2 * Block driver for Connectix / Microsoft Virtual PC images
4 * Copyright (c) 2005 Alex Beregszaszi
5 * Copyright (c) 2009 Kevin Wolf <kwolf@suse.de>
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
25 #include "qemu-common.h"
26 #include "block_int.h"
28 #include "migration.h"
30 /**************************************************************/
32 #define HEADER_SIZE 512
42 // Seconds since Jan 1, 2000 0:00:00 (UTC)
43 #define VHD_TIMESTAMP_BASE 946684800
47 char creator
[8]; // "conectix"
51 // Offset of next header structure, 0xFFFFFFFF if none
54 // Seconds since Jan 1, 2000 0:00:00 (UTC)
57 char creator_app
[4]; // "vpc "
60 char creator_os
[4]; // "Wi2k"
71 // Checksum of the Hard Disk Footer ("one's complement of the sum of all
72 // the bytes in the footer without the checksum field")
75 // UUID used to identify a parent hard disk (backing file)
78 uint8_t in_saved_state
;
81 struct vhd_dyndisk_header
{
82 char magic
[8]; // "cxsparse"
84 // Offset of next header structure, 0xFFFFFFFF if none
87 // Offset of the Block Allocation Table (BAT)
88 uint64_t table_offset
;
91 uint32_t max_table_entries
; // 32bit/entry
93 // 2 MB by default, must be a power of two
97 uint8_t parent_uuid
[16];
98 uint32_t parent_timestamp
;
101 // Backing file name (in UTF-16)
102 uint8_t parent_name
[512];
107 uint32_t data_length
;
109 uint64_t data_offset
;
113 typedef struct BDRVVPCState
{
115 uint8_t footer_buf
[HEADER_SIZE
];
116 uint64_t free_data_block_offset
;
117 int max_table_entries
;
120 uint64_t last_bitmap_offset
;
123 uint32_t bitmap_size
;
126 uint8_t *pageentry_u8
;
127 uint32_t *pageentry_u32
;
128 uint16_t *pageentry_u16
;
130 uint64_t last_bitmap
;
133 Error
*migration_blocker
;
136 static uint32_t vpc_checksum(uint8_t* buf
, size_t size
)
141 for (i
= 0; i
< size
; i
++)
148 static int vpc_probe(const uint8_t *buf
, int buf_size
, const char *filename
)
150 if (buf_size
>= 8 && !strncmp((char *)buf
, "conectix", 8))
155 static int vpc_open(BlockDriverState
*bs
, int flags
)
157 BDRVVPCState
*s
= bs
->opaque
;
159 struct vhd_footer
* footer
;
160 struct vhd_dyndisk_header
* dyndisk_header
;
161 uint8_t buf
[HEADER_SIZE
];
164 int disk_type
= VHD_DYNAMIC
;
166 if (bdrv_pread(bs
->file
, 0, s
->footer_buf
, HEADER_SIZE
) != HEADER_SIZE
)
169 footer
= (struct vhd_footer
*) s
->footer_buf
;
170 if (strncmp(footer
->creator
, "conectix", 8)) {
171 int64_t offset
= bdrv_getlength(bs
->file
);
172 if (offset
< HEADER_SIZE
) {
175 /* If a fixed disk, the footer is found only at the end of the file */
176 if (bdrv_pread(bs
->file
, offset
-HEADER_SIZE
, s
->footer_buf
, HEADER_SIZE
)
180 if (strncmp(footer
->creator
, "conectix", 8)) {
183 disk_type
= VHD_FIXED
