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, 1 if *result is assigned, and negative
353 static int get_cluster_offset(BlockDriverState
*bs
,
354 uint64_t offset
, int allocate
,
356 int n_start
, int n_end
, uint64_t *result
)
358 BDRVQcowState
*s
= bs
->opaque
;
359 int min_index
, i
, j
, l1_index
, l2_index
, ret
;
360 uint64_t l2_offset
, *l2_table
, cluster_offset
, tmp
;
365 l1_index
= offset
>> (s
->l2_bits
+ s
->cluster_bits
);
366 l2_offset
= s
->l1_table
[l1_index
];
371 /* allocate a new l2 entry */
372 l2_offset
= bdrv_getlength(bs
->file
->bs
);
373 /* round to cluster size */
374 l2_offset
= (l2_offset
+ s
->cluster_size
- 1) & ~(s
->cluster_size
- 1);
375 /* update the L1 entry */
376 s
->l1_table
[l1_index
] = l2_offset
;
377 tmp
= cpu_to_be64(l2_offset
);
378 ret
= bdrv_pwrite_sync(bs
->file
,
379 s
->l1_table_offset
+ l1_index
* sizeof(tmp
),
386 for(i
= 0; i
< L2_CACHE_SIZE
; i
++) {
387 if (l2_offset
== s
->l2_cache_offsets
[i
]) {
388 /* increment the hit count */
389 if (++s
->l2_cache_counts
[i
] == 0xffffffff) {
390 for(j
= 0; j
< L2_CACHE_SIZE
; j
++) {
391 s
->l2_cache_counts
[j
] >>= 1;
394 l2_table
= s
->l2_cache
+ (i
<< s
->l2_bits
);
398 /* not found: load a new entry in the least used one */
400 min_count
= 0xffffffff;
401 for(i
= 0; i
< L2_CACHE_SIZE
; i
++) {
402 if (s
->l2_cache_counts
[i
] < min_count
) {
403 min_count
= s
->l2_cache_counts
[i
];
407 l2_table
= s
->l2_cache
+ (min_index
<< s
->l2_bits
);
409 memset(l2_table
, 0, s
->l2_size
* sizeof(uint64_t));
410 ret
= bdrv_pwrite_sync(bs
->file
, l2_offset
, l2_table
,
411 s
->l2_size
* sizeof(uint64_t));
416 ret
= bdrv_pread(bs
->file
, l2_offset
, l2_table
,
417 s
->l2_size
* sizeof(uint64_t));
422 s
->l2_cache_offsets
[min_index
] = l2_offset
;
423 s
->l2_cache_counts
[min_index
] = 1;
425 l2_index
= (offset
>> s
->cluster_bits
) & (s
->l2_size
- 1);
426 cluster_offset
= be64_to_cpu(l2_table
[l2_index
]);
427 if (!cluster_offset
||
428 ((cluster_offset
& QCOW_OFLAG_COMPRESSED
) && allocate
== 1)) {
431 /* allocate a new cluster */
432 if ((cluster_offset
& QCOW_OFLAG_COMPRESSED
) &&
433 (n_end
- n_start
) < s
->cluster_sectors
) {
434 /* if the cluster is already compressed, we must
435 decompress it in the case it is not completely
437 if (decompress_cluster(bs
, cluster_offset
) < 0) {
440 cluster_offset
= bdrv_getlength(bs
->file
->bs
);
441 cluster_offset
= (cluster_offset
+ s
->cluster_size
- 1) &
442 ~(s
->cluster_size
- 1);
443 /* write the cluster content */
444 ret
= bdrv_pwrite(bs
->file
, cluster_offset
, s
->cluster_cache
,
450 cluster_offset
= bdrv_getlength(bs
->file
->bs
);
452 /* round to cluster size */
453 cluster_offset
= (cluster_offset
+ s
->cluster_size
- 1) &
454 ~(s
->cluster_size
- 1);
455 bdrv_truncate(bs
->file
, cluster_offset
+ s
->cluster_size
,
456 PREALLOC_MODE_OFF
, NULL
