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-common.h"
53 #include "block/block_int.h"
54 #include "qemu/module.h"
55 #include "migration/migration.h"
56 #include "block/coroutine.h"
58 #if defined(CONFIG_UUID)
59 #include <uuid/uuid.h>
61 /* TODO: move uuid emulation to some central place in QEMU. */
62 #include "sysemu/sysemu.h" /* UUID_FMT */
63 typedef unsigned char uuid_t[16];
66 /* Code configuration options. */
68 /* Enable debug messages. */
69 //~ #define CONFIG_VDI_DEBUG
71 /* Support write operations on VDI images. */
72 #define CONFIG_VDI_WRITE
74 /* Support non-standard block (cluster) size. This is untested.
75 * Maybe it will be needed for very large images.
77 //~ #define CONFIG_VDI_BLOCK_SIZE
79 /* Support static (fixed, pre-allocated) images. */
80 #define CONFIG_VDI_STATIC_IMAGE
82 /* Command line option for static images. */
83 #define BLOCK_OPT_STATIC "static"
86 #define MiB (KiB * KiB)
88 #define SECTOR_SIZE 512
89 #define DEFAULT_CLUSTER_SIZE (1 * MiB)
91 #if defined(CONFIG_VDI_DEBUG)
92 #define logout(fmt, ...) \
93 fprintf(stderr, "vdi\t%-24s" fmt, __func__, ##__VA_ARGS__)
95 #define logout(fmt, ...) ((void)0)
98 /* Image signature. */
99 #define VDI_SIGNATURE 0xbeda107f
102 #define VDI_VERSION_1_1 0x00010001
105 #define VDI_TYPE_DYNAMIC 1
106 #define VDI_TYPE_STATIC 2
108 /* Innotek / SUN images use these strings in header.text:
109 * "<<< innotek VirtualBox Disk Image >>>\n"
110 * "<<< Sun xVM VirtualBox Disk Image >>>\n"
111 * "<<< Sun VirtualBox Disk Image >>>\n"
112 * The value does not matter, so QEMU created images use a different text.
114 #define VDI_TEXT "<<< QEMU VM Virtual Disk Image >>>\n"
116 /* A never-allocated block; semantically arbitrary content. */
117 #define VDI_UNALLOCATED 0xffffffffU
119 /* A discarded (no longer allocated) block; semantically zero-filled. */
120 #define VDI_DISCARDED 0xfffffffeU
122 #define VDI_IS_ALLOCATED(X) ((X) < VDI_DISCARDED)
124 /* The bmap will take up VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) bytes; since
125 * the bmap is read and written in a single operation, its size needs to be
126 * limited to INT_MAX; furthermore, when opening an image, the bmap size is
127 * rounded up to be aligned on BDRV_SECTOR_SIZE.
128 * Therefore this should satisfy the following:
129 * VDI_BLOCKS_IN_IMAGE_MAX * sizeof(uint32_t) + BDRV_SECTOR_SIZE == INT_MAX + 1
130 * (INT_MAX + 1 is the first value not representable as an int)
131 * This guarantees that any value below or equal to the constant will, when
132 * multiplied by sizeof(uint32_t) and rounded up to a BDRV_SECTOR_SIZE boundary,
133 * still be below or equal to INT_MAX. */
134 #define VDI_BLOCKS_IN_IMAGE_MAX \
135 ((unsigned)((INT_MAX + 1u - BDRV_SECTOR_SIZE) / sizeof(uint32_t)))
136 #define VDI_DISK_SIZE_MAX ((uint64_t)VDI_BLOCKS_IN_IMAGE_MAX * \
137 (uint64_t)DEFAULT_CLUSTER_SIZE)
139 #if !defined(CONFIG_UUID)
140 static inline void uuid_generate(uuid_t out)
142 memset(out, 0, sizeof(uuid_t));
145 static inline int uuid_is_null(const uuid_t uu)
147 uuid_t null_uuid = { 0 };
148 return memcmp(uu, null_uuid, sizeof(uuid_t)) == 0;
151 # if defined(CONFIG_VDI_DEBUG)
152 static inline void uuid_unparse(const uuid_t uu, char *out)
154 snprintf(out, 37, UUID_FMT,
155 uu[0], uu[1], uu[2], uu[3], uu[4], uu[5], uu[6], uu[7],
156 uu[8], uu[9], uu[10], uu[11], uu[12], uu[13], uu[14], uu[15]);
