
7b88e48b | 616 | acb->hd_iov.iov_len = acb->cur_nr_sectors * 512; |
c87c0672 | 617 | qemu_iovec_init_external(&acb->hd_qiov, &acb->hd_iov, 1); |
8252278a | 618 | BLKDBG_EVENT(s->hd, BLKDBG_WRITE_AIO); |
c87c0672 AL |
619 | acb->hd_aiocb = bdrv_aio_writev(s->hd, |
620 | (acb->cluster_offset >> 9) + index_in_cluster, | |
7b88e48b | 621 | &acb->hd_qiov, acb->cur_nr_sectors, |
c87c0672 | 622 | qcow_aio_write_cb, acb); |
171e3d6b KW |
623 | if (acb->hd_aiocb == NULL) { |
624 | ret = -EIO; | |
c644db3d | 625 | goto fail; |
171e3d6b | 626 | } |
f141eafe AL |
627 | |
628 | return; | |
629 | ||
c644db3d KW |
630 | fail: |
631 | if (acb->l2meta.nb_clusters != 0) { | |
632 | QLIST_REMOVE(&acb->l2meta, next_in_flight); | |
633 | } | |
f141eafe AL |
634 | done: |
635 | if (acb->qiov->niov > 1) | |
636 | qemu_vfree(acb->orig_buf); | |
637 | acb->common.cb(acb->common.opaque, ret); | |
638 | qemu_aio_release(acb); | |
585f8587 FB |
639 | } |
640 | ||
f141eafe AL |
641 | static BlockDriverAIOCB *qcow_aio_writev(BlockDriverState *bs, |
642 | int64_t sector_num, QEMUIOVector *qiov, int nb_sectors, | |
ce1a14dc | 643 | BlockDriverCompletionFunc *cb, void *opaque) |
585f8587 | 644 | { |
585f8587 | 645 | BDRVQcowState *s = bs->opaque; |
ce1a14dc | 646 | QCowAIOCB *acb; |
3b46e624 | 647 | |
585f8587 FB |
648 | s->cluster_cache_offset = -1; /* disable compressed cache */ |
649 | ||
f141eafe | 650 | acb = qcow_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, 1); |
ce1a14dc PB |
651 | if (!acb) |
652 | return NULL; | |
3b46e624 | 653 | |
585f8587 | 654 | qcow_aio_write_cb(acb, 0); |
ce1a14dc | 655 | return &acb->common; |
585f8587 FB |
656 | } |
657 | ||
585f8587 FB |
658 | static void qcow_close(BlockDriverState *bs) |
659 | { | |
660 | BDRVQcowState *s = bs->opaque; | |
661 | qemu_free(s->l1_table); | |
662 | qemu_free(s->l2_cache); | |
663 | qemu_free(s->cluster_cache); | |
664 | qemu_free(s->cluster_data); | |
ed6ccf0f | 665 | qcow2_refcount_close(bs); |
585f8587 FB |
666 | bdrv_delete(s->hd); |
667 | } | |
668 | ||
756e6736 KW |
669 | /* |
670 | * Updates the variable length parts of the qcow2 header, i.e. the backing file | |
671 | * name and all extensions. qcow2 was not designed to allow such changes, so if | |
672 | * we run out of space (we can only use the first cluster) this function may | |
673 | * fail. | |
674 | * | |
675 | * Returns 0 on success, -errno in error cases. | |
676 | */ | |
677 | static int qcow2_update_ext_header(BlockDriverState *bs, | |
678 | const char *backing_file, const char *backing_fmt) | |
679 | { | |
680 | size_t backing_file_len = 0; | |
681 | size_t backing_fmt_len = 0; | |
682 | BDRVQcowState *s = bs->opaque; | |
683 | QCowExtension ext_backing_fmt = {0, 0}; | |
684 | int ret; | |
685 | ||
686 | /* Backing file format doesn't make sense without a backing file */ | |
687 | if (backing_fmt && !backing_file) { | |
688 | return -EINVAL; | |
689 | } | |
690 | ||
691 | /* Prepare the backing file format extension if needed */ | |
692 | if (backing_fmt) { | |
693 | ext_backing_fmt.len = cpu_to_be32(strlen(backing_fmt)); | |
694 | ext_backing_fmt.magic = cpu_to_be32(QCOW_EXT_MAGIC_BACKING_FORMAT); | |
695 | backing_fmt_len = ((sizeof(ext_backing_fmt) | |
696 | + strlen(backing_fmt) + 7) & ~7); | |
697 | } | |
698 | ||