);
457 /* if encrypted, we must initialize the cluster
458 content which won't be written */
460 (n_end
- n_start
) < s
->cluster_sectors
) {
463 start_sect
= (offset
& ~(s
->cluster_size
- 1)) >> 9;
464 for(i
= 0; i
< s
->cluster_sectors
; i
++) {
465 if (i
< n_start
|| i
>= n_end
) {
466 memset(s
->cluster_data
, 0x00, 512);
467 if (qcrypto_block_encrypt(s
->crypto
, start_sect
+ i
,
473 ret
= bdrv_pwrite(bs
->file
,
474 cluster_offset
+ i
* 512,
475 s
->cluster_data
, 512);
482 } else if (allocate
== 2) {
483 cluster_offset
|= QCOW_OFLAG_COMPRESSED
|
484 (uint64_t)compressed_size
<< (63 - s
->cluster_bits
);
487 /* update L2 table */
488 tmp
= cpu_to_be64(cluster_offset
);
489 l2_table
[l2_index
] = tmp
;
490 ret
= bdrv_pwrite_sync(bs
->file
, l2_offset
+ l2_index
* sizeof(tmp
),
496 *result
= cluster_offset
;
500 static int64_t coroutine_fn
qcow_co_get_block_status(BlockDriverState
*bs
,
501 int64_t sector_num
, int nb_sectors
, int *pnum
, BlockDriverState
**file
)
503 BDRVQcowState
*s
= bs
->opaque
;
504 int index_in_cluster
, n
, ret
;
505 uint64_t cluster_offset
;
507 qemu_co_mutex_lock(&s
->lock
);
508 ret
= get_cluster_offset(bs
, sector_num
<< 9, 0, 0, 0, 0, &cluster_offset
);
509 qemu_co_mutex_unlock(&s
->lock
);
513 index_in_cluster
= sector_num
& (s
->cluster_sectors
- 1);
514 n
= s
->cluster_sectors
- index_in_cluster
;
518 if (!cluster_offset
) {
521 if ((cluster_offset
& QCOW_OFLAG_COMPRESSED
) || s
->crypto
) {
522 return BDRV_BLOCK_DATA
;
524 cluster_offset
|= (index_in_cluster
<< BDRV_SECTOR_BITS
);
525 *file
= bs
->file
->bs
;
526 return BDRV_BLOCK_DATA
| BDRV_BLOCK_OFFSET_VALID
| cluster_offset
;
529 static int decompress_buffer(uint8_t *out_buf
, int out_buf_size
,
530 const uint8_t *buf
, int buf_size
)
532 z_stream strm1
, *strm
= &strm1
;
535 memset(strm
, 0, sizeof(*strm
));
537 strm
->next_in
= (uint8_t *)buf
;
538 strm
->avail_in
= buf_size
;
539 strm
->next_out
= out_buf
;
540 strm
->avail_out
= out_buf_size
;
542 ret
= inflateInit2(strm
, -12);
545 ret
= inflate(strm
, Z_FINISH
);
546 out_len
= strm
->next_out
- out_buf
;
547 if ((ret
!= Z_STREAM_END
&& ret
!= Z_BUF_ERROR
) ||
548 out_len
!= out_buf_size
) {
556 static int decompress_cluster(BlockDriverState
*bs
, uint64_t cluster_offset
)
558 BDRVQcowState
*s
= bs
->opaque
;
562 coffset
= cluster_offset
& s
->cluster_offset_mask
;
563 if (s
->cluster_cache_offset
!= coffset
) {
564 csize
= cluster_offset
>> (63 - s
->cluster_bits
);
565 csize
&= (s
->cluster_size
- 1);
566 ret
= bdrv_pread(bs
->file
, coffset
, s
->cluster_data
, csize
);
569 if (decompress_buffer(s
->cluster_cache
, s
->cluster_size
,
570 s
->cluster_data
, csize
) < 0) {
573 s
->cluster_cache_offset
= coffset
;
578 static coroutine_fn
int qcow_co_readv(BlockDriverState
*bs
, int64_t sector_num
,
579 int nb_sectors
, QEMUIOVector
*qiov
)
581 BDRVQcowState
