2 * Block driver for the Virtual Disk Image (VDI) format
4 * Copyright (c) 2009, 2012 Stefan Weil
6 * This program is free software: you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation, either version 2 of the License, or
9 * (at your option) version 3 or any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program. If not, see <http://www.gnu.org/licenses/>.
20 * http://forums.virtualbox.org/viewtopic.php?t=8046
22 * This driver supports create / read / write operations on VDI images.
24 * Todo (see also TODO in code):
26 * Some features like snapshots are still missing.
28 * Deallocation of zero-filled blocks and shrinking images are missing, too
29 * (might be added to common block layer).
31 * Allocation of blocks could be optimized (less writes to block map and
34 * Read and write of adjacent blocks could be done in one operation
35 * (current code uses one operation per block (1 MiB).
37 * The code is not thread safe (missing locks for changes in header and
38 * block table, no problem with current QEMU).
42 * Blocks (VDI documentation) correspond to clusters (QEMU).
43 * QEMU's backing files could be implemented using VDI snapshot files (TODO).
44 * VDI snapshot files may also contain the complete machine state.
45 * Maybe this machine state can be converted to QEMU PC machine snapshot data.
47 * The driver keeps a block cache (little endian entries) in memory.
48 * For the standard block size (1 MiB), a 1 TiB disk will use 4 MiB RAM,
49 * so this seems to be reasonable.
52 #include "qemu/osdep.h"
53 #include "qapi/error.h"
54 #include "qapi/qmp/qdict.h"
55 #include "qapi/qobject-input-visitor.h"
56 #include "qapi/qapi-visit-block-core.h"
57 #include "block/block_int.h"
58 #include "sysemu/block-backend.h"
59 #include "qemu/module.h"
60 #include "qemu/option.h"
61 #include "qemu/bswap.h"
62 #include "migration/blocker.h"
63 #include "qemu/coroutine.h"
64 #include "qemu/cutils.h"
65 #include "qemu/uuid.h"
67 /* Code configuration options. */
69 /* Enable debug messages. */
70 //~ #define CONFIG_VDI_DEBUG
72 /* Support write operations on VDI images. */
73 #define CONFIG_VDI_WRITE
75 /* Support non-standard block (cluster) size. This is untested.
76 * Maybe it will be needed for very large images.
78 //~ #define CONFIG_VDI_BLOCK_SIZE
80 /* Support static (fixed, pre-allocated) images. */
81 #define CONFIG_VDI_STATIC_IMAGE
83 /* Command line option for static images. */
84 #define BLOCK_OPT_STATIC "static"
87 #define MiB (KiB * KiB)
89 #define SECTOR_SIZE 512
90 #define DEFAULT_CLUSTER_SIZE (1 * MiB)
92 #if defined(CONFIG_VDI_DEBUG)
98 #define logout(fmt, ...) \
101 fprintf(stderr, "vdi\t%-24s" fmt, __func__, ##__VA_ARGS__); \
105 /* Image signature. */
106 #define VDI_SIGNATURE 0xbeda107f
109 #define VDI_VERSION_1_1 0x00010001
112 #define VDI_TYPE_DYNAMIC 1
113 #define VDI_TYPE_STATIC 2
115 /* Innotek / SUN images use these strings in header.text:
116 * "<<< innotek VirtualBox Disk Image >>>\n"
117 * "<<< Sun xVM VirtualBox Disk Image >>>\n"
118 * "<<< Sun VirtualBox Disk Image >>>\n"
119 * The value does not matter, so QEMU created images use a different text.
121 #define VDI_TEXT "<<< QEMU VM Virtual Disk Image >>>\n"
123 /* A never-allocated block; semantically arbitrary content. */
124 #define VDI_UNALLOCATED 0xffffffffU
126 /* A discarded (no longer allocated) block; semantically zero-filled. */
127 #define VDI_DISCARDED 0xfffffffeU
129 #define VDI_IS_ALLOCATED(X) ((X) < VDI_DISCARDED)
131 /* The bmap will take up VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) bytes; since
132 * the bmap is read and written in a single operation, its size needs to be
133 * limited to INT_MAX; furthermore, when opening an image, the bmap size is
134 * rounded up to be aligned on BDRV_SECTOR_SIZE.
