2 * Block driver for the QCOW format
4 * Copyright (c) 2004-2006 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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
24 #include "qemu/osdep.h"
25 #include "qapi/error.h"
26 #include "qemu-common.h"
27 #include "qemu/error-report.h"
28 #include "block/block_int.h"
29 #include "sysemu/block-backend.h"
30 #include "qemu/module.h"
31 #include "qemu/bswap.h"
33 #include "qapi/qmp/qerror.h"
34 #include "qapi/qmp/qstring.h"
35 #include "crypto/block.h"
36 #include "migration/blocker.h"
37 #include "block/crypto.h"
39 /**************************************************************/
40 /* QEMU COW block driver with compression and encryption support */
42 #define QCOW_MAGIC (('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb)
43 #define QCOW_VERSION 1
45 #define QCOW_CRYPT_NONE 0
46 #define QCOW_CRYPT_AES 1
48 #define QCOW_OFLAG_COMPRESSED (1LL << 63)
50 typedef struct QCowHeader
{
53 uint64_t backing_file_offset
;
54 uint32_t backing_file_size
;
56 uint64_t size
; /* in bytes */
60 uint32_t crypt_method
;
61 uint64_t l1_table_offset
;
62 } QEMU_PACKED QCowHeader
;
64 #define L2_CACHE_SIZE 16
66 typedef struct BDRVQcowState
{
73 uint64_t cluster_offset_mask
;
74 uint64_t l1_table_offset
;
77 uint64_t l2_cache_offsets
[L2_CACHE_SIZE
];
78 uint32_t l2_cache_counts
[L2_CACHE_SIZE
];
79 uint8_t *cluster_cache
;
80 uint8_t *cluster_data
;
81 uint64_t cluster_cache_offset
;
82 QCryptoBlock
*crypto
; /* Disk encryption format driver */
83 uint32_t crypt_method_header
;
85 Error
*migration_blocker
;
88 static int decompress_cluster(BlockDriverState
*bs
, uint64_t cluster_offset
);
90 static int qcow_probe(const uint8_t *buf
, int buf_size
, const char *filename
)
92 const QCowHeader
*cow_header
= (const void *)buf
;
94 if (buf_size
>= sizeof(QCowHeader
) &&
95 be32_to_cpu(cow_header
->magic
) == QCOW_MAGIC
&&
96 be32_to_cpu(cow_header
->version
) == QCOW_VERSION
)
102 static QemuOptsList qcow_runtime_opts
= {
104 .head
= QTAILQ_HEAD_INITIALIZER(qcow_runtime_opts
.head
),
106 BLOCK_CRYPTO_OPT_DEF_QCOW_KEY_SECRET("encrypt."),
107 { /* end of list */ }
111 static int qcow_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
114 BDRVQcowState
*s
= bs
->opaque
;
115 unsigned int len
, i
, shift
;
118 Error
*local_err
= NULL
;
119 QCryptoBlockOpenOptions
*crypto_opts
= NULL
;
120 unsigned int cflags
= 0;
121 QDict
*encryptopts
= NULL
;
122 const char *encryptfmt
;
124 qdict_extract_subqdict(options
, &encryptopts
, "encrypt.");
125 encryptfmt
= qdict_get_try_str(encryptopts
, "format");
127 bs
->file
= bdrv_open_child(NULL
, options
, "file", bs
, &child_file
,
134 ret
= bdrv_pread(bs
->file
, 0, &header
, sizeof(header
));
138 be32_to_cpus(&header
.magic
);
139 be32_to_cpus(&header
.version
);
140 be64_to_cpus(&header
.backing_file_offset
);
141 be32_to_cpus(&header
.backing_file_size
);
142 be32_to_cpus(&header
.mtime
);
143 be64_to_cpus(&header
.size
);
144 be32_to_cpus(&header
.crypt_method
);
145 be64_to_cpus(&header
.l1_table_offset
);
147 if (header
.magic
!= QCOW_MAGIC
) {
148 error_setg(errp