;
186 checksum
= be32_to_cpu(footer
->checksum
);
187 footer
->checksum
= 0;
188 if (vpc_checksum(s
->footer_buf
, HEADER_SIZE
) != checksum
)
189 fprintf(stderr
, "block-vpc: The header checksum of '%s' is "
190 "incorrect.\n", bs
->filename
);
192 /* Write 'checksum' back to footer, or else will leave it with zero. */
193 footer
->checksum
= be32_to_cpu(checksum
);
195 // The visible size of a image in Virtual PC depends on the geometry
196 // rather than on the size stored in the footer (the size in the footer
197 // is too large usually)
198 bs
->total_sectors
= (int64_t)
199 be16_to_cpu(footer
->cyls
) * footer
->heads
* footer
->secs_per_cyl
;
201 if (bs
->total_sectors
>= 65535 * 16 * 255) {
206 if (disk_type
== VHD_DYNAMIC
) {
207 if (bdrv_pread(bs
->file
, be64_to_cpu(footer
->data_offset
), buf
,
208 HEADER_SIZE
) != HEADER_SIZE
) {
212 dyndisk_header
= (struct vhd_dyndisk_header
*) buf
;
214 if (strncmp(dyndisk_header
->magic
, "cxsparse", 8)) {
218 s
->block_size
= be32_to_cpu(dyndisk_header
->block_size
);
219 s
->bitmap_size
= ((s
->block_size
/ (8 * 512)) + 511) & ~511;
221 s
->max_table_entries
= be32_to_cpu(dyndisk_header
->max_table_entries
);
222 s
->pagetable
= g_malloc(s
->max_table_entries
* 4);
224 s
->bat_offset
= be64_to_cpu(dyndisk_header
->table_offset
);
225 if (bdrv_pread(bs
->file
, s
->bat_offset
, s
->pagetable
,
226 s
->max_table_entries
* 4) != s
->max_table_entries
* 4) {
230 s
->free_data_block_offset
=
231 (s
->bat_offset
+ (s
->max_table_entries
* 4) + 511) & ~511;
233 for (i
= 0; i
< s
->max_table_entries
; i
++) {
234 be32_to_cpus(&s
->pagetable
[i
]);
235 if (s
->pagetable
[i
] != 0xFFFFFFFF) {
236 int64_t next
= (512 * (int64_t) s
->pagetable
[i
]) +
237 s
->bitmap_size
+ s
->block_size
;
239 if (next
> s
->free_data_block_offset
) {
240 s
->free_data_block_offset
= next
;
245 s
->last_bitmap_offset
= (int64_t) -1;
248 s
->pageentry_u8
= g_malloc(512);
249 s
->pageentry_u32
= s
->pageentry_u8
;
250 s
->pageentry_u16
= s
->pageentry_u8
;
251 s
->last_pagetable
= -1;
255 qemu_co_mutex_init(&s
->lock
);
257 /* Disable migration when VHD images are used */
258 error_set(&s
->migration_blocker
,
259 QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
260 "vpc", bs
->device_name
, "live migration");
261 migrate_add_blocker(s
->migration_blocker
);
268 static int vpc_reopen_prepare(BDRVReopenState
*state
,
269 BlockReopenQueue
*queue
, Error
**errp
)
275 * Returns the absolute byte offset of the given sector in the image file.
276 * If the sector is not allocated, -1 is returned instead.
278 * The parameter write must be 1 if the offset will be used for a write
279 * operation (the block bitmaps is updated then), 0 otherwise.