135 * Therefore this should satisfy the following:
136 * VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) + BDRV_SECTOR_SIZE == INT_MAX + 1
137 * (INT_MAX + 1 is the first value not representable as an int)
138 * This guarantees that any value below or equal to the constant will, when
139 * multiplied by sizeof(uint32_t) and rounded up to a BDRV_SECTOR_SIZE boundary,
140 * still be below or equal to INT_MAX. */
141 #define VDI_BLOCKS_IN_IMAGE_MAX \
142 ((unsigned)((INT_MAX + 1u - BDRV_SECTOR_SIZE) / sizeof(uint32_t)))
143 #define VDI_DISK_SIZE_MAX ((uint64_t)VDI_BLOCKS_IN_IMAGE_MAX * \
144 (uint64_t)DEFAULT_CLUSTER_SIZE)
146 static QemuOptsList vdi_create_opts
;
152 uint32_t header_size
;
154 uint32_t image_flags
;
155 char description
[256];
156 uint32_t offset_bmap
;
157 uint32_t offset_data
;
158 uint32_t cylinders
; /* disk geometry, unused here */
159 uint32_t heads
; /* disk geometry, unused here */
160 uint32_t sectors
; /* disk geometry, unused here */
161 uint32_t sector_size
;
165 uint32_t block_extra
; /* unused here */
166 uint32_t blocks_in_image
;
167 uint32_t blocks_allocated
;
169 QemuUUID uuid_last_snap
;
171 QemuUUID uuid_parent
;
173 } QEMU_PACKED VdiHeader
;
176 /* The block map entries are little endian (even in memory). */
178 /* Size of block (bytes). */
180 /* First sector of block map. */
181 uint32_t bmap_sector
;
182 /* VDI header (converted to host endianness). */
187 Error
*migration_blocker
;
190 static void vdi_header_to_cpu(VdiHeader
*header
)
192 le32_to_cpus(&header
->signature
);
193 le32_to_cpus(&header
->version
);
194 le32_to_cpus(&header
->header_size
);
195 le32_to_cpus(&header
->image_type
);
196 le32_to_cpus(&header
->image_flags
);
197 le32_to_cpus(&header
->offset_bmap
);
198 le32_to_cpus(&header
->offset_data
);
199 le32_to_cpus(&header
->cylinders
);
200 le32_to_cpus(&header
->heads
);
201 le32_to_cpus(&header
->sectors
);
202 le32_to_cpus(&header
->sector_size
);
203 le64_to_cpus(&header
->disk_size
);
204 le32_to_cpus(&header
->block_size
);
205 le32_to_cpus(&header
->block_extra
);
206 le32_to_cpus(&header
->blocks_in_image
);
207 le32_to_cpus(&header
->blocks_allocated
);
208 qemu_uuid_bswap(&header
->uuid_image
);
209 qemu_uuid_bswap(&header
->uuid_last_snap
);
210 qemu_uuid_bswap(&header
->uuid_link
);
211 qemu_uuid_bswap(&header
->uuid_parent
);
214 static void vdi_header_to_le(VdiHeader
*header
)
216 cpu_to_le32s(&header
->signature
);
217 cpu_to_le32s(&header
->version
);
218 cpu_to_le32s(&header
->header_size
);
219 cpu_to_le32s(&header
->image_type
);
220 cpu_to_le32s(&header
->image_flags
);
221 cpu_to_le32s(&header
->offset_bmap
);
222 cpu_to_le32s(&header
->offset_data
);
223 cpu_to_le32s(&header
->cylinders
);
224 cpu_to_le32s(&header
->heads
);
225 cpu_to_le32s(&header
->sectors
);
226 cpu_to_le32s(&header
->sector_size
);
227 cpu_to_le64s(&header
->disk_size
);
228 cpu_to_le32s(&header
->block_size
);
229 cpu_to_le32s(&header
->block_extra
);
230 cpu_to_le32s(&header
->blocks_in_image
);
231 cpu_to_le32s(&header
->blocks_allocated
);
232 qemu_uuid_bswap(&header
->uuid_image
);
233 qemu_uuid_bswap(&header
->uuid_last_snap
);
234 qemu_uuid_bswap(&header
->uuid_link
);
235 qemu_uuid_bswap(&header
->uuid_parent
);
238 #if defined(CONFIG_VDI_DEBUG)
239 static void vdi_header_print(VdiHeader
*header
)
242 logout("text %s", header
->text
);
243 logout("signature 0x%08x\n", header
->signature
);
244 logout("header size 0x%04x\n", header
->header_size
);
245 logout("image type 0x%04x\n", header
->image_type
);
246 logout("image flags 0x%04x\n", header
->image_flags
);
247 logout("description %s\n", header
->description
);
248 logout("offset bmap 0x%04x\n", header
->offset_bmap
);
