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 /* Number of GrainTableEntries per GrainTable */
75 uint32_t num_gtes_per_gt
;
78 uint64_t grain_offset
;
81 uint16_t compressAlgorithm
;
82 } QEMU_PACKED VMDK4Header
;
84 #define L2_CACHE_SIZE 16
86 typedef struct VmdkExtent
{
87 BlockDriverState
*file
;
95 int64_t flat_start_offset
;
96 int64_t l1_table_offset
;
97 int64_t l1_backup_table_offset
;
99 uint32_t *l1_backup_table
;
100 unsigned int l1_size
;
101 uint32_t l1_entry_sectors
;
103 unsigned int l2_size
;
105 uint32_t l2_cache_offsets
[L2_CACHE_SIZE
];
106 uint32_t l2_cache_counts
[L2_CACHE_SIZE
];
108 int64_t cluster_sectors
;
111 typedef struct BDRVVmdkState
{
113 uint64_t desc_offset
;
117 /* Extent array with num_extents entries, ascend ordered by address */
119 Error
*migration_blocker
;
122 typedef struct VmdkMetaData
{
124 unsigned int l1_index
;
125 unsigned int l2_index
;
126 unsigned int l2_offset
;
128 uint32_t *l2_cache_entry
;
131 typedef struct VmdkGrainMarker
{
135 } QEMU_PACKED VmdkGrainMarker
;
138 MARKER_END_OF_STREAM
= 0,
139 MARKER_GRAIN_TABLE
= 1,
140 MARKER_GRAIN_DIRECTORY
= 2,
144 static int vmdk_probe(const uint8_t *buf
, int buf_size
, const char *filename
)
151 magic
= be32_to_cpu(*(uint32_t *)buf
);
152 if (magic
== VMDK3_MAGIC
||
153 magic
== VMDK4_MAGIC
) {
156 const char *p
= (const char *)buf
;
157 const char *end
= p
+ buf_size
;
160 /* skip comment line */
161 while (p
< end
&& *p
!= '\n') {
168 while (p
< end
&& *p
== ' ') {
171 /* skip '\r' if windows line endings used. */
172 if (p
< end
&& *p
== '\r') {
175 /* only accept blank lines before 'version=' line */
176 if (p
== end
|| *p
!= '\n') {
182 if (end
- p
>= strlen("version=X\n")) {
183 if (strncmp("version=1\n", p
, strlen("version=1\n")) == 0 ||
184 strncmp("version=2\n", p
, strlen("version=2\n")) == 0) {
188 if (end
- p
>= strlen("version=X\r\n")) {
189 if (strncmp("version=1\r\n", p
, strlen("version=1\r\n")) == 0 ||
190 strncmp("version=2\r\n", p
, strlen("version=2\r\n")) == 0) {
202 #define SECTOR_SIZE 512
203 #define DESC_SIZE (20 * SECTOR_SIZE) /* 20 sectors of 512 bytes each */
204 #define BUF_SIZE 4096
205 #define HEADER_SIZE 512 /* first sector of 512 bytes */
207 static void vmdk_free_extents(BlockDriverState
*bs
)
210 BDRVVmdkState
*s
= bs
->opaque
;
213 for (i
= 0; i
< s
->num_extents
; i
++) {
217 g_free(e
->l1_backup_table
);
218 if (e
->file
!= bs
->file
) {
225 static void vmdk_free_last_extent(BlockDriverState
*bs
)
227 BDRVVmdkState
*s
= bs
->opaque
;
229 if (s
->num_extents
== 0) {
233 s
->extents
= g_realloc(s
->extents
, s
->num_extents
* sizeof(VmdkExtent
));
236 static uint32_t vmdk_read_cid(BlockDriverState
*bs
, int parent
)
238 char desc
[DESC_SIZE
];
239 uint32_t cid
= 0xffffffff;
240 const char *p_name
, *cid_str
;
242 BDRVVmdkState
*s
= bs
->opaque
;
245 ret
= bdrv_pread(bs
->file
, s
->desc_offset
, desc
, DESC_SIZE
);
251 cid_str
= "parentCID";
252 cid_str_size
= sizeof("parentCID");
255 cid_str_size
= sizeof("CID");
258 desc
[DESC_SIZE
- 1] = '\0';
259 p_name
= strstr(desc
, cid_str
);
260 if (p_name
!= NULL
) {
261 p_name
+= cid_str_size
;
262 sscanf(p_name
, "%x", &cid
);
268 static int vmdk_write_cid(BlockDriverState
*bs
, uint32_t cid
)
270 char desc
[DESC_SIZE
], tmp_desc
[DESC_SIZE
];
271 char *p_name
, *tmp_str
;
272 BDRVVmdkState
*s
= bs
->opaque
;
275 ret
= bdrv_pread(bs
->file
, s
->desc_offset
, desc
, DESC_SIZE
);
280 desc
[DESC_SIZE
- 1] = '\0';
281 tmp_str
= strstr(desc
, "parentCID");
282 if (tmp_str
== NULL
) {
286 pstrcpy(tmp_desc
, sizeof(tmp_desc
), tmp_str
);
287 p_name
= strstr(desc
, "CID");
288 if (p_name
!= NULL
) {
289 p_name
+= sizeof("CID");
290 snprintf(p_name
, sizeof(desc
) - (p_name
- desc
), "%x\n", cid
);
291 pstrcat(desc
, sizeof(desc
), tmp_desc
);
294 ret
= bdrv_pwrite_sync(bs
->file
, s
->desc_offset
, desc
, DESC_SIZE
);
302 static int vmdk_is_cid_valid(BlockDriverState
*bs
)
305 BDRVVmdkState
*s
= bs
->opaque
;
306 BlockDriverState
*p_bs
= bs
->backing_hd
;
310 cur_pcid
= vmdk_read_cid(p_bs
, 0);
311 if (s
->parent_cid
!= cur_pcid
) {
321 /* Queue extents, if any, for reopen() */
322 static int vmdk_reopen_prepare(BDRVReopenState
*state
,
323 BlockReopenQueue
*queue
, Error
**errp
)
330 assert(state
!= NULL
);
331 assert(state
->bs
!= NULL
);
334 error_setg(errp
, "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
,
397 BDRVVmdkState
*s
= bs
->opaque
;
399 if (cluster_sectors
> 0x200000) {
400 /* 0x200000 * 512Bytes = 1GB for one cluster is unrealistic */
401 error_setg(errp
, "Invalid granularity, image may be corrupt");
404 if (l1_size
> 512 * 1024 * 1024) {
405 /* Although with big capacity and small l1_entry_sectors, we can get a
406 * big l1_size, we don't want unbounded value to allocate the table.