281 static inline int64_t get_sector_offset(BlockDriverState
*bs
,
282 int64_t sector_num
, int write
)
284 BDRVVPCState
*s
= bs
->opaque
;
285 uint64_t offset
= sector_num
* 512;
286 uint64_t bitmap_offset
, block_offset
;
287 uint32_t pagetable_index
, pageentry_index
;
289 pagetable_index
= offset
/ s
->block_size
;
290 pageentry_index
= (offset
% s
->block_size
) / 512;
292 if (pagetable_index
>= s
->max_table_entries
|| s
->pagetable
[pagetable_index
] == 0xffffffff)
293 return -1; // not allocated
295 bitmap_offset
= 512 * (uint64_t) s
->pagetable
[pagetable_index
];
296 block_offset
= bitmap_offset
+ s
->bitmap_size
+ (512 * pageentry_index
);
298 // We must ensure that we don't write to any sectors which are marked as
299 // unused in the bitmap. We get away with setting all bits in the block
300 // bitmap each time we write to a new block. This might cause Virtual PC to
301 // miss sparse read optimization, but it's not a problem in terms of
303 if (write
&& (s
->last_bitmap_offset
!= bitmap_offset
)) {
304 uint8_t bitmap
[s
->bitmap_size
];
306 s
->last_bitmap_offset
= bitmap_offset
;
307 memset(bitmap
, 0xff, s
->bitmap_size
);
308 bdrv_pwrite_sync(bs
->file
, bitmap_offset
, bitmap
, s
->bitmap_size
);
311 // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
312 // sector_num, pagetable_index, pageentry_index,
313 // bitmap_offset, block_offset);
315 // disabled by reason
318 if (bitmap_offset
!= s
->last_bitmap
)
320 lseek(s
->fd
, bitmap_offset
, SEEK_SET
);
322 s
->last_bitmap
= bitmap_offset
;
324 // Scary! Bitmap is stored as big endian 32bit entries,
325 // while we used to look it up byte by byte
326 read(s
->fd
, s
->pageentry_u8
, 512);
327 for (i
= 0; i
< 128; i
++)
328 be32_to_cpus(&s
->pageentry_u32
[i
]);
331 if ((s
->pageentry_u8
[pageentry_index
/ 8] >> (pageentry_index
% 8)) & 1)
334 lseek(s
->fd
, bitmap_offset
+ (pageentry_index
/ 8), SEEK_SET
);
336 read(s
->fd
, &bitmap_entry
, 1);
338 if ((bitmap_entry
>> (pageentry_index
% 8)) & 1)
339 return -1; // not allocated
347 * Writes the footer to the end of the image file. This is needed when the
348 * file grows as it overwrites the old footer
350 * Returns 0 on success and < 0 on error
352 static int rewrite_footer(BlockDriverState
* bs
)
355 BDRVVPCState
*s
= bs
->opaque
;
356 int64_t offset
= s
->free_data_block_offset
;
358 ret
= bdrv_pwrite_sync(bs
->file
, offset
, s
->footer_buf
, HEADER_SIZE
);
366 * Allocates a new block. This involves writing a new footer and updating
367 * the Block Allocation Table to use the space at the old end of the image
368 * file (overwriting the old footer)
370 * Returns the sectors' offset in the image file on success and < 0 on error
372 static int64_t alloc_block(BlockDriverState
* bs
, int64_t sector_num
)
374 BDRVVPCState
*s
= bs
->opaque
;
376 uint32_t index
, bat_value
;
378 uint8_t bitmap
[s
->bitmap_size
];
380 // Check if sector_num is valid
381 if ((sector_num
< 0) || (sector_num
> bs
->total_sectors
))
384 // Write entry into in-memory BAT
385 index
= (sector_num
* 512) / s
->block_size
;
386 if (s
->pagetable
[index
] != 0xFFFFFFFF)
389 s
->pagetable
[index
] = s
->free_data_block_offset
/ 512;
391 // Initialize the block's bitmap