249 logout("offset data 0x%04x\n", header
->offset_data
);
250 logout("cylinders 0x%04x\n", header
->cylinders
);
251 logout("heads 0x%04x\n", header
->heads
);
252 logout("sectors 0x%04x\n", header
->sectors
);
253 logout("sector size 0x%04x\n", header
->sector_size
);
254 logout("image size 0x%" PRIx64
" B (%" PRIu64
" MiB)\n",
255 header
->disk_size
, header
->disk_size
/ MiB
);
256 logout("block size 0x%04x\n", header
->block_size
);
257 logout("block extra 0x%04x\n", header
->block_extra
);
258 logout("blocks tot. 0x%04x\n", header
->blocks_in_image
);
259 logout("blocks all. 0x%04x\n", header
->blocks_allocated
);
260 uuid_unparse(header
->uuid_image
, uuid
);
261 logout("uuid image %s\n", uuid
);
262 uuid_unparse(header
->uuid_last_snap
, uuid
);
263 logout("uuid snap %s\n", uuid
);
264 uuid_unparse(header
->uuid_link
, uuid
);
265 logout("uuid link %s\n", uuid
);
266 uuid_unparse(header
->uuid_parent
, uuid
);
267 logout("uuid parent %s\n", uuid
);
271 static int coroutine_fn
vdi_co_check(BlockDriverState
*bs
, BdrvCheckResult
*res
,
274 /* TODO: additional checks possible. */
275 BDRVVdiState
*s
= (BDRVVdiState
*)bs
->opaque
;
276 uint32_t blocks_allocated
= 0;
285 bmap
= g_try_new(uint32_t, s
->header
.blocks_in_image
);
286 if (s
->header
.blocks_in_image
&& bmap
== NULL
) {
291 memset(bmap
, 0xff, s
->header
.blocks_in_image
* sizeof(uint32_t));
293 /* Check block map and value of blocks_allocated. */
294 for (block
= 0; block
< s
->header
.blocks_in_image
; block
++) {
295 uint32_t bmap_entry
= le32_to_cpu(s
->bmap
[block
]);
296 if (VDI_IS_ALLOCATED(bmap_entry
)) {
297 if (bmap_entry
< s
->header
.blocks_in_image
) {
299 if (!VDI_IS_ALLOCATED(bmap
[bmap_entry
])) {
300 bmap
[bmap_entry
] = bmap_entry
;
302 fprintf(stderr
, "ERROR: block index %" PRIu32
303 " also used by %" PRIu32
"\n", bmap
[bmap_entry
], bmap_entry
);
307 fprintf(stderr
, "ERROR: block index %" PRIu32
308 " too large, is %" PRIu32
"\n", block
, bmap_entry
);
313 if (blocks_allocated
!= s
->header
.blocks_allocated
) {
314 fprintf(stderr
, "ERROR: allocated blocks mismatch, is %" PRIu32
315 ", should be %" PRIu32
"\n",
316 blocks_allocated
, s
->header
.blocks_allocated
);
325 static int vdi_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
327 /* TODO: vdi_get_info would be needed for machine snapshots.
328 vm_state_offset is still missing. */
329 BDRVVdiState
*s
= (BDRVVdiState
*)bs
->opaque
;
331 bdi
->cluster_size
= s
->block_size
;
332 bdi
->vm_state_offset
= 0;
333 bdi
->unallocated_blocks_are_zero
= true;
337 static int vdi_make_empty(BlockDriverState
*bs
)
339 /* TODO: missing code. */
341 /* The return value for missing code must be 0, see block.c. */
345 static int vdi_probe(const uint8_t *buf
, int buf_size
, const char *filename
)
347 const VdiHeader
*header
= (const VdiHeader
*)buf
;
352 if (buf_size
< sizeof(*header
)) {
353 /* Header too small, no VDI. */
354 } else if (le32_to_cpu(header
->signature
) == VDI_SIGNATURE
) {
359 logout("no vdi image\n");
361 logout("%s", header
->text
);
367 static int vdi_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
370 BDRVVdiState
*s
= bs
->opaque
;
374 Error
*local_err
= NULL
;
376 bs
->file
= bdrv_open_child(NULL
, options
, "file", bs
, &child_file
,
384 ret
= bdrv_read(bs
->file
, 0, (uint8_t *)&header
, 1);
389 vdi_header_to_cpu(&header
);
390 #if defined(CONFIG_VDI_DEBUG)
391 vdi_header_print(&header
);
394 if (header
.disk_size
> VDI_DISK_SIZE_MAX
) {
395 error_setg(errp
, "Unsupported VDI image size (size is 0x%" PRIx64
396 ", max supported is 0x%" PRIx64
")",
397 header
.disk_size
, VDI_DISK_SIZE_MAX
);
402 if (header
.disk_size
% SECTOR_SIZE
!= 0) {
403 /* 'VBoxManage convertfromraw' can create images with odd disk sizes.