699 | /* Check if we can fit the new header into the first cluster */ | |
700 | if (backing_file) { | |
701 | backing_file_len = strlen(backing_file); | |
702 | } | |
703 | ||
704 | size_t header_size = sizeof(QCowHeader) + backing_file_len | |
705 | + backing_fmt_len; | |
706 | ||
707 | if (header_size > s->cluster_size) { | |
708 | return -ENOSPC; | |
709 | } | |
710 | ||
711 | /* Rewrite backing file name and qcow2 extensions */ | |
712 | size_t ext_size = header_size - sizeof(QCowHeader); | |
713 | uint8_t buf[ext_size]; | |
714 | size_t offset = 0; | |
715 | size_t backing_file_offset = 0; | |
716 | ||
717 | if (backing_file) { | |
718 | if (backing_fmt) { | |
719 | int padding = backing_fmt_len - | |
720 | (sizeof(ext_backing_fmt) + strlen(backing_fmt)); | |
721 | ||
722 | memcpy(buf + offset, &ext_backing_fmt, sizeof(ext_backing_fmt)); | |
723 | offset += sizeof(ext_backing_fmt); | |
724 | ||
725 | memcpy(buf + offset, backing_fmt, strlen(backing_fmt)); | |
726 | offset += strlen(backing_fmt); | |
727 | ||
728 | memset(buf + offset, 0, padding); | |
729 | offset += padding; | |
730 | } | |
731 | ||
732 | memcpy(buf + offset, backing_file, backing_file_len); | |
733 | backing_file_offset = sizeof(QCowHeader) + offset; | |
734 | } | |
735 | ||
736 | ret = bdrv_pwrite(s->hd, sizeof(QCowHeader), buf, ext_size); | |
737 | if (ret < 0) { | |
738 | goto fail; | |
739 | } | |
740 | ||
741 | /* Update header fields */ | |
742 | uint64_t be_backing_file_offset = cpu_to_be64(backing_file_offset); | |
743 | uint32_t be_backing_file_size = cpu_to_be32(backing_file_len); | |
744 | ||
745 | ret = bdrv_pwrite(s->hd, offsetof(QCowHeader, backing_file_offset), | |
746 | &be_backing_file_offset, sizeof(uint64_t)); | |
747 | if (ret < 0) { | |
748 | goto fail; | |
749 | } | |
750 | ||
751 | ret = bdrv_pwrite(s->hd, offsetof(QCowHeader, backing_file_size), | |
752 | &be_backing_file_size, sizeof(uint32_t)); | |
753 | if (ret < 0) { | |
754 | goto fail; | |
755 | } | |
756 | ||
757 | ret = 0; | |
758 | fail: | |
759 | return ret; | |
760 | } | |
761 | ||
762 | static int qcow2_change_backing_file(BlockDriverState *bs, | |
763 | const char *backing_file, const char *backing_fmt) | |
764 | { | |
765 | return qcow2_update_ext_header(bs, backing_file, backing_fmt); | |
766 | } | |
767 | ||
73c632ed KW |
768 | static int get_bits_from_size(size_t size) |
769 | { | |
770 | int res = 0; | |
771 | ||
772 | if (size == 0) { | |
773 | return -1; | |
774 | } | |
775 | ||
776 | while (size != 1) { | |
777 | /* Not a power of two */ | |
778 | if (size & 1) { | |
779 | return -1; | |
780 | } | |
781 | ||
782 | size >>= 1; | |
783 | res++; | |
784 | } | |
785 | ||
786 | return res; | |
787 | } | |
788 | ||
a35e1c17 KW |
789 | |
790 | static int preallocate(BlockDriverState *bs) | |
791 | { | |
792 | BDRVQcowState *s = bs->opaque; | |
a35e1c17 KW |
793 | uint64_t nb_sectors; |
794 | uint64_t offset; | |
795 | int num; | |
148da7ea | 796 | int ret; |
a35e1c17 KW |
797 | QCowL2Meta meta; |
798 | ||
799 | nb_sectors = bdrv_getlength(bs) >> 9; | |
800 | offset = 0; | |
72cf2d4f | 801 | QLIST_INIT(&meta.dependent_requests); |
148da7ea | 802 | meta.cluster_offset = 0; |
a35e1c17 KW |
803 | |
804 | while (nb_sectors) { | |
805 | num = MIN(nb_sectors, INT_MAX >> 9); | |