165 uint32_t header_size;
167 uint32_t image_flags;
168 char description[256];
169 uint32_t offset_bmap;
170 uint32_t offset_data;
171 uint32_t cylinders; /* disk geometry, unused here */
172 uint32_t heads; /* disk geometry, unused here */
173 uint32_t sectors; /* disk geometry, unused here */
174 uint32_t sector_size;
178 uint32_t block_extra; /* unused here */
179 uint32_t blocks_in_image;
180 uint32_t blocks_allocated;
182 uuid_t uuid_last_snap;
186 } QEMU_PACKED VdiHeader;
189 /* The block map entries are little endian (even in memory). */
191 /* Size of block (bytes). */
193 /* Size of block (sectors). */
194 uint32_t block_sectors;
195 /* First sector of block map. */
196 uint32_t bmap_sector;
197 /* VDI header (converted to host endianness). */
202 Error *migration_blocker;
205 /* Change UUID from little endian (IPRT = VirtualBox format) to big endian
206 * format (network byte order, standard, see RFC 4122) and vice versa.
208 static void uuid_convert(uuid_t uuid)
210 bswap32s((uint32_t *)&uuid[0]);
211 bswap16s((uint16_t *)&uuid[4]);
212 bswap16s((uint16_t *)&uuid[6]);
215 static void vdi_header_to_cpu(VdiHeader *header)
217 le32_to_cpus(&header->signature);
218 le32_to_cpus(&header->version);
219 le32_to_cpus(&header->header_size);
220 le32_to_cpus(&header->image_type);
221 le32_to_cpus(&header->image_flags);
222 le32_to_cpus(&header->offset_bmap);
223 le32_to_cpus(&header->offset_data);
224 le32_to_cpus(&header->cylinders);
225 le32_to_cpus(&header->heads);
226 le32_to_cpus(&header->sectors);
227 le32_to_cpus(&header->sector_size);
228 le64_to_cpus(&header->disk_size);
229 le32_to_cpus(&header->block_size);
230 le32_to_cpus(&header->block_extra);
231 le32_to_cpus(&header->blocks_in_image);
232 le32_to_cpus(&header->blocks_allocated);
233 uuid_convert(header->uuid_image);
234 uuid_convert(header->uuid_last_snap);
235 uuid_convert(header->uuid_link);
236 uuid_convert(header->uuid_parent);
239 static void vdi_header_to_le(VdiHeader *header)
241 cpu_to_le32s(&header->signature);
242 cpu_to_le32s(&header->version);
243 cpu_to_le32s(&header->header_size);
244 cpu_to_le32s(&header->image_type);
245 cpu_to_le32s(&header->image_flags);
246 cpu_to_le32s(&header->offset_bmap);
247 cpu_to_le32s(&header->offset_data);
248 cpu_to_le32s(&header->cylinders);
249 cpu_to_le32s(&header->heads);
250 cpu_to_le32s(&header->sectors);
251 cpu_to_le32s(&header->sector_size);
252 cpu_to_le64s(&header->disk_size);
253 cpu_to_le32s(&header->block_size);
254 cpu_to_le32s(&header->block_extra);
255 cpu_to_le32s(&header->blocks_in_image);
256 cpu_to_le32s(&header->blocks_allocated);
257 uuid_convert(header->uuid_image);
258 uuid_convert(header->uuid_last_snap);
259 uuid_convert(header->uuid_link);
260 uuid_convert(header->uuid_parent);
263 #if defined(CONFIG_VDI_DEBUG)
264 static void vdi_header_print(VdiHeader *header)
267 logout("text %s", header->text);
268 logout("signature 0x%08x\n", header->signature);
269 logout("header size 0x%04x\n", header->header_size);
270 logout("image type 0x%04x\n", header->image_type);
271 logout("image flags 0x%04x\n", header->image_flags);
272 logout("description %s\n", header->description);
273 logout("offset bmap 0x%04x\n", header->offset_bmap);
274 logout("offset data 0x%04x\n", header->offset_data);
275 logout("cylinders 0x%04x\n", header->cylinders);
276 logout("heads 0x%04x\n", header->heads);