404 We accept them but round the disk size to the next multiple of
406 logout("odd disk size %" PRIu64
" B, round up\n", header
.disk_size
);
407 header
.disk_size
= ROUND_UP(header
.disk_size
, SECTOR_SIZE
);
410 if (header
.signature
!= VDI_SIGNATURE
) {
411 error_setg(errp
, "Image not in VDI format (bad signature %08" PRIx32
412 ")", header
.signature
);
415 } else if (header
.version
!= VDI_VERSION_1_1
) {
416 error_setg(errp
, "unsupported VDI image (version %" PRIu32
".%" PRIu32
417 ")", header
.version
>> 16, header
.version
& 0xffff);
420 } else if (header
.offset_bmap
% SECTOR_SIZE
!= 0) {
421 /* We only support block maps which start on a sector boundary. */
422 error_setg(errp
, "unsupported VDI image (unaligned block map offset "
423 "0x%" PRIx32
")", header
.offset_bmap
);
426 } else if (header
.offset_data
% SECTOR_SIZE
!= 0) {
427 /* We only support data blocks which start on a sector boundary. */
428 error_setg(errp
, "unsupported VDI image (unaligned data offset 0x%"
429 PRIx32
")", header
.offset_data
);
432 } else if (header
.sector_size
!= SECTOR_SIZE
) {
433 error_setg(errp
, "unsupported VDI image (sector size %" PRIu32
434 " is not %u)", header
.sector_size
, SECTOR_SIZE
);
437 } else if (header
.block_size
!= DEFAULT_CLUSTER_SIZE
) {
438 error_setg(errp
, "unsupported VDI image (block size %" PRIu32
439 " is not %u)", header
.block_size
, DEFAULT_CLUSTER_SIZE
);
442 } else if (header
.disk_size
>
443 (uint64_t)header
.blocks_in_image
* header
.block_size
) {
444 error_setg(errp
, "unsupported VDI image (disk size %" PRIu64
", "
445 "image bitmap has room for %" PRIu64
")",
447 (uint64_t)header
.blocks_in_image
* header
.block_size
);
450 } else if (!qemu_uuid_is_null(&header
.uuid_link
)) {
451 error_setg(errp
, "unsupported VDI image (non-NULL link UUID)");
454 } else if (!qemu_uuid_is_null(&header
.uuid_parent
)) {
455 error_setg(errp
, "unsupported VDI image (non-NULL parent UUID)");
458 } else if (header
.blocks_in_image
> VDI_BLOCKS_IN_IMAGE_MAX
) {
459 error_setg(errp
, "unsupported VDI image "
460 "(too many blocks %u, max is %u)",
461 header
.blocks_in_image
, VDI_BLOCKS_IN_IMAGE_MAX
);
466 bs
->total_sectors
= header
.disk_size
/ SECTOR_SIZE
;
468 s
->block_size
= header
.block_size
;
469 s
->bmap_sector
= header
.offset_bmap
/ SECTOR_SIZE
;
472 bmap_size
= header
.blocks_in_image
* sizeof(uint32_t);
473 bmap_size
= DIV_ROUND_UP(bmap_size
, SECTOR_SIZE
);
474 s
->bmap
= qemu_try_blockalign(bs
->file
->bs
, bmap_size
* SECTOR_SIZE
);
475 if (s
->bmap
== NULL
) {
480 ret
= bdrv_read(bs
->file
, s
->bmap_sector
, (uint8_t *)s
->bmap
,
486 /* Disable migration when vdi images are used */
487 error_setg(&s
->migration_blocker
, "The vdi format used by node '%s' "
488 "does not support live migration",
489 bdrv_get_device_or_node_name(bs
));
490 ret
= migrate_add_blocker(s
->migration_blocker
, &local_err
);
492 error_propagate(errp
, local_err
);
493 error_free(s
->migration_blocker
);
497 qemu_co_rwlock_init(&s