*s
= bs
->opaque
;
582 int index_in_cluster
;
584 uint64_t cluster_offset
;
586 QEMUIOVector hd_qiov
;
590 if (qiov
->niov
> 1) {
591 buf
= orig_buf
= qemu_try_blockalign(bs
, qiov
->size
);
597 buf
= (uint8_t *)qiov
->iov
->iov_base
;
600 qemu_co_mutex_lock(&s
->lock
);
602 while (nb_sectors
!= 0) {
603 /* prepare next request */
604 ret
= get_cluster_offset(bs
, sector_num
<< 9,
605 0, 0, 0, 0, &cluster_offset
);
609 index_in_cluster
= sector_num
& (s
->cluster_sectors
- 1);
610 n
= s
->cluster_sectors
- index_in_cluster
;
611 if (n
> nb_sectors
) {
615 if (!cluster_offset
) {
617 /* read from the base image */
618 hd_iov
.iov_base
= (void *)buf
;
619 hd_iov
.iov_len
= n
* 512;
620 qemu_iovec_init_external(&hd_qiov
, &hd_iov
, 1);
621 qemu_co_mutex_unlock(&s
->lock
);
622 ret
= bdrv_co_readv(bs
->backing
, sector_num
, n
, &hd_qiov
);
623 qemu_co_mutex_lock(&s
->lock
);
628 /* Note: in this case, no need to wait */
629 memset(buf
, 0, 512 * n
);
631 } else if (cluster_offset
& QCOW_OFLAG_COMPRESSED
) {
632 /* add AIO support for compressed blocks ? */
633 if (decompress_cluster(bs
, cluster_offset
) < 0) {
638 s
->cluster_cache
+ index_in_cluster
* 512, 512 * n
);
640 if ((cluster_offset
& 511) != 0) {
644 hd_iov
.iov_base
= (void *)buf
;
645 hd_iov
.iov_len
= n
* 512;
646 qemu_iovec_init_external(&hd_qiov
, &hd_iov
, 1);
647 qemu_co_mutex_unlock(&s
->lock
);
648 ret
= bdrv_co_readv(bs
->file
,
649 (cluster_offset
>> 9) + index_in_cluster
,
651 qemu_co_mutex_lock(&s
->lock
);
657 if (qcrypto_block_decrypt(s
->crypto
, sector_num
, buf
,
658 n
* BDRV_SECTOR_SIZE
, NULL
) < 0) {
671 qemu_co_mutex_unlock(&s
->lock
);
673 if (qiov
->niov
> 1) {
674 qemu_iovec_from_buf(qiov
, 0, orig_buf
, qiov
->size
);
675 qemu_vfree(orig_buf
);
681 static coroutine_fn
int qcow_co_writev(BlockDriverState
*bs
, int64_t sector_num
,
682 int nb_sectors
, QEMUIOVector
*qiov
)
684 BDRVQcowState
*s
= bs
->opaque
;
685 int index_in_cluster
;
686 uint64_t cluster_offset
;
689 QEMUIOVector hd_qiov
;
693 s
->cluster_cache_offset
= -1; /* disable compressed cache */
695 /* We must always copy the iov when encrypting, so we
696 * don't modify the original data buffer during encryption */
697 if (bs
->encrypted
|| qiov
->niov
> 1) {
698 buf
= orig_buf
= qemu_try_blockalign(bs
, qiov
->size
);
702 qemu_iovec_to_buf(qiov
, 0, buf
, qiov
->size
);
705 buf
= (uint8_t *)qiov
->iov
->iov_base
;
708 qemu_co_mutex_lock(&s
->lock
);
710 while (nb_sectors
!= 0) {
712 index_in_cluster
= sector_num
& (s
->cluster_sectors
- 1);
713 n
= s
->cluster_sectors
- index_in_cluster
;
714 if (n
> nb_sectors
) {
717 ret
= get_cluster_offset(bs
, sector_num
<< 9, 1, 0,
719 index_in_cluster
+ n
, &cluster_offset
);
723 if (!cluster_offset
|| (cluster_offset
& 511) != 0) {
729 if (qcrypto_block_encrypt(s
->crypto
, sector_num
, buf
,
730 n
* BDRV_SECTOR_SIZE
