2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "config-host.h"
25 #include "qemu-common.h"
27 #include "block_int.h"
29 #include "qemu-objects.h"
32 #include <sys/types.h>
34 #include <sys/ioctl.h>
35 #include <sys/queue.h>
45 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
46 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
47 BlockDriverCompletionFunc
*cb
, void *opaque
);
48 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
49 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
50 BlockDriverCompletionFunc
*cb
, void *opaque
);
51 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
52 BlockDriverCompletionFunc
*cb
, void *opaque
);
53 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
54 uint8_t *buf
, int nb_sectors
);
55 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
56 const uint8_t *buf
, int nb_sectors
);
58 BlockDriverState
*bdrv_first
;
60 static BlockDriver
*first_drv
;
62 /* If non-zero, use only whitelisted block drivers */
63 static int use_bdrv_whitelist
;
65 int path_is_absolute(const char *path
)
69 /* specific case for names like: "\\.\d:" */
70 if (*path
== '/' || *path
== '\\')
73 p
= strchr(path
, ':');
79 return (*p
== '/' || *p
== '\\');
85 /* if filename is absolute, just copy it to dest. Otherwise, build a
86 path to it by considering it is relative to base_path. URL are
88 void path_combine(char *dest
, int dest_size
,
89 const char *base_path
,
97 if (path_is_absolute(filename
)) {
98 pstrcpy(dest
, dest_size
, filename
);
100 p
= strchr(base_path
, ':');
105 p1
= strrchr(base_path
, '/');
109 p2
= strrchr(base_path
, '\\');
121 if (len
> dest_size
- 1)
123 memcpy(dest
, base_path
, len
);
125 pstrcat(dest
, dest_size
, filename
);
129 void bdrv_register(BlockDriver
*bdrv
)
131 if (!bdrv
->bdrv_aio_readv
) {
132 /* add AIO emulation layer */
133 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
134 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
135 } else if (!bdrv
->bdrv_read
) {
136 /* add synchronous IO emulation layer */
137 bdrv
->bdrv_read
= bdrv_read_em
;
138 bdrv
->bdrv_write
= bdrv_write_em
;
141 if (!bdrv
->bdrv_aio_flush
)
142 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
144 bdrv
->next
= first_drv
;
148 /* create a new block device (by default it is empty) */
149 BlockDriverState
*bdrv_new(const char *device_name
)
151 BlockDriverState
**pbs
, *bs
;
153 bs
= qemu_mallocz(sizeof(BlockDriverState
));
154 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
155 if (device_name
[0] != '\0') {
156 /* insert at the end */
165 BlockDriver
*bdrv_find_format(const char *format_name
)
168 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
169 if (!strcmp(drv1
->format_name
, format_name
))
175 static int bdrv_is_whitelisted(BlockDriver
*drv
)
177 static const char *whitelist
[] = {
178 CONFIG_BDRV_WHITELIST
183 return 1; /* no whitelist, anything goes */
185 for (p
= whitelist
; *p
; p
++) {
186 if (!strcmp(drv
->format_name
, *p
)) {
193 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
195 BlockDriver
*drv
= bdrv_find_format(format_name
);
196 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
199 int bdrv_create(BlockDriver
*drv
, const char* filename
,
200 QEMUOptionParameter
*options
)
202 if (!drv
->bdrv_create
)
205 return drv
->bdrv_create(filename
, options
);
209 void get_tmp_filename(char *filename
, int size
)
211 char temp_dir
[MAX_PATH
];
213 GetTempPath(MAX_PATH
, temp_dir
);
214 GetTempFileName(temp_dir
, "qem", 0, filename
);
217 void get_tmp_filename(char *filename
, int size
)
221 /* XXX: race condition possible */
222 tmpdir
= getenv("TMPDIR");
225 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
226 fd
= mkstemp(filename
);
232 static int is_windows_drive_prefix(const char *filename
)
234 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
235 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
239 int is_windows_drive(const char *filename
)
241 if (is_windows_drive_prefix(filename
) &&
244 if (strstart(filename
, "\\\\.\\", NULL
) ||
245 strstart(filename
, "//./", NULL
))
251 static BlockDriver
*find_protocol(const char *filename
)
259 if (is_windows_drive(filename
) ||
260 is_windows_drive_prefix(filename
))
261 return bdrv_find_format("raw");
263 p
= strchr(filename
, ':');
265 return bdrv_find_format("raw");
267 if (len
> sizeof(protocol
) - 1)
268 len
= sizeof(protocol
) - 1;
269 memcpy(protocol
, filename
, len
);
270 protocol
[len
] = '\0';
271 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
272 if (drv1
->protocol_name
&&
273 !strcmp(drv1
->protocol_name
, protocol
))
280 * Detect host devices. By convention, /dev/cdrom[N] is always
281 * recognized as a host CDROM.