->bmap_lock
);
508 static int vdi_reopen_prepare(BDRVReopenState
*state
,
509 BlockReopenQueue
*queue
, Error
**errp
)
514 static int coroutine_fn
vdi_co_block_status(BlockDriverState
*bs
,
516 int64_t offset
, int64_t bytes
,
517 int64_t *pnum
, int64_t *map
,
518 BlockDriverState
**file
)
520 BDRVVdiState
*s
= (BDRVVdiState
*)bs
->opaque
;
521 size_t bmap_index
= offset
/ s
->block_size
;
522 size_t index_in_block
= offset
% s
->block_size
;
523 uint32_t bmap_entry
= le32_to_cpu(s
->bmap
[bmap_index
]);
526 logout("%p, %" PRId64
", %" PRId64
", %p\n", bs
, offset
, bytes
, pnum
);
527 *pnum
= MIN(s
->block_size
- index_in_block
, bytes
);
528 result
= VDI_IS_ALLOCATED(bmap_entry
);
533 *map
= s
->header
.offset_data
+ (uint64_t)bmap_entry
* s
->block_size
+
535 *file
= bs
->file
->bs
;
536 return BDRV_BLOCK_DATA
| BDRV_BLOCK_OFFSET_VALID
;
539 static int coroutine_fn
540 vdi_co_preadv(BlockDriverState
*bs
, uint64_t offset
, uint64_t bytes
,
541 QEMUIOVector
*qiov
, int flags
)
543 BDRVVdiState
*s
= bs
->opaque
;
544 QEMUIOVector local_qiov
;
546 uint32_t block_index
;
547 uint32_t offset_in_block
;
549 uint64_t bytes_done
= 0;
554 qemu_iovec_init(&local_qiov
, qiov
->niov
);
556 while (ret
>= 0 && bytes
> 0) {
557 block_index
= offset
/ s
->block_size
;
558 offset_in_block
= offset
% s
->block_size
;
559 n_bytes
= MIN(bytes
, s
->block_size
- offset_in_block
);
561 logout("will read %u bytes starting at offset %" PRIu64
"\n",
564 /* prepare next AIO request */
565 qemu_co_rwlock_rdlock(&s
->bmap_lock
);
566 bmap_entry
= le32_to_cpu(s
->bmap
[block_index
]);
567 qemu_co_rwlock_unlock(&s
->bmap_lock
);
568 if (!VDI_IS_ALLOCATED(bmap_entry
)) {
569 /* Block not allocated, return zeros, no need to wait. */
570 qemu_iovec_memset(qiov
, bytes_done
, 0, n_bytes
);
573 uint64_t data_offset
= s
->header
.offset_data
+
574 (uint64_t)bmap_entry
* s
->block_size
+
577 qemu_iovec_reset(&local_qiov
);
578 qemu_iovec_concat(&local_qiov
, qiov
, bytes_done
, n_bytes
);
580 ret
= bdrv_co_preadv(bs
->file
, data_offset
, n_bytes
,
583 logout("%u bytes read\n", n_bytes
);
587 bytes_done
+= n_bytes
;
590 qemu_iovec_destroy(&local_qiov
);
595 static int coroutine_fn
596 vdi_co_pwritev(BlockDriverState
*bs
, uint64_t offset
, uint64_t bytes
,
597 QEMUIOVector
*qiov
, int flags
)
599 BDRVVdiState
*s
= bs
->opaque
;
600 QEMUIOVector local_qiov
;
602 uint32_t block_index
;
603 uint32_t offset_in_block
;
605 uint64_t data_offset
;
606 uint32_t bmap_first
= VDI_UNALLOCATED
;
607 uint32_t bmap_last
= VDI_UNALLOCATED
;
608 uint8_t *block
= NULL
;
609 uint64_t bytes_done
= 0;
614 qemu_iovec_init(&local_qiov
, qiov
->niov
);
616 while (ret
>= 0 && bytes
> 0) {
617 block_index
= offset
/ s
->block_size
;
618 offset_in_block
= offset
% s
->block_size
;
619 n_bytes
= MIN(bytes
, s
->block_size
- offset_in_block
);
621 logout("will write %u bytes starting at offset %" PRIu64
"\n",
624 /* prepare next AIO request */
625 qemu_co_rwlock_rdlock(&s
->bmap_lock
);
626 bmap_entry