, "Image not in qcow format");
152 if (header
.version
!= QCOW_VERSION
) {
153 error_setg(errp
, "Unsupported qcow version %" PRIu32
, header
.version
);
158 if (header
.size
<= 1) {
159 error_setg(errp
, "Image size is too small (must be at least 2 bytes)");
163 if (header
.cluster_bits
< 9 || header
.cluster_bits
> 16) {
164 error_setg(errp
, "Cluster size must be between 512 and 64k");
169 /* l2_bits specifies number of entries; storing a uint64_t in each entry,
170 * so bytes = num_entries << 3. */
171 if (header
.l2_bits
< 9 - 3 || header
.l2_bits
> 16 - 3) {
172 error_setg(errp
, "L2 table size must be between 512 and 64k");
177 s
->crypt_method_header
= header
.crypt_method
;
178 if (s
->crypt_method_header
) {
179 if (bdrv_uses_whitelist() &&
180 s
->crypt_method_header
== QCOW_CRYPT_AES
) {
182 "Use of AES-CBC encrypted qcow images is no longer "
183 "supported in system emulators");
184 error_append_hint(errp
,
185 "You can use 'qemu-img convert' to convert your "
186 "image to an alternative supported format, such "
187 "as unencrypted qcow, or raw with the LUKS "
188 "format instead.\n");
192 if (s
->crypt_method_header
== QCOW_CRYPT_AES
) {
193 if (encryptfmt
&& !g_str_equal(encryptfmt
, "aes")) {
195 "Header reported 'aes' encryption format but "
196 "options specify '%s'", encryptfmt
);
200 qdict_del(encryptopts
, "format");
201 crypto_opts
= block_crypto_open_opts_init(
202 Q_CRYPTO_BLOCK_FORMAT_QCOW
, encryptopts
, errp
);
208 if (flags
& BDRV_O_NO_IO
) {
209 cflags
|= QCRYPTO_BLOCK_OPEN_NO_IO
;
211 s
->crypto
= qcrypto_block_open(crypto_opts
, "encrypt.",
212 NULL
, NULL
, cflags
, errp
);
218 error_setg(errp
, "invalid encryption method in qcow header");
222 bs
->encrypted
= true;
225 error_setg(errp
, "No encryption in image header, but options "
226 "specified format '%s'", encryptfmt
);
231 s
->cluster_bits
= header
.cluster_bits
;
232 s
->cluster_size
= 1 << s
->cluster_bits
;
233 s
->cluster_sectors
= 1 << (s
->cluster_bits
- 9);
234 s
->l2_bits
= header
.l2_bits
;
235 s
->l2_size
= 1 << s
->l2_bits
;
236 bs
->total_sectors
= header
.size
/ 512;
237 s
->cluster_offset_mask
= (1LL << (63 - s
->cluster_bits
)) - 1;
239 /* read the level 1 table */
240 shift
= s
->cluster_bits
+ s
->l2_bits
;
241 if (header
.size
> UINT64_MAX
- (1LL << shift
)) {
242 error_setg(errp
, "Image too large");
246 uint64_t l1_size
= (header
.size
+ (1LL << shift
) - 1) >> shift
;
247 if (l1_size
> INT_MAX
/ sizeof(uint64_t)) {
248 error_setg(errp
, "Image too large");
252 s
->l1_size
= l1_size
;
255 s
->l1_table_offset
= header
.l1_table_offset
;
256 s
->l1_table
= g_try_new(uint64_t, s
->l1_size
);
257 if (s
->l1_table
== NULL
) {
258 error_setg(errp
, "Could not allocate memory for L1 table");
263 ret
= bdrv_pread(bs
->file
, s
->l1_table_offset
, s
->l1_table
,
264 s
->l1_size
* sizeof(uint64_t));
269 for(i
= 0;i
< s
->l1_size
; i
++) {
270 be64_to_cpus(&s
->l1_table
[i
]);
273 /* alloc L2 cache (max. 64k * 16 * 8 = 8 MB) */
275 qemu_try_blockalign(bs
->file
->bs
,
276 s
->l2_size
* L2_CACHE_SIZE
* sizeof(uint64_t));
277 if (s
->l2_cache