277 logout("sectors 0x%04x\n", header->sectors);
278 logout("sector size 0x%04x\n", header->sector_size);
279 logout("image size 0x%" PRIx64 " B (%" PRIu64 " MiB)\n",
280 header->disk_size, header->disk_size / MiB);
281 logout("block size 0x%04x\n", header->block_size);
282 logout("block extra 0x%04x\n", header->block_extra);
283 logout("blocks tot. 0x%04x\n", header->blocks_in_image);
284 logout("blocks all. 0x%04x\n", header->blocks_allocated);
285 uuid_unparse(header->uuid_image, uuid);
286 logout("uuid image %s\n", uuid);
287 uuid_unparse(header->uuid_last_snap, uuid);
288 logout("uuid snap %s\n", uuid);
289 uuid_unparse(header->uuid_link, uuid);
290 logout("uuid link %s\n", uuid);
291 uuid_unparse(header->uuid_parent, uuid);
292 logout("uuid parent %s\n", uuid);
296 static int vdi_check(BlockDriverState *bs, BdrvCheckResult *res,
299 /* TODO: additional checks possible. */
300 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
301 uint32_t blocks_allocated = 0;
310 bmap = g_try_new(uint32_t, s->header.blocks_in_image);
311 if (s->header.blocks_in_image && bmap == NULL) {
316 memset(bmap, 0xff, s->header.blocks_in_image * sizeof(uint32_t));
318 /* Check block map and value of blocks_allocated. */
319 for (block = 0; block < s->header.blocks_in_image; block++) {
320 uint32_t bmap_entry = le32_to_cpu(s->bmap[block]);
321 if (VDI_IS_ALLOCATED(bmap_entry)) {
322 if (bmap_entry < s->header.blocks_in_image) {
324 if (!VDI_IS_ALLOCATED(bmap[bmap_entry])) {
325 bmap[bmap_entry] = bmap_entry;
327 fprintf(stderr, "ERROR: block index %" PRIu32
328 " also used by %" PRIu32 "\n", bmap[bmap_entry], bmap_entry);
332 fprintf(stderr, "ERROR: block index %" PRIu32
333 " too large, is %" PRIu32 "\n", block, bmap_entry);
338 if (blocks_allocated != s->header.blocks_allocated) {
339 fprintf(stderr, "ERROR: allocated blocks mismatch, is %" PRIu32
340 ", should be %" PRIu32 "\n",
341 blocks_allocated, s->header.blocks_allocated);
350 static int vdi_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
352 /* TODO: vdi_get_info would be needed for machine snapshots.
353 vm_state_offset is still missing. */
354 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
356 bdi->cluster_size = s->block_size;
357 bdi->vm_state_offset = 0;
358 bdi->unallocated_blocks_are_zero = true;
362 static int vdi_make_empty(BlockDriverState *bs)
364 /* TODO: missing code. */
366 /* The return value for missing code must be 0, see block.c. */
370 static int vdi_probe(const uint8_t *buf, int buf_size, const char *filename)
372 const VdiHeader *header = (const VdiHeader *)buf;
377 if (buf_size < sizeof(*header)) {
378 /* Header too small, no VDI. */
379 } else if (le32_to_cpu(header->signature) == VDI_SIGNATURE) {
384 logout("no vdi image\n");
386 logout("%s", header->text);
392 static int vdi_open(BlockDriverState *bs, QDict *options, int flags,
395 BDRVVdiState *s = bs->opaque;
402 ret = bdrv_read(bs->file, 0, (uint8_t *)&header, 1);
407 vdi_header_to_cpu(&header);
408 #if defined(CONFIG_VDI_DEBUG)
409 vdi_header_print(&header);
412 if (header.disk_size > VDI_DISK_SIZE_MAX) {
413 error_setg(errp, "Unsupported VDI image size (size is 0x%" PRIx64
414 ", max supported is 0x%" PRIx64 ")",
415 header.disk_size, VDI_DISK_SIZE_MAX);
420 if (header.disk_size % SECTOR_SIZE != 0) {
421 /* 'VBoxManage convertfromraw' can create images with odd disk sizes.