407 * Limit it to 512M, which is 16PB for default cluster and L2 table
409 error_setg(errp
, "L1 size too big");
413 s
->extents
= g_realloc(s
->extents
,
414 (s
->num_extents
+ 1) * sizeof(VmdkExtent
));
415 extent
= &s
->extents
[s
->num_extents
];
418 memset(extent
, 0, sizeof(VmdkExtent
));
421 extent
->sectors
= sectors
;
422 extent
->l1_table_offset
= l1_offset
;
423 extent
->l1_backup_table_offset
= l1_backup_offset
;
424 extent
->l1_size
= l1_size
;
425 extent
->l1_entry_sectors
= l2_size
* cluster_sectors
;
426 extent
->l2_size
= l2_size
;
427 extent
->cluster_sectors
= flat
? sectors
: cluster_sectors
;
429 if (s
->num_extents
> 1) {
430 extent
->end_sector
= (*(extent
- 1)).end_sector
+ extent
->sectors
;
432 extent
->end_sector
= extent
->sectors
;
434 bs
->total_sectors
= extent
->end_sector
;
436 *new_extent
= extent
;
441 static int vmdk_init_tables(BlockDriverState
*bs
, VmdkExtent
*extent
,
447 /* read the L1 table */
448 l1_size
= extent
->l1_size
* sizeof(uint32_t);
449 extent
->l1_table
= g_malloc(l1_size
);
450 ret
= bdrv_pread(extent
->file
,
451 extent
->l1_table_offset
,
455 error_setg_errno(errp
, -ret
,
456 "Could not read l1 table from extent '%s'",
457 extent
->file
->filename
);
460 for (i
= 0; i
< extent
->l1_size
; i
++) {
461 le32_to_cpus(&extent
->l1_table
[i
]);
464 if (extent
->l1_backup_table_offset
) {
465 extent
->l1_backup_table
= g_malloc(l1_size
);
466 ret
= bdrv_pread(extent
->file
,
467 extent
->l1_backup_table_offset
,
468 extent
->l1_backup_table
,
471 error_setg_errno(errp
, -ret
,
472 "Could not read l1 backup table from extent '%s'",
473 extent
->file
->filename
);
476 for (i
= 0; i
< extent
->l1_size
; i
++) {
477 le32_to_cpus(&extent
->l1_backup_table
[i
]);
482 g_malloc(extent
->l2_size
* L2_CACHE_SIZE
* sizeof(uint32_t));
485 g_free(extent
->l1_backup_table
);
487 g_free(extent
->l1_table
);
491 static int vmdk_open_vmfs_sparse(BlockDriverState
*bs
,
492 BlockDriverState
*file
,
493 int flags
, Error
**errp
)
500 ret
= bdrv_pread(file
, sizeof(magic
), &header
, sizeof(header
));
502 error_setg_errno(errp
, -ret
,
503 "Could not read header from file '%s'",
507 ret
= vmdk_add_extent(bs
, file
, false,
508 le32_to_cpu(header
.disk_sectors
),
509 le32_to_cpu(header
.l1dir_offset
) << 9,
511 le32_to_cpu(header
.l1dir_size
),
513 le32_to_cpu(header
.granularity
),
519 ret
= vmdk_init_tables(bs
, extent
, errp
);
521 /* free extent allocated by vmdk_add_extent */
522 vmdk_free_last_extent(bs
);
527 static int vmdk_open_desc_file(BlockDriverState
*bs
, int flags
,
528 uint64_t desc_offset
, Error
**errp
);
530 static int vmdk_open_vmdk4(BlockDriverState
*bs
,
531 BlockDriverState
*file
,
532 int flags
, Error
**errp
)
536 uint32_t l1_size
, l1_entry_sectors
;
539 int64_t l1_backup_offset
= 0;
541 ret
= bdrv_pread(file
, sizeof(magic
), &header
, sizeof(header
));
543 error_setg_errno(errp
, -ret
,
544 "Could not read header from file '%s'",
547 if (header
.capacity
== 0) {
548 uint64_t desc_offset
= le64_to_cpu(header
.desc_offset
);
550 return vmdk_open_desc_file(bs
, flags
, desc_offset
<< 9, errp
);
554 if (le64_to_cpu(header
.gd_offset
) == VMDK4_GD_AT_END
) {
556 * The footer takes precedence over the header, so read it in. The
557 * footer starts at offset -1024 from the end: One sector for the
558 * footer, and another one for the end-of-stream marker.