148da7ea | 806 | ret = qcow2_alloc_cluster_offset(bs, offset, 0, num, &num, &meta); |
a35e1c17 | 807 | |
148da7ea | 808 | if (ret < 0) { |
a35e1c17 KW |
809 | return -1; |
810 | } | |
811 | ||
148da7ea KW |
812 | if (qcow2_alloc_cluster_link_l2(bs, &meta) < 0) { |
813 | qcow2_free_any_clusters(bs, meta.cluster_offset, meta.nb_clusters); | |
a35e1c17 KW |
814 | return -1; |
815 | } | |
816 | ||
f214978a KW |
817 | /* There are no dependent requests, but we need to remove our request |
818 | * from the list of in-flight requests */ | |
819 | run_dependent_requests(&meta); | |
820 | ||
a35e1c17 KW |
821 | /* TODO Preallocate data if requested */ |
822 | ||
823 | nb_sectors -= num; | |
824 | offset += num << 9; | |
825 | } | |
826 | ||
827 | /* | |
828 | * It is expected that the image file is large enough to actually contain | |
829 | * all of the allocated clusters (otherwise we get failing reads after | |
830 | * EOF). Extend the image to the last allocated sector. | |
831 | */ | |
148da7ea | 832 | if (meta.cluster_offset != 0) { |
ea80b906 KW |
833 | uint8_t buf[512]; |
834 | memset(buf, 0, 512); | |
148da7ea | 835 | bdrv_write(s->hd, (meta.cluster_offset >> 9) + num - 1, buf, 1); |
a35e1c17 KW |
836 | } |
837 | ||
838 | return 0; | |
839 | } | |
840 | ||
20d97356 BS |
841 | static int qcow_make_empty(BlockDriverState *bs) |
842 | { | |
843 | #if 0 | |
844 | /* XXX: not correct */ | |
845 | BDRVQcowState *s = bs->opaque; | |
846 | uint32_t l1_length = s->l1_size * sizeof(uint64_t); | |
847 | int ret; | |
848 | ||
849 | memset(s->l1_table, 0, l1_length); | |
850 | if (bdrv_pwrite(s->hd, s->l1_table_offset, s->l1_table, l1_length) < 0) | |
851 | return -1; | |
852 | ret = bdrv_truncate(s->hd, s->l1_table_offset + l1_length); | |
853 | if (ret < 0) | |
854 | return ret; | |
855 | ||
856 | l2_cache_reset(bs); | |
857 | #endif | |
858 | return 0; | |
859 | } | |
860 | ||
861 | /* XXX: put compressed sectors first, then all the cluster aligned | |
862 | tables to avoid losing bytes in alignment */ | |
863 | static int qcow_write_compressed(BlockDriverState *bs, int64_t sector_num, | |
864 | const uint8_t *buf, int nb_sectors) | |
865 | { | |
866 | BDRVQcowState *s = bs->opaque; | |
867 | z_stream strm; | |
868 | int ret, out_len; | |
869 | uint8_t *out_buf; | |
870 | uint64_t cluster_offset; | |
871 | ||
872 | if (nb_sectors == 0) { | |
873 | /* align end of file to a sector boundary to ease reading with | |
874 | sector based I/Os */ | |
875 | cluster_offset = bdrv_getlength(s->hd); | |
876 | cluster_offset = (cluster_offset + 511) & ~511; | |
877 | bdrv_truncate(s->hd, cluster_offset); | |
878 | return 0; | |
879 | } | |
880 | ||
881 | if (nb_sectors != s->cluster_sectors) | |
882 | return -EINVAL; | |
883 | ||
884 | out_buf = qemu_malloc(s->cluster_size + (s->cluster_size / 1000) + 128); | |
885 | ||
886 | /* best compression, small window, no zlib header */ | |
887 | memset(&strm, 0, sizeof(strm)); | |
888 | ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, | |
889 | Z_DEFLATED, -12, | |
890 | 9, Z_DEFAULT_STRATEGY); | |
891 | if (ret != 0) { | |
892 | qemu_free(out_buf); | |
893 | return -1; | |
894 | } | |
895 | ||
896 | strm.avail_in = s->cluster_size; | |
897 | strm.next_in = (uint8_t *)buf; | |
898 | strm.avail_out = s->cluster_size; | |