422 We accept them but round the disk size to the next multiple of
424 logout("odd disk size %" PRIu64 " B, round up\n", header.disk_size);
425 header.disk_size = ROUND_UP(header.disk_size, SECTOR_SIZE);
428 if (header.signature != VDI_SIGNATURE) {
429 error_setg(errp, "Image not in VDI format (bad signature %08" PRIx32
430 ")", header.signature);
433 } else if (header.version != VDI_VERSION_1_1) {
434 error_setg(errp, "unsupported VDI image (version %" PRIu32 ".%" PRIu32
435 ")", header.version >> 16, header.version & 0xffff);
438 } else if (header.offset_bmap % SECTOR_SIZE != 0) {
439 /* We only support block maps which start on a sector boundary. */
440 error_setg(errp, "unsupported VDI image (unaligned block map offset "
441 "0x%" PRIx32 ")", header.offset_bmap);
444 } else if (header.offset_data % SECTOR_SIZE != 0) {
445 /* We only support data blocks which start on a sector boundary. */
446 error_setg(errp, "unsupported VDI image (unaligned data offset 0x%"
447 PRIx32 ")", header.offset_data);
450 } else if (header.sector_size != SECTOR_SIZE) {
451 error_setg(errp, "unsupported VDI image (sector size %" PRIu32
452 " is not %u)", header.sector_size, SECTOR_SIZE);
455 } else if (header.block_size != DEFAULT_CLUSTER_SIZE) {
456 error_setg(errp, "unsupported VDI image (block size %" PRIu32
457 " is not %u)", header.block_size, DEFAULT_CLUSTER_SIZE);
460 } else if (header.disk_size >
461 (uint64_t)header.blocks_in_image * header.block_size) {
462 error_setg(errp, "unsupported VDI image (disk size %" PRIu64 ", "
463 "image bitmap has room for %" PRIu64 ")",
465 (uint64_t)header.blocks_in_image * header.block_size);
468 } else if (!uuid_is_null(header.uuid_link)) {
469 error_setg(errp, "unsupported VDI image (non-NULL link UUID)");
472 } else if (!uuid_is_null(header.uuid_parent)) {
473 error_setg(errp, "unsupported VDI image (non-NULL parent UUID)");
476 } else if (header.blocks_in_image > VDI_BLOCKS_IN_IMAGE_MAX) {
477 error_setg(errp, "unsupported VDI image "
478 "(too many blocks %u, max is %u)",
479 header.blocks_in_image, VDI_BLOCKS_IN_IMAGE_MAX);
484 bs->total_sectors = header.disk_size / SECTOR_SIZE;
486 s->block_size = header.block_size;
487 s->block_sectors = header.block_size / SECTOR_SIZE;
488 s->bmap_sector = header.offset_bmap / SECTOR_SIZE;
491 bmap_size = header.blocks_in_image * sizeof(uint32_t);
492 bmap_size = DIV_ROUND_UP(bmap_size, SECTOR_SIZE);
493 s->bmap = qemu_try_blockalign(bs->file, bmap_size * SECTOR_SIZE);
494 if (s->bmap == NULL) {
499 ret = bdrv_read(bs->file, s->bmap_sector, (uint8_t *)s->bmap, bmap_size);
504 /* Disable migration when vdi images are used */
505 error_setg(&s->migration_blocker, "The vdi format used by node '%s' "
506 "does not support live migration",
507 bdrv_get_device_or_node_name(bs));
508 migrate_add_blocker(s->migration_blocker);
510 qemu_co_mutex_init(&s->write_lock);
521 static int vdi_reopen_prepare(BDRVReopenState *state,
522 BlockReopenQueue *queue, Error **errp)
527 static int64_t coroutine_fn vdi_co_get_block_status(BlockDriverState *bs,
528 int64_t sector_num, int nb_sectors, int *pnum)
530 /* TODO: Check for too large sector_num (in bdrv_is_allocated or here). */
531 BDRVVdiState *s = (BDRVVdiState *)bs->opaque;
532 size_t bmap_index = sector_num / s->block_sectors;
533 size_t sector_in_block = sector_num % s->block_sectors;