283 static BlockDriver
*find_hdev_driver(const char *filename
)
285 int score_max
= 0, score
;
286 BlockDriver
*drv
= NULL
, *d
;
288 for (d
= first_drv
; d
; d
= d
->next
) {
289 if (d
->bdrv_probe_device
) {
290 score
= d
->bdrv_probe_device(filename
);
291 if (score
> score_max
) {
301 static BlockDriver
*find_image_format(const char *filename
)
303 int ret
, score
, score_max
;
304 BlockDriver
*drv1
, *drv
;
306 BlockDriverState
*bs
;
308 drv
= find_protocol(filename
);
309 /* no need to test disk image formats for vvfat */
310 if (drv
&& strcmp(drv
->format_name
, "vvfat") == 0)
313 ret
= bdrv_file_open(&bs
, filename
, 0);
316 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
323 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
324 if (drv1
->bdrv_probe
) {
325 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
326 if (score
> score_max
) {
335 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
337 BlockDriverState
*bs
;
341 drv
= find_protocol(filename
);
347 ret
= bdrv_open(bs
, filename
, flags
, drv
);
357 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
361 char tmp_filename
[PATH_MAX
];
362 char backing_filename
[PATH_MAX
];
364 bs
->is_temporary
= 0;
367 bs
->open_flags
= flags
;
368 /* buffer_alignment defaulted to 512, drivers can change this value */
369 bs
->buffer_alignment
= 512;
371 if (flags
& BDRV_O_SNAPSHOT
) {
372 BlockDriverState
*bs1
;
375 BlockDriver
*bdrv_qcow2
;
376 QEMUOptionParameter
*options
;
378 /* if snapshot, we create a temporary backing file and open it
379 instead of opening 'filename' directly */
381 /* if there is a backing file, use it */
383 ret
= bdrv_open(bs1
, filename
, 0, drv
);
388 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
390 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
395 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
397 /* Real path is meaningless for protocols */
399 snprintf(backing_filename
, sizeof(backing_filename
),
401 else if (!realpath(filename
, backing_filename
))
404 bdrv_qcow2
= bdrv_find_format("qcow2");
405 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
407 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
408 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
410 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
414 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
419 filename
= tmp_filename
;
421 bs
->is_temporary
= 1;
424 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
427 drv
= find_hdev_driver(filename
);
429 drv
= find_image_format(filename
);
435 goto unlink_and_fail
;
437 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
439 goto unlink_and_fail
;
443 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
446 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
447 * write cache to the guest. We do need the fdatasync to flush
448 * out transactions for block allocations, and we maybe have a
449 * volatile write cache in our backing device to deal with.
451 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
452 bs
->enable_write_cache
= 1;
455 * Clear flags that are internal to the block layer before opening the
458 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
461 * Snapshots should be writeable.