= le32_to_cpu(s
->bmap
[block_index
]);
627 if (!VDI_IS_ALLOCATED(bmap_entry
)) {
628 /* Allocate new block and write to it. */
629 uint64_t data_offset
;
630 qemu_co_rwlock_upgrade(&s
->bmap_lock
);
631 bmap_entry
= le32_to_cpu(s
->bmap
[block_index
]);
632 if (VDI_IS_ALLOCATED(bmap_entry
)) {
633 /* A concurrent allocation did the work for us. */
634 qemu_co_rwlock_downgrade(&s
->bmap_lock
);
635 goto nonallocating_write
;
638 bmap_entry
= s
->header
.blocks_allocated
;
639 s
->bmap
[block_index
] = cpu_to_le32(bmap_entry
);
640 s
->header
.blocks_allocated
++;
641 data_offset
= s
->header
.offset_data
+
642 (uint64_t)bmap_entry
* s
->block_size
;
644 block
= g_malloc(s
->block_size
);
645 bmap_first
= block_index
;
647 bmap_last
= block_index
;
648 /* Copy data to be written to new block and zero unused parts. */
649 memset(block
, 0, offset_in_block
);
650 qemu_iovec_to_buf(qiov
, bytes_done
, block
+ offset_in_block
,
652 memset(block
+ offset_in_block
+ n_bytes
, 0,
653 s
->block_size
- n_bytes
- offset_in_block
);
655 /* Write the new block under CoRwLock write-side protection,
656 * so this full-cluster write does not overlap a partial write
657 * of the same cluster, issued from the "else" branch.
659 ret
= bdrv_pwrite(bs
->file
, data_offset
, block
, s
->block_size
);
660 qemu_co_rwlock_unlock(&s
->bmap_lock
);
663 data_offset
= s
->header
.offset_data
+
664 (uint64_t)bmap_entry
* s
->block_size
+
666 qemu_co_rwlock_unlock(&s
->bmap_lock
);
668 qemu_iovec_reset(&local_qiov
);
669 qemu_iovec_concat(&local_qiov
, qiov
, bytes_done
, n_bytes
);
671 ret
= bdrv_co_pwritev(bs
->file
, data_offset
, n_bytes
,
677 bytes_done
+= n_bytes
;
679 logout("%u bytes written\n", n_bytes
);
682 qemu_iovec_destroy(&local_qiov
);
684 logout("finished data write\n");
690 /* One or more new blocks were allocated. */
691 VdiHeader
*header
= (VdiHeader
*) block
;
696 logout("now writing modified header\n");
697 assert(VDI_IS_ALLOCATED(bmap_first
));
699 vdi_header_to_le(header
);
700 ret
= bdrv_write(bs
->file
, 0, block
, 1);
708 logout("now writing modified block map entry %u...%u\n",
709 bmap_first
, bmap_last
);
710 /* Write modified sectors from block map. */
711 bmap_first
/= (SECTOR_SIZE
/ sizeof(uint32_t));
712 bmap_last
/= (SECTOR_SIZE
/ sizeof(uint32_t));
713 n_sectors
= bmap_last
- bmap_first
+ 1;
714 offset
= s
->bmap_sector
+ bmap_first
;
715 base
= ((uint8_t *)&s
->bmap
[0]) + bmap_first
* SECTOR_SIZE
;
716 logout("will write %u block map sectors starting from entry %u\n",
717 n_sectors
, bmap_first
);
718 ret
= bdrv_write(bs
->file
, offset
, base
, n_sectors
);
724 static int coroutine_fn
vdi_co_do_create(BlockdevCreateOptions
*create_options
,
725 size_t block_size
, Error
**errp
)
727 BlockdevCreateOptionsVdi
*vdi_opts
;
731 uint32_t image_type
= VDI_TYPE_DYNAMIC
;
736 BlockDriverState
*bs_file
= NULL
;
737 BlockBackend
*blk
= NULL
;
738 uint32_t *bmap
= NULL
;
740 assert(create_options
->driver
== BLOCKDEV_DRIVER_VDI
);
741 vdi_opts