392 memset(bitmap
, 0xff, s
->bitmap_size
);
393 ret
= bdrv_pwrite_sync(bs
->file
, s
->free_data_block_offset
, bitmap
,
399 // Write new footer (the old one will be overwritten)
400 s
->free_data_block_offset
+= s
->block_size
+ s
->bitmap_size
;
401 ret
= rewrite_footer(bs
);
405 // Write BAT entry to disk
406 bat_offset
= s
->bat_offset
+ (4 * index
);
407 bat_value
= be32_to_cpu(s
->pagetable
[index
]);
408 ret
= bdrv_pwrite_sync(bs
->file
, bat_offset
, &bat_value
, 4);
412 return get_sector_offset(bs
, sector_num
, 0);
415 s
->free_data_block_offset
-= (s
->block_size
+ s
->bitmap_size
);
419 static int vpc_read(BlockDriverState
*bs
, int64_t sector_num
,
420 uint8_t *buf
, int nb_sectors
)
422 BDRVVPCState
*s
= bs
->opaque
;
425 int64_t sectors
, sectors_per_block
;
426 struct vhd_footer
*footer
= (struct vhd_footer
*) s
->footer_buf
;
428 if (cpu_to_be32(footer
->type
) == VHD_FIXED
) {
429 return bdrv_read(bs
->file
, sector_num
, buf
, nb_sectors
);
431 while (nb_sectors
> 0) {
432 offset
= get_sector_offset(bs
, sector_num
, 0);
434 sectors_per_block
= s
->block_size
>> BDRV_SECTOR_BITS
;
435 sectors
= sectors_per_block
- (sector_num
% sectors_per_block
);
436 if (sectors
> nb_sectors
) {
437 sectors
= nb_sectors
;
441 memset(buf
, 0, sectors
* BDRV_SECTOR_SIZE
);
443 ret
= bdrv_pread(bs
->file
, offset
, buf
,
444 sectors
* BDRV_SECTOR_SIZE
);
445 if (ret
!= sectors
* BDRV_SECTOR_SIZE
) {
450 nb_sectors
-= sectors
;
451 sector_num
+= sectors
;
452 buf
+= sectors
* BDRV_SECTOR_SIZE
;
457 static coroutine_fn
int vpc_co_read(BlockDriverState
*bs
, int64_t sector_num
,
458 uint8_t *buf
, int nb_sectors
)
461 BDRVVPCState
*s
= bs
->opaque
;
462 qemu_co_mutex_lock(&s
->lock
);
463 ret
= vpc_read(bs
, sector_num
, buf
, nb_sectors
);
464 qemu_co_mutex_unlock(&s
->lock
);
468 static int vpc_write(BlockDriverState
*bs
, int64_t sector_num
,
469 const uint8_t *buf
, int nb_sectors
)
471 BDRVVPCState
*s
= bs
->opaque
;
473 int64_t sectors
, sectors_per_block
;
475 struct vhd_footer
*footer
= (struct vhd_footer
*) s
->footer_buf
;
477 if (cpu_to_be32(footer
->type
) == VHD_FIXED
) {
478 return bdrv_write(bs
->file
, sector_num
, buf
, nb_sectors
);
480 while (nb_sectors
> 0) {
481 offset
= get_sector_offset(bs
, sector_num
, 1);
483 sectors_per_block
= s
->block_size
>> BDRV_SECTOR_BITS
;
484 sectors
= sectors_per_block
- (sector_num
% sectors_per_block
);
485 if (sectors
> nb_sectors
) {
486 sectors
= nb_sectors
;
490 offset
= alloc_block(bs
, sector_num
);
495 ret
= bdrv_pwrite(bs
->file
, offset
, buf
, sectors
* BDRV_SECTOR_SIZE
);
496 if (ret
!= sectors
* BDRV_SECTOR_SIZE
) {
500 nb_sectors
-= sectors
;
501 sector_num
+= sectors
;
502 buf
+= sectors
* BDRV_SECTOR_SIZE
;
508 static coroutine_fn
int vpc_co_write(BlockDriverState
*bs
, int64_t sector_num
,
509 const uint8_t *buf
, int nb_sectors
)
512 BDRVVPCState
*s
= bs
->opaque
;
513 qemu_co_mutex_lock(&s
->lock
);
514 ret
= vpc_write(bs
, sector_num
, buf
, nb_sectors
);
515 qemu_co_mutex_unlock(&s
->lock
);
520 * Calculates the number of cylinders, heads and sectors per cylinder
521 * based on a given number of sectors. This is the algorithm described
522 * in the VHD specification.