534 int n_sectors = s->block_sectors - sector_in_block;
535 uint32_t bmap_entry = le32_to_cpu(s->bmap[bmap_index]);
539 logout("%p, %" PRId64 ", %d, %p\n", bs, sector_num, nb_sectors, pnum);
540 if (n_sectors > nb_sectors) {
541 n_sectors = nb_sectors;
544 result = VDI_IS_ALLOCATED(bmap_entry);
549 offset = s->header.offset_data +
550 (uint64_t)bmap_entry * s->block_size +
551 sector_in_block * SECTOR_SIZE;
552 return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | offset;
555 static int vdi_co_read(BlockDriverState *bs,
556 int64_t sector_num, uint8_t *buf, int nb_sectors)
558 BDRVVdiState *s = bs->opaque;
560 uint32_t block_index;
561 uint32_t sector_in_block;
567 while (ret >= 0 && nb_sectors > 0) {
568 block_index = sector_num / s->block_sectors;
569 sector_in_block = sector_num % s->block_sectors;
570 n_sectors = s->block_sectors - sector_in_block;
571 if (n_sectors > nb_sectors) {
572 n_sectors = nb_sectors;
575 logout("will read %u sectors starting at sector %" PRIu64 "\n",
576 n_sectors, sector_num);
578 /* prepare next AIO request */
579 bmap_entry = le32_to_cpu(s->bmap[block_index]);
580 if (!VDI_IS_ALLOCATED(bmap_entry)) {
581 /* Block not allocated, return zeros, no need to wait. */
582 memset(buf, 0, n_sectors * SECTOR_SIZE);
585 uint64_t offset = s->header.offset_data / SECTOR_SIZE +
586 (uint64_t)bmap_entry * s->block_sectors +
588 ret = bdrv_read(bs->file, offset, buf, n_sectors);
590 logout("%u sectors read\n", n_sectors);
592 nb_sectors -= n_sectors;
593 sector_num += n_sectors;
594 buf += n_sectors * SECTOR_SIZE;
600 static int vdi_co_write(BlockDriverState *bs,
601 int64_t sector_num, const uint8_t *buf, int nb_sectors)
603 BDRVVdiState *s = bs->opaque;
605 uint32_t block_index;
606 uint32_t sector_in_block;
608 uint32_t bmap_first = VDI_UNALLOCATED;
609 uint32_t bmap_last = VDI_UNALLOCATED;
610 uint8_t *block = NULL;
615 while (ret >= 0 && nb_sectors > 0) {
616 block_index = sector_num / s->block_sectors;
617 sector_in_block = sector_num % s->block_sectors;
618 n_sectors = s->block_sectors - sector_in_block;
619 if (n_sectors > nb_sectors) {
620 n_sectors = nb_sectors;
623 logout("will write %u sectors starting at sector %" PRIu64 "\n",
624 n_sectors, sector_num);
626 /* prepare next AIO request */
627 bmap_entry = le32_to_cpu(s->bmap[block_index]);
628 if (!VDI_IS_ALLOCATED(bmap_entry)) {
629 /* Allocate new block and write to it. */
631 bmap_entry = s->header.blocks_allocated;
632 s->bmap[block_index] = cpu_to_le32(bmap_entry);
633 s->header.blocks_allocated++;
634 offset = s->header.offset_data / SECTOR_SIZE +
635 (uint64_t)bmap_entry * s->block_sectors;
637 block = g_malloc(s->block_size);
638 bmap_first = block_index;
640 bmap_last = block_index;
641 /* Copy data to be written to new block and zero unused parts. */
642 memset(block, 0, sector_in_block * SECTOR_SIZE);
643 memcpy(block + sector_in_block * SECTOR_SIZE,
644 buf, n_sectors * SECTOR_SIZE);
645 memset(block + (sector_in_block + n_sectors) * SECTOR_SIZE, 0,
646 (s->block_sectors - n_sectors - sector_in_block) * SECTOR_SIZE);
648 /* Note that this coroutine does not yield anywhere from reading the
649 * bmap entry until here, so in regards to all the coroutines trying
650 * to write to this cluster, the one doing the allocation will
651 * always be the first to try to acquire the lock.