= &create_options
->u
.vdi
;
745 /* Read out options. */
746 bytes
= vdi_opts
->size
;
747 if (vdi_opts
->q_static
) {
748 image_type
= VDI_TYPE_STATIC
;
750 #ifndef CONFIG_VDI_STATIC_IMAGE
751 if (image_type
== VDI_TYPE_STATIC
) {
753 error_setg(errp
, "Statically allocated images cannot be created in "
758 #ifndef CONFIG_VDI_BLOCK_SIZE
759 if (block_size
!= DEFAULT_CLUSTER_SIZE
) {
762 "A non-default cluster size is not supported in this build");
767 if (bytes
> VDI_DISK_SIZE_MAX
) {
769 error_setg(errp
, "Unsupported VDI image size (size is 0x%" PRIx64
770 ", max supported is 0x%" PRIx64
")",
771 bytes
, VDI_DISK_SIZE_MAX
);
775 bs_file
= bdrv_open_blockdev_ref(vdi_opts
->file
, errp
);
781 blk
= blk_new(BLK_PERM_WRITE
| BLK_PERM_RESIZE
, BLK_PERM_ALL
);
782 ret
= blk_insert_bs(blk
, bs_file
, errp
);
787 blk_set_allow_write_beyond_eof(blk
, true);
789 /* We need enough blocks to store the given disk size,
790 so always round up. */
791 blocks
= DIV_ROUND_UP(bytes
, block_size
);
793 bmap_size
= blocks
* sizeof(uint32_t);
794 bmap_size
= ROUND_UP(bmap_size
, SECTOR_SIZE
);
796 memset(&header
, 0, sizeof(header
));
797 pstrcpy(header
.text
, sizeof(header
.text
), VDI_TEXT
);
798 header
.signature
= VDI_SIGNATURE
;
799 header
.version
= VDI_VERSION_1_1
;
800 header
.header_size
= 0x180;
801 header
.image_type
= image_type
;
802 header
.offset_bmap
= 0x200;
803 header
.offset_data
= 0x200 + bmap_size
;
804 header
.sector_size
= SECTOR_SIZE
;
805 header
.disk_size
= bytes
;
806 header
.block_size
= block_size
;
807 header
.blocks_in_image
= blocks
;
808 if (image_type
== VDI_TYPE_STATIC
) {
809 header
.blocks_allocated
= blocks
;
811 qemu_uuid_generate(&header
.uuid_image
);
812 qemu_uuid_generate(&header
.uuid_last_snap
);
813 /* There is no need to set header.uuid_link or header.uuid_parent here. */
814 #if defined(CONFIG_VDI_DEBUG)
815 vdi_header_print(&header
);
817 vdi_header_to_le(&header
);
818 ret
= blk_pwrite(blk
, offset
, &header
, sizeof(header
), 0);
820 error_setg(errp
, "Error writing header");
823 offset
+= sizeof(header
);
826 bmap
= g_try_malloc0(bmap_size
);
829 error_setg(errp
, "Could not allocate bmap");
832 for (i
= 0; i
< blocks
; i
++) {
833 if (image_type
== VDI_TYPE_STATIC
) {
836 bmap
[i
] = VDI_UNALLOCATED
;
839 ret
= blk_pwrite(blk
, offset
, bmap
, bmap_size
, 0);
841 error_setg(errp
, "Error writing bmap");
847 if (image_type
== VDI_TYPE_STATIC
) {
848 ret
= blk_truncate(blk
, offset
+ blocks
* block_size
,
849 PREALLOC_MODE_OFF
, errp
);
851 error_prepend(errp
, "Failed to statically allocate file");
863 static int coroutine_fn
vdi_co_create(BlockdevCreateOptions
*create_options
,
866 return vdi_co_do_create(create_options
, DEFAULT_CLUSTER_SIZE
, errp
);
869 static int coroutine_fn
vdi_co_create_opts(const char *filename
, QemuOpts
*opts
,
873 BlockdevCreateOptions
*create_options
= NULL
;
874 BlockDriverState
*bs_file
= NULL
;
875 uint64_t block_size
= DEFAULT_CLUSTER_SIZE
;
877 Error
*local_err
= NULL
