2 * Block driver for the VMDK format
4 * Copyright (c) 2004 Fabrice Bellard
5 * Copyright (c) 2005 Filip Navara
7 * Permission is hereby granted, free of charge, to any person obtaining a copy
8 * of this software and associated documentation files (the "Software"), to deal
9 * in the Software without restriction, including without limitation the rights
10 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11 * copies of the Software, and to permit persons to whom the Software is
12 * furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included in
15 * all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
26 #include "qemu-common.h"
27 #include "block/block_int.h"
28 #include "qemu/module.h"
29 #include "migration/migration.h"
32 #define VMDK3_MAGIC (('C' << 24) | ('O' << 16) | ('W' << 8) | 'D')
33 #define VMDK4_MAGIC (('K' << 24) | ('D' << 16) | ('M' << 8) | 'V')
34 #define VMDK4_COMPRESSION_DEFLATE 1
35 #define VMDK4_FLAG_NL_DETECT (1 << 0)
36 #define VMDK4_FLAG_RGD (1 << 1)
37 /* Zeroed-grain enable bit */
38 #define VMDK4_FLAG_ZERO_GRAIN (1 << 2)
39 #define VMDK4_FLAG_COMPRESS (1 << 16)
40 #define VMDK4_FLAG_MARKER (1 << 17)
41 #define VMDK4_GD_AT_END 0xffffffffffffffffULL
43 #define VMDK_GTE_ZEROED 0x1
45 /* VMDK internal error codes */
47 #define VMDK_ERROR (-1)
48 /* Cluster not allocated */
49 #define VMDK_UNALLOC (-2)
50 #define VMDK_ZEROED (-3)
52 #define BLOCK_OPT_ZEROED_GRAIN "zeroed_grain"
57 uint32_t disk_sectors
;
59 uint32_t l1dir_offset
;
61 uint32_t file_sectors
;
64 uint32_t sectors_per_track
;
65 } QEMU_PACKED VMDK3Header
;
74 uint32_t num_gtes_per_gte
;
77 uint64_t grain_offset
;
80 uint16_t compressAlgorithm
;
81 } QEMU_PACKED VMDK4Header
;
83 #define L2_CACHE_SIZE 16
85 typedef struct VmdkExtent
{
86 BlockDriverState
*file
;
94 int64_t flat_start_offset
;
95 int64_t l1_table_offset
;
96 int64_t l1_backup_table_offset
;
98 uint32_t *l1_backup_table
;
100 uint32_t l1_entry_sectors
;
102 unsigned int l2_size
;
104 uint32_t l2_cache_offsets
[L2_CACHE_SIZE
];
105 uint32_t l2_cache_counts
[L2_CACHE_SIZE
];
107 unsigned int cluster_sectors
;
110 typedef struct BDRVVmdkState
{
112 uint64_t desc_offset
;
116 /* Extent array with num_extents entries, ascend ordered by address */
118 Error
*migration_blocker
;
121 typedef struct VmdkMetaData
{
123 unsigned int l1_index
;
124 unsigned int l2_index
;
125 unsigned int l2_offset
;
127 uint32_t *l2_cache_entry
;
130 typedef struct VmdkGrainMarker
{
134 } QEMU_PACKED VmdkGrainMarker
;
137 MARKER_END_OF_STREAM
= 0,
138 MARKER_GRAIN_TABLE
= 1,
139 MARKER_GRAIN_DIRECTORY
= 2,
143 static int vmdk_probe(const uint8_t *buf
, int buf_size
, const char *filename
)
150 magic
= be32_to_cpu(*(uint32_t *)buf
);
151 if (magic
== VMDK3_MAGIC
||
152 magic
== VMDK4_MAGIC
) {
155 const char *p
= (const char *)buf
;
156 const char *end
= p
+ buf_size
;
159 /* skip comment line */
160 while (p
< end
&& *p
!= '\n') {
167 while (p
< end
&& *p
== ' ') {
170 /* skip '\r' if windows line endings used. */
171 if (p
< end
&& *p
== '\r') {
174 /* only accept blank lines before 'version=' line */
175 if (p
== end
|| *p
!= '\n') {
181 if (end
- p
>= strlen("version=X\n")) {
182 if (strncmp("version=1\n", p
, strlen("version=1\n")) == 0 ||
183 strncmp("version=2\n", p
, strlen("version=2\n")) == 0) {
187 if (end
- p
>= strlen("version=X\r\n")) {
188 if (strncmp("version=1\r\n", p
, strlen("version=1\r\n")) == 0 ||
189 strncmp("version=2\r\n", p
, strlen("version=2\r\n")) == 0) {
201 #define SECTOR_SIZE 512
202 #define DESC_SIZE (20 * SECTOR_SIZE) /* 20 sectors of 512 bytes each */
203 #define BUF_SIZE 4096
204 #define HEADER_SIZE 512 /* first sector of 512 bytes */
206 static void vmdk_free_extents(BlockDriverState
*bs
)
209 BDRVVmdkState
*s
= bs
->opaque
;
212 for (i
= 0; i
< s
->num_extents
; i
++) {
216 g_free(e
->l1_backup_table
);
217 if (e
->file
!= bs
->file
) {
218 bdrv_delete(e
->file
);
224 static void vmdk_free_last_extent(BlockDriverState
*bs
)
226 BDRVVmdkState
*s
= bs
->opaque
;
228 if (s
->num_extents
== 0) {
232 s
->extents
= g_realloc(s
->extents
, s
->num_extents
* sizeof(VmdkExtent
));
235 static uint32_t vmdk_read_cid(BlockDriverState
*bs
, int parent
)
237 char desc
[DESC_SIZE
];
238 uint32_t cid
= 0xffffffff;
239 const char *p_name
, *cid_str
;
241 BDRVVmdkState
*s
= bs
->opaque
;
244 ret
= bdrv_pread(bs
->file
, s
->desc_offset
, desc
, DESC_SIZE
);
250 cid_str
= "parentCID";
251 cid_str_size
= sizeof("parentCID");
254 cid_str_size
= sizeof("CID");
257 desc
[DESC_SIZE
- 1] = '\0';
258 p_name
= strstr(desc
, cid_str
);
259 if (p_name
!= NULL
) {
260 p_name
+= cid_str_size
;
261 sscanf(p_name
, "%x", &cid
);
267 static int vmdk_write_cid(BlockDriverState
*bs
, uint32_t cid
)
269 char desc
[DESC_SIZE
], tmp_desc
[DESC_SIZE
];
270 char *p_name
, *tmp_str
;
271 BDRVVmdkState
*s
= bs
->opaque
;
274 ret
= bdrv_pread(bs
->file
, s
->desc_offset
, desc
, DESC_SIZE
);
279 desc
[DESC_SIZE