652 * Therefore, it is also the first that will actually be able to
653 * acquire the lock and thus the padded cluster is written before
654 * the other coroutines can write to the affected area. */
655 qemu_co_mutex_lock(&s->write_lock);
656 ret = bdrv_write(bs->file, offset, block, s->block_sectors);
657 qemu_co_mutex_unlock(&s->write_lock);
659 uint64_t offset = s->header.offset_data / SECTOR_SIZE +
660 (uint64_t)bmap_entry * s->block_sectors +
662 qemu_co_mutex_lock(&s->write_lock);
663 /* This lock is only used to make sure the following write operation
664 * is executed after the write issued by the coroutine allocating
665 * this cluster, therefore we do not need to keep it locked.
666 * As stated above, the allocating coroutine will always try to lock
667 * the mutex before all the other concurrent accesses to that
668 * cluster, therefore at this point we can be absolutely certain
669 * that that write operation has returned (there may be other writes
670 * in flight, but they do not concern this very operation). */
671 qemu_co_mutex_unlock(&s->write_lock);
672 ret = bdrv_write(bs->file, offset, buf, n_sectors);
675 nb_sectors -= n_sectors;
676 sector_num += n_sectors;
677 buf += n_sectors * SECTOR_SIZE;
679 logout("%u sectors written\n", n_sectors);
682 logout("finished data write\n");
688 /* One or more new blocks were allocated. */
689 VdiHeader *header = (VdiHeader *) block;
693 logout("now writing modified header\n");
694 assert(VDI_IS_ALLOCATED(bmap_first));
696 vdi_header_to_le(header);
697 ret = bdrv_write(bs->file, 0, block, 1);
705 logout("now writing modified block map entry %u...%u\n",
706 bmap_first, bmap_last);
707 /* Write modified sectors from block map. */
708 bmap_first /= (SECTOR_SIZE / sizeof(uint32_t));
709 bmap_last /= (SECTOR_SIZE / sizeof(uint32_t));
710 n_sectors = bmap_last - bmap_first + 1;
711 offset = s->bmap_sector + bmap_first;
712 base = ((uint8_t *)&s->bmap[0]) + bmap_first * SECTOR_SIZE;
713 logout("will write %u block map sectors starting from entry %u\n",
714 n_sectors, bmap_first);
715 ret = bdrv_write(bs->file, offset, base, n_sectors);
721 static int vdi_create(const char *filename, QemuOpts *opts, Error **errp)
726 size_t block_size = DEFAULT_CLUSTER_SIZE;
727 uint32_t image_type = VDI_TYPE_DYNAMIC;
732 Error *local_err = NULL;
733 BlockDriverState *bs = NULL;
734 uint32_t *bmap = NULL;
738 /* Read out options. */
739 bytes = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
741 #if defined(CONFIG_VDI_BLOCK_SIZE)
742 /* TODO: Additional checks (SECTOR_SIZE * 2^n, ...). */
743 block_size = qemu_opt_get_size_del(opts,
744 BLOCK_OPT_CLUSTER_SIZE,
745 DEFAULT_CLUSTER_SIZE);
747 #if defined(CONFIG_VDI_STATIC_IMAGE)
748 if (qemu_opt_get_bool_del(opts, BLOCK_OPT_STATIC, false)) {
749 image_type = VDI_TYPE_STATIC;
753 if (bytes > VDI_DISK_SIZE_MAX) {
755 error_setg(errp, "Unsupported VDI image size (size is 0x%" PRIx64
756 ", max supported is 0x%" PRIx64 ")",
757 bytes, VDI_DISK_SIZE_MAX);
761 ret = bdrv_create_file(filename, opts, &local_err);
763 error_propagate(errp, local_err);
766 ret = bdrv_open(&bs, filename, NULL, NULL, BDRV_O_RDWR | BDRV_O_PROTOCOL,
769 error_propagate(errp, local_err);