, NULL
) < 0) {
736 hd_iov
.iov_base
= (void *)buf
;
737 hd_iov
.iov_len
= n
* 512;
738 qemu_iovec_init_external(&hd_qiov
, &hd_iov
, 1);
739 qemu_co_mutex_unlock(&s
->lock
);
740 ret
= bdrv_co_writev(bs
->file
,
741 (cluster_offset
>> 9) + index_in_cluster
,
743 qemu_co_mutex_lock(&s
->lock
);
753 qemu_co_mutex_unlock(&s
->lock
);
755 qemu_vfree(orig_buf
);
760 static void qcow_close(BlockDriverState
*bs
)
762 BDRVQcowState
*s
= bs
->opaque
;
764 qcrypto_block_free(s
->crypto
);
767 qemu_vfree(s
->l2_cache
);
768 g_free(s
->cluster_cache
);
769 g_free(s
->cluster_data
);
771 migrate_del_blocker(s
->migration_blocker
);
772 error_free(s
->migration_blocker
);
775 static int qcow_create(const char *filename
, QemuOpts
*opts
, Error
**errp
)
777 int header_size
, backing_filename_len
, l1_size
, shift
, i
;
780 int64_t total_size
= 0;
781 char *backing_file
= NULL
;
782 Error
*local_err
= NULL
;
784 BlockBackend
*qcow_blk
;
785 char *encryptfmt
= NULL
;
787 QDict
*encryptopts
= NULL
;
788 QCryptoBlockCreateOptions
*crypto_opts
= NULL
;
789 QCryptoBlock
*crypto
= NULL
;
791 /* Read out options */
792 total_size
= ROUND_UP(qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0),
794 if (total_size
== 0) {
795 error_setg(errp
, "Image size is too small, cannot be zero length");
800 backing_file
= qemu_opt_get_del(opts
, BLOCK_OPT_BACKING_FILE
);
801 encryptfmt
= qemu_opt_get_del(opts
, BLOCK_OPT_ENCRYPT_FORMAT
);
803 if (qemu_opt_get(opts
, BLOCK_OPT_ENCRYPT
)) {
804 error_setg(errp
, "Options " BLOCK_OPT_ENCRYPT
" and "
805 BLOCK_OPT_ENCRYPT_FORMAT
" are mutually exclusive");
809 } else if (qemu_opt_get_bool_del(opts
, BLOCK_OPT_ENCRYPT
, false)) {
810 encryptfmt
= g_strdup("aes");
813 ret
= bdrv_create_file(filename
, opts
, &local_err
);
815 error_propagate(errp
, local_err
);
819 qcow_blk
= blk_new_open(filename
, NULL
, NULL
,
820 BDRV_O_RDWR
| BDRV_O_RESIZE
| BDRV_O_PROTOCOL
,
822 if (qcow_blk
== NULL
) {
823 error_propagate(errp
, local_err
);
828 blk_set_allow_write_beyond_eof(qcow_blk
, true);
830 ret
= blk_truncate(qcow_blk
, 0, PREALLOC_MODE_OFF
, errp
);
835 memset(&header
, 0, sizeof(header
));
836 header
.magic
= cpu_to_be32(QCOW_MAGIC
);
837 header
.version
= cpu_to_be32(QCOW_VERSION
);
838 header
.size
= cpu_to_be64(total_size
);
839 header_size
= sizeof(header
);
840 backing_filename_len
= 0;
842 if (strcmp(backing_file
, "fat:")) {
843 header
.backing_file_offset
= cpu_to_be64(header_size
);
844 backing_filename_len
= strlen(backing_file
);
845 header
.backing_file_size
= cpu_to_be32(backing_filename_len
);
846 header_size
+= backing_filename_len
;
848 /* special backing file for vvfat */
849 g_free(backing_file
);
852 header
.cluster_bits
= 9; /* 512 byte cluster to avoid copying
853 unmodified sectors */
854 header
.l2_bits
= 12; /* 32 KB L2 tables */
856 header
.cluster_bits