524 * Note that the geometry doesn't always exactly match total_sectors but
527 * Returns 0 on success, -EFBIG if the size is larger than 127 GB
529 static int calculate_geometry(int64_t total_sectors
, uint16_t* cyls
,
530 uint8_t* heads
, uint8_t* secs_per_cyl
)
532 uint32_t cyls_times_heads
;
534 if (total_sectors
> 65535 * 16 * 255)
537 if (total_sectors
> 65535 * 16 * 63) {
540 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
543 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
544 *heads
= (cyls_times_heads
+ 1023) / 1024;
549 if (cyls_times_heads
>= (*heads
* 1024) || *heads
> 16) {
552 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
555 if (cyls_times_heads
>= (*heads
* 1024)) {
558 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
562 *cyls
= cyls_times_heads
/ *heads
;
567 static int create_dynamic_disk(int fd
, uint8_t *buf
, int64_t total_sectors
)
569 struct vhd_dyndisk_header
* dyndisk_header
=
570 (struct vhd_dyndisk_header
*) buf
;
571 size_t block_size
, num_bat_entries
;
575 // Write the footer (twice: at the beginning and at the end)
576 block_size
= 0x200000;
577 num_bat_entries
= (total_sectors
+ block_size
/ 512) / (block_size
/ 512);
579 if (write(fd
, buf
, HEADER_SIZE
) != HEADER_SIZE
) {
583 if (lseek(fd
, 1536 + ((num_bat_entries
* 4 + 511) & ~511), SEEK_SET
) < 0) {
586 if (write(fd
, buf
, HEADER_SIZE
) != HEADER_SIZE
) {
590 // Write the initial BAT
591 if (lseek(fd
, 3 * 512, SEEK_SET
) < 0) {
595 memset(buf
, 0xFF, 512);
596 for (i
= 0; i
< (num_bat_entries
* 4 + 511) / 512; i
++) {
597 if (write(fd
, buf
, 512) != 512) {
602 // Prepare the Dynamic Disk Header
603 memset(buf
, 0, 1024);
605 memcpy(dyndisk_header
->magic
, "cxsparse", 8);
608 * Note: The spec is actually wrong here for data_offset, it says
609 * 0xFFFFFFFF, but MS tools expect all 64 bits to be set.
611 dyndisk_header
->data_offset
= be64_to_cpu(0xFFFFFFFFFFFFFFFFULL
);
612 dyndisk_header
->table_offset
= be64_to_cpu(3 * 512);
613 dyndisk_header
->version
= be32_to_cpu(0x00010000);
614 dyndisk_header
->block_size
= be32_to_cpu(block_size
);
615 dyndisk_header
->max_table_entries
= be32_to_cpu(num_bat_entries
);
617 dyndisk_header
->checksum
= be32_to_cpu(vpc_checksum(buf
, 1024));
620 if (lseek(fd
, 512, SEEK_SET
) < 0) {
624 if (write(fd
, buf
, 1024) != 1024) {
633 static int create_fixed_disk(int fd
, uint8_t *buf
, int64_t total_size
)
637 /* Add footer to total size */
639 if (ftruncate(fd
, total_size
) != 0) {
643 if (lseek(fd
, -512, SEEK_END
) < 0) {
646 if (write(fd
, buf
, HEADER_SIZE
) != HEADER_SIZE
) {
656 static int vpc_create(const char *filename
, QEMUOptionParameter
*options
)
659 struct vhd_footer
*footer
= (struct vhd_footer
*) buf
;
660 QEMUOptionParameter
*disk_type_param
;
664 uint8_t secs_per_cyl
= 0;
665 int64_t total_sectors
;
670 /* Read out options */
671 total_size
= get_option_parameter(options
, BLOCK_OPT_SIZE
)->value
.n
;
673 disk_type_param
= get_option_parameter(options
, BLOCK_OPT_SUBFMT
);
674 if (disk_type_param
&& disk_type_param
->value
.s
) {
675 if (!strcmp(disk_type_param
->value
.s
, "dynamic")) {
676 disk_type
= VHD_DYNAMIC
;
677 } else if (!strcmp(disk_type_param
->value
.s
, "fixed")) {
678 disk_type
= VHD_FIXED
;
683 disk_type
= VHD_DYNAMIC
;
686 /* Create the file */
687 fd
= qemu_open(filename
, O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
, 0644);
693 * Calculate matching total_size and geometry. Increase the number of
694 * sectors requested until we get enough (or fail). This ensures that
695 * qemu-img convert doesn't truncate images, but rather rounds up.