899 | strm.next_out = out_buf; | |
900 | ||
901 | ret = deflate(&strm, Z_FINISH); | |
902 | if (ret != Z_STREAM_END && ret != Z_OK) { | |
903 | qemu_free(out_buf); | |
904 | deflateEnd(&strm); | |
905 | return -1; | |
906 | } | |
907 | out_len = strm.next_out - out_buf; | |
908 | ||
909 | deflateEnd(&strm); | |
910 | ||
911 | if (ret != Z_STREAM_END || out_len >= s->cluster_size) { | |
912 | /* could not compress: write normal cluster */ | |
913 | bdrv_write(bs, sector_num, buf, s->cluster_sectors); | |
914 | } else { | |
915 | cluster_offset = qcow2_alloc_compressed_cluster_offset(bs, | |
916 | sector_num << 9, out_len); | |
917 | if (!cluster_offset) | |
918 | return -1; | |
919 | cluster_offset &= s->cluster_offset_mask; | |
920 | BLKDBG_EVENT(s->hd, BLKDBG_WRITE_COMPRESSED); | |
921 | if (bdrv_pwrite(s->hd, cluster_offset, out_buf, out_len) != out_len) { | |
922 | qemu_free(out_buf); | |
923 | return -1; | |
924 | } | |
925 | } | |
926 | ||
927 | qemu_free(out_buf); | |
928 | return 0; | |
929 | } | |
930 | ||
931 | static void qcow_flush(BlockDriverState *bs) | |
932 | { | |
933 | BDRVQcowState *s = bs->opaque; | |
934 | bdrv_flush(s->hd); | |
935 | } | |
936 | ||
937 | static BlockDriverAIOCB *qcow_aio_flush(BlockDriverState *bs, | |
938 | BlockDriverCompletionFunc *cb, void *opaque) | |
939 | { | |
940 | BDRVQcowState *s = bs->opaque; | |
941 | ||
942 | return bdrv_aio_flush(s->hd, cb, opaque); | |
943 | } | |
944 | ||
945 | static int64_t qcow_vm_state_offset(BDRVQcowState *s) | |
946 | { | |
947 | return (int64_t)s->l1_vm_state_index << (s->cluster_bits + s->l2_bits); | |
948 | } | |
949 | ||
950 | static int qcow_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) | |
951 | { | |
952 | BDRVQcowState *s = bs->opaque; | |
953 | bdi->cluster_size = s->cluster_size; | |
954 | bdi->vm_state_offset = qcow_vm_state_offset(s); | |
955 | return 0; | |
956 | } | |
957 | ||
958 | ||
959 | static int qcow_check(BlockDriverState *bs) | |
960 | { | |
961 | return qcow2_check_refcounts(bs); | |
962 | } | |
963 | ||
964 | #if 0 | |
965 | static void dump_refcounts(BlockDriverState *bs) | |
966 | { | |
967 | BDRVQcowState *s = bs->opaque; | |
968 | int64_t nb_clusters, k, k1, size; | |
969 | int refcount; | |
970 | ||
971 | size = bdrv_getlength(s->hd); | |
972 | nb_clusters = size_to_clusters(s, size); | |
973 | for(k = 0; k < nb_clusters;) { | |
974 | k1 = k; | |
975 | refcount = get_refcount(bs, k); | |
976 | k++; | |
977 | while (k < nb_clusters && get_refcount(bs, k) == refcount) | |
978 | k++; | |
979 | printf("%lld: refcount=%d nb=%lld\n", k, refcount, k - k1); | |
980 | } | |
981 | } | |
982 | #endif | |
983 | ||
984 | static int qcow_save_vmstate(BlockDriverState *bs, const uint8_t *buf, | |
985 | int64_t pos, int size) | |
986 | { | |
987 | BDRVQcowState *s = bs->opaque; | |
988 | int growable = bs->growable; | |
989 | int ret; | |
990 | ||
991 | BLKDBG_EVENT(s->hd, BLKDBG_VMSTATE_SAVE); | |
992 | bs->growable = 1; | |
993 | ret = bdrv_pwrite(bs, qcow_vm_state_offset(s) + pos, buf, size); | |
994 | bs->growable = growable; | |
995 | ||
996 | return ret; | |
997 | } | |
998 | ||
999 | static int qcow_load_vmstate(BlockDriverState *bs, uint8_t *buf, | |
1000 | int64_t pos, int size) | |
1001 | { | |