= 12; /* 4 KB clusters */
857 header
.l2_bits
= 9; /* 4 KB L2 tables */
859 header_size
= (header_size
+ 7) & ~7;
860 shift
= header
.cluster_bits
+ header
.l2_bits
;
861 l1_size
= (total_size
+ (1LL << shift
) - 1) >> shift
;
863 header
.l1_table_offset
= cpu_to_be64(header_size
);
865 options
= qemu_opts_to_qdict(opts
, NULL
);
866 qdict_extract_subqdict(options
, &encryptopts
, "encrypt.");
869 if (!g_str_equal(encryptfmt
, "aes")) {
870 error_setg(errp
, "Unknown encryption format '%s', expected 'aes'",
875 header
.crypt_method
= cpu_to_be32(QCOW_CRYPT_AES
);
877 crypto_opts
= block_crypto_create_opts_init(
878 Q_CRYPTO_BLOCK_FORMAT_QCOW
, encryptopts
, errp
);
884 crypto
= qcrypto_block_create(crypto_opts
, "encrypt.",
885 NULL
, NULL
, NULL
, errp
);
891 header
.crypt_method
= cpu_to_be32(QCOW_CRYPT_NONE
);
894 /* write all the data */
895 ret
= blk_pwrite(qcow_blk
, 0, &header
, sizeof(header
), 0);
896 if (ret
!= sizeof(header
)) {
901 ret
= blk_pwrite(qcow_blk
, sizeof(header
),
902 backing_file
, backing_filename_len
, 0);
903 if (ret
!= backing_filename_len
) {
908 tmp
= g_malloc0(BDRV_SECTOR_SIZE
);
909 for (i
= 0; i
< DIV_ROUND_UP(sizeof(uint64_t) * l1_size
, BDRV_SECTOR_SIZE
);
911 ret
= blk_pwrite(qcow_blk
, header_size
+ BDRV_SECTOR_SIZE
* i
,
912 tmp
, BDRV_SECTOR_SIZE
, 0);
913 if (ret
!= BDRV_SECTOR_SIZE
) {
924 QDECREF(encryptopts
);
926 qcrypto_block_free(crypto
);
927 qapi_free_QCryptoBlockCreateOptions(crypto_opts
);
928 g_free(backing_file
);
932 static int qcow_make_empty(BlockDriverState
*bs
)
934 BDRVQcowState
*s
= bs
->opaque
;
935 uint32_t l1_length
= s
->l1_size
* sizeof(uint64_t);
938 memset(s
->l1_table
, 0, l1_length
);
939 if (bdrv_pwrite_sync(bs
->file
, s
->l1_table_offset
, s
->l1_table
,
942 ret
= bdrv_truncate(bs
->file
, s
->l1_table_offset
+ l1_length
,
943 PREALLOC_MODE_OFF
, NULL
);
947 memset(s
->l2_cache
, 0, s
->l2_size
* L2_CACHE_SIZE
* sizeof(uint64_t));
948 memset(s
->l2_cache_offsets
, 0, L2_CACHE_SIZE
* sizeof(uint64_t));
949 memset(s
->l2_cache_counts
, 0, L2_CACHE_SIZE
* sizeof(uint32_t));
954 /* XXX: put compressed sectors first, then all the cluster aligned
955 tables to avoid losing bytes in alignment */
956 static coroutine_fn
int
957 qcow_co_pwritev_compressed(BlockDriverState
*bs
, uint64_t offset
,
958 uint64_t bytes
, QEMUIOVector
*qiov
)
960 BDRVQcowState
*s
= bs
->opaque
;
961 QEMUIOVector hd_qiov
;
965 uint8_t *buf
, *out_buf
;
966 uint64_t cluster_offset
;
968 buf
= qemu_blockalign(bs
, s
->cluster_size
);
969 if (bytes
!= s
->cluster_size
) {
970 if (bytes
> s
->cluster_size
||
971 offset
+ bytes
!= bs
->total_sectors
<< BDRV_SECTOR_BITS
)
976 /* Zero-pad last write if image size is not cluster aligned */
977 memset(buf
+ bytes
, 0, s
->cluster_size
- bytes
);
979 qemu_iovec_to_buf(qiov
, 0, buf
, qiov
->size
);
981 out_buf
= g_malloc(s
->cluster_size