697 total_sectors
= total_size
/ BDRV_SECTOR_SIZE
;
698 for (i
= 0; total_sectors
> (int64_t)cyls
* heads
* secs_per_cyl
; i
++) {
699 if (calculate_geometry(total_sectors
+ i
, &cyls
, &heads
,
707 total_sectors
= (int64_t) cyls
* heads
* secs_per_cyl
;
709 /* Prepare the Hard Disk Footer */
710 memset(buf
, 0, 1024);
712 memcpy(footer
->creator
, "conectix", 8);
713 /* TODO Check if "qemu" creator_app is ok for VPC */
714 memcpy(footer
->creator_app
, "qemu", 4);
715 memcpy(footer
->creator_os
, "Wi2k", 4);
717 footer
->features
= be32_to_cpu(0x02);
718 footer
->version
= be32_to_cpu(0x00010000);
719 if (disk_type
== VHD_DYNAMIC
) {
720 footer
->data_offset
= be64_to_cpu(HEADER_SIZE
);
722 footer
->data_offset
= be64_to_cpu(0xFFFFFFFFFFFFFFFFULL
);
724 footer
->timestamp
= be32_to_cpu(time(NULL
) - VHD_TIMESTAMP_BASE
);
726 /* Version of Virtual PC 2007 */
727 footer
->major
= be16_to_cpu(0x0005);
728 footer
->minor
= be16_to_cpu(0x0003);
729 if (disk_type
== VHD_DYNAMIC
) {
730 footer
->orig_size
= be64_to_cpu(total_sectors
* 512);
731 footer
->size
= be64_to_cpu(total_sectors
* 512);
733 footer
->orig_size
= be64_to_cpu(total_size
);
734 footer
->size
= be64_to_cpu(total_size
);
736 footer
->cyls
= be16_to_cpu(cyls
);
737 footer
->heads
= heads
;
738 footer
->secs_per_cyl
= secs_per_cyl
;
740 footer
->type
= be32_to_cpu(disk_type
);
742 /* TODO uuid is missing */
744 footer
->checksum
= be32_to_cpu(vpc_checksum(buf
, HEADER_SIZE
));
746 if (disk_type
== VHD_DYNAMIC
) {
747 ret
= create_dynamic_disk(fd
, buf
, total_sectors
);
749 ret
= create_fixed_disk(fd
, buf
, total_size
);
757 static void vpc_close(BlockDriverState
*bs
)
759 BDRVVPCState
*s
= bs
->opaque
;
760 g_free(s
->pagetable
);
762 g_free(s
->pageentry_u8
);
765 migrate_del_blocker(s
->migration_blocker
);
766 error_free(s
->migration_blocker
);
769 static QEMUOptionParameter vpc_create_options
[] = {
771 .name
= BLOCK_OPT_SIZE
,
773 .help
= "Virtual disk size"
776 .name
= BLOCK_OPT_SUBFMT
,
779 "Type of virtual hard disk format. Supported formats are "
780 "{dynamic (default) | fixed} "
785 static BlockDriver bdrv_vpc
= {
786 .format_name
= "vpc",
787 .instance_size
= sizeof(BDRVVPCState
),
789 .bdrv_probe
= vpc_probe
,
790 .bdrv_open
= vpc_open
,
791 .bdrv_close
= vpc_close
,
792 .bdrv_reopen_prepare
= vpc_reopen_prepare
,
793 .bdrv_create
= vpc_create
,
795 .bdrv_read
= vpc_co_read
,
796 .bdrv_write
= vpc_co_write
,
798 .create_options
= vpc_create_options
,
801 static void bdrv_vpc_init(void)
803 bdrv_register(&bdrv_vpc
);
806 block_init(bdrv_vpc_init
);