773 /* We need enough blocks to store the given disk size,
774 so always round up. */
775 blocks = DIV_ROUND_UP(bytes, block_size);
777 bmap_size = blocks * sizeof(uint32_t);
778 bmap_size = ROUND_UP(bmap_size, SECTOR_SIZE);
780 memset(&header, 0, sizeof(header));
781 pstrcpy(header.text, sizeof(header.text), VDI_TEXT);
782 header.signature = VDI_SIGNATURE;
783 header.version = VDI_VERSION_1_1;
784 header.header_size = 0x180;
785 header.image_type = image_type;
786 header.offset_bmap = 0x200;
787 header.offset_data = 0x200 + bmap_size;
788 header.sector_size = SECTOR_SIZE;
789 header.disk_size = bytes;
790 header.block_size = block_size;
791 header.blocks_in_image = blocks;
792 if (image_type == VDI_TYPE_STATIC) {
793 header.blocks_allocated = blocks;
795 uuid_generate(header.uuid_image);
796 uuid_generate(header.uuid_last_snap);
797 /* There is no need to set header.uuid_link or header.uuid_parent here. */
798 #if defined(CONFIG_VDI_DEBUG)
799 vdi_header_print(&header);
801 vdi_header_to_le(&header);
802 ret = bdrv_pwrite_sync(bs, offset, &header, sizeof(header));
804 error_setg(errp, "Error writing header to %s", filename);
807 offset += sizeof(header);
810 bmap = g_try_malloc0(bmap_size);
813 error_setg(errp, "Could not allocate bmap");
816 for (i = 0; i < blocks; i++) {
817 if (image_type == VDI_TYPE_STATIC) {
820 bmap[i] = VDI_UNALLOCATED;
823 ret = bdrv_pwrite_sync(bs, offset, bmap, bmap_size);
825 error_setg(errp, "Error writing bmap to %s", filename);
831 if (image_type == VDI_TYPE_STATIC) {
832 ret = bdrv_truncate(bs, offset + blocks * block_size);
834 error_setg(errp, "Failed to statically allocate %s", filename);
845 static void vdi_close(BlockDriverState *bs)
847 BDRVVdiState *s = bs->opaque;
851 migrate_del_blocker(s->migration_blocker);
852 error_free(s->migration_blocker);
855 static QemuOptsList vdi_create_opts = {
856 .name = "vdi-create-opts",
857 .head = QTAILQ_HEAD_INITIALIZER(vdi_create_opts.head),
860 .name = BLOCK_OPT_SIZE,
861 .type = QEMU_OPT_SIZE,
862 .help = "Virtual disk size"
864 #if defined(CONFIG_VDI_BLOCK_SIZE)
866 .name = BLOCK_OPT_CLUSTER_SIZE,
867 .type = QEMU_OPT_SIZE,
868 .help = "VDI cluster (block) size",
869 .def_value_str = stringify(DEFAULT_CLUSTER_SIZE)
872 #if defined(CONFIG_VDI_STATIC_IMAGE)
874 .name = BLOCK_OPT_STATIC,
875 .type = QEMU_OPT_BOOL,
876 .help = "VDI static (pre-allocated) image",
877 .def_value_str = "off"
880 /* TODO: An additional option to set UUID values might be useful. */
881 { /* end of list */ }
885 static BlockDriver bdrv_vdi = {
886 .format_name = "vdi",
887 .instance_size = sizeof(BDRVVdiState),
888 .bdrv_probe = vdi_probe,
889 .bdrv_open = vdi_open,
890 .bdrv_close = vdi_close,
891 .bdrv_reopen_prepare = vdi_reopen_prepare,
892 .bdrv_create = vdi_create,
893 .bdrv_has_zero_init = bdrv_has_zero_init_1,
894 .bdrv_co_get_block_status = vdi_co_get_block_status,
895 .bdrv_make_empty = vdi_make_empty,
897 .bdrv_read = vdi_co_read,
898 #if defined(CONFIG_VDI_WRITE)
899 .bdrv_write = vdi_co_write,
902 .bdrv_get_info = vdi_get_info,
904 .create_opts = &vdi_create_opts,
905 .bdrv_check = vdi_check,
908 static void bdrv_vdi_init(void)
911 bdrv_register(&bdrv_vdi);
914 block_init(bdrv_vdi_init);