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"
29 /**************************************************************/
31 #define HEADER_SIZE 512
41 // Seconds since Jan 1, 2000 0:00:00 (UTC)
42 #define VHD_TIMESTAMP_BASE 946684800
46 char creator
[8]; // "conectix"
50 // Offset of next header structure, 0xFFFFFFFF if none
53 // Seconds since Jan 1, 2000 0:00:00 (UTC)
56 char creator_app
[4]; // "vpc "
59 char creator_os
[4]; // "Wi2k"
70 // Checksum of the Hard Disk Footer ("one's complement of the sum of all
71 // the bytes in the footer without the checksum field")
74 // UUID used to identify a parent hard disk (backing file)
77 uint8_t in_saved_state
;
80 struct vhd_dyndisk_header
{
81 char magic
[8]; // "cxsparse"
83 // Offset of next header structure, 0xFFFFFFFF if none
86 // Offset of the Block Allocation Table (BAT)
87 uint64_t table_offset
;
90 uint32_t max_table_entries
; // 32bit/entry
92 // 2 MB by default, must be a power of two
96 uint8_t parent_uuid
[16];
97 uint32_t parent_timestamp
;
100 // Backing file name (in UTF-16)
101 uint8_t parent_name
[512];
106 uint32_t data_length
;
108 uint64_t data_offset
;
112 typedef struct BDRVVPCState
{
114 uint8_t footer_buf
[HEADER_SIZE
];
115 uint64_t free_data_block_offset
;
116 int max_table_entries
;
119 uint64_t last_bitmap_offset
;
122 uint32_t bitmap_size
;
125 uint8_t *pageentry_u8
;
126 uint32_t *pageentry_u32
;
127 uint16_t *pageentry_u16
;
129 uint64_t last_bitmap
;
133 static uint32_t vpc_checksum(uint8_t* buf
, size_t size
)
138 for (i
= 0; i
< size
; i
++)
145 static int vpc_probe(const uint8_t *buf
, int buf_size
, const char *filename
)
147 if (buf_size
>= 8 && !strncmp((char *)buf
, "conectix", 8))
152 static int vpc_open(BlockDriverState
*bs
, int flags
)
154 BDRVVPCState
*s
= bs
->opaque
;
156 struct vhd_footer
* footer
;
157 struct vhd_dyndisk_header
* dyndisk_header
;
158 uint8_t buf
[HEADER_SIZE
];
162 if (bdrv_pread(bs
->file
, 0, s
->footer_buf
, HEADER_SIZE
) != HEADER_SIZE
)
165 footer
= (struct vhd_footer
*) s
->footer_buf
;
166 if (strncmp(footer
->creator
, "conectix", 8))
169 checksum
= be32_to_cpu(footer
->checksum
);
170 footer
->checksum
= 0;
171 if (vpc_checksum(s
->footer_buf
, HEADER_SIZE
) != checksum
)
172 fprintf(stderr
, "block-vpc: The header checksum of '%s' is "
173 "incorrect.\n", bs
->filename
);
175 // The visible size of a image in Virtual PC depends on the geometry
176 // rather than on the size stored in the footer (the size in the footer
177 // is too large usually)
178 bs
->total_sectors
= (int64_t)
179 be16_to_cpu(footer
->cyls
) * footer
->heads
* footer
->secs_per_cyl
;
181 if (bs
->total_sectors
>= 65535 * 16 * 255) {
186 if (bdrv_pread(bs
->file
, be64_to_cpu(footer
->data_offset
), buf
, HEADER_SIZE
)
190 dyndisk_header
= (struct vhd_dyndisk_header
*) buf
;
192 if (strncmp(dyndisk_header
->magic
, "cxsparse", 8))
196 s
->block_size
= be32_to_cpu(dyndisk_header
->block_size
);
197 s
->bitmap_size