- 1] = '\0';
280 tmp_str
= strstr(desc
, "parentCID");
281 if (tmp_str
== NULL
) {
285 pstrcpy(tmp_desc
, sizeof(tmp_desc
), tmp_str
);
286 p_name
= strstr(desc
, "CID");
287 if (p_name
!= NULL
) {
288 p_name
+= sizeof("CID");
289 snprintf(p_name
, sizeof(desc
) - (p_name
- desc
), "%x\n", cid
);
290 pstrcat(desc
, sizeof(desc
), tmp_desc
);
293 ret
= bdrv_pwrite_sync(bs
->file
, s
->desc_offset
, desc
, DESC_SIZE
);
301 static int vmdk_is_cid_valid(BlockDriverState
*bs
)
304 BDRVVmdkState
*s
= bs
->opaque
;
305 BlockDriverState
*p_bs
= bs
->backing_hd
;
309 cur_pcid
= vmdk_read_cid(p_bs
, 0);
310 if (s
->parent_cid
!= cur_pcid
) {
320 /* Queue extents, if any, for reopen() */
321 static int vmdk_reopen_prepare(BDRVReopenState
*state
,
322 BlockReopenQueue
*queue
, Error
**errp
)
329 assert(state
!= NULL
);
330 assert(state
->bs
!= NULL
);
333 error_set(errp
, ERROR_CLASS_GENERIC_ERROR
,
334 "No reopen queue for VMDK extents");
338 s
= state
->bs
->opaque
;
342 for (i
= 0; i
< s
->num_extents
; i
++) {
344 if (e
->file
!= state
->bs
->file
) {
345 bdrv_reopen_queue(queue
, e
->file
, state
->flags
);
354 static int vmdk_parent_open(BlockDriverState
*bs
)
357 char desc
[DESC_SIZE
+ 1];
358 BDRVVmdkState
*s
= bs
->opaque
;
361 desc
[DESC_SIZE
] = '\0';
362 ret
= bdrv_pread(bs
->file
, s
->desc_offset
, desc
, DESC_SIZE
);
367 p_name
= strstr(desc
, "parentFileNameHint");
368 if (p_name
!= NULL
) {
371 p_name
+= sizeof("parentFileNameHint") + 1;
372 end_name
= strchr(p_name
, '\"');
373 if (end_name
== NULL
) {
376 if ((end_name
- p_name
) > sizeof(bs
->backing_file
) - 1) {
380 pstrcpy(bs
->backing_file
, end_name
- p_name
+ 1, p_name
);
386 /* Create and append extent to the extent array. Return the added VmdkExtent
387 * address. return NULL if allocation failed. */
388 static int vmdk_add_extent(BlockDriverState
*bs
,
389 BlockDriverState
*file
, bool flat
, int64_t sectors
,
390 int64_t l1_offset
, int64_t l1_backup_offset
,
392 int l2_size
, uint64_t cluster_sectors
,
393 VmdkExtent
**new_extent
)
396 BDRVVmdkState
*s
= bs
->opaque
;
398 if (cluster_sectors
> 0x200000) {
399 /* 0x200000 * 512Bytes = 1GB for one cluster is unrealistic */
400 error_report("invalid granularity, image may be corrupt");
404 s
->extents
= g_realloc(s
->extents
,
405 (s
->num_extents
+ 1) * sizeof(VmdkExtent
));
406 extent
= &s
->extents
[s
->num_extents
];
409 memset(extent
, 0, sizeof(VmdkExtent
));
412 extent
->sectors
= sectors
;
413 extent
->l1_table_offset
= l1_offset
;
414 extent
->l1_backup_table_offset
= l1_backup_offset
;
415 extent
->l1_size
= l1_size
;
416 extent
->l1_entry_sectors
= l2_size
* cluster_sectors
;
417 extent
->l2_size
= l2_size
;
418 extent
->cluster_sectors
= cluster_sectors
;
420 if (s
->num_extents
> 1) {
421 extent
->end_sector
= (*(extent
- 1)).end_sector
+ extent
->sectors
;
423 extent
->end_sector
= extent
->sectors
;
425 bs
->total_sectors
= extent
->end_sector
;
427 *new_extent
= extent
;
432 static int vmdk_init_tables(BlockDriverState
*bs
, VmdkExtent
*extent
)
437 /* read the L1 table */
438 l1_size
= extent
->l1_size
* sizeof(uint32_t);
439 extent
->l1_table
= g_malloc(l1_size
);
440 ret
= bdrv_pread(extent
->file
,
441 extent
->l1_table_offset
,
447 for (i
= 0; i
< extent
->l1_size
; i
++) {
448 le32_to_cpus(&extent
->l1_table
[i
]);
451 if (extent
->l1_backup_table_offset
) {
452 extent
->l1_backup_table
= g_malloc(l1_size
);
453 ret
= bdrv_pread(extent
->file
,
454 extent
->l1_backup_table_offset
,
455 extent
->l1_backup_table
,
460 for (i
= 0; i
< extent
->l1_size
; i
++) {
461 le32_to_cpus(&extent
->l1_backup_table
[i
]);
466 g_malloc(extent
->l2_size
* L2_CACHE_SIZE
* sizeof(uint32_t));
469 g_free(extent
->l1_backup_table
);
471 g_free(extent
->l1_table
);
475 static int vmdk_open_vmdk3(BlockDriverState
*bs
,
476 BlockDriverState
*file
,
484 ret
= bdrv_pread(file
, sizeof(magic
), &header
, sizeof(header
));
489 ret
= vmdk_add_extent(bs
,
491 le32_to_cpu(header
.disk_sectors
),
492 le32_to_cpu(header
.l1dir_offset
) << 9,
494 le32_to_cpu(header
.granularity
),
499 ret
= vmdk_init_tables(bs
, extent
);
501 /* free extent allocated by vmdk_add_extent */
502 vmdk_free_last_extent(bs
);
507 static int vmdk_open_desc_file(BlockDriverState
*bs
, int flags
,
508 uint64_t desc_offset
);
510 static int vmdk_open_vmdk4(BlockDriverState
*bs
,
511 BlockDriverState
*file
,
516 uint32_t l1_size
, l1_entry_sectors
;
519 int64_t l1_backup_offset
= 0;
521 ret
= bdrv_pread(file
, sizeof(magic
), &header
, sizeof(header
));
525 if (header
.capacity
== 0) {
526 uint64_t desc_offset
= le64_to_cpu(header
.desc_offset
);
528 return vmdk_open_desc_file(bs
, flags
, desc_offset
<< 9);
532 if (le64_to_cpu(header
.gd_offset
) == VMDK4_GD_AT_END
) {
534 * The footer takes precedence over the header, so read it in. The
535 * footer starts at offset -1024 from the end: One sector for the
536 * footer, and another one for the end-of-stream marker.