== NULL
) {
278 error_setg(errp
, "Could not allocate L2 table cache");
282 s
->cluster_cache
= g_malloc(s
->cluster_size
);
283 s
->cluster_data
= g_malloc(s
->cluster_size
);
284 s
->cluster_cache_offset
= -1;
286 /* read the backing file name */
287 if (header
.backing_file_offset
!= 0) {
288 len
= header
.backing_file_size
;
289 if (len
> 1023 || len
>= sizeof(bs
->backing_file
)) {
290 error_setg(errp
, "Backing file name too long");
294 ret
= bdrv_pread(bs
->file
, header
.backing_file_offset
,
295 bs
->backing_file
, len
);
299 bs
->backing_file
[len
] = '\0';
302 /* Disable migration when qcow images are used */
303 error_setg(&s
->migration_blocker
, "The qcow format used by node '%s' "
304 "does not support live migration",
305 bdrv_get_device_or_node_name(bs
));
306 ret
= migrate_add_blocker(s
->migration_blocker
, &local_err
);
308 error_propagate(errp
, local_err
);
309 error_free(s
->migration_blocker
);
313 QDECREF(encryptopts
);
314 qapi_free_QCryptoBlockOpenOptions(crypto_opts
);
315 qemu_co_mutex_init(&s
->lock
);
320 qemu_vfree(s
->l2_cache
);
321 g_free(s
->cluster_cache
);
322 g_free(s
->cluster_data
);
323 qcrypto_block_free(s
->crypto
);
324 QDECREF(encryptopts
);
325 qapi_free_QCryptoBlockOpenOptions(crypto_opts
);
330 /* We have nothing to do for QCOW reopen, stubs just return
332 static int qcow_reopen_prepare(BDRVReopenState
*state
,
333 BlockReopenQueue
*queue
, Error
**errp
)
343 * 1 to allocate a normal cluster (for sector indexes 'n_start' to
346 * 2 to allocate a compressed cluster of size
347 * 'compressed_size'. 'compressed_size' must be > 0 and <
350 * return 0 if not allocated.
352 static uint64_t get_cluster_offset(BlockDriverState
*bs
,
353 uint64_t offset
, int allocate
,
355 int n_start
, int n_end
)
357 BDRVQcowState
*s
= bs
->opaque
;
358 int min_index
, i
, j
, l1_index
, l2_index
;
359 uint64_t l2_offset
, *l2_table
, cluster_offset
, tmp
;
363 l1_index
= offset
>> (s
->l2_bits
+ s
->cluster_bits
);
364 l2_offset
= s
->l1_table
[l1_index
];
369 /* allocate a new l2 entry */
370 l2_offset
= bdrv_getlength(bs
->file
->bs
);
371 /* round to cluster size */
372 l2_offset
= (l2_offset
+ s
->cluster_size
- 1) & ~(s
->cluster_size
- 1);
373 /* update the L1 entry */
374 s
->l1_table
[l1_index
] = l2_offset
;
375 tmp
= cpu_to_be64(l2_offset
);
376 if (bdrv_pwrite_sync(bs
->file
,
377 s
->l1_table_offset
+ l1_index
* sizeof(tmp
),
378 &tmp
, sizeof(tmp
)) < 0)
382 for(i
= 0; i
< L2_CACHE_SIZE
; i
++) {
383 if (l2_offset
== s
->l2_cache_offsets
[i
]) {
384 /* increment the hit count */
385 if (++s
->l2_cache_counts
[i
] == 0xffffffff) {
386 for(j
= 0; j
< L2_CACHE_SIZE
; j
++) {
387 s
->l2_cache_counts
[j
] >>= 1;
390 l2_table
= s
->l2_cache
+ (i
<< s
->l2_bits
);
394 /* not found: load a new entry in the least used one */
396 min_count
= 0xffffffff;
397 for(i
= 0; i
< L2_CACHE_SIZE
; i
++) {
398 if (s
->l2_cache_counts
[i
] < min_count
) {
399 min_count
= s
->l2_cache_counts
[i
];
403 l2_table
= s
->l2_cache
+ (min_index
<< s
->l2_bits
);
405 memset(l2_table
, 0, s
->l2_size
* sizeof(uint64_t));