= ((s
->block_size
/ (8 * 512)) + 511) & ~511;
199 s
->max_table_entries
= be32_to_cpu(dyndisk_header
->max_table_entries
);
200 s
->pagetable
= g_malloc(s
->max_table_entries
* 4);
202 s
->bat_offset
= be64_to_cpu(dyndisk_header
->table_offset
);
203 if (bdrv_pread(bs
->file
, s
->bat_offset
, s
->pagetable
,
204 s
->max_table_entries
* 4) != s
->max_table_entries
* 4)
207 s
->free_data_block_offset
=
208 (s
->bat_offset
+ (s
->max_table_entries
* 4) + 511) & ~511;
210 for (i
= 0; i
< s
->max_table_entries
; i
++) {
211 be32_to_cpus(&s
->pagetable
[i
]);
212 if (s
->pagetable
[i
] != 0xFFFFFFFF) {
213 int64_t next
= (512 * (int64_t) s
->pagetable
[i
]) +
214 s
->bitmap_size
+ s
->block_size
;
216 if (next
> s
->free_data_block_offset
)
217 s
->free_data_block_offset
= next
;
221 s
->last_bitmap_offset
= (int64_t) -1;
224 s
->pageentry_u8
= g_malloc(512);
225 s
->pageentry_u32
= s
->pageentry_u8
;
226 s
->pageentry_u16
= s
->pageentry_u8
;
227 s
->last_pagetable
= -1;
230 qemu_co_mutex_init(&s
->lock
);
237 * Returns the absolute byte offset of the given sector in the image file.
238 * If the sector is not allocated, -1 is returned instead.
240 * The parameter write must be 1 if the offset will be used for a write
241 * operation (the block bitmaps is updated then), 0 otherwise.
243 static inline int64_t get_sector_offset(BlockDriverState
*bs
,
244 int64_t sector_num
, int write
)
246 BDRVVPCState
*s
= bs
->opaque
;
247 uint64_t offset
= sector_num
* 512;
248 uint64_t bitmap_offset
, block_offset
;
249 uint32_t pagetable_index
, pageentry_index
;
251 pagetable_index
= offset
/ s
->block_size
;
252 pageentry_index
= (offset
% s
->block_size
) / 512;
254 if (pagetable_index
>= s
->max_table_entries
|| s
->pagetable
[pagetable_index
] == 0xffffffff)
255 return -1; // not allocated
257 bitmap_offset
= 512 * (uint64_t) s
->pagetable
[pagetable_index
];
258 block_offset
= bitmap_offset
+ s
->bitmap_size
+ (512 * pageentry_index
);
260 // We must ensure that we don't write to any sectors which are marked as
261 // unused in the bitmap. We get away with setting all bits in the block
262 // bitmap each time we write to a new block. This might cause Virtual PC to
263 // miss sparse read optimization, but it's not a problem in terms of
265 if (write
&& (s
->last_bitmap_offset
!= bitmap_offset
)) {
266 uint8_t bitmap
[s
->bitmap_size
];
268 s
->last_bitmap_offset
= bitmap_offset
;
269 memset(bitmap
, 0xff, s
->bitmap_size
);
270 bdrv_pwrite_sync(bs
->file
, bitmap_offset
, bitmap
, s
->bitmap_size
);
273 // printf("sector: %" PRIx64 ", index: %x, offset: %x, bioff: %" PRIx64 ", bloff: %" PRIx64 "\n",
274 // sector_num, pagetable_index, pageentry_index,
275 // bitmap_offset, block_offset);
277 // disabled by reason
280 if (bitmap_offset
!= s
->last_bitmap
)
282 lseek(s
->fd
, bitmap_offset
, SEEK_SET
);
284 s
->last_bitmap
= bitmap_offset
;
286 // Scary! Bitmap is stored as big endian 32bit entries,