543 uint8_t pad
[512 - 16];
544 } QEMU_PACKED footer_marker
;
548 uint8_t pad
[512 - 4 - sizeof(VMDK4Header
)];
554 uint8_t pad
[512 - 16];
555 } QEMU_PACKED eos_marker
;
556 } QEMU_PACKED footer
;
558 ret
= bdrv_pread(file
,
559 bs
->file
->total_sectors
* 512 - 1536,
560 &footer
, sizeof(footer
));
565 /* Some sanity checks for the footer */
566 if (be32_to_cpu(footer
.magic
) != VMDK4_MAGIC
||
567 le32_to_cpu(footer
.footer_marker
.size
) != 0 ||
568 le32_to_cpu(footer
.footer_marker
.type
) != MARKER_FOOTER
||
569 le64_to_cpu(footer
.eos_marker
.val
) != 0 ||
570 le32_to_cpu(footer
.eos_marker
.size
) != 0 ||
571 le32_to_cpu(footer
.eos_marker
.type
) != MARKER_END_OF_STREAM
)
576 header
= footer
.header
;
579 if (le32_to_cpu(header
.version
) >= 3) {
581 snprintf(buf
, sizeof(buf
), "VMDK version %d",
582 le32_to_cpu(header
.version
));
583 qerror_report(QERR_UNKNOWN_BLOCK_FORMAT_FEATURE
,
584 bs
->device_name
, "vmdk", buf
);
588 if (le32_to_cpu(header
.num_gtes_per_gte
) > 512) {
589 error_report("L2 table size too big");
593 l1_entry_sectors
= le32_to_cpu(header
.num_gtes_per_gte
)
594 * le64_to_cpu(header
.granularity
);
595 if (l1_entry_sectors
== 0) {
598 l1_size
= (le64_to_cpu(header
.capacity
) + l1_entry_sectors
- 1)
600 if (le32_to_cpu(header
.flags
) & VMDK4_FLAG_RGD
) {
601 l1_backup_offset
= le64_to_cpu(header
.rgd_offset
) << 9;
603 ret
= vmdk_add_extent(bs
, file
, false,
604 le64_to_cpu(header
.capacity
),
605 le64_to_cpu(header
.gd_offset
) << 9,
608 le32_to_cpu(header
.num_gtes_per_gte
),
609 le64_to_cpu(header
.granularity
),
615 le16_to_cpu(header
.compressAlgorithm
) == VMDK4_COMPRESSION_DEFLATE
;
616 extent
->has_marker
= le32_to_cpu(header
.flags
) & VMDK4_FLAG_MARKER
;
617 extent
->version
= le32_to_cpu(header
.version
);
618 extent
->has_zero_grain
= le32_to_cpu(header
.flags
) & VMDK4_FLAG_ZERO_GRAIN
;
619 ret
= vmdk_init_tables(bs
, extent
);
621 /* free extent allocated by vmdk_add_extent */
622 vmdk_free_last_extent(bs
);
627 /* find an option value out of descriptor file */
628 static int vmdk_parse_description(const char *desc
, const char *opt_name
,
629 char *buf
, int buf_size
)
631 char *opt_pos
, *opt_end
;
632 const char *end
= desc
+ strlen(desc
);
634 opt_pos
= strstr(desc
, opt_name
);
638 /* Skip "=\"" following opt_name */
639 opt_pos
+= strlen(opt_name
) + 2;
640 if (opt_pos
>= end
) {
644 while (opt_end
< end
&& *opt_end
!= '"') {
647 if (opt_end
== end
|| buf_size
< opt_end
- opt_pos
+ 1) {
650 pstrcpy(buf
, opt_end
- opt_pos
+ 1, opt_pos
);
654 /* Open an extent file and append to bs array */
655 static int vmdk_open_sparse(BlockDriverState
*bs
,
656 BlockDriverState
*file
,
661 if (bdrv_pread(file
, 0, &magic
, sizeof(magic
)) != sizeof(magic
)) {
665 magic
= be32_to_cpu(magic
);
668 return vmdk_open_vmdk3(bs
, file
, flags
);
671 return vmdk_open_vmdk4(bs
, file
, flags
);
679 static int vmdk_parse_extents(const char *desc
, BlockDriverState
*bs
,
680 const char *desc_file_path
)
686 const char *p
= desc
;
689 char extent_path
[PATH_MAX
];
690 BlockDriverState
*extent_file
;
693 /* parse extent line:
694 * RW [size in sectors] FLAT "file-name.vmdk" OFFSET
696 * RW [size in sectors] SPARSE "file-name.vmdk"
699 ret
= sscanf(p
, "%10s %" SCNd64
" %10s \"%511[^\n\r\"]\" %" SCNd64
,
700 access
, §ors
, type
, fname
, &flat_offset
);
701 if (ret
< 4 || strcmp(access
, "RW")) {
703 } else if (!strcmp(type
, "FLAT")) {
704 if (ret
!= 5 || flat_offset
< 0) {
707 } else if (ret
!= 4) {
712 (strcmp(type
, "FLAT") && strcmp(type
, "SPARSE")) ||
713 (strcmp(access
, "RW"))) {
717 path_combine(extent_path
, sizeof(extent_path
),
718 desc_file_path
, fname
);
719 ret
= bdrv_file_open(&extent_file
, extent_path
, NULL
, bs
->open_flags
);
724 /* save to extents array */
725 if (!strcmp(type
, "FLAT")) {
729 ret
= vmdk_add_extent(bs
, extent_file
, true, sectors
,
730 0, 0, 0, 0, sectors
, &extent
);
734 extent
->flat_start_offset
= flat_offset
<< 9;
735 } else if (!strcmp(type
, "SPARSE")) {
737 ret
= vmdk_open_sparse(bs
, extent_file
, bs
->open_flags
);
739 bdrv_delete(extent_file
);
744 "VMDK: Not supported extent type \"%s\""".\n", type
);
748 /* move to next line */
749 while (*p
&& *p
!= '\n') {
757 static int vmdk_open_desc_file(BlockDriverState
*bs
, int flags
,
758 uint64_t desc_offset
)
763 BDRVVmdkState
*s
= bs
->opaque
;
766 size
= bdrv_getlength(bs
->file
);
771 size
= MIN(size
, 1 << 20); /* avoid unbounded allocation */
772 buf
= g_malloc0(size
+ 1);
774 ret
= bdrv_pread(bs
->file
, desc_offset
, buf
, size
);
778 if (vmdk_parse_description(buf
, "createType", ct
, sizeof(ct
))) {
782 if (strcmp(ct
, "monolithicFlat") &&
783 strcmp(ct
, "twoGbMaxExtentSparse") &&
784 strcmp(ct
, "twoGbMaxExtentFlat")) {
786 "VMDK: Not supported image type \"%s\""".\n", ct
);
791 ret
= vmdk_parse_extents(buf
, bs
, bs
->file
->filename
);
797 static int vmdk_open(BlockDriverState
*bs
, QDict
*options
, int flags
)
800 BDRVVmdkState
*s
= bs
->opaque
;
802 if (vmdk_open_sparse(bs
, bs
->file
, flags
) == 0) {
803 s
->desc_offset
= 0x200;
805 ret
= vmdk_open_desc_file(bs
, flags
, 0);
810 /* try to open parent images, if exist */
811 ret
= vmdk_parent_open(bs
);
815 s
->parent_cid
= vmdk_read_cid(bs
, 1);
816 qemu_co_mutex_init(&s
->lock
);
818 /* Disable migration when VMDK images are used */
819 error_set(&s
->migration_blocker
,
820 QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
821 "vmdk", bs
->device_name
, "live migration");
822 migrate_add_blocker(s
->migration_blocker
);
827 vmdk_free_extents(bs
);
831 static int get_whole_cluster(BlockDriverState
*bs
,
833 uint64_t cluster_offset
,
837 /* 128 sectors * 512 bytes each = grain size 64KB */
838 uint8_t whole_grain
[extent
->cluster_sectors
* 512];
840 /* we will be here if it's first write on non-exist grain(cluster).