565 uint8_t pad
[512 - 16];
566 } QEMU_PACKED footer_marker
;
570 uint8_t pad
[512 - 4 - sizeof(VMDK4Header
)];
576 uint8_t pad
[512 - 16];
577 } QEMU_PACKED eos_marker
;
578 } QEMU_PACKED footer
;
580 ret
= bdrv_pread(file
,
581 bs
->file
->total_sectors
* 512 - 1536,
582 &footer
, sizeof(footer
));
587 /* Some sanity checks for the footer */
588 if (be32_to_cpu(footer
.magic
) != VMDK4_MAGIC
||
589 le32_to_cpu(footer
.footer_marker
.size
) != 0 ||
590 le32_to_cpu(footer
.footer_marker
.type
) != MARKER_FOOTER
||
591 le64_to_cpu(footer
.eos_marker
.val
) != 0 ||
592 le32_to_cpu(footer
.eos_marker
.size
) != 0 ||
593 le32_to_cpu(footer
.eos_marker
.type
) != MARKER_END_OF_STREAM
)
598 header
= footer
.header
;
601 if (le32_to_cpu(header
.version
) >= 3) {
603 snprintf(buf
, sizeof(buf
), "VMDK version %d",
604 le32_to_cpu(header
.version
));
605 qerror_report(QERR_UNKNOWN_BLOCK_FORMAT_FEATURE
,
606 bs
->device_name
, "vmdk", buf
);
610 if (le32_to_cpu(header
.num_gtes_per_gt
) > 512) {
611 error_report("L2 table size too big");
615 l1_entry_sectors
= le32_to_cpu(header
.num_gtes_per_gt
)
616 * le64_to_cpu(header
.granularity
);
617 if (l1_entry_sectors
== 0) {
620 l1_size
= (le64_to_cpu(header
.capacity
) + l1_entry_sectors
- 1)
622 if (le32_to_cpu(header
.flags
) & VMDK4_FLAG_RGD
) {
623 l1_backup_offset
= le64_to_cpu(header
.rgd_offset
) << 9;
625 ret
= vmdk_add_extent(bs
, file
, false,
626 le64_to_cpu(header
.capacity
),
627 le64_to_cpu(header
.gd_offset
) << 9,
630 le32_to_cpu(header
.num_gtes_per_gt
),
631 le64_to_cpu(header
.granularity
),
638 le16_to_cpu(header
.compressAlgorithm
) == VMDK4_COMPRESSION_DEFLATE
;
639 extent
->has_marker
= le32_to_cpu(header
.flags
) & VMDK4_FLAG_MARKER
;
640 extent
->version
= le32_to_cpu(header
.version
);
641 extent
->has_zero_grain
= le32_to_cpu(header
.flags
) & VMDK4_FLAG_ZERO_GRAIN
;
642 ret
= vmdk_init_tables(bs
, extent
, errp
);
644 /* free extent allocated by vmdk_add_extent */
645 vmdk_free_last_extent(bs
);
650 /* find an option value out of descriptor file */
651 static int vmdk_parse_description(const char *desc
, const char *opt_name
,
652 char *buf
, int buf_size
)
654 char *opt_pos
, *opt_end
;
655 const char *end
= desc
+ strlen(desc
);
657 opt_pos
= strstr(desc
, opt_name
);
661 /* Skip "=\"" following opt_name */
662 opt_pos
+= strlen(opt_name
) + 2;
663 if (opt_pos
>= end
) {
667 while (opt_end
< end
&& *opt_end
!= '"') {
670 if (opt_end
== end
|| buf_size
< opt_end
- opt_pos
+ 1) {
673 pstrcpy(buf
, opt_end
- opt_pos
+ 1, opt_pos
);
677 /* Open an extent file and append to bs array */
678 static int vmdk_open_sparse(BlockDriverState
*bs
,
679 BlockDriverState
*file
,
680 int flags
, Error
**errp
)
684 if (bdrv_pread(file
, 0, &magic
, sizeof(magic
)) != sizeof(magic
)) {
688 magic
= be32_to_cpu(magic
);
691 return vmdk_open_vmfs_sparse(bs
, file
, flags
, errp
);
694 return vmdk_open_vmdk4(bs
, file
, flags
, errp
);
702 static int vmdk_parse_extents(const char *desc
, BlockDriverState
*bs
,
703 const char *desc_file_path
, Error
**errp
)
709 const char *p
= desc
;
712 char extent_path
[PATH_MAX
];
713 BlockDriverState
*extent_file
;
716 /* parse extent line:
717 * RW [size in sectors] FLAT "file-name.vmdk" OFFSET
719 * RW [size in sectors] SPARSE "file-name.vmdk"
722 ret
= sscanf(p
, "%10s %" SCNd64
" %10s \"%511[^\n\r\"]\" %" SCNd64
,
723 access
, §ors
, type
, fname
, &flat_offset
);
724 if (ret
< 4 || strcmp(access
, "RW")) {
726 } else if (!strcmp(type
, "FLAT")) {
727 if (ret
!= 5 || flat_offset
< 0) {
728 error_setg(errp
, "Invalid extent lines: \n%s", p
);
731 } else if (ret
!= 4) {
732 error_setg(errp
, "Invalid extent lines: \n%s", p
);
737 (strcmp(type
, "FLAT") && strcmp(type
, "SPARSE") &&
738 strcmp(type
, "VMFS") && strcmp(type
, "VMFSSPARSE")) ||
739 (strcmp(access
, "RW"))) {
743 path_combine(extent_path
, sizeof(extent_path
),
744 desc_file_path
, fname
);
745 ret
= bdrv_file_open(&extent_file
, extent_path
, NULL
, bs
->open_flags
,
751 /* save to extents array */
752 if (!strcmp(type
, "FLAT") || !strcmp(type
, "VMFS")) {
756 ret
= vmdk_add_extent(bs
, extent_file
, true, sectors
,
757 0, 0, 0, 0, 0, &extent
, errp
);
761 extent
->flat_start_offset
= flat_offset
<< 9;
762 } else if (!strcmp(type
, "SPARSE") || !strcmp(type
, "VMFSSPARSE")) {
763 /* SPARSE extent and VMFSSPARSE extent are both "COWD" sparse file*/
764 ret
= vmdk_open_sparse(bs
, extent_file
, bs
->open_flags
, errp
);
766 bdrv_unref(extent_file
);
770 error_setg(errp
, "Unsupported extent type '%s'", type
);
774 /* move to next line */
775 while (*p
&& *p
!= '\n') {
783 static int vmdk_open_desc_file(BlockDriverState
*bs
, int flags
,
784 uint64_t desc_offset
, Error
**errp
)
789 BDRVVmdkState
*s
= bs
->opaque
;
792 size
= bdrv_getlength(bs
->file
);
797 size
= MIN(size
, 1 << 20); /* avoid unbounded allocation */
798 buf
= g_malloc0(size
+ 1);
800 ret
= bdrv_pread(bs
->file
, desc_offset
, buf
, size
);
804 if (vmdk_parse_description(buf
, "createType", ct
, sizeof(ct
))) {
808 if (strcmp(ct
, "monolithicFlat") &&
809 strcmp(ct
, "vmfs") &&
810 strcmp(ct
, "vmfsSparse") &&
811 strcmp(ct
, "twoGbMaxExtentSparse") &&
812 strcmp(ct
, "twoGbMaxExtentFlat")) {
813 error_setg(errp
, "Unsupported image type '%s'", ct
);
818 ret
= vmdk_parse_extents(buf
, bs
, bs
->file
->filename
, errp
);
824 static int vmdk_open(BlockDriverState
*bs
, QDict
*options
, int flags
,
828 BDRVVmdkState
*s
= bs
->opaque
;
830 if (vmdk_open_sparse(bs
, bs
->file
, flags
, errp
) == 0) {
831 s
->desc_offset
= 0x200;
833 ret
= vmdk_open_desc_file(bs
, flags
, 0, errp
);
838 /* try to open parent images, if exist */
839 ret
= vmdk_parent_open(bs
);
843 s
->parent_cid
= vmdk_read_cid(bs
, 1);
844 qemu_co_mutex_init(&s
->lock
);
846 /* Disable migration when VMDK images are used */
847 error_set(&s
->migration_blocker
,
848 QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
849 "vmdk", bs
->device_name
, "live migration");
850 migrate_add_blocker(s
->migration_blocker
);
855 vmdk_free_extents(bs
);
859 static int get_whole_cluster(BlockDriverState
*bs
,
861 uint64_t cluster_offset
,
866 uint8_t *whole_grain
= NULL
;
868 /* we will be here if it's first write on non-exist grain(cluster).