1002 | BDRVQcowState *s = bs->opaque; | |
1003 | int growable = bs->growable; | |
1004 | int ret; | |
1005 | ||
1006 | BLKDBG_EVENT(s->hd, BLKDBG_VMSTATE_LOAD); | |
1007 | bs->growable = 1; | |
1008 | ret = bdrv_pread(bs, qcow_vm_state_offset(s) + pos, buf, size); | |
1009 | bs->growable = growable; | |
1010 | ||
1011 | return ret; | |
1012 | } | |
1013 | ||
1014 | static QEMUOptionParameter qcow_create_options[] = { | |
1015 | { | |
1016 | .name = BLOCK_OPT_SIZE, | |
1017 | .type = OPT_SIZE, | |
1018 | .help = "Virtual disk size" | |
1019 | }, | |
1020 | { | |
1021 | .name = BLOCK_OPT_BACKING_FILE, | |
1022 | .type = OPT_STRING, | |
1023 | .help = "File name of a base image" | |
1024 | }, | |
1025 | { | |
1026 | .name = BLOCK_OPT_BACKING_FMT, | |
1027 | .type = OPT_STRING, | |
1028 | .help = "Image format of the base image" | |
1029 | }, | |
1030 | { | |
1031 | .name = BLOCK_OPT_ENCRYPT, | |
1032 | .type = OPT_FLAG, | |
1033 | .help = "Encrypt the image" | |
1034 | }, | |
1035 | { | |
1036 | .name = BLOCK_OPT_CLUSTER_SIZE, | |
1037 | .type = OPT_SIZE, | |
1038 | .help = "qcow2 cluster size" | |
1039 | }, | |
1040 | { | |
1041 | .name = BLOCK_OPT_PREALLOC, | |
1042 | .type = OPT_STRING, | |
1043 | .help = "Preallocation mode (allowed values: off, metadata)" | |
1044 | }, | |
1045 | { NULL } | |
1046 | }; | |
1047 | ||
1048 | static BlockDriver bdrv_qcow2 = { | |
1049 | .format_name = "qcow2", | |
1050 | .instance_size = sizeof(BDRVQcowState), | |
1051 | .bdrv_probe = qcow_probe, | |
1052 | .bdrv_open = qcow_open, | |
1053 | .bdrv_close = qcow_close, | |
1054 | .bdrv_create = qcow_create, | |
1055 | .bdrv_flush = qcow_flush, | |
1056 | .bdrv_is_allocated = qcow_is_allocated, | |
1057 | .bdrv_set_key = qcow_set_key, | |
1058 | .bdrv_make_empty = qcow_make_empty, | |
1059 | ||
1060 | .bdrv_aio_readv = qcow_aio_readv, | |
1061 | .bdrv_aio_writev = qcow_aio_writev, | |
1062 | .bdrv_aio_flush = qcow_aio_flush, | |
1063 | .bdrv_write_compressed = qcow_write_compressed, | |
1064 | ||
1065 | .bdrv_snapshot_create = qcow2_snapshot_create, | |
1066 | .bdrv_snapshot_goto = qcow2_snapshot_goto, | |
1067 | .bdrv_snapshot_delete = qcow2_snapshot_delete, | |
1068 | .bdrv_snapshot_list = qcow2_snapshot_list, | |
1069 | .bdrv_get_info = qcow_get_info, | |
1070 | ||
1071 | .bdrv_save_vmstate = qcow_save_vmstate, | |
1072 | .bdrv_load_vmstate = qcow_load_vmstate, | |
1073 | ||
1074 | .bdrv_change_backing_file = qcow2_change_backing_file, | |
1075 | ||
1076 | .create_options = qcow_create_options, | |
1077 | .bdrv_check = qcow_check, | |
1078 | }; | |
1079 | ||
f965509c AL |
1080 | static int qcow_create2(const char *filename, int64_t total_size, |
1081 | const char *backing_file, const char *backing_format, | |
a35e1c17 | 1082 | int flags, size_t cluster_size, int prealloc) |
585f8587 | 1083 | { |
f965509c | 1084 | |
585f8587 | 1085 | int fd, header_size, backing_filename_len, l1_size, i, shift, l2_bits; |
4768fa90 | 1086 | int ref_clusters, reftable_clusters, backing_format_len = 0; |
e1c7f0e3 | 1087 | int rounded_ext_bf_len = 0; |
585f8587 FB |
1088 | QCowHeader header; |
1089 | uint64_t tmp, offset; | |
4768fa90 | 1090 | uint64_t old_ref_clusters; |
585f8587 | 1091 | QCowCreateState s1, *s = &s1; |
f965509c | 1092 | QCowExtension ext_bf = {0, 0}; |
db89119d | 1093 | int ret; |