287 // while we used to look it up byte by byte
288 read(s
->fd
, s
->pageentry_u8
, 512);
289 for (i
= 0; i
< 128; i
++)
290 be32_to_cpus(&s
->pageentry_u32
[i
]);
293 if ((s
->pageentry_u8
[pageentry_index
/ 8] >> (pageentry_index
% 8)) & 1)
296 lseek(s
->fd
, bitmap_offset
+ (pageentry_index
/ 8), SEEK_SET
);
298 read(s
->fd
, &bitmap_entry
, 1);
300 if ((bitmap_entry
>> (pageentry_index
% 8)) & 1)
301 return -1; // not allocated
309 * Writes the footer to the end of the image file. This is needed when the
310 * file grows as it overwrites the old footer
312 * Returns 0 on success and < 0 on error
314 static int rewrite_footer(BlockDriverState
* bs
)
317 BDRVVPCState
*s
= bs
->opaque
;
318 int64_t offset
= s
->free_data_block_offset
;
320 ret
= bdrv_pwrite_sync(bs
->file
, offset
, s
->footer_buf
, HEADER_SIZE
);
328 * Allocates a new block. This involves writing a new footer and updating
329 * the Block Allocation Table to use the space at the old end of the image
330 * file (overwriting the old footer)
332 * Returns the sectors' offset in the image file on success and < 0 on error
334 static int64_t alloc_block(BlockDriverState
* bs
, int64_t sector_num
)
336 BDRVVPCState
*s
= bs
->opaque
;
338 uint32_t index
, bat_value
;
340 uint8_t bitmap
[s
->bitmap_size
];
342 // Check if sector_num is valid
343 if ((sector_num
< 0) || (sector_num
> bs
->total_sectors
))
346 // Write entry into in-memory BAT
347 index
= (sector_num
* 512) / s
->block_size
;
348 if (s
->pagetable
[index
] != 0xFFFFFFFF)
351 s
->pagetable
[index
] = s
->free_data_block_offset
/ 512;
353 // Initialize the block's bitmap
354 memset(bitmap
, 0xff, s
->bitmap_size
);
355 bdrv_pwrite_sync(bs
->file
, s
->free_data_block_offset
, bitmap
,
358 // Write new footer (the old one will be overwritten)
359 s
->free_data_block_offset
+= s
->block_size
+ s
->bitmap_size
;
360 ret
= rewrite_footer(bs
);
364 // Write BAT entry to disk
365 bat_offset
= s
->bat_offset
+ (4 * index
);
366 bat_value
= be32_to_cpu(s
->pagetable
[index
]);
367 ret
= bdrv_pwrite_sync(bs
->file
, bat_offset
, &bat_value
, 4);
371 return get_sector_offset(bs
, sector_num
, 0);
374 s
->free_data_block_offset
-= (s
->block_size
+ s
->bitmap_size
);
378 static int vpc_read(BlockDriverState
*bs
, int64_t sector_num
,
379 uint8_t *buf
, int nb_sectors
)
381 BDRVVPCState
*s
= bs
->opaque
;
384 int64_t sectors
, sectors_per_block
;
386 while (nb_sectors
> 0) {
387 offset
= get_sector_offset(bs
, sector_num
, 0);
389 sectors_per_block
= s
->block_size
>> BDRV_SECTOR_BITS
;
390 sectors
= sectors_per_block
- (sector_num
% sectors_per_block
);
391 if (sectors
> nb_sectors
) {
392 sectors
= nb_sectors
;
396 memset(buf
, 0, sectors
* BDRV_SECTOR_SIZE
);
398 ret
= bdrv_pread(bs
->file
, offset
, buf
,
399 sectors
* BDRV_SECTOR_SIZE
);
400 if (ret
!= sectors
* BDRV_SECTOR_SIZE
) {
405 nb_sectors
-= sectors
;
406 sector_num
+= sectors
;
407 buf
+= sectors
* BDRV_SECTOR_SIZE
;
412 static coroutine_fn