869 * try to read from parent image, if exist */
870 if (bs
->backing_hd
) {
872 qemu_blockalign(bs
, extent
->cluster_sectors
<< BDRV_SECTOR_BITS
);
873 if (!vmdk_is_cid_valid(bs
)) {
878 /* floor offset to cluster */
879 offset
-= offset
% (extent
->cluster_sectors
* 512);
880 ret
= bdrv_read(bs
->backing_hd
, offset
>> 9, whole_grain
,
881 extent
->cluster_sectors
);
887 /* Write grain only into the active image */
888 ret
= bdrv_write(extent
->file
, cluster_offset
, whole_grain
,
889 extent
->cluster_sectors
);
896 qemu_vfree(whole_grain
);
900 static int vmdk_L2update(VmdkExtent
*extent
, VmdkMetaData
*m_data
)
903 QEMU_BUILD_BUG_ON(sizeof(offset
) != sizeof(m_data
->offset
));
904 offset
= cpu_to_le32(m_data
->offset
);
905 /* update L2 table */
906 if (bdrv_pwrite_sync(
908 ((int64_t)m_data
->l2_offset
* 512)
909 + (m_data
->l2_index
* sizeof(m_data
->offset
)),
910 &offset
, sizeof(offset
)) < 0) {
913 /* update backup L2 table */
914 if (extent
->l1_backup_table_offset
!= 0) {
915 m_data
->l2_offset
= extent
->l1_backup_table
[m_data
->l1_index
];
916 if (bdrv_pwrite_sync(
918 ((int64_t)m_data
->l2_offset
* 512)
919 + (m_data
->l2_index
* sizeof(m_data
->offset
)),
920 &offset
, sizeof(offset
)) < 0) {
924 if (m_data
->l2_cache_entry
) {
925 *m_data
->l2_cache_entry
= offset
;
931 static int get_cluster_offset(BlockDriverState
*bs
,
933 VmdkMetaData
*m_data
,
936 uint64_t *cluster_offset
)
938 unsigned int l1_index
, l2_offset
, l2_index
;
940 uint32_t min_count
, *l2_table
;
947 *cluster_offset
= extent
->flat_start_offset
;
951 offset
-= (extent
->end_sector
- extent
->sectors
) * SECTOR_SIZE
;
952 l1_index
= (offset
>> 9) / extent
->l1_entry_sectors
;
953 if (l1_index
>= extent
->l1_size
) {
956 l2_offset
= extent
->l1_table
[l1_index
];
960 for (i
= 0; i
< L2_CACHE_SIZE
; i
++) {
961 if (l2_offset
== extent
->l2_cache_offsets
[i
]) {
962 /* increment the hit count */
963 if (++extent
->l2_cache_counts
[i
] == 0xffffffff) {
964 for (j
= 0; j
< L2_CACHE_SIZE
; j
++) {
965 extent
->l2_cache_counts
[j
] >>= 1;
968 l2_table
= extent
->l2_cache
+ (i
* extent
->l2_size
);
972 /* not found: load a new entry in the least used one */
974 min_count
= 0xffffffff;
975 for (i
= 0; i
< L2_CACHE_SIZE
; i
++) {
976 if (extent
->l2_cache_counts
[i
] < min_count
) {
977 min_count
= extent
->l2_cache_counts
[i
];
981 l2_table
= extent
->l2_cache
+ (min_index
* extent
->l2_size
);
984 (int64_t)l2_offset
* 512,
986 extent
->l2_size
* sizeof(uint32_t)
987 ) != extent
->l2_size
* sizeof(uint32_t)) {
991 extent
->l2_cache_offsets
[min_index
] = l2_offset
;
992 extent
->l2_cache_counts
[min_index
] = 1;
994 l2_index
= ((offset
>> 9) / extent
->cluster_sectors
) % extent
->l2_size
;
995 *cluster_offset
= le32_to_cpu(l2_table
[l2_index
]);
999 m_data
->l1_index
= l1_index
;
1000 m_data
->l2_index
= l2_index
;
1001 m_data
->offset
= *cluster_offset
;
1002 m_data
->l2_offset
= l2_offset
;
1003 m_data
->l2_cache_entry
= &l2_table
[l2_index
];
1005 if (extent
->has_zero_grain
&& *cluster_offset
== VMDK_GTE_ZEROED
) {
1009 if (!*cluster_offset
|| zeroed
) {
1011 return zeroed
? VMDK_ZEROED
: VMDK_UNALLOC
;
1014 /* Avoid the L2 tables update for the images that have snapshots. */
1015 *cluster_offset
= bdrv_getlength(extent
->file
);
1016 if (!extent
->compressed
) {
1019 *cluster_offset
+ (extent
->cluster_sectors
<< 9)
1023 *cluster_offset
>>= 9;
1024 l2_table
[l2_index
] = cpu_to_le32(*cluster_offset
);
1026 /* First of all we write grain itself, to avoid race condition
1027 * that may to corrupt the image.
1028 * This problem may occur because of insufficient space on host disk
1029 * or inappropriate VM shutdown.