);
983 /* best compression, small window, no zlib header */
984 memset(&strm
, 0, sizeof(strm
));
985 ret
= deflateInit2(&strm
, Z_DEFAULT_COMPRESSION
,
987 9, Z_DEFAULT_STRATEGY
);
993 strm
.avail_in
= s
->cluster_size
;
994 strm
.next_in
= (uint8_t *)buf
;
995 strm
.avail_out
= s
->cluster_size
;
996 strm
.next_out
= out_buf
;
998 ret
= deflate(&strm
, Z_FINISH
);
999 if (ret
!= Z_STREAM_END
&& ret
!= Z_OK
) {
1004 out_len
= strm
.next_out
- out_buf
;
1008 if (ret
!= Z_STREAM_END
|| out_len
>= s
->cluster_size
) {
1009 /* could not compress: write normal cluster */
1010 ret
= qcow_co_writev(bs
, offset
>> BDRV_SECTOR_BITS
,
1011 bytes
>> BDRV_SECTOR_BITS
, qiov
);
1017 qemu_co_mutex_lock(&s
->lock
);
1018 ret
= get_cluster_offset(bs
, offset
, 2, out_len
, 0, 0, &cluster_offset
);
1019 qemu_co_mutex_unlock(&s
->lock
);
1023 if (cluster_offset
== 0) {
1027 cluster_offset
&= s
->cluster_offset_mask
;
1029 iov
= (struct iovec
) {
1030 .iov_base
= out_buf
,
1033 qemu_iovec_init_external(&hd_qiov
, &iov
, 1);
1034 ret
= bdrv_co_pwritev(bs
->file
, cluster_offset
, out_len
, &hd_qiov
, 0);
1046 static int qcow_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1048 BDRVQcowState
*s
= bs
->opaque
;
1049 bdi
->cluster_size
= s
->cluster_size
;
1053 static QemuOptsList qcow_create_opts
= {
1054 .name
= "qcow-create-opts",
1055 .head
= QTAILQ_HEAD_INITIALIZER(qcow_create_opts
.head
),
1058 .name
= BLOCK_OPT_SIZE
,
1059 .type
= QEMU_OPT_SIZE
,
1060 .help
= "Virtual disk size"
1063 .name
= BLOCK_OPT_BACKING_FILE
,
1064 .type
= QEMU_OPT_STRING
,
1065 .help
= "File name of a base image"
1068 .name
= BLOCK_OPT_ENCRYPT
,
1069 .type
= QEMU_OPT_BOOL
,
1070 .help
= "Encrypt the image with format 'aes'. (Deprecated "
1071 "in favor of " BLOCK_OPT_ENCRYPT_FORMAT
"=aes)",
1074 .name
= BLOCK_OPT_ENCRYPT_FORMAT
,
1075 .type
= QEMU_OPT_STRING
,
1076 .help
= "Encrypt the image, format choices: 'aes'",
1078 BLOCK_CRYPTO_OPT_DEF_QCOW_KEY_SECRET("encrypt."),
1079 { /* end of list */ }
1083 static BlockDriver bdrv_qcow
= {
1084 .format_name
= "qcow",
1085 .instance_size
= sizeof(BDRVQcowState
),
1086 .bdrv_probe
= qcow_probe
,
1087 .bdrv_open
= qcow_open
,
1088 .bdrv_close
= qcow_close
,
1089 .bdrv_child_perm
= bdrv_format_default_perms
,
1090 .bdrv_reopen_prepare
= qcow_reopen_prepare
,
1091 .bdrv_create
= qcow_create
,
1092 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
1093 .supports_backing
= true,
1095 .bdrv_co_readv
= qcow_co_readv
,
1096 .bdrv_co_writev
= qcow_co_writev
,
1097 .bdrv_co_get_block_status
= qcow_co_get_block_status
,
1099 .bdrv_make_empty
= qcow_make_empty
,
1100 .bdrv_co_pwritev_compressed
= qcow_co_pwritev_compressed
,
1101 .bdrv_get_info
= qcow_get_info
,
1103 .create_opts
= &qcow_create_opts
,
1106 static void bdrv_qcow_init(void)
1108 bdrv_register(&bdrv_qcow
);
1111 block_init(bdrv_qcow_init
);