463 if (bs
->is_temporary
) {
464 open_flags
|= BDRV_O_RDWR
;
467 ret
= drv
->bdrv_open(bs
, filename
, open_flags
);
472 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
473 if (drv
->bdrv_getlength
) {
474 bs
->total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
477 if (bs
->is_temporary
) {
481 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
482 /* if there is a backing file, use it */
483 BlockDriver
*back_drv
= NULL
;
484 bs
->backing_hd
= bdrv_new("");
485 path_combine(backing_filename
, sizeof(backing_filename
),
486 filename
, bs
->backing_file
);
487 if (bs
->backing_format
[0] != '\0')
488 back_drv
= bdrv_find_format(bs
->backing_format
);
490 /* backing files always opened read-only */
491 open_flags
&= ~BDRV_O_RDWR
;
493 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, open_flags
, back_drv
);
498 if (bs
->is_temporary
) {
499 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
501 /* base image inherits from "parent" */
502 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
506 if (!bdrv_key_required(bs
)) {
507 /* call the change callback */
508 bs
->media_changed
= 1;
510 bs
->change_cb(bs
->change_opaque
);
515 qemu_free(bs
->opaque
);
519 if (bs
->is_temporary
)
524 void bdrv_close(BlockDriverState
*bs
)
528 bdrv_delete(bs
->backing_hd
);
529 bs
->drv
->bdrv_close(bs
);
530 qemu_free(bs
->opaque
);
532 if (bs
->is_temporary
) {
533 unlink(bs
->filename
);
539 /* call the change callback */
540 bs
->media_changed
= 1;
542 bs
->change_cb(bs
->change_opaque
);
546 void bdrv_delete(BlockDriverState
*bs
)
548 BlockDriverState
**pbs
;
551 while (*pbs
!= bs
&& *pbs
!= NULL
)
561 * Run consistency checks on an image
563 * Returns the number of errors or -errno when an internal error occurs
565 int bdrv_check(BlockDriverState
*bs
)
567 if (bs
->drv
->bdrv_check
== NULL
) {
571 return bs
->drv
->bdrv_check(bs
);
574 /* commit COW file into the raw image */
575 int bdrv_commit(BlockDriverState
*bs
)
577 BlockDriver
*drv
= bs
->drv
;
578 int64_t i
, total_sectors
;
579 int n
, j
, ro
, open_flags
;
580 int ret
= 0, rw_ret
= 0;
581 unsigned char sector
[512];
583 BlockDriverState
*bs_rw
, *bs_ro
;
588 if (!bs
->backing_hd
) {
592 if (bs
->backing_hd
->keep_read_only
) {
596 ro
= bs
->backing_hd
->read_only
;
597 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
598 open_flags
= bs
->backing_hd
->open_flags
;
602 bdrv_delete(bs
->backing_hd
);
603 bs
->backing_hd
= NULL
;
604 bs_rw
= bdrv_new("");
605 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, NULL
);
608 /* try to re-open read-only */
609 bs_ro
= bdrv_new("");
610 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
613 /* drive not functional anymore */
617 bs
->backing_hd
= bs_ro
;
620 bs
->backing_hd
= bs_rw
;
623 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
624 for (i
= 0; i
< total_sectors
;) {
625 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
626 for(j
= 0; j
< n
; j
++) {
627 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
632 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
643 if (drv
->bdrv_make_empty
) {
644 ret
= drv
->bdrv_make_empty(bs
);
649 * Make sure all data we wrote to the backing device is actually
653 bdrv_flush(bs
->backing_hd
);
659 bdrv_delete(bs
->backing_hd
);
660 bs
->backing_hd
= NULL
;
661 bs_ro
= bdrv_new("");
662 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
665 /* drive not functional anymore */
669 bs
->backing_hd
= bs_ro
;
670 bs
->backing_hd
->keep_read_only
= 0;
679 * -EINVAL - backing format specified, but no file
680 * -ENOSPC - can't update the backing file because no space is left in the
682 * -ENOTSUP - format driver doesn't support changing the backing file
684 int bdrv_change_backing_file(BlockDriverState
*bs
,
685 const char *backing_file
, const char *backing_fmt
)
687 BlockDriver
*drv
= bs
->drv
;
689 if (drv
->bdrv_change_backing_file
!= NULL
) {
690 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
696 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
701 if (!bdrv_is_inserted(bs
))
707 len
= bdrv_getlength(bs
);
712 if ((offset
> len
) || (len
- offset
< size
))
718 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
721 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
724 /* return < 0 if error. See bdrv_write() for the return codes */
725 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
726 uint8_t *buf
, int nb_sectors
)
728 BlockDriver
*drv
= bs
->drv
;
732 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
735 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
738 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
739 int nb_sectors
, int dirty
)
742 unsigned long val
, idx
, bit
;
744 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
745 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
747 for (; start
<= end
; start
++) {
748 idx
= start
/ (sizeof(unsigned long) * 8);
749 bit
= start
% (sizeof(unsigned long) * 8);
750 val
= bs
->dirty_bitmap
[idx
];
752 if (!(val
& (1 << bit
))) {
757 if (val
& (1 << bit
)) {
762 bs
->dirty_bitmap
[idx
] = val
;
766 /* Return < 0 if error. Important errors are:
767 -EIO generic I/O error (may happen for all errors)
768 -ENOMEDIUM No media inserted.