841 * try to read from parent image, if exist */
842 if (bs
->backing_hd
) {
845 if (!vmdk_is_cid_valid(bs
)) {
849 /* floor offset to cluster */
850 offset
-= offset
% (extent
->cluster_sectors
* 512);
851 ret
= bdrv_read(bs
->backing_hd
, offset
>> 9, whole_grain
,
852 extent
->cluster_sectors
);
857 /* Write grain only into the active image */
858 ret
= bdrv_write(extent
->file
, cluster_offset
, whole_grain
,
859 extent
->cluster_sectors
);
867 static int vmdk_L2update(VmdkExtent
*extent
, VmdkMetaData
*m_data
)
870 QEMU_BUILD_BUG_ON(sizeof(offset
) != sizeof(m_data
->offset
));
871 offset
= cpu_to_le32(m_data
->offset
);
872 /* update L2 table */
873 if (bdrv_pwrite_sync(
875 ((int64_t)m_data
->l2_offset
* 512)
876 + (m_data
->l2_index
* sizeof(m_data
->offset
)),
877 &offset
, sizeof(offset
)) < 0) {
880 /* update backup L2 table */
881 if (extent
->l1_backup_table_offset
!= 0) {
882 m_data
->l2_offset
= extent
->l1_backup_table
[m_data
->l1_index
];
883 if (bdrv_pwrite_sync(
885 ((int64_t)m_data
->l2_offset
* 512)
886 + (m_data
->l2_index
* sizeof(m_data
->offset
)),
887 &offset
, sizeof(offset
)) < 0) {
891 if (m_data
->l2_cache_entry
) {
892 *m_data
->l2_cache_entry
= offset
;
898 static int get_cluster_offset(BlockDriverState
*bs
,
900 VmdkMetaData
*m_data
,
903 uint64_t *cluster_offset
)
905 unsigned int l1_index
, l2_offset
, l2_index
;
907 uint32_t min_count
, *l2_table
;
914 *cluster_offset
= extent
->flat_start_offset
;
918 offset
-= (extent
->end_sector
- extent
->sectors
) * SECTOR_SIZE
;
919 l1_index
= (offset
>> 9) / extent
->l1_entry_sectors
;
920 if (l1_index
>= extent
->l1_size
) {
923 l2_offset
= extent
->l1_table
[l1_index
];
927 for (i
= 0; i
< L2_CACHE_SIZE
; i
++) {
928 if (l2_offset
== extent
->l2_cache_offsets
[i
]) {
929 /* increment the hit count */
930 if (++extent
->l2_cache_counts
[i
] == 0xffffffff) {
931 for (j
= 0; j
< L2_CACHE_SIZE
; j
++) {
932 extent
->l2_cache_counts
[j
] >>= 1;
935 l2_table
= extent
->l2_cache
+ (i
* extent
->l2_size
);
939 /* not found: load a new entry in the least used one */
941 min_count
= 0xffffffff;
942 for (i
= 0; i
< L2_CACHE_SIZE
; i
++) {
943 if (extent
->l2_cache_counts
[i
] < min_count
) {
944 min_count
= extent
->l2_cache_counts
[i
];
948 l2_table
= extent
->l2_cache
+ (min_index
* extent
->l2_size
);
951 (int64_t)l2_offset
* 512,
953 extent
->l2_size
* sizeof(uint32_t)
954 ) != extent
->l2_size
* sizeof(uint32_t)) {
958 extent
->l2_cache_offsets
[min_index
] = l2_offset
;
959 extent
->l2_cache_counts
[min_index
] = 1;
961 l2_index
= ((offset
>> 9) / extent
->cluster_sectors
) % extent
->l2_size
;
962 *cluster_offset
= le32_to_cpu(l2_table
[l2_index
]);
966 m_data
->l1_index
= l1_index
;
967 m_data
->l2_index
= l2_index
;
968 m_data
->offset
= *cluster_offset
;
969 m_data
->l2_offset
= l2_offset
;
970 m_data
->l2_cache_entry
= &l2_table
[l2_index
];
972 if (extent
->has_zero_grain
&& *cluster_offset
== VMDK_GTE_ZEROED
) {
976 if (!*cluster_offset
|| zeroed
) {
978 return zeroed
? VMDK_ZEROED
: VMDK_UNALLOC
;
981 /* Avoid the L2 tables update for the images that have snapshots. */
982 *cluster_offset
= bdrv_getlength(extent
->file
);
983 if (!extent
->compressed
) {
986 *cluster_offset
+ (extent
->cluster_sectors
<< 9)
990 *cluster_offset
>>= 9;
991 l2_table
[l2_index
] = cpu_to_le32(*cluster_offset
);
993 /* First of all we write grain itself, to avoid race condition
994 * that may to corrupt the image.
995 * This problem may occur because of insufficient space on host disk
996 * or inappropriate VM shutdown.