1031 if (get_whole_cluster(
1032 bs
, extent
, *cluster_offset
, offset
, allocate
) == -1) {
1037 m_data
->offset
= *cluster_offset
;
1040 *cluster_offset
<<= 9;
1044 static VmdkExtent
*find_extent(BDRVVmdkState
*s
,
1045 int64_t sector_num
, VmdkExtent
*start_hint
)
1047 VmdkExtent
*extent
= start_hint
;
1050 extent
= &s
->extents
[0];
1052 while (extent
< &s
->extents
[s
->num_extents
]) {
1053 if (sector_num
< extent
->end_sector
) {
1061 static int64_t coroutine_fn
vmdk_co_get_block_status(BlockDriverState
*bs
,
1062 int64_t sector_num
, int nb_sectors
, int *pnum
)
1064 BDRVVmdkState
*s
= bs
->opaque
;
1065 int64_t index_in_cluster
, n
, ret
;
1069 extent
= find_extent(s
, sector_num
, NULL
);
1073 qemu_co_mutex_lock(&s
->lock
);
1074 ret
= get_cluster_offset(bs
, extent
, NULL
,
1075 sector_num
* 512, 0, &offset
);
1076 qemu_co_mutex_unlock(&s
->lock
);
1086 ret
= BDRV_BLOCK_ZERO
;
1089 ret
= BDRV_BLOCK_DATA
;
1090 if (extent
->file
== bs
->file
) {
1091 ret
|= BDRV_BLOCK_OFFSET_VALID
| offset
;
1097 index_in_cluster
= sector_num
% extent
->cluster_sectors
;
1098 n
= extent
->cluster_sectors
- index_in_cluster
;
1099 if (n
> nb_sectors
) {
1106 static int vmdk_write_extent(VmdkExtent
*extent
, int64_t cluster_offset
,
1107 int64_t offset_in_cluster
, const uint8_t *buf
,
1108 int nb_sectors
, int64_t sector_num
)
1111 VmdkGrainMarker
*data
= NULL
;
1113 const uint8_t *write_buf
= buf
;
1114 int write_len
= nb_sectors
* 512;
1116 if (extent
->compressed
) {
1117 if (!extent
->has_marker
) {
1121 buf_len
= (extent
->cluster_sectors
<< 9) * 2;
1122 data
= g_malloc(buf_len
+ sizeof(VmdkGrainMarker
));
1123 if (compress(data
->data
, &buf_len
, buf
, nb_sectors
<< 9) != Z_OK
||
1128 data
->lba
= sector_num
;
1129 data
->size
= buf_len
;
1130 write_buf
= (uint8_t *)data
;
1131 write_len
= buf_len
+ sizeof(VmdkGrainMarker
);
1133 ret
= bdrv_pwrite(extent
->file
,
1134 cluster_offset
+ offset_in_cluster
,
1137 if (ret
!= write_len
) {
1138 ret
= ret
< 0 ? ret
: -EIO
;
1147 static int vmdk_read_extent(VmdkExtent
*extent
, int64_t cluster_offset
,
1148 int64_t offset_in_cluster
, uint8_t *buf
,
1152 int cluster_bytes
, buf_bytes
;
1153 uint8_t *cluster_buf
, *compressed_data
;
1154 uint8_t *uncomp_buf
;
1156 VmdkGrainMarker
*marker
;
1160 if (!extent
->compressed
) {
1161 ret
= bdrv_pread(extent
->file
,
1162 cluster_offset
+ offset_in_cluster
,
1163 buf
, nb_sectors
* 512);
1164 if (ret
== nb_sectors
* 512) {
1170 cluster_bytes
= extent
->cluster_sectors
* 512;
1171 /* Read two clusters in case GrainMarker + compressed data > one cluster */
1172 buf_bytes
= cluster_bytes
* 2;
1173 cluster_buf
= g_malloc(buf_bytes
);
1174 uncomp_buf
= g_malloc(cluster_bytes
);
1175 ret
= bdrv_pread(extent
->file
,
1177 cluster_buf
, buf_bytes
);
1181 compressed_data
= cluster_buf
;
1182 buf_len
= cluster_bytes
;
1183 data_len
= cluster_bytes
;
1184 if (extent
->has_marker
) {
1185 marker
= (VmdkGrainMarker
*)cluster_buf
;
1186 compressed_data
= marker
->data
;
1187 data_len
= le32_to_cpu(marker
->size
);
1189 if (!data_len
|| data_len
> buf_bytes
) {
1193 ret
= uncompress(uncomp_buf
, &buf_len
, compressed_data
, data_len
);
1199 if (offset_in_cluster
< 0 ||
1200 offset_in_cluster
+ nb_sectors
* 512 > buf_len
) {
1204 memcpy(buf
, uncomp_buf
+ offset_in_cluster
, nb_sectors
* 512);
1209 g_free(cluster_buf
);
1213 static int vmdk_read(BlockDriverState
*bs
, int64_t sector_num
,
1214 uint8_t *buf
, int nb_sectors
)
1216 BDRVVmdkState
*s
= bs
->opaque
;
1218 uint64_t n
, index_in_cluster
;
1219 uint64_t extent_begin_sector
, extent_relative_sector_num
;
1220 VmdkExtent
*extent
= NULL
;
1221 uint64_t cluster_offset
;
1223 while (nb_sectors
> 0) {
1224 extent
= find_extent(s
, sector_num
, extent
);
1228 ret
= get_cluster_offset(
1230 sector_num
<< 9, 0, &cluster_offset
);
1231 extent_begin_sector
= extent
->end_sector
- extent
->sectors
;
1232 extent_relative_sector_num
= sector_num
- extent_begin_sector
;
1233 index_in_cluster
= extent_relative_sector_num
% extent
->cluster_sectors
;
1234 n
= extent
->cluster_sectors
- index_in_cluster
;
1235 if (n
> nb_sectors
) {
1238 if (ret
!= VMDK_OK
) {
1239 /* if not allocated, try to read from parent image, if exist */
1240 if (bs
->backing_hd
&& ret
!= VMDK_ZEROED
) {
1241 if (!vmdk_is_cid_valid(bs
)) {
1244 ret
= bdrv_read(bs
->backing_hd
, sector_num
, buf
, n
);
1249 memset(buf
, 0, 512 * n
);
1252 ret
= vmdk_read_extent(extent
,
1253 cluster_offset
, index_in_cluster
* 512,
1266 static coroutine_fn
int vmdk_co_read(BlockDriverState
*bs
, int64_t sector_num
,
1267 uint8_t *buf
, int nb_sectors
)
1270 BDRVVmdkState
*s
= bs
->opaque
;
1271 qemu_co_mutex_lock(&s
->lock
);
1272 ret
= vmdk_read(bs
, sector_num
, buf
, nb_sectors
);
1273 qemu_co_mutex_unlock(&s
->lock
);
1279 * @zeroed: buf is ignored (data is zero), use zeroed_grain GTE feature
1280 * if possible, otherwise return -ENOTSUP.
1281 * @zero_dry_run: used for zeroed == true only, don't update L2 table, just try
1282 * with each cluster. By dry run we can find if the zero write
1283 * is possible without modifying image data.
1285 * Returns: error code with 0 for success.