769 -EINVAL Invalid sector number or nb_sectors
770 -EACCES Trying to write a read-only device
772 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
773 const uint8_t *buf
, int nb_sectors
)
775 BlockDriver
*drv
= bs
->drv
;
780 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
783 if (bs
->dirty_bitmap
) {
784 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
787 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
790 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
791 void *buf
, int count1
)
793 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
794 int len
, nb_sectors
, count
;
799 /* first read to align to sector start */
800 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
803 sector_num
= offset
>> BDRV_SECTOR_BITS
;
805 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
807 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
815 /* read the sectors "in place" */
816 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
817 if (nb_sectors
> 0) {
818 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
820 sector_num
+= nb_sectors
;
821 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
826 /* add data from the last sector */
828 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
830 memcpy(buf
, tmp_buf
, count
);
835 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
836 const void *buf
, int count1
)
838 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
839 int len
, nb_sectors
, count
;
844 /* first write to align to sector start */
845 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
848 sector_num
= offset
>> BDRV_SECTOR_BITS
;
850 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
852 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
853 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
862 /* write the sectors "in place" */
863 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
864 if (nb_sectors
> 0) {
865 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
867 sector_num
+= nb_sectors
;
868 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
873 /* add data from the last sector */
875 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
877 memcpy(tmp_buf
, buf
, count
);
878 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
885 * Truncate file to 'offset' bytes (needed only for file protocols)
887 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
889 BlockDriver
*drv
= bs
->drv
;
892 if (!drv
->bdrv_truncate
)
896 return drv
->bdrv_truncate(bs
, offset
);
900 * Length of a file in bytes. Return < 0 if error or unknown.
902 int64_t bdrv_getlength(BlockDriverState
*bs
)
904 BlockDriver
*drv
= bs
->drv
;
907 if (!drv
->bdrv_getlength
) {
909 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
911 return drv
->bdrv_getlength(bs
);
914 /* return 0 as number of sectors if no device present or error */
915 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
918 length
= bdrv_getlength(bs
);
922 length
= length
>> BDRV_SECTOR_BITS
;
923 *nb_sectors_ptr
= length
;
927 uint8_t boot_ind
; /* 0x80 - active */
928 uint8_t head
; /* starting head */
929 uint8_t sector
; /* starting sector */
930 uint8_t cyl
; /* starting cylinder */
931 uint8_t sys_ind
; /* What partition type */
932 uint8_t end_head
; /* end head */
933 uint8_t end_sector
; /* end sector */
934 uint8_t end_cyl
; /* end cylinder */
935 uint32_t start_sect
; /* starting sector counting from 0 */
936 uint32_t nr_sects
; /* nr of sectors in partition */
937 } __attribute__((packed
));
939 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
940 static int guess_disk_lchs(BlockDriverState
*bs
,
941 int *pcylinders
, int *pheads
, int *psectors
)
944 int ret
, i
, heads
, sectors
, cylinders
;
949 bdrv_get_geometry(bs
, &nb_sectors
);
951 ret
= bdrv_read(bs
, 0, buf
, 1);
954 /* test msdos magic */
955 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
957 for(i
= 0; i
< 4; i
++) {
958 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
959 nr_sects
= le32_to_cpu(p
->nr_sects
);
960 if (nr_sects
&& p
->end_head
) {
961 /* We make the assumption that the partition terminates on
962 a cylinder boundary */
963 heads
= p
->end_head
+ 1;
964 sectors
= p
->end_sector
& 63;
967 cylinders
= nb_sectors
/ (heads
* sectors
);
968 if (cylinders
< 1 || cylinders
> 16383)
972 *pcylinders
= cylinders
;
974 printf("guessed geometry: LCHS=%d %d %d\n",
975 cylinders
, heads
, sectors
);
983 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