998 if (get_whole_cluster(
999 bs
, extent
, *cluster_offset
, offset
, allocate
) == -1) {
1004 m_data
->offset
= *cluster_offset
;
1007 *cluster_offset
<<= 9;
1011 static VmdkExtent
*find_extent(BDRVVmdkState
*s
,
1012 int64_t sector_num
, VmdkExtent
*start_hint
)
1014 VmdkExtent
*extent
= start_hint
;
1017 extent
= &s
->extents
[0];
1019 while (extent
< &s
->extents
[s
->num_extents
]) {
1020 if (sector_num
< extent
->end_sector
) {
1028 static int coroutine_fn
vmdk_co_is_allocated(BlockDriverState
*bs
,
1029 int64_t sector_num
, int nb_sectors
, int *pnum
)
1031 BDRVVmdkState
*s
= bs
->opaque
;
1032 int64_t index_in_cluster
, n
, ret
;
1036 extent
= find_extent(s
, sector_num
, NULL
);
1040 qemu_co_mutex_lock(&s
->lock
);
1041 ret
= get_cluster_offset(bs
, extent
, NULL
,
1042 sector_num
* 512, 0, &offset
);
1043 qemu_co_mutex_unlock(&s
->lock
);
1045 ret
= (ret
== VMDK_OK
|| ret
== VMDK_ZEROED
);
1047 index_in_cluster
= sector_num
% extent
->cluster_sectors
;
1048 n
= extent
->cluster_sectors
- index_in_cluster
;
1049 if (n
> nb_sectors
) {
1056 static int vmdk_write_extent(VmdkExtent
*extent
, int64_t cluster_offset
,
1057 int64_t offset_in_cluster
, const uint8_t *buf
,
1058 int nb_sectors
, int64_t sector_num
)
1061 VmdkGrainMarker
*data
= NULL
;
1063 const uint8_t *write_buf
= buf
;
1064 int write_len
= nb_sectors
* 512;
1066 if (extent
->compressed
) {
1067 if (!extent
->has_marker
) {
1071 buf_len
= (extent
->cluster_sectors
<< 9) * 2;
1072 data
= g_malloc(buf_len
+ sizeof(VmdkGrainMarker
));
1073 if (compress(data
->data
, &buf_len
, buf
, nb_sectors
<< 9) != Z_OK
||
1078 data
->lba
= sector_num
;
1079 data
->size
= buf_len
;
1080 write_buf
= (uint8_t *)data
;
1081 write_len
= buf_len
+ sizeof(VmdkGrainMarker
);
1083 ret
= bdrv_pwrite(extent
->file
,
1084 cluster_offset
+ offset_in_cluster
,
1087 if (ret
!= write_len
) {
1088 ret
= ret
< 0 ? ret
: -EIO
;
1097 static int vmdk_read_extent(VmdkExtent
*extent
, int64_t cluster_offset
,
1098 int64_t offset_in_cluster
, uint8_t *buf
,
1102 int cluster_bytes
, buf_bytes
;
1103 uint8_t *cluster_buf
, *compressed_data
;
1104 uint8_t *uncomp_buf
;
1106 VmdkGrainMarker
*marker
;
1110 if (!extent
->compressed
) {
1111 ret
= bdrv_pread(extent
->file
,
1112 cluster_offset
+ offset_in_cluster
,
1113 buf
, nb_sectors
* 512);
1114 if (ret
== nb_sectors
* 512) {
1120 cluster_bytes
= extent
->cluster_sectors
* 512;
1121 /* Read two clusters in case GrainMarker + compressed data > one cluster */
1122 buf_bytes
= cluster_bytes
* 2;
1123 cluster_buf
= g_malloc(buf_bytes
);
1124 uncomp_buf
= g_malloc(cluster_bytes
);
1125 ret
= bdrv_pread(extent
->file
,
1127 cluster_buf
, buf_bytes
);
1131 compressed_data
= cluster_buf
;
1132 buf_len
= cluster_bytes
;
1133 data_len
= cluster_bytes
;
1134 if (extent
->has_marker
) {
1135 marker
= (VmdkGrainMarker
*)cluster_buf
;
1136 compressed_data
= marker
->data
;
1137 data_len
= le32_to_cpu(marker
->size
);
1139 if (!data_len
|| data_len
> buf_bytes
) {
1143 ret
= uncompress(uncomp_buf
, &buf_len
, compressed_data
, data_len
);
1149 if (offset_in_cluster
< 0 ||
1150 offset_in_cluster
+ nb_sectors
* 512 > buf_len
) {
1154 memcpy(buf
, uncomp_buf
+ offset_in_cluster
, nb_sectors
* 512);
1159 g_free(cluster_buf
);
1163 static int vmdk_read(BlockDriverState
*bs
, int64_t sector_num
,
1164 uint8_t *buf
, int nb_sectors
)
1166 BDRVVmdkState
*s
= bs
->opaque
;
1168 uint64_t n
, index_in_cluster
;
1169 uint64_t extent_begin_sector
, extent_relative_sector_num
;
1170 VmdkExtent
*extent
= NULL
;
1171 uint64_t cluster_offset
;
1173 while (nb_sectors
> 0) {
1174 extent
= find_extent(s
, sector_num
, extent
);
1178 ret
= get_cluster_offset(
1180 sector_num
<< 9, 0, &cluster_offset
);
1181 extent_begin_sector
= extent
->end_sector
- extent
->sectors
;
1182 extent_relative_sector_num
= sector_num
- extent_begin_sector
;
1183 index_in_cluster
= extent_relative_sector_num
% extent
->cluster_sectors
;
1184 n
= extent
->cluster_sectors
- index_in_cluster
;
1185 if (n
> nb_sectors
) {
1188 if (ret
!= VMDK_OK
) {
1189 /* if not allocated, try to read from parent image, if exist */
1190 if (bs
->backing_hd
&& ret
!= VMDK_ZEROED
) {
1191 if (!vmdk_is_cid_valid(bs
)) {
1194 ret
= bdrv_read(bs
->backing_hd
, sector_num
, buf
, n
);
1199 memset(buf
, 0, 512 * n
);
1202 ret
= vmdk_read_extent(extent
,
1203 cluster_offset
, index_in_cluster
* 512,
1216 static coroutine_fn
int vmdk_co_read(BlockDriverState
*bs
, int64_t sector_num
,
1217 uint8_t *buf
, int nb_sectors
)
1220 BDRVVmdkState
*s
= bs
->opaque
;
1221 qemu_co_mutex_lock(&s
->lock
);
1222 ret
= vmdk_read(bs
, sector_num
, buf
, nb_sectors
);
1223 qemu_co_mutex_unlock(&s
->lock
);
1229 * @zeroed: buf is ignored (data is zero), use zeroed_grain GTE feature
1230 * if possible, otherwise return -ENOTSUP.