3b46e624 | 1094 | |
585f8587 FB |
1095 | memset(s, 0, sizeof(*s)); |
1096 | ||
1097 | fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644); | |
1098 | if (fd < 0) | |
bef57da5 | 1099 | return -errno; |
585f8587 FB |
1100 | memset(&header, 0, sizeof(header)); |
1101 | header.magic = cpu_to_be32(QCOW_MAGIC); | |
1102 | header.version = cpu_to_be32(QCOW_VERSION); | |
1103 | header.size = cpu_to_be64(total_size * 512); | |
1104 | header_size = sizeof(header); | |
1105 | backing_filename_len = 0; | |
1106 | if (backing_file) { | |
f965509c AL |
1107 | if (backing_format) { |
1108 | ext_bf.magic = QCOW_EXT_MAGIC_BACKING_FORMAT; | |
1109 | backing_format_len = strlen(backing_format); | |
e1c7f0e3 KW |
1110 | ext_bf.len = backing_format_len; |
1111 | rounded_ext_bf_len = (sizeof(ext_bf) + ext_bf.len + 7) & ~7; | |
1112 | header_size += rounded_ext_bf_len; | |
f965509c | 1113 | } |
585f8587 FB |
1114 | header.backing_file_offset = cpu_to_be64(header_size); |
1115 | backing_filename_len = strlen(backing_file); | |
1116 | header.backing_file_size = cpu_to_be32(backing_filename_len); | |
1117 | header_size += backing_filename_len; | |
1118 | } | |
73c632ed KW |
1119 | |
1120 | /* Cluster size */ | |
1121 | s->cluster_bits = get_bits_from_size(cluster_size); | |
1122 | if (s->cluster_bits < MIN_CLUSTER_BITS || | |
1123 | s->cluster_bits > MAX_CLUSTER_BITS) | |
1124 | { | |
1125 | fprintf(stderr, "Cluster size must be a power of two between " | |
1126 | "%d and %dk\n", | |
1127 | 1 << MIN_CLUSTER_BITS, | |
1128 | 1 << (MAX_CLUSTER_BITS - 10)); | |
1129 | return -EINVAL; | |
1130 | } | |
585f8587 | 1131 | s->cluster_size = 1 << s->cluster_bits; |
73c632ed | 1132 | |
585f8587 FB |
1133 | header.cluster_bits = cpu_to_be32(s->cluster_bits); |
1134 | header_size = (header_size + 7) & ~7; | |
ec36ba14 | 1135 | if (flags & BLOCK_FLAG_ENCRYPT) { |
585f8587 FB |
1136 | header.crypt_method = cpu_to_be32(QCOW_CRYPT_AES); |
1137 | } else { | |
1138 | header.crypt_method = cpu_to_be32(QCOW_CRYPT_NONE); | |
1139 | } | |
1140 | l2_bits = s->cluster_bits - 3; | |
1141 | shift = s->cluster_bits + l2_bits; | |
1142 | l1_size = (((total_size * 512) + (1LL << shift) - 1) >> shift); | |
1143 | offset = align_offset(header_size, s->cluster_size); | |
1144 | s->l1_table_offset = offset; | |
1145 | header.l1_table_offset = cpu_to_be64(s->l1_table_offset); | |
1146 | header.l1_size = cpu_to_be32(l1_size); | |
15e6690a | 1147 | offset += align_offset(l1_size * sizeof(uint64_t), s->cluster_size); |
585f8587 | 1148 | |
4768fa90 KW |
1149 | /* count how many refcount blocks needed */ |
1150 | ||
1151 | #define NUM_CLUSTERS(bytes) \ | |
1152 | (((bytes) + (s->cluster_size) - 1) / (s->cluster_size)) | |
1153 | ||
1154 | ref_clusters = NUM_CLUSTERS(NUM_CLUSTERS(offset) * sizeof(uint16_t)); | |
1155 | ||
1156 | do { | |
1157 | uint64_t image_clusters; | |
1158 | old_ref_clusters = ref_clusters; | |
1159 | ||
1160 | /* Number of clusters used for the refcount table */ | |
1161 | reftable_clusters = NUM_CLUSTERS(ref_clusters * sizeof(uint64_t)); | |
1162 | ||
1163 | /* Number of clusters that the whole image will have */ | |
1164 | image_clusters = NUM_CLUSTERS(offset) + ref_clusters | |
1165 | + reftable_clusters; | |
1166 | ||
1167 | /* Number of refcount blocks needed for the image */ | |