406 if (bdrv_pwrite_sync(bs
->file
, l2_offset
, l2_table
,
407 s
->l2_size
* sizeof(uint64_t)) < 0)
410 if (bdrv_pread(bs
->file
, l2_offset
, l2_table
,
411 s
->l2_size
* sizeof(uint64_t)) !=
412 s
->l2_size
* sizeof(uint64_t))
415 s
->l2_cache_offsets
[min_index
] = l2_offset
;
416 s
->l2_cache_counts
[min_index
] = 1;
418 l2_index
= (offset
>> s
->cluster_bits
) & (s
->l2_size
- 1);
419 cluster_offset
= be64_to_cpu(l2_table
[l2_index
]);
420 if (!cluster_offset
||
421 ((cluster_offset
& QCOW_OFLAG_COMPRESSED
) && allocate
== 1)) {
424 /* allocate a new cluster */
425 if ((cluster_offset
& QCOW_OFLAG_COMPRESSED
) &&
426 (n_end
- n_start
) < s
->cluster_sectors
) {
427 /* if the cluster is already compressed, we must
428 decompress it in the case it is not completely
430 if (decompress_cluster(bs
, cluster_offset
) < 0)
432 cluster_offset
= bdrv_getlength(bs
->file
->bs
);
433 cluster_offset
= (cluster_offset
+ s
->cluster_size
- 1) &
434 ~(s
->cluster_size
- 1);
435 /* write the cluster content */
436 if (bdrv_pwrite(bs
->file
, cluster_offset
, s
->cluster_cache
,
441 cluster_offset
= bdrv_getlength(bs
->file
->bs
);
443 /* round to cluster size */
444 cluster_offset
= (cluster_offset
+ s
->cluster_size
- 1) &
445 ~(s
->cluster_size
- 1);
446 bdrv_truncate(bs
->file
, cluster_offset
+ s
->cluster_size
, NULL
);
447 /* if encrypted, we must initialize the cluster
448 content which won't be written */
450 (n_end
- n_start
) < s
->cluster_sectors
) {
453 start_sect
= (offset
& ~(s
->cluster_size
- 1)) >> 9;
454 for(i
= 0; i
< s
->cluster_sectors
; i
++) {
455 if (i
< n_start
|| i
>= n_end
) {
457 memset(s
->cluster_data
, 0x00, 512);
458 if (qcrypto_block_encrypt(s
->crypto
, start_sect
+ i
,
466 if (bdrv_pwrite(bs
->file
,
467 cluster_offset
+ i
* 512,
468 s
->cluster_data
, 512) != 512)
473 } else if (allocate
== 2) {
474 cluster_offset
|= QCOW_OFLAG_COMPRESSED
|
475 (uint64_t)compressed_size
<< (63 - s
->cluster_bits
);
478 /* update L2 table */
479 tmp
= cpu_to_be64(cluster_offset
);
480 l2_table
[l2_index
] = tmp
;
481 if (bdrv_pwrite_sync(bs
->file
, l2_offset
+ l2_index
* sizeof(tmp
),
482 &tmp
, sizeof(tmp
)) < 0)
485 return cluster_offset
;
488 static int64_t coroutine_fn
qcow_co_get_block_status(BlockDriverState
*bs
,
489 int64_t sector_num
, int nb_sectors
, int *pnum
, BlockDriverState
**file
)
491 BDRVQcowState
*s
= bs
->opaque
;
492 int index_in_cluster
, n
;
493 uint64_t cluster_offset
;
495 qemu_co_mutex_lock(&s
->lock
);
496 cluster_offset
= get_cluster_offset(bs
, sector_num
<< 9, 0, 0, 0, 0);
497 qemu_co_mutex_unlock(&s
->lock
);
498 index_in_cluster
= sector_num
& (s
->cluster_sectors
- 1);
499 n
= s
->cluster_sectors
- index_in_cluster
;
503 if (!cluster_offset
) {
506 if ((cluster_offset
& QCOW_OFLAG_COMPRESSED
) || s
->crypto
) {
507 return BDRV_BLOCK_DATA
;
509 cluster_offset
|= (index_in_cluster
<< BDRV_SECTOR_BITS
);
510 *file
= bs
->file
->bs
;
511 return BDRV_BLOCK_DATA
| BDRV_BLOCK_OFFSET_VALID
| cluster_offset
;
514 static int decompress_buffer(uint8_t *out_buf
, int out_buf_size
,
515 const uint8_t *buf
, int buf_size