1231 * @zero_dry_run: used for zeroed == true only, don't update L2 table, just try
1232 * with each cluster. By dry run we can find if the zero write
1233 * is possible without modifying image data.
1235 * Returns: error code with 0 for success.
1237 static int vmdk_write(BlockDriverState
*bs
, int64_t sector_num
,
1238 const uint8_t *buf
, int nb_sectors
,
1239 bool zeroed
, bool zero_dry_run
)
1241 BDRVVmdkState
*s
= bs
->opaque
;
1242 VmdkExtent
*extent
= NULL
;
1244 int64_t index_in_cluster
;
1245 uint64_t extent_begin_sector
, extent_relative_sector_num
;
1246 uint64_t cluster_offset
;
1247 VmdkMetaData m_data
;
1249 if (sector_num
> bs
->total_sectors
) {
1251 "(VMDK) Wrong offset: sector_num=0x%" PRIx64
1252 " total_sectors=0x%" PRIx64
"\n",
1253 sector_num
, bs
->total_sectors
);
1257 while (nb_sectors
> 0) {
1258 extent
= find_extent(s
, sector_num
, extent
);
1262 ret
= get_cluster_offset(
1266 sector_num
<< 9, !extent
->compressed
,
1268 if (extent
->compressed
) {
1269 if (ret
== VMDK_OK
) {
1270 /* Refuse write to allocated cluster for streamOptimized */
1272 "VMDK: can't write to allocated cluster"
1273 " for streamOptimized\n");
1277 ret
= get_cluster_offset(
1285 if (ret
== VMDK_ERROR
) {
1288 extent_begin_sector
= extent
->end_sector
- extent
->sectors
;
1289 extent_relative_sector_num
= sector_num
- extent_begin_sector
;
1290 index_in_cluster
= extent_relative_sector_num
% extent
->cluster_sectors
;
1291 n
= extent
->cluster_sectors
- index_in_cluster
;
1292 if (n
> nb_sectors
) {
1296 /* Do zeroed write, buf is ignored */
1297 if (extent
->has_zero_grain
&&
1298 index_in_cluster
== 0 &&
1299 n
>= extent
->cluster_sectors
) {
1300 n
= extent
->cluster_sectors
;
1301 if (!zero_dry_run
) {
1302 m_data
.offset
= VMDK_GTE_ZEROED
;
1303 /* update L2 tables */
1304 if (vmdk_L2update(extent
, &m_data
) != VMDK_OK
) {
1312 ret
= vmdk_write_extent(extent
,
1313 cluster_offset
, index_in_cluster
* 512,
1314 buf
, n
, sector_num
);
1319 /* update L2 tables */
1320 if (vmdk_L2update(extent
, &m_data
) != VMDK_OK
) {
1329 /* update CID on the first write every time the virtual disk is
1331 if (!s
->cid_updated
) {
1332 ret
= vmdk_write_cid(bs
, time(NULL
));
1336 s
->cid_updated
= true;
1342 static coroutine_fn
int vmdk_co_write(BlockDriverState
*bs
, int64_t sector_num
,
1343 const uint8_t *buf
, int nb_sectors
)
1346 BDRVVmdkState
*s
= bs
->opaque
;
1347 qemu_co_mutex_lock(&s
->lock
);
1348 ret
= vmdk_write(bs
, sector_num
, buf
, nb_sectors
, false, false);
1349 qemu_co_mutex_unlock(&s
->lock
);
1353 static int coroutine_fn
vmdk_co_write_zeroes(BlockDriverState
*bs
,
1358 BDRVVmdkState
*s
= bs
->opaque
;
1359 qemu_co_mutex_lock(&s
->lock
);
1360 /* write zeroes could fail if sectors not aligned to cluster, test it with
1361 * dry_run == true before really updating image */
1362 ret
= vmdk_write(bs
, sector_num
, NULL
, nb_sectors
, true, true);
1364 ret
= vmdk_write(bs
, sector_num
, NULL
, nb_sectors
, true, false);
1366 qemu_co_mutex_unlock(&s
->lock
);
1371 static int vmdk_create_extent(const char *filename
, int64_t filesize
,
1372 bool flat
, bool compress
, bool zeroed_grain
)
1377 uint32_t tmp
, magic
, grains
, gd_size
, gt_size
, gt_count
;
1379 fd
= qemu_open(filename
,
1380 O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
| O_LARGEFILE
,
1386 ret
= ftruncate(fd
, filesize
);
1392 magic
= cpu_to_be32(VMDK4_MAGIC
);
1393 memset(&header
, 0, sizeof(header
));
1394 header
.version
= zeroed_grain
? 2 : 1;
1395 header
.flags
= VMDK4_FLAG_RGD
| VMDK4_FLAG_NL_DETECT
1396 | (compress
? VMDK4_FLAG_COMPRESS
| VMDK4_FLAG_MARKER
: 0)
1397 | (zeroed_grain
? VMDK4_FLAG_ZERO_GRAIN
: 0);
1398 header
.compressAlgorithm
= compress
? VMDK4_COMPRESSION_DEFLATE
: 0;
1399 header
.capacity
= filesize
/ 512;
1400 header
.granularity
= 128;
1401 header
.num_gtes_per_gte
= 512;
1403 grains
= (filesize
/ 512 + header
.granularity
- 1) / header
.granularity
;
1404 gt_size
= ((header
.num_gtes_per_gte
* sizeof(uint32_t)) + 511) >> 9;
1406 (grains
+ header
.num_gtes_per_gte
- 1) / header
.num_gtes_per_gte
;
1407 gd_size
= (gt_count
* sizeof(uint32_t) + 511) >> 9;
1409 header
.desc_offset
= 1;
1410 header
.desc_size
= 20;
1411 header
.rgd_offset
= header
.desc_offset
+ header
.desc_size
;
1412 header
.gd_offset
= header
.rgd_offset
+ gd_size
+ (gt_size
* gt_count
);
1413 header
.grain_offset
=
1414 ((header
.gd_offset
+ gd_size
+ (gt_size
* gt_count
) +
1415 header
.granularity
- 1) / header
.granularity
) *
1417 /* swap endianness for all header fields */
1418 header
.version
= cpu_to_le32(header
.version
);
1419 header
.flags
= cpu_to_le32(header
.flags
);
1420 header
.capacity
= cpu_to_le64(header
.capacity
);
1421 header
.granularity
= cpu_to_le64(header
.granularity
);
1422 header