1168 | ref_clusters = NUM_CLUSTERS(image_clusters * sizeof(uint16_t)); | |
1169 | ||
1170 | } while (ref_clusters != old_ref_clusters); | |
1171 | ||
1172 | s->refcount_table = qemu_mallocz(reftable_clusters * s->cluster_size); | |
3b46e624 | 1173 | |
585f8587 FB |
1174 | s->refcount_table_offset = offset; |
1175 | header.refcount_table_offset = cpu_to_be64(offset); | |
4768fa90 KW |
1176 | header.refcount_table_clusters = cpu_to_be32(reftable_clusters); |
1177 | offset += (reftable_clusters * s->cluster_size); | |
585f8587 | 1178 | s->refcount_block_offset = offset; |
2d2431f0 | 1179 | |
2d2431f0 AL |
1180 | for (i=0; i < ref_clusters; i++) { |
1181 | s->refcount_table[i] = cpu_to_be64(offset); | |
1182 | offset += s->cluster_size; | |
1183 | } | |
1184 | ||
1185 | s->refcount_block = qemu_mallocz(ref_clusters * s->cluster_size); | |
585f8587 FB |
1186 | |
1187 | /* update refcounts */ | |
ed6ccf0f KW |
1188 | qcow2_create_refcount_update(s, 0, header_size); |
1189 | qcow2_create_refcount_update(s, s->l1_table_offset, | |
1190 | l1_size * sizeof(uint64_t)); | |
4768fa90 KW |
1191 | qcow2_create_refcount_update(s, s->refcount_table_offset, |
1192 | reftable_clusters * s->cluster_size); | |
ed6ccf0f KW |
1193 | qcow2_create_refcount_update(s, s->refcount_block_offset, |
1194 | ref_clusters * s->cluster_size); | |
3b46e624 | 1195 | |
585f8587 | 1196 | /* write all the data */ |
db89119d KS |
1197 | ret = qemu_write_full(fd, &header, sizeof(header)); |
1198 | if (ret != sizeof(header)) { | |
bef57da5 | 1199 | ret = -errno; |
db89119d KS |
1200 | goto exit; |
1201 | } | |
585f8587 | 1202 | if (backing_file) { |
f965509c AL |
1203 | if (backing_format_len) { |
1204 | char zero[16]; | |
e1c7f0e3 | 1205 | int padding = rounded_ext_bf_len - (ext_bf.len + sizeof(ext_bf)); |
f965509c AL |
1206 | |
1207 | memset(zero, 0, sizeof(zero)); | |
1208 | cpu_to_be32s(&ext_bf.magic); | |
1209 | cpu_to_be32s(&ext_bf.len); | |
db89119d KS |
1210 | ret = qemu_write_full(fd, &ext_bf, sizeof(ext_bf)); |
1211 | if (ret != sizeof(ext_bf)) { | |
bef57da5 | 1212 | ret = -errno; |
db89119d KS |
1213 | goto exit; |
1214 | } | |
1215 | ret = qemu_write_full(fd, backing_format, backing_format_len); | |
1216 | if (ret != backing_format_len) { | |
bef57da5 | 1217 | ret = -errno; |
db89119d KS |
1218 | goto exit; |
1219 | } | |
e1c7f0e3 | 1220 | if (padding > 0) { |
db89119d KS |
1221 | ret = qemu_write_full(fd, zero, padding); |
1222 | if (ret != padding) { | |
bef57da5 | 1223 | ret = -errno; |
db89119d KS |
1224 | goto exit; |
1225 | } | |
f965509c AL |
1226 | } |
1227 | } | |
db89119d KS |
1228 | ret = qemu_write_full(fd, backing_file, backing_filename_len); |
1229 | if (ret != backing_filename_len) { | |
bef57da5 | 1230 | ret = -errno; |
db89119d KS |
1231 | goto exit; |
1232 | } | |
585f8587 FB |
1233 | } |
1234 | lseek(fd, s->l1_table_offset, SEEK_SET); | |
1235 | tmp = 0; | |
1236 | for(i = 0;i < l1_size; i++) { | |
db89119d KS |
1237 | ret = qemu_write_full(fd, &tmp, sizeof(tmp)); |
1238 | if (ret != sizeof(tmp)) { | |
bef57da5 | 1239 | ret = -errno; |
db89119d KS |
1240 | goto exit; |
1241 | } | |
585f8587 FB |
1242 | } |
1243 | lseek(fd, s->refcount_table_offset, SEEK_SET); | |
4768fa90 KW |
1244 | ret = qemu_write_full(fd, s->refcount_table, |