int vpc_co_read(BlockDriverState
*bs
, int64_t sector_num
,
413 uint8_t *buf
, int nb_sectors
)
416 BDRVVPCState
*s
= bs
->opaque
;
417 qemu_co_mutex_lock(&s
->lock
);
418 ret
= vpc_read(bs
, sector_num
, buf
, nb_sectors
);
419 qemu_co_mutex_unlock(&s
->lock
);
423 static int vpc_write(BlockDriverState
*bs
, int64_t sector_num
,
424 const uint8_t *buf
, int nb_sectors
)
426 BDRVVPCState
*s
= bs
->opaque
;
428 int64_t sectors
, sectors_per_block
;
431 while (nb_sectors
> 0) {
432 offset
= get_sector_offset(bs
, sector_num
, 1);
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 offset
= alloc_block(bs
, sector_num
);
446 ret
= bdrv_pwrite(bs
->file
, offset
, buf
, sectors
* BDRV_SECTOR_SIZE
);
447 if (ret
!= sectors
* BDRV_SECTOR_SIZE
) {
451 nb_sectors
-= sectors
;
452 sector_num
+= sectors
;
453 buf
+= sectors
* BDRV_SECTOR_SIZE
;
459 static coroutine_fn
int vpc_co_write(BlockDriverState
*bs
, int64_t sector_num
,
460 const uint8_t *buf
, int nb_sectors
)
463 BDRVVPCState
*s
= bs
->opaque
;
464 qemu_co_mutex_lock(&s
->lock
);
465 ret
= vpc_write(bs
, sector_num
, buf
, nb_sectors
);
466 qemu_co_mutex_unlock(&s
->lock
);
470 static coroutine_fn
int vpc_co_flush(BlockDriverState
*bs
)
472 return bdrv_co_flush(bs
->file
);
476 * Calculates the number of cylinders, heads and sectors per cylinder
477 * based on a given number of sectors. This is the algorithm described
478 * in the VHD specification.
480 * Note that the geometry doesn't always exactly match total_sectors but
483 * Returns 0 on success, -EFBIG if the size is larger than 127 GB
485 static int calculate_geometry(int64_t total_sectors
, uint16_t* cyls
,
486 uint8_t* heads
, uint8_t* secs_per_cyl
)
488 uint32_t cyls_times_heads
;
490 if (total_sectors
> 65535 * 16 * 255)
493 if (total_sectors
> 65535 * 16 * 63) {
496 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
499 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
500 *heads
= (cyls_times_heads
+ 1023) / 1024;
505 if (cyls_times_heads
>= (*heads
* 1024) || *heads
> 16) {
508 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
511 if (cyls_times_heads
>= (*heads
* 1024)) {
514 cyls_times_heads
= total_sectors
/ *secs_per_cyl
;
518 *cyls
= cyls_times_heads
/ *heads
;
523 static int vpc_create(const char *filename
, QEMUOptionParameter
*options
)
526 struct vhd_footer
* footer
= (struct vhd_footer
*) buf
;
527 struct vhd_dyndisk_header
* dyndisk_header
=
528 (struct vhd_dyndisk_header
*) buf
;
532 uint8_t secs_per_cyl
= 0;
533 size_t block_size
, num_bat_entries
;
534 int64_t total_sectors
= 0;
538 total_sectors
= get_option_parameter(options
, BLOCK_OPT_SIZE
)->value
.n
/
542 fd
= open(filename
, O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
, 0644);
546 /* Calculate matching total_size and geometry. Increase the number of
547 sectors requested until we get enough (or fail). */
548 for (i
= 0; total_sectors
> (int64_t)cyls
* heads
* secs_per_cyl
; i
++) {
549 if (calculate_geometry(total_sectors