.num_gtes_per_gte
= cpu_to_le32(header
.num_gtes_per_gte
);
1423 header
.desc_offset
= cpu_to_le64(header
.desc_offset
);
1424 header
.desc_size
= cpu_to_le64(header
.desc_size
);
1425 header
.rgd_offset
= cpu_to_le64(header
.rgd_offset
);
1426 header
.gd_offset
= cpu_to_le64(header
.gd_offset
);
1427 header
.grain_offset
= cpu_to_le64(header
.grain_offset
);
1428 header
.compressAlgorithm
= cpu_to_le16(header
.compressAlgorithm
);
1430 header
.check_bytes
[0] = 0xa;
1431 header
.check_bytes
[1] = 0x20;
1432 header
.check_bytes
[2] = 0xd;
1433 header
.check_bytes
[3] = 0xa;
1435 /* write all the data */
1436 ret
= qemu_write_full(fd
, &magic
, sizeof(magic
));
1437 if (ret
!= sizeof(magic
)) {
1441 ret
= qemu_write_full(fd
, &header
, sizeof(header
));
1442 if (ret
!= sizeof(header
)) {
1447 ret
= ftruncate(fd
, le64_to_cpu(header
.grain_offset
) << 9);
1453 /* write grain directory */
1454 lseek(fd
, le64_to_cpu(header
.rgd_offset
) << 9, SEEK_SET
);
1455 for (i
= 0, tmp
= le64_to_cpu(header
.rgd_offset
) + gd_size
;
1456 i
< gt_count
; i
++, tmp
+= gt_size
) {
1457 ret
= qemu_write_full(fd
, &tmp
, sizeof(tmp
));
1458 if (ret
!= sizeof(tmp
)) {
1464 /* write backup grain directory */
1465 lseek(fd
, le64_to_cpu(header
.gd_offset
) << 9, SEEK_SET
);
1466 for (i
= 0, tmp
= le64_to_cpu(header
.gd_offset
) + gd_size
;
1467 i
< gt_count
; i
++, tmp
+= gt_size
) {
1468 ret
= qemu_write_full(fd
, &tmp
, sizeof(tmp
));
1469 if (ret
!= sizeof(tmp
)) {
1481 static int filename_decompose(const char *filename
, char *path
, char *prefix
,
1482 char *postfix
, size_t buf_len
)
1486 if (filename
== NULL
|| !strlen(filename
)) {
1487 fprintf(stderr
, "Vmdk: no filename provided.\n");
1490 p
= strrchr(filename
, '/');
1492 p
= strrchr(filename
, '\\');
1495 p
= strrchr(filename
, ':');
1499 if (p
- filename
>= buf_len
) {
1502 pstrcpy(path
, p
- filename
+ 1, filename
);
1507 q
= strrchr(p
, '.');
1509 pstrcpy(prefix
, buf_len
, p
);
1512 if (q
- p
>= buf_len
) {
1515 pstrcpy(prefix
, q
- p
+ 1, p
);
1516 pstrcpy(postfix
, buf_len
, q
);
1521 static int vmdk_create(const char *filename
, QEMUOptionParameter
*options
)
1524 char desc
[BUF_SIZE
];
1525 int64_t total_size
= 0, filesize
;
1526 const char *adapter_type
= NULL
;
1527 const char *backing_file
= NULL
;
1528 const char *fmt
= NULL
;
1531 bool flat
, split
, compress
;
1532 char ext_desc_lines
[BUF_SIZE
] = "";
1533 char path
[PATH_MAX
], prefix
[PATH_MAX
], postfix
[PATH_MAX
];
1534 const int64_t split_size
= 0x80000000; /* VMDK has constant split size */
1535 const char *desc_extent_line
;
1536 char parent_desc_line
[BUF_SIZE
] = "";
1537 uint32_t parent_cid
= 0xffffffff;
1538 uint32_t number_heads
= 16;
1539 bool zeroed_grain
= false;
1540 const char desc_template
[] =
1541 "# Disk DescriptorFile\n"
1545 "createType=\"%s\"\n"
1548 "# Extent description\n"
1551 "# The Disk Data Base\n"
1554 "ddb.virtualHWVersion = \"%d\"\n"
1555 "ddb.geometry.cylinders = \"%" PRId64
"\"\n"
1556 "ddb.geometry.heads = \"%d\"\n"
1557 "ddb.geometry.sectors = \"63\"\n"
1558 "ddb.adapterType = \"%s\"\n";
1560 if (filename_decompose(filename
, path
, prefix
, postfix
, PATH_MAX
)) {
1563 /* Read out options */
1564 while (options
&& options
->name
) {
1565 if (!strcmp(options
->name
, BLOCK_OPT_SIZE
)) {
1566 total_size
= options
->value
.n
;
1567 } else if (!strcmp(options
->name
, BLOCK_OPT_ADAPTER_TYPE
)) {
1568 adapter_type
= options
->value
.s
;
1569 } else if (!strcmp(options
->name
, BLOCK_OPT_BACKING_FILE
)) {
1570 backing_file
= options
->value
.s
;
1571 } else if (!strcmp(options
->name
, BLOCK_OPT_COMPAT6
)) {
1572 flags
|= options
->value
.n
? BLOCK_FLAG_COMPAT6
: 0;
1573 } else if (!strcmp(options
->name
, BLOCK_OPT_SUBFMT
)) {
1574 fmt
= options
->value
.s
;
1575 } else if (!strcmp(options
->name
, BLOCK_OPT_ZEROED_GRAIN
)) {
1576 zeroed_grain
|= options
->value
.n
;
1580 if (!adapter_type
) {
1581 adapter_type
= "ide";
1582 } else if (strcmp(adapter_type
, "ide") &&
1583 strcmp(adapter_type
, "buslogic") &&
1584 strcmp(adapter_type
, "lsilogic") &&
1585 strcmp(adapter_type
, "legacyESX")) {
1586 fprintf(stderr
, "VMDK: Unknown adapter type: '%s'.\n", adapter_type
);
1589 if (strcmp(adapter_type
, "ide") != 0) {
1590 /* that's the number of heads with which vmware operates when
1591 creating, exporting, etc. vmdk files with a non-ide adapter type */
1595 /* Default format to monolithicSparse */
1596 fmt
= "monolithicSparse";
1597 } else if (strcmp(fmt
, "monolithicFlat") &&
1598 strcmp(fmt
, "monolithicSparse") &&
1599 strcmp(fmt
, "twoGbMaxExtentSparse") &&
1600 strcmp(fmt
, "twoGbMaxExtentFlat") &&
1601 strcmp(fmt
, "streamOptimized")) {
1602 fprintf(stderr
, "VMDK: Unknown subformat: %s\n", fmt
);
1605 split
= !(strcmp(fmt