;
880 /* Since CONFIG_VDI_BLOCK_SIZE is disabled by default,
881 * cluster-size is not part of the QAPI schema; therefore we have
882 * to parse it before creating the QAPI object. */
883 #if defined(CONFIG_VDI_BLOCK_SIZE)
884 block_size
= qemu_opt_get_size_del(opts
,
885 BLOCK_OPT_CLUSTER_SIZE
,
886 DEFAULT_CLUSTER_SIZE
);
887 if (block_size
< BDRV_SECTOR_SIZE
|| block_size
> UINT32_MAX
||
888 !is_power_of_2(block_size
))
890 error_setg(errp
, "Invalid cluster size");
896 qdict
= qemu_opts_to_qdict_filtered(opts
, NULL
, &vdi_create_opts
, true);
898 ret
= bdrv_create_file(filename
, opts
, errp
);
903 bs_file
= bdrv_open(filename
, NULL
, NULL
,
904 BDRV_O_RDWR
| BDRV_O_RESIZE
| BDRV_O_PROTOCOL
, errp
);
910 qdict_put_str(qdict
, "driver", "vdi");
911 qdict_put_str(qdict
, "file", bs_file
->node_name
);
913 /* Get the QAPI object */
914 v
= qobject_input_visitor_new_keyval(QOBJECT(qdict
));
915 visit_type_BlockdevCreateOptions(v
, NULL
, &create_options
, &local_err
);
919 error_propagate(errp
, local_err
);
924 assert(create_options
->driver
== BLOCKDEV_DRIVER_VDI
);
925 create_options
->u
.vdi
.size
= ROUND_UP(create_options
->u
.vdi
.size
,
928 ret
= vdi_co_do_create(create_options
, block_size
, errp
);
931 qapi_free_BlockdevCreateOptions(create_options
);
936 static void vdi_close(BlockDriverState
*bs
)
938 BDRVVdiState
*s
= bs
->opaque
;
942 migrate_del_blocker(s
->migration_blocker
);
943 error_free(s
->migration_blocker
);
946 static QemuOptsList vdi_create_opts
= {
947 .name
= "vdi-create-opts",
948 .head
= QTAILQ_HEAD_INITIALIZER(vdi_create_opts
.head
),
951 .name
= BLOCK_OPT_SIZE
,
952 .type
= QEMU_OPT_SIZE
,
953 .help
= "Virtual disk size"
955 #if defined(CONFIG_VDI_BLOCK_SIZE)
957 .name
= BLOCK_OPT_CLUSTER_SIZE
,
958 .type
= QEMU_OPT_SIZE
,
959 .help
= "VDI cluster (block) size",
960 .def_value_str
= stringify(DEFAULT_CLUSTER_SIZE
)
963 #if defined(CONFIG_VDI_STATIC_IMAGE)
965 .name
= BLOCK_OPT_STATIC
,
966 .type
= QEMU_OPT_BOOL
,
967 .help
= "VDI static (pre-allocated) image",
968 .def_value_str
= "off"
971 /* TODO: An additional option to set UUID values might be useful. */
972 { /* end of list */ }
976 static BlockDriver bdrv_vdi
= {
977 .format_name
= "vdi",
978 .instance_size
= sizeof(BDRVVdiState
),
979 .bdrv_probe
= vdi_probe
,
980 .bdrv_open
= vdi_open
,
981 .bdrv_close
= vdi_close
,
982 .bdrv_reopen_prepare
= vdi_reopen_prepare
,
983 .bdrv_child_perm
= bdrv_format_default_perms
,
984 .bdrv_co_create_opts
= vdi_co_create_opts
,
985 .bdrv_co_create
= vdi_co_create
,
986 .bdrv_has_zero_init
= bdrv_has_zero_init_1
,
987 .bdrv_co_block_status
= vdi_co_block_status
,
988 .bdrv_make_empty
= vdi_make_empty
,
990 .bdrv_co_preadv
= vdi_co_preadv
,
991 #if defined(CONFIG_VDI_WRITE)
992 .bdrv_co_pwritev
= vdi_co_pwritev
,
995 .bdrv_get_info
= vdi_get_info
,
997 .create_opts
= &vdi_create_opts
,
998 .bdrv_co_check
= vdi_co_check
,
1001 static void bdrv_vdi_init(void)
1004 bdrv_register(&bdrv_vdi
);
1007 block_init(bdrv_vdi_init
);