1287 static int vmdk_write(BlockDriverState
*bs
, int64_t sector_num
,
1288 const uint8_t *buf
, int nb_sectors
,
1289 bool zeroed
, bool zero_dry_run
)
1291 BDRVVmdkState
*s
= bs
->opaque
;
1292 VmdkExtent
*extent
= NULL
;
1294 int64_t index_in_cluster
;
1295 uint64_t extent_begin_sector
, extent_relative_sector_num
;
1296 uint64_t cluster_offset
;
1297 VmdkMetaData m_data
;
1299 if (sector_num
> bs
->total_sectors
) {
1300 error_report("Wrong offset: sector_num=0x%" PRIx64
1301 " total_sectors=0x%" PRIx64
"\n",
1302 sector_num
, bs
->total_sectors
);
1306 while (nb_sectors
> 0) {
1307 extent
= find_extent(s
, sector_num
, extent
);
1311 ret
= get_cluster_offset(
1315 sector_num
<< 9, !extent
->compressed
,
1317 if (extent
->compressed
) {
1318 if (ret
== VMDK_OK
) {
1319 /* Refuse write to allocated cluster for streamOptimized */
1320 error_report("Could not write to allocated cluster"
1321 " for streamOptimized");
1325 ret
= get_cluster_offset(
1333 if (ret
== VMDK_ERROR
) {
1336 extent_begin_sector
= extent
->end_sector
- extent
->sectors
;
1337 extent_relative_sector_num
= sector_num
- extent_begin_sector
;
1338 index_in_cluster
= extent_relative_sector_num
% extent
->cluster_sectors
;
1339 n
= extent
->cluster_sectors
- index_in_cluster
;
1340 if (n
> nb_sectors
) {
1344 /* Do zeroed write, buf is ignored */
1345 if (extent
->has_zero_grain
&&
1346 index_in_cluster
== 0 &&
1347 n
>= extent
->cluster_sectors
) {
1348 n
= extent
->cluster_sectors
;
1349 if (!zero_dry_run
) {
1350 m_data
.offset
= VMDK_GTE_ZEROED
;
1351 /* update L2 tables */
1352 if (vmdk_L2update(extent
, &m_data
) != VMDK_OK
) {
1360 ret
= vmdk_write_extent(extent
,
1361 cluster_offset
, index_in_cluster
* 512,
1362 buf
, n
, sector_num
);
1367 /* update L2 tables */
1368 if (vmdk_L2update(extent
, &m_data
) != VMDK_OK
) {
1377 /* update CID on the first write every time the virtual disk is
1379 if (!s
->cid_updated
) {
1380 ret
= vmdk_write_cid(bs
, time(NULL
));
1384 s
->cid_updated
= true;
1390 static coroutine_fn
int vmdk_co_write(BlockDriverState
*bs
, int64_t sector_num
,
1391 const uint8_t *buf
, int nb_sectors
)
1394 BDRVVmdkState
*s
= bs
->opaque
;
1395 qemu_co_mutex_lock(&s
->lock
);
1396 ret
= vmdk_write(bs
, sector_num
, buf
, nb_sectors
, false, false);
1397 qemu_co_mutex_unlock(&s
->lock
);
1401 static int coroutine_fn
vmdk_co_write_zeroes(BlockDriverState
*bs
,
1406 BDRVVmdkState
*s
= bs
->opaque
;
1407 qemu_co_mutex_lock(&s
->lock
);
1408 /* write zeroes could fail if sectors not aligned to cluster, test it with
1409 * dry_run == true before really updating image */
1410 ret
= vmdk_write(bs
, sector_num
, NULL
, nb_sectors
, true, true);
1412 ret
= vmdk_write(bs
, sector_num
, NULL
, nb_sectors
, true, false);
1414 qemu_co_mutex_unlock(&s
->lock
);
1418 static int vmdk_create_extent(const char *filename
, int64_t filesize
,
1419 bool flat
, bool compress
, bool zeroed_grain
)
1424 uint32_t tmp
, magic
, grains
, gd_size
, gt_size
, gt_count
;
1426 fd
= qemu_open(filename
,
1427 O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
| O_LARGEFILE
,
1433 ret
= ftruncate(fd
, filesize
);
1439 magic
= cpu_to_be32(VMDK4_MAGIC
);
1440 memset(&header
, 0, sizeof(header
));
1441 header
.version
= zeroed_grain
? 2 : 1;
1442 header
.flags
= VMDK4_FLAG_RGD
| VMDK4_FLAG_NL_DETECT
1443 | (compress
? VMDK4_FLAG_COMPRESS
| VMDK4_FLAG_MARKER
: 0)
1444 | (zeroed_grain
? VMDK4_FLAG_ZERO_GRAIN
: 0);
1445 header
.compressAlgorithm
= compress
? VMDK4_COMPRESSION_DEFLATE
: 0;
1446 header
.capacity
= filesize
/ 512;
1447 header
.granularity
= 128;
1448 header
.num_gtes_per_gt
= 512;
1450 grains
= (filesize
/ 512 + header
.granularity
- 1) / header
.granularity
;
1451 gt_size
= ((header
.num_gtes_per_gt
* sizeof(uint32_t)) + 511) >> 9;
1453 (grains
+ header
.num_gtes_per_gt
- 1) / header
.num_gtes_per_gt
;
1454 gd_size
= (gt_count
* sizeof(uint32_t) + 511) >> 9;
1456 header
.desc_offset
= 1;
1457 header
.desc_size
= 20;
1458 header
.rgd_offset
= header
.desc_offset
+ header
.desc_size
;
1459 header
.gd_offset
= header
.rgd_offset
+ gd_size
+ (gt_size
* gt_count
);
1460 header
.grain_offset
=
1461 ((header
.gd_offset
+ gd_size
+ (gt_size
* gt_count
) +
1462 header
.granularity
- 1) / header
.granularity
) *
1464 /* swap endianness for all header fields */
1465 header
.version
= cpu_to_le32(header
.version
);
1466 header
.flags
= cpu_to_le32(header
.flags
);
1467 header
.capacity
= cpu_to_le64(header
.capacity
);
1468 header
.granularity
= cpu_to_le64(header
.granularity
);
1469 header
.num_gtes_per_gt
= cpu_to_le32(header
.num_gtes_per_gt
);
1470 header
.desc_offset
= cpu_to_le64(header
.desc_offset
);
1471 header
.desc_size
= cpu_to_le64(header