)
517 z_stream strm1
, *strm
= &strm1
;
520 memset(strm
, 0, sizeof(*strm
));
522 strm
->next_in
= (uint8_t *)buf
;
523 strm
->avail_in
= buf_size
;
524 strm
->next_out
= out_buf
;
525 strm
->avail_out
= out_buf_size
;
527 ret
= inflateInit2(strm
, -12);
530 ret
= inflate(strm
, Z_FINISH
);
531 out_len
= strm
->next_out
- out_buf
;
532 if ((ret
!= Z_STREAM_END
&& ret
!= Z_BUF_ERROR
) ||
533 out_len
!= out_buf_size
) {
541 static int decompress_cluster(BlockDriverState
*bs
, uint64_t cluster_offset
)
543 BDRVQcowState
*s
= bs
->opaque
;
547 coffset
= cluster_offset
& s
->cluster_offset_mask
;
548 if (s
->cluster_cache_offset
!= coffset
) {
549 csize
= cluster_offset
>> (63 - s
->cluster_bits
);
550 csize
&= (s
->cluster_size
- 1);
551 ret
= bdrv_pread(bs
->file
, coffset
, s
->cluster_data
, csize
);
554 if (decompress_buffer(s
->cluster_cache
, s
->cluster_size
,
555 s
->cluster_data
, csize
) < 0) {
558 s
->cluster_cache_offset
= coffset
;
563 static coroutine_fn
int qcow_co_readv(BlockDriverState
*bs
, int64_t sector_num
,
564 int nb_sectors
, QEMUIOVector
*qiov
)
566 BDRVQcowState
*s
= bs
->opaque
;
567 int index_in_cluster
;
569 uint64_t cluster_offset
;
571 QEMUIOVector hd_qiov
;
576 if (qiov
->niov
> 1) {
577 buf
= orig_buf
= qemu_try_blockalign(bs
, qiov
->size
);
583 buf
= (uint8_t *)qiov
->iov
->iov_base
;
586 qemu_co_mutex_lock(&s
->lock
);
588 while (nb_sectors
!= 0) {
589 /* prepare next request */
590 cluster_offset
= get_cluster_offset(bs
, sector_num
<< 9,
592 index_in_cluster
= sector_num
& (s
->cluster_sectors
- 1);
593 n
= s
->cluster_sectors
- index_in_cluster
;
594 if (n
> nb_sectors
) {
598 if (!cluster_offset
) {
600 /* read from the base image */
601 hd_iov
.iov_base
= (void *)buf
;
602 hd_iov
.iov_len
= n
* 512;
603 qemu_iovec_init_external(&hd_qiov
, &hd_iov
, 1);
604 qemu_co_mutex_unlock(&s
->lock
);
605 ret
= bdrv_co_readv(bs
->backing
, sector_num
, n
, &hd_qiov
);
606 qemu_co_mutex_lock(&s
->lock
);
611 /* Note: in this case, no need to wait */
612 memset(buf
, 0, 512 * n
);
614 } else if (cluster_offset
& QCOW_OFLAG_COMPRESSED
) {
615 /* add AIO support for compressed blocks ? */
616 if (decompress_cluster(bs
, cluster_offset
) < 0) {
620 s
->cluster_cache
+ index_in_cluster
* 512, 512 * n
);
622 if ((cluster_offset
& 511) != 0) {
625 hd_iov
.iov_base
= (void *)buf
;
626 hd_iov
.iov_len
= n
* 512;
627 qemu_iovec_init_external(&hd_qiov
, &hd_iov
, 1);
628 qemu_co_mutex_unlock(&s
->lock
);
629 ret
= bdrv_co_readv(bs
->file
,
630 (cluster_offset
>> 9) + index_in_cluster
,
632 qemu_co_mutex_lock(&s
->lock
);
638 if (qcrypto_block_decrypt(s
->crypto
, sector_num
, buf
,
639 n
* BDRV_SECTOR_SIZE
, &err
) < 0) {
652 qemu_co_mutex_unlock(&s
->lock
);
654 if (qiov
->niov
> 1) {
655 qemu_iovec_from_buf(qiov
, 0, orig_buf
, qiov
->size
);
656 qemu_vfree(orig_buf
);
667 static coroutine_fn
int qcow_co_writev(BlockDriverState
*bs
, int64_t sector_num
,
668 int nb_sectors
, QEMUIOVector
*qiov
)
670 BDRVQcowState
*s
= bs
->opaque
;
671 int index_in_cluster
;
672 uint64_t cluster_offset
;
675 QEMUIOVector hd_qiov
;