1245 | reftable_clusters * s->cluster_size); | |
1246 | if (ret != reftable_clusters * s->cluster_size) { | |
bef57da5 | 1247 | ret = -errno; |
db89119d KS |
1248 | goto exit; |
1249 | } | |
3b46e624 | 1250 | |
585f8587 | 1251 | lseek(fd, s->refcount_block_offset, SEEK_SET); |
db89119d KS |
1252 | ret = qemu_write_full(fd, s->refcount_block, |
1253 | ref_clusters * s->cluster_size); | |
6f745bda | 1254 | if (ret != ref_clusters * s->cluster_size) { |
bef57da5 | 1255 | ret = -errno; |
db89119d KS |
1256 | goto exit; |
1257 | } | |
585f8587 | 1258 | |
db89119d KS |
1259 | ret = 0; |
1260 | exit: | |
585f8587 FB |
1261 | qemu_free(s->refcount_table); |
1262 | qemu_free(s->refcount_block); | |
1263 | close(fd); | |
a35e1c17 KW |
1264 | |
1265 | /* Preallocate metadata */ | |
6f745bda | 1266 | if (ret == 0 && prealloc) { |
a35e1c17 KW |
1267 | BlockDriverState *bs; |
1268 | bs = bdrv_new(""); | |
d6e9098e | 1269 | bdrv_open(bs, filename, BDRV_O_CACHE_WB | BDRV_O_RDWR, &bdrv_qcow2); |
a35e1c17 KW |
1270 | preallocate(bs); |
1271 | bdrv_close(bs); | |
1272 | } | |
1273 | ||
db89119d | 1274 | return ret; |
585f8587 FB |
1275 | } |
1276 | ||
0e7e1989 KW |
1277 | static int qcow_create(const char *filename, QEMUOptionParameter *options) |
1278 | { | |
1279 | const char *backing_file = NULL; | |
1280 | const char *backing_fmt = NULL; | |
1281 | uint64_t sectors = 0; | |
1282 | int flags = 0; | |
9ccb258e | 1283 | size_t cluster_size = 65536; |
a35e1c17 | 1284 | int prealloc = 0; |
0e7e1989 KW |
1285 | |
1286 | /* Read out options */ | |
1287 | while (options && options->name) { | |
1288 | if (!strcmp(options->name, BLOCK_OPT_SIZE)) { | |
1289 | sectors = options->value.n / 512; | |
1290 | } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FILE)) { | |
1291 | backing_file = options->value.s; | |
1292 | } else if (!strcmp(options->name, BLOCK_OPT_BACKING_FMT)) { | |
1293 | backing_fmt = options->value.s; | |
1294 | } else if (!strcmp(options->name, BLOCK_OPT_ENCRYPT)) { | |
1295 | flags |= options->value.n ? BLOCK_FLAG_ENCRYPT : 0; | |
73c632ed KW |
1296 | } else if (!strcmp(options->name, BLOCK_OPT_CLUSTER_SIZE)) { |
1297 | if (options->value.n) { | |
1298 | cluster_size = options->value.n; | |
1299 | } | |
a35e1c17 KW |
1300 | } else if (!strcmp(options->name, BLOCK_OPT_PREALLOC)) { |
1301 | if (!options->value.s || !strcmp(options->value.s, "off")) { | |
1302 | prealloc = 0; | |
1303 | } else if (!strcmp(options->value.s, "metadata")) { | |
1304 | prealloc = 1; | |
1305 | } else { | |
1306 | fprintf(stderr, "Invalid preallocation mode: '%s'\n", | |
1307 | options->value.s); | |
1308 | return -EINVAL; | |
1309 | } | |
0e7e1989 KW |
1310 | } |
1311 | options++; | |
1312 | } | |
1313 | ||
a35e1c17 KW |
1314 | if (backing_file && prealloc) { |
1315 | fprintf(stderr, "Backing file and preallocation cannot be used at " | |
1316 | "the same time\n"); | |
1317 | return -EINVAL; | |
1318 | } | |
1319 | ||
73c632ed | 1320 | return qcow_create2(filename, sectors, backing_file, backing_fmt, flags, |
a35e1c17 | 1321 | cluster_size, prealloc); |
f965509c AL |
1322 | } |
1323 | ||
5efa9d5a AL |
1324 | static void bdrv_qcow2_init(void) |
1325 | { | |
1326 | bdrv_register(&bdrv_qcow2); | |
1327 | } | |
1328 | ||
1329 | block_init(bdrv_qcow2_init); |