, "twoGbMaxExtentFlat") &&
1606 strcmp(fmt
, "twoGbMaxExtentSparse"));
1607 flat
= !(strcmp(fmt
, "monolithicFlat") &&
1608 strcmp(fmt
, "twoGbMaxExtentFlat"));
1609 compress
= !strcmp(fmt
, "streamOptimized");
1611 desc_extent_line
= "RW %lld FLAT \"%s\" 0\n";
1613 desc_extent_line
= "RW %lld SPARSE \"%s\"\n";
1615 if (flat
&& backing_file
) {
1616 /* not supporting backing file for flat image */
1620 BlockDriverState
*bs
= bdrv_new("");
1621 ret
= bdrv_open(bs
, backing_file
, NULL
, 0, NULL
);
1626 if (strcmp(bs
->drv
->format_name
, "vmdk")) {
1630 parent_cid
= vmdk_read_cid(bs
, 0);
1632 snprintf(parent_desc_line
, sizeof(parent_desc_line
),
1633 "parentFileNameHint=\"%s\"", backing_file
);
1636 /* Create extents */
1637 filesize
= total_size
;
1638 while (filesize
> 0) {
1639 char desc_line
[BUF_SIZE
];
1640 char ext_filename
[PATH_MAX
];
1641 char desc_filename
[PATH_MAX
];
1642 int64_t size
= filesize
;
1644 if (split
&& size
> split_size
) {
1648 snprintf(desc_filename
, sizeof(desc_filename
), "%s-%c%03d%s",
1649 prefix
, flat
? 'f' : 's', ++idx
, postfix
);
1651 snprintf(desc_filename
, sizeof(desc_filename
), "%s-flat%s",
1654 snprintf(desc_filename
, sizeof(desc_filename
), "%s%s",
1657 snprintf(ext_filename
, sizeof(ext_filename
), "%s%s",
1658 path
, desc_filename
);
1660 if (vmdk_create_extent(ext_filename
, size
,
1661 flat
, compress
, zeroed_grain
)) {
1666 /* Format description line */
1667 snprintf(desc_line
, sizeof(desc_line
),
1668 desc_extent_line
, size
/ 512, desc_filename
);
1669 pstrcat(ext_desc_lines
, sizeof(ext_desc_lines
), desc_line
);
1671 /* generate descriptor file */
1672 snprintf(desc
, sizeof(desc
), desc_template
,
1673 (unsigned int)time(NULL
),
1678 (flags
& BLOCK_FLAG_COMPAT6
? 6 : 4),
1679 total_size
/ (int64_t)(63 * number_heads
* 512), number_heads
,
1681 if (split
|| flat
) {
1682 fd
= qemu_open(filename
,
1683 O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
| O_LARGEFILE
,
1686 fd
= qemu_open(filename
,
1687 O_WRONLY
| O_BINARY
| O_LARGEFILE
,
1693 /* the descriptor offset = 0x200 */
1694 if (!split
&& !flat
&& 0x200 != lseek(fd
, 0x200, SEEK_SET
)) {
1698 ret
= qemu_write_full(fd
, desc
, strlen(desc
));
1699 if (ret
!= strlen(desc
)) {
1709 static void vmdk_close(BlockDriverState
*bs
)
1711 BDRVVmdkState
*s
= bs
->opaque
;
1713 vmdk_free_extents(bs
);
1715 migrate_del_blocker(s
->migration_blocker
);
1716 error_free(s
->migration_blocker
);
1719 static coroutine_fn
int vmdk_co_flush(BlockDriverState
*bs
)
1721 BDRVVmdkState
*s
= bs
->opaque
;
1725 for (i
= 0; i
< s
->num_extents
; i
++) {
1726 err
= bdrv_co_flush(s
->extents
[i
].file
);
1734 static int64_t vmdk_get_allocated_file_size(BlockDriverState
*bs
)
1739 BDRVVmdkState
*s
= bs
->opaque
;
1741 ret
= bdrv_get_allocated_file_size(bs
->file
);
1745 for (i
= 0; i
< s
->num_extents
; i
++) {
1746 if (s
->extents
[i
].file
== bs
->file
) {
1749 r
= bdrv_get_allocated_file_size(s
->extents
[i
].file
);
1758 static int vmdk_has_zero_init(BlockDriverState
*bs
)
1761 BDRVVmdkState
*s
= bs
->opaque
;
1763 /* If has a flat extent and its underlying storage doesn't have zero init,
1765 for (i
= 0; i
< s
->num_extents
; i
++) {
1766 if (s
->extents
[i
].flat
) {
1767 if (!bdrv_has_zero_init(s
->extents
[i
].file
)) {
1775 static QEMUOptionParameter vmdk_create_options
[] = {
1777 .name
= BLOCK_OPT_SIZE
,
1779 .help
= "Virtual disk size"
1782 .name
= BLOCK_OPT_ADAPTER_TYPE
,
1784 .help
= "Virtual adapter type, can be one of "
1785 "ide (default), lsilogic, buslogic or legacyESX"
1788 .name
= BLOCK_OPT_BACKING_FILE
,
1790 .help
= "File name of a base image"
1793 .name
= BLOCK_OPT_COMPAT6
,
1795 .help
= "VMDK version 6 image"
1798 .name
= BLOCK_OPT_SUBFMT
,
1801 "VMDK flat extent format, can be one of "
1802 "{monolithicSparse (default) | monolithicFlat | twoGbMaxExtentSparse | twoGbMaxExtentFlat | streamOptimized} "
1805 .name
= BLOCK_OPT_ZEROED_GRAIN
,
1807 .help
= "Enable efficient zero writes using the zeroed-grain GTE feature"
1812 static BlockDriver bdrv_vmdk
= {
1813 .format_name
= "vmdk",
1814 .instance_size
= sizeof(BDRVVmdkState
),
1815 .bdrv_probe
= vmdk_probe
,
1816 .bdrv_open
= vmdk_open
,
1817 .bdrv_reopen_prepare
= vmdk_reopen_prepare
,
1818 .bdrv_read
= vmdk_co_read
,
1819 .bdrv_write
= vmdk_co_write
,
1820 .bdrv_co_write_zeroes
= vmdk_co_write_zeroes
,
1821 .bdrv_close
= vmdk_close
,
1822 .bdrv_create
= vmdk_create
,
1823 .bdrv_co_flush_to_disk
= vmdk_co_flush
,
1824 .bdrv_co_is_allocated
= vmdk_co_is_allocated
,
1825 .bdrv_get_allocated_file_size
= vmdk_get_allocated_file_size
,
1826 .bdrv_has_zero_init
= vmdk_has_zero_init
,
1828 .create_options
= vmdk_create_options
,
1831 static void bdrv_vmdk_init(void)
1833 bdrv_register(&bdrv_vmdk
);
1836 block_init(bdrv_vmdk_init
);