679 s
->cluster_cache_offset
= -1; /* disable compressed cache */
681 /* We must always copy the iov when encrypting, so we
682 * don't modify the original data buffer during encryption */
683 if (bs
->encrypted
|| qiov
->niov
> 1) {
684 buf
= orig_buf
= qemu_try_blockalign(bs
, qiov
->size
);
688 qemu_iovec_to_buf(qiov
, 0, buf
, qiov
->size
);
691 buf
= (uint8_t *)qiov
->iov
->iov_base
;
694 qemu_co_mutex_lock(&s
->lock
);
696 while (nb_sectors
!= 0) {
698 index_in_cluster
= sector_num
& (s
->cluster_sectors
- 1);
699 n
= s
->cluster_sectors
- index_in_cluster
;
700 if (n
> nb_sectors
) {
703 cluster_offset
= get_cluster_offset(bs
, sector_num
<< 9, 1, 0,
705 index_in_cluster
+ n
);
706 if (!cluster_offset
|| (cluster_offset
& 511) != 0) {
713 if (qcrypto_block_encrypt(s
->crypto
, sector_num
, buf
,
714 n
* BDRV_SECTOR_SIZE
, &err
) < 0) {
721 hd_iov
.iov_base
= (void *)buf
;
722 hd_iov
.iov_len
= n
* 512;
723 qemu_iovec_init_external(&hd_qiov
, &hd_iov
, 1);
724 qemu_co_mutex_unlock(&s
->lock
);
725 ret
= bdrv_co_writev(bs
->file
,
726 (cluster_offset
>> 9) + index_in_cluster
,
728 qemu_co_mutex_lock(&s
->lock
);
738 qemu_co_mutex_unlock(&s
->lock
);
740 qemu_vfree(orig_buf
);
745 static void qcow_close(BlockDriverState
*bs
)
747 BDRVQcowState
*s
= bs
->opaque
;
749 qcrypto_block_free(s
->crypto
);
752 qemu_vfree(s
->l2_cache
);
753 g_free(s
->cluster_cache
);
754 g_free(s
->cluster_data
);
756 migrate_del_blocker(s
->migration_blocker
);
757 error_free(s
->migration_blocker
);
760 static int qcow_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
762 int header_size
, backing_filename_len
, l1_size
, shift
, i
;
765 int64_t total_size
= 0;
766 char *backing_file
= NULL
;
767 Error
*local_err
= NULL
;
769 BlockBackend
*qcow_blk
;
770 const char *encryptfmt
= NULL
;
772 QDict
*encryptopts
= NULL
;
773 QCryptoBlockCreateOptions
*crypto_opts
= NULL
;
774 QCryptoBlock
*crypto
= NULL
;
776 /* Read out options */
777 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
779 if (total_size
== 0) {
780 error_setg(errp
, "Image size is too small, cannot be zero length");
785 backing_file
= qemu_opt_get_del(opts
, BLOCK_OPT_BACKING_FILE
);
786 encryptfmt
= qemu_opt_get_del(opts
, BLOCK_OPT_ENCRYPT_FORMAT
);
788 if (qemu_opt_get(opts
, BLOCK_OPT_ENCRYPT
)) {
789 error_setg(errp
, "Options " BLOCK_OPT_ENCRYPT
" and "
790 BLOCK_OPT_ENCRYPT_FORMAT
" are mutually exclusive");
794 } else if (qemu_opt_get_bool_del(opts
, BLOCK_OPT_ENCRYPT
, false)) {
798 ret
= bdrv_create_file(filename
, opts
, &local_err
);
800 error_propagate(errp
, local_err
);
804 qcow_blk
= blk_new_open(filename
, NULL
, NULL
,
805 BDRV_O_RDWR
| BDRV_O_RESIZE
| BDRV_O_PROTOCOL
,
807 if (qcow_blk
== NULL
) {
808 error_propagate(errp
, local_err
);
813 blk_set_allow_write_beyond_eof(qcow_blk
, true);
815 ret
= blk_truncate(qcow_blk
, 0, errp
);
820 memset(&header
, 0, sizeof(header
));
821 header
.magic
= cpu_to_be32(QCOW_MAGIC
);
822 header
.version
= cpu_to_be32(QCOW_VERSION
);
823 header
.size
= cpu_to_be64(total_size
);
824 header_size
= sizeof(header
);
825 backing_filename_len