.desc_size
);
1472 header
.rgd_offset
= cpu_to_le64(header
.rgd_offset
);
1473 header
.gd_offset
= cpu_to_le64(header
.gd_offset
);
1474 header
.grain_offset
= cpu_to_le64(header
.grain_offset
);
1475 header
.compressAlgorithm
= cpu_to_le16(header
.compressAlgorithm
);
1477 header
.check_bytes
[0] = 0xa;
1478 header
.check_bytes
[1] = 0x20;
1479 header
.check_bytes
[2] = 0xd;
1480 header
.check_bytes
[3] = 0xa;
1482 /* write all the data */
1483 ret
= qemu_write_full(fd
, &magic
, sizeof(magic
));
1484 if (ret
!= sizeof(magic
)) {
1488 ret
= qemu_write_full(fd
, &header
, sizeof(header
));
1489 if (ret
!= sizeof(header
)) {
1494 ret
= ftruncate(fd
, le64_to_cpu(header
.grain_offset
) << 9);
1500 /* write grain directory */
1501 lseek(fd
, le64_to_cpu(header
.rgd_offset
) << 9, SEEK_SET
);
1502 for (i
= 0, tmp
= le64_to_cpu(header
.rgd_offset
) + gd_size
;
1503 i
< gt_count
; i
++, tmp
+= gt_size
) {
1504 ret
= qemu_write_full(fd
, &tmp
, sizeof(tmp
));
1505 if (ret
!= sizeof(tmp
)) {
1511 /* write backup grain directory */
1512 lseek(fd
, le64_to_cpu(header
.gd_offset
) << 9, SEEK_SET
);
1513 for (i
= 0, tmp
= le64_to_cpu(header
.gd_offset
) + gd_size
;
1514 i
< gt_count
; i
++, tmp
+= gt_size
) {
1515 ret
= qemu_write_full(fd
, &tmp
, sizeof(tmp
));
1516 if (ret
!= sizeof(tmp
)) {
1528 static int filename_decompose(const char *filename
, char *path
, char *prefix
,
1529 char *postfix
, size_t buf_len
, Error
**errp
)
1533 if (filename
== NULL
|| !strlen(filename
)) {
1534 error_setg(errp
, "No filename provided");
1537 p
= strrchr(filename
, '/');
1539 p
= strrchr(filename
, '\\');
1542 p
= strrchr(filename
, ':');
1546 if (p
- filename
>= buf_len
) {
1549 pstrcpy(path
, p
- filename
+ 1, filename
);
1554 q
= strrchr(p
, '.');
1556 pstrcpy(prefix
, buf_len
, p
);
1559 if (q
- p
>= buf_len
) {
1562 pstrcpy(prefix
, q
- p
+ 1, p
);
1563 pstrcpy(postfix
, buf_len
, q
);
1568 static int vmdk_create(const char *filename
, QEMUOptionParameter
*options
,
1572 char desc
[BUF_SIZE
];
1573 int64_t total_size
= 0, filesize
;
1574 const char *adapter_type
= NULL
;
1575 const char *backing_file
= NULL
;
1576 const char *fmt
= NULL
;
1579 bool flat
, split
, compress
;
1580 char ext_desc_lines
[BUF_SIZE
] = "";
1581 char path
[PATH_MAX
], prefix
[PATH_MAX
], postfix
[PATH_MAX
];
1582 const int64_t split_size
= 0x80000000; /* VMDK has constant split size */
1583 const char *desc_extent_line
;
1584 char parent_desc_line
[BUF_SIZE
] = "";
1585 uint32_t parent_cid
= 0xffffffff;
1586 uint32_t number_heads
= 16;
1587 bool zeroed_grain
= false;
1588 const char desc_template
[] =
1589 "# Disk DescriptorFile\n"
1593 "createType=\"%s\"\n"
1596 "# Extent description\n"
1599 "# The Disk Data Base\n"
1602 "ddb.virtualHWVersion = \"%d\"\n"
1603 "ddb.geometry.cylinders = \"%" PRId64
"\"\n"
1604 "ddb.geometry.heads = \"%d\"\n"
1605 "ddb.geometry.sectors = \"63\"\n"
1606 "ddb.adapterType = \"%s\"\n";
1608 if (filename_decompose(filename
, path
, prefix
, postfix
, PATH_MAX
, errp
)) {
1611 /* Read out options */
1612 while (options
&& options
->name
) {
1613 if (!strcmp(options
->name
, BLOCK_OPT_SIZE
)) {
1614 total_size
= options
->value
.n
;
1615 } else if (!strcmp(options
->name
, BLOCK_OPT_ADAPTER_TYPE
)) {
1616 adapter_type
= options
->value
.s
;
1617 } else if (!strcmp(options
->name
, BLOCK_OPT_BACKING_FILE
)) {
1618 backing_file
= options
->value
.s
;
1619 } else if (!strcmp(options
->name
, BLOCK_OPT_COMPAT6
)) {
1620 flags
|= options
->value
.n
? BLOCK_FLAG_COMPAT6
: 0;
1621 } else if (!strcmp(options
->name
, BLOCK_OPT_SUBFMT
)) {
1622 fmt
= options
->value
.s
;
1623 } else if (!strcmp(options
->name
, BLOCK_OPT_ZEROED_GRAIN
)) {
1624 zeroed_grain
|= options
->value
.n
;
1628 if (!adapter_type
) {
1629 adapter_type
= "ide";
1630 } else if (strcmp(adapter_type
, "ide") &&
1631 strcmp(adapter_type
, "buslogic") &&
1632 strcmp(adapter_type
, "lsilogic") &&
1633 strcmp(adapter_type
, "legacyESX")) {
1634 error_setg(errp
, "Unknown adapter type: '%s'", adapter_type
);
1637 if (strcmp(adapter_type
, "ide") != 0) {
1638 /* that's the number of heads with which vmware operates when
1639 creating, exporting, etc. vmdk files with a non-ide adapter type */
1643 /* Default format to monolithicSparse */
1644 fmt
= "monolithicSparse";
1645 } else if (strcmp(fmt
, "monolithicFlat") &&
1646 strcmp(fmt
, "monolithicSparse") &&
1647 strcmp(fmt
, "twoGbMaxExtentSparse") &&
1648 strcmp(fmt
, "twoGbMaxExtentFlat") &&
1649 strcmp(fmt
, "streamOptimized")) {
1650 error_setg(errp
, "Unknown subformat: '%s'", fmt
);
1653 split
= !(strcmp(fmt
, "twoGbMaxExtentFlat") &&
1654 strcmp(fmt
, "twoGbMaxExtentSparse"));
1655 flat
= !(strcmp(fmt