+ i
,
550 &cyls
, &heads
, &secs_per_cyl
)) {
555 total_sectors
= (int64_t) cyls
* heads
* secs_per_cyl
;
557 // Prepare the Hard Disk Footer
558 memset(buf
, 0, 1024);
560 memcpy(footer
->creator
, "conectix", 8);
561 // TODO Check if "qemu" creator_app is ok for VPC
562 memcpy(footer
->creator_app
, "qemu", 4);
563 memcpy(footer
->creator_os
, "Wi2k", 4);
565 footer
->features
= be32_to_cpu(0x02);
566 footer
->version
= be32_to_cpu(0x00010000);
567 footer
->data_offset
= be64_to_cpu(HEADER_SIZE
);
568 footer
->timestamp
= be32_to_cpu(time(NULL
) - VHD_TIMESTAMP_BASE
);
570 // Version of Virtual PC 2007
571 footer
->major
= be16_to_cpu(0x0005);
572 footer
->minor
=be16_to_cpu(0x0003);
574 footer
->orig_size
= be64_to_cpu(total_sectors
* 512);
575 footer
->size
= be64_to_cpu(total_sectors
* 512);
577 footer
->cyls
= be16_to_cpu(cyls
);
578 footer
->heads
= heads
;
579 footer
->secs_per_cyl
= secs_per_cyl
;
581 footer
->type
= be32_to_cpu(VHD_DYNAMIC
);
583 // TODO uuid is missing
585 footer
->checksum
= be32_to_cpu(vpc_checksum(buf
, HEADER_SIZE
));
587 // Write the footer (twice: at the beginning and at the end)
588 block_size
= 0x200000;
589 num_bat_entries
= (total_sectors
+ block_size
/ 512) / (block_size
/ 512);
591 if (write(fd
, buf
, HEADER_SIZE
) != HEADER_SIZE
) {
595 if (lseek(fd
, 1536 + ((num_bat_entries
* 4 + 511) & ~511), SEEK_SET
) < 0) {
598 if (write(fd
, buf
, HEADER_SIZE
) != HEADER_SIZE
) {
602 // Write the initial BAT
603 if (lseek(fd
, 3 * 512, SEEK_SET
) < 0) {
607 memset(buf
, 0xFF, 512);
608 for (i
= 0; i
< (num_bat_entries
* 4 + 511) / 512; i
++) {
609 if (write(fd
, buf
, 512) != 512) {
615 // Prepare the Dynamic Disk Header
616 memset(buf
, 0, 1024);
618 memcpy(dyndisk_header
->magic
, "cxsparse", 8);
620 dyndisk_header
->data_offset
= be64_to_cpu(0xFFFFFFFF);
621 dyndisk_header
->table_offset
= be64_to_cpu(3 * 512);
622 dyndisk_header
->version
= be32_to_cpu(0x00010000);
623 dyndisk_header
->block_size
= be32_to_cpu(block_size
);
624 dyndisk_header
->max_table_entries
= be32_to_cpu(num_bat_entries
);
626 dyndisk_header
->checksum
= be32_to_cpu(vpc_checksum(buf
, 1024));
629 if (lseek(fd
, 512, SEEK_SET
) < 0) {
633 if (write(fd
, buf
, 1024) != 1024) {
643 static void vpc_close(BlockDriverState
*bs
)
645 BDRVVPCState
*s
= bs
->opaque
;
646 g_free(s
->pagetable
);
648 g_free(s
->pageentry_u8
);
652 static QEMUOptionParameter vpc_create_options
[] = {
654 .name
= BLOCK_OPT_SIZE
,
656 .help
= "Virtual disk size"
661 static BlockDriver bdrv_vpc
= {
662 .format_name
= "vpc",
663 .instance_size
= sizeof(BDRVVPCState
),
664 .bdrv_probe
= vpc_probe
,
665 .bdrv_open
= vpc_open
,
666 .bdrv_read
= vpc_co_read
,
667 .bdrv_write
= vpc_co_write
,
668 .bdrv_co_flush
= vpc_co_flush
,
669 .bdrv_close
= vpc_close
,
670 .bdrv_create
= vpc_create
,
672 .create_options
= vpc_create_options
,
675 static void bdrv_vpc_init(void)
677 bdrv_register(&bdrv_vpc
);
680 block_init(bdrv_vpc_init
);