= 0;
827 if (strcmp(backing_file
, "fat:")) {
828 header
.backing_file_offset
= cpu_to_be64(header_size
);
829 backing_filename_len
= strlen(backing_file
);
830 header
.backing_file_size
= cpu_to_be32(backing_filename_len
);
831 header_size
+= backing_filename_len
;
833 /* special backing file for vvfat */
834 g_free(backing_file
);
837 header
.cluster_bits
= 9; /* 512 byte cluster to avoid copying
838 unmodified sectors */
839 header
.l2_bits
= 12; /* 32 KB L2 tables */
841 header
.cluster_bits
= 12; /* 4 KB clusters */
842 header
.l2_bits
= 9; /* 4 KB L2 tables */
844 header_size
= (header_size
+ 7) & ~7;
845 shift
= header
.cluster_bits
+ header
.l2_bits
;
846 l1_size
= (total_size
+ (1LL << shift
) - 1) >> shift
;
848 header
.l1_table_offset
= cpu_to_be64(header_size
);
850 options
= qemu_opts_to_qdict(opts
, NULL
);
851 qdict_extract_subqdict(options
, &encryptopts
, "encrypt.");
854 if (!g_str_equal(encryptfmt
, "aes")) {
855 error_setg(errp
, "Unknown encryption format '%s', expected 'aes'",
860 header
.crypt_method
= cpu_to_be32(QCOW_CRYPT_AES
);
862 crypto_opts
= block_crypto_create_opts_init(
863 Q_CRYPTO_BLOCK_FORMAT_QCOW
, encryptopts
, errp
);
869 crypto
= qcrypto_block_create(crypto_opts
, "encrypt.",
870 NULL
, NULL
, NULL
, errp
);
876 header
.crypt_method
= cpu_to_be32(QCOW_CRYPT_NONE
);
879 /* write all the data */
880 ret
= blk_pwrite(qcow_blk
, 0, &header
, sizeof(header
), 0);
881 if (ret
!= sizeof(header
)) {
886 ret
= blk_pwrite(qcow_blk
, sizeof(header
),
887 backing_file
, backing_filename_len
, 0);
888 if (ret
!= backing_filename_len
) {
893 tmp
= g_malloc0(BDRV_SECTOR_SIZE
);
894 for (i
= 0; i
< DIV_ROUND_UP(sizeof(uint64_t) * l1_size
, BDRV_SECTOR_SIZE
);
896 ret
= blk_pwrite(qcow_blk
, header_size
+ BDRV_SECTOR_SIZE
* i
,
897 tmp
, BDRV_SECTOR_SIZE
, 0);
898 if (ret
!= BDRV_SECTOR_SIZE
) {
909 QDECREF(encryptopts
);
910 qcrypto_block_free(crypto
);
911 qapi_free_QCryptoBlockCreateOptions(crypto_opts
);
912 g_free(backing_file
);
916 static int qcow_make_empty(BlockDriverState
*bs
)
918 BDRVQcowState
*s
= bs
->opaque
;
919 uint32_t l1_length
= s
->l1_size
* sizeof(uint64_t);
922 memset(s
->l1_table
, 0, l1_length
);
923 if (bdrv_pwrite_sync(bs
->file
, s
->l1_table_offset
, s
->l1_table
,
926 ret
= bdrv_truncate(bs
->file
, s
->l1_table_offset
+ l1_length
, NULL
);
930 memset(s
->l2_cache
, 0, s
->l2_size
* L2_CACHE_SIZE
* sizeof(uint64_t));
931 memset(s
->l2_cache_offsets
, 0, L2_CACHE_SIZE
* sizeof(uint64_t));
932 memset(s
->l2_cache_counts
, 0, L2_CACHE_SIZE
* sizeof(uint32_t));
937 /* XXX: put compressed sectors first, then all the cluster aligned
938 tables to avoid losing bytes in alignment */
939 static coroutine_fn
int
940 qcow_co_pwritev_compressed(BlockDriverState
*bs
, uint64_t offset
,
941 uint64_t bytes
, QEMUIOVector
*qiov
)
943 BDRVQcowState
*s
= bs
->opaque
;
944 QEMUIOVector hd_qiov
;
948 uint8_t *buf
, *out_buf
;
949 uint64_t cluster_offset
;
951 buf
= qemu_blockalign(bs
, s
->cluster_size
);
952 if (bytes
!= s
->cluster_size
) {
953 if (bytes
> s
->cluster_size
||
954 offset
+ bytes
!= bs