, "monolithicFlat") &&
1656 strcmp(fmt
, "twoGbMaxExtentFlat"));
1657 compress
= !strcmp(fmt
, "streamOptimized");
1659 desc_extent_line
= "RW %lld FLAT \"%s\" 0\n";
1661 desc_extent_line
= "RW %lld SPARSE \"%s\"\n";
1663 if (flat
&& backing_file
) {
1664 error_setg(errp
, "Flat image can't have backing file");
1668 BlockDriverState
*bs
= bdrv_new("");
1669 ret
= bdrv_open(bs
, backing_file
, NULL
, 0, NULL
, errp
);
1674 if (strcmp(bs
->drv
->format_name
, "vmdk")) {
1678 parent_cid
= vmdk_read_cid(bs
, 0);
1680 snprintf(parent_desc_line
, sizeof(parent_desc_line
),
1681 "parentFileNameHint=\"%s\"", backing_file
);
1684 /* Create extents */
1685 filesize
= total_size
;
1686 while (filesize
> 0) {
1687 char desc_line
[BUF_SIZE
];
1688 char ext_filename
[PATH_MAX
];
1689 char desc_filename
[PATH_MAX
];
1690 int64_t size
= filesize
;
1692 if (split
&& size
> split_size
) {
1696 snprintf(desc_filename
, sizeof(desc_filename
), "%s-%c%03d%s",
1697 prefix
, flat
? 'f' : 's', ++idx
, postfix
);
1699 snprintf(desc_filename
, sizeof(desc_filename
), "%s-flat%s",
1702 snprintf(desc_filename
, sizeof(desc_filename
), "%s%s",
1705 snprintf(ext_filename
, sizeof(ext_filename
), "%s%s",
1706 path
, desc_filename
);
1708 if (vmdk_create_extent(ext_filename
, size
,
1709 flat
, compress
, zeroed_grain
)) {
1714 /* Format description line */
1715 snprintf(desc_line
, sizeof(desc_line
),
1716 desc_extent_line
, size
/ 512, desc_filename
);
1717 pstrcat(ext_desc_lines
, sizeof(ext_desc_lines
), desc_line
);
1719 /* generate descriptor file */
1720 snprintf(desc
, sizeof(desc
), desc_template
,
1721 (unsigned int)time(NULL
),
1726 (flags
& BLOCK_FLAG_COMPAT6
? 6 : 4),
1727 total_size
/ (int64_t)(63 * number_heads
* 512), number_heads
,
1729 if (split
|| flat
) {
1730 fd
= qemu_open(filename
,
1731 O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
| O_LARGEFILE
,
1734 fd
= qemu_open(filename
,
1735 O_WRONLY
| O_BINARY
| O_LARGEFILE
,
1741 /* the descriptor offset = 0x200 */
1742 if (!split
&& !flat
&& 0x200 != lseek(fd
, 0x200, SEEK_SET
)) {
1746 ret
= qemu_write_full(fd
, desc
, strlen(desc
));
1747 if (ret
!= strlen(desc
)) {
1757 static void vmdk_close(BlockDriverState
*bs
)
1759 BDRVVmdkState
*s
= bs
->opaque
;
1761 vmdk_free_extents(bs
);
1763 migrate_del_blocker(s
->migration_blocker
);
1764 error_free(s
->migration_blocker
);
1767 static coroutine_fn
int vmdk_co_flush(BlockDriverState
*bs
)
1769 BDRVVmdkState
*s
= bs
->opaque
;
1773 for (i
= 0; i
< s
->num_extents
; i
++) {
1774 err
= bdrv_co_flush(s
->extents
[i
].file
);
1782 static int64_t vmdk_get_allocated_file_size(BlockDriverState
*bs
)
1787 BDRVVmdkState
*s
= bs
->opaque
;
1789 ret
= bdrv_get_allocated_file_size(bs
->file
);
1793 for (i
= 0; i
< s
->num_extents
; i
++) {
1794 if (s
->extents
[i
].file
== bs
->file
) {
1797 r
= bdrv_get_allocated_file_size(s
->extents
[i
].file
);
1806 static int vmdk_has_zero_init(BlockDriverState
*bs
)
1809 BDRVVmdkState
*s
= bs
->opaque
;
1811 /* If has a flat extent and its underlying storage doesn't have zero init,
1813 for (i
= 0; i
< s
->num_extents
; i
++) {
1814 if (s
->extents
[i
].flat
) {
1815 if (!bdrv_has_zero_init(s
->extents
[i
].file
)) {
1823 static QEMUOptionParameter vmdk_create_options
[] = {
1825 .name
= BLOCK_OPT_SIZE
,
1827 .help
= "Virtual disk size"
1830 .name
= BLOCK_OPT_ADAPTER_TYPE
,
1832 .help
= "Virtual adapter type, can be one of "
1833 "ide (default), lsilogic, buslogic or legacyESX"
1836 .name
= BLOCK_OPT_BACKING_FILE
,
1838 .help
= "File name of a base image"
1841 .name
= BLOCK_OPT_COMPAT6
,
1843 .help
= "VMDK version 6 image"
1846 .name
= BLOCK_OPT_SUBFMT
,
1849 "VMDK flat extent format, can be one of "
1850 "{monolithicSparse (default) | monolithicFlat | twoGbMaxExtentSparse | twoGbMaxExtentFlat | streamOptimized} "
1853 .name
= BLOCK_OPT_ZEROED_GRAIN
,
1855 .help
= "Enable efficient zero writes using the zeroed-grain GTE feature"
1860 static BlockDriver bdrv_vmdk
= {
1861 .format_name
= "vmdk",
1862 .instance_size
= sizeof(BDRVVmdkState
),
1863 .bdrv_probe
= vmdk_probe
,
1864 .bdrv_open
= vmdk_open
,
1865 .bdrv_reopen_prepare
= vmdk_reopen_prepare
,
1866 .bdrv_read
= vmdk_co_read
,
1867 .bdrv_write
= vmdk_co_write
,
1868 .bdrv_co_write_zeroes
= vmdk_co_write_zeroes
,
1869 .bdrv_close
= vmdk_close
,
1870 .bdrv_create
= vmdk_create
,
1871 .bdrv_co_flush_to_disk
= vmdk_co_flush
,
1872 .bdrv_co_get_block_status
= vmdk_co_get_block_status
,
1873 .bdrv_get_allocated_file_size
= vmdk_get_allocated_file_size
,
1874 .bdrv_has_zero_init
= vmdk_has_zero_init
,
1876 .create_options
= vmdk_create_options
,
1879 static void bdrv_vmdk_init(void)
1881 bdrv_register(&bdrv_vmdk
);
1884 block_init(bdrv_vmdk_init
);