->total_sectors
<< BDRV_SECTOR_BITS
)
959 /* Zero-pad last write if image size is not cluster aligned */
960 memset(buf
+ bytes
, 0, s
->cluster_size
- bytes
);
962 qemu_iovec_to_buf(qiov
, 0, buf
, qiov
->size
);
964 out_buf
= g_malloc(s
->cluster_size
);
966 /* best compression, small window, no zlib header */
967 memset(&strm
, 0, sizeof(strm
));
968 ret
= deflateInit2(&strm
, Z_DEFAULT_COMPRESSION
,
970 9, Z_DEFAULT_STRATEGY
);
976 strm
.avail_in
= s
->cluster_size
;
977 strm
.next_in
= (uint8_t *)buf
;
978 strm
.avail_out
= s
->cluster_size
;
979 strm
.next_out
= out_buf
;
981 ret
= deflate(&strm
, Z_FINISH
);
982 if (ret
!= Z_STREAM_END
&& ret
!= Z_OK
) {
987 out_len
= strm
.next_out
- out_buf
;
991 if (ret
!= Z_STREAM_END
|| out_len
>= s
->cluster_size
) {
992 /* could not compress: write normal cluster */
993 ret
= qcow_co_writev(bs
, offset
>> BDRV_SECTOR_BITS
,
994 bytes
>> BDRV_SECTOR_BITS
, qiov
);
1000 qemu_co_mutex_lock(&s
->lock
);
1001 cluster_offset
= get_cluster_offset(bs
, offset
, 2, out_len
, 0, 0);
1002 qemu_co_mutex_unlock(&s
->lock
);
1003 if (cluster_offset
== 0) {
1007 cluster_offset
&= s
->cluster_offset_mask
;
1009 iov
= (struct iovec
) {
1010 .iov_base
= out_buf
,
1013 qemu_iovec_init_external(&hd_qiov
, &iov
, 1);
1014 ret
= bdrv_co_pwritev(bs
->file
, cluster_offset
, out_len
, &hd_qiov
, 0);
1026 static int qcow_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1028 BDRVQcowState
*s
= bs
->opaque
;
1029 bdi
->cluster_size
= s
->cluster_size
;
1033 static QemuOptsList qcow_create_opts
= {
1034 .name
= "qcow-create-opts",
1035 .head
= QTAILQ_HEAD_INITIALIZER(qcow_create_opts
.head
),
1038 .name
= BLOCK_OPT_SIZE
,
1039 .type
= QEMU_OPT_SIZE
,
1040 .help
= "Virtual disk size"
1043 .name
= BLOCK_OPT_BACKING_FILE
,
1044 .type
= QEMU_OPT_STRING
,
1045 .help
= "File name of a base image"
1048 .name
= BLOCK_OPT_ENCRYPT
,
1049 .type
= QEMU_OPT_BOOL
,
1050 .help
= "Encrypt the image with format 'aes'. (Deprecated "
1051 "in favor of " BLOCK_OPT_ENCRYPT_FORMAT
"=aes)",
1054 .name
= BLOCK_OPT_ENCRYPT_FORMAT
,
1055 .type
= QEMU_OPT_STRING
,
1056 .help
= "Encrypt the image, format choices: 'aes'",
1058 BLOCK_CRYPTO_OPT_DEF_QCOW_KEY_SECRET("encrypt."),
1059 { /* end of list */ }
1063 static BlockDriver bdrv_qcow
= {
1064 .format_name
= "qcow",
1065 .instance_size
= sizeof(BDRVQcowState
),
1066 .bdrv_probe
= qcow_probe
,
1067 .bdrv_open
= qcow_open
,
1068 .bdrv_close
= qcow_close
,
1069 .bdrv_child_perm
= bdrv_format_default_perms
,
1070 .bdrv_reopen_prepare
= qcow_reopen_prepare
,
1071 .bdrv_create
= qcow_create
,
1072 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1073 .supports_backing
= true,
1075 .bdrv_co_readv
= qcow_co_readv
,
1076 .bdrv_co_writev
= qcow_co_writev
,
1077 .bdrv_co_get_block_status
= qcow_co_get_block_status
,
1079 .bdrv_make_empty
= qcow_make_empty
,
1080 .bdrv_co_pwritev_compressed
= qcow_co_pwritev_compressed
,
1081 .bdrv_get_info
= qcow_get_info
,
1083 .create_opts
= &qcow_create_opts
,
1086 static void bdrv_qcow_init(void)
1088 bdrv_register(&bdrv_qcow
);
1091 block_init(bdrv_qcow_init
);