985 int translation
, lba_detected
= 0;
986 int cylinders
, heads
, secs
;
989 /* if a geometry hint is available, use it */
990 bdrv_get_geometry(bs
, &nb_sectors
);
991 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
992 translation
= bdrv_get_translation_hint(bs
);
993 if (cylinders
!= 0) {
998 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1000 /* if heads > 16, it means that a BIOS LBA
1001 translation was active, so the default
1002 hardware geometry is OK */
1004 goto default_geometry
;
1009 /* disable any translation to be in sync with
1010 the logical geometry */
1011 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1012 bdrv_set_translation_hint(bs
,
1013 BIOS_ATA_TRANSLATION_NONE
);
1018 /* if no geometry, use a standard physical disk geometry */
1019 cylinders
= nb_sectors
/ (16 * 63);
1021 if (cylinders
> 16383)
1023 else if (cylinders
< 2)
1028 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1029 if ((*pcyls
* *pheads
) <= 131072) {
1030 bdrv_set_translation_hint(bs
,
1031 BIOS_ATA_TRANSLATION_LARGE
);
1033 bdrv_set_translation_hint(bs
,
1034 BIOS_ATA_TRANSLATION_LBA
);
1038 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1042 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1043 int cyls
, int heads
, int secs
)
1050 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1053 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1054 type
== BDRV_TYPE_FLOPPY
));
1057 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1059 bs
->translation
= translation
;
1062 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1063 int *pcyls
, int *pheads
, int *psecs
)
1066 *pheads
= bs
->heads
;
1070 int bdrv_get_type_hint(BlockDriverState
*bs
)
1075 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1077 return bs
->translation
;
1080 int bdrv_is_removable(BlockDriverState
*bs
)
1082 return bs
->removable
;
1085 int bdrv_is_read_only(BlockDriverState
*bs
)
1087 return bs
->read_only
;
1090 int bdrv_is_sg(BlockDriverState
*bs
)
1095 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1097 return bs
->enable_write_cache
;
1100 /* XXX: no longer used */
1101 void bdrv_set_change_cb(BlockDriverState
*bs
,
1102 void (*change_cb
)(void *opaque
), void *opaque
)
1104 bs
->change_cb
= change_cb
;
1105 bs
->change_opaque
= opaque
;
1108 int bdrv_is_encrypted(BlockDriverState
*bs
)
1110 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1112 return bs
->encrypted
;
1115 int bdrv_key_required(BlockDriverState
*bs
)
1117 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1119 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1121 return (bs
->encrypted
&& !bs
->valid_key
);
1124 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1127 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1128 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1134 if (!bs
->encrypted
) {
1136 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1139 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1142 } else if (!bs
->valid_key
) {
1144 /* call the change callback now, we skipped it on open */
1145 bs
->media_changed
= 1;
1147 bs
->change_cb(bs
->change_opaque
);
1152 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1157 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1161 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1166 for (drv
= first_drv
; drv
!= NULL
; drv
= drv
->next
) {
1167 it(opaque
, drv
->format_name
);
1171 BlockDriverState
*bdrv_find(const char *name
)
1173 BlockDriverState
*bs
;
1175 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1176 if (!strcmp(name
, bs
->device_name
))
1182 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1184 BlockDriverState
*bs
;
1186 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1191 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1193 return bs
->device_name
;
1196 void bdrv_flush(BlockDriverState
*bs
)
1198 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1199 bs
->drv
->bdrv_flush(bs
);
1202 void bdrv_flush_all(void)
1204 BlockDriverState
*bs
;
1206 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
)
1207 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1208 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
)))
1213 * Returns true iff the specified sector is present in the disk image. Drivers
1214 * not implementing the functionality are assumed to not support backing files,
1215 * hence all their sectors are reported as allocated.
1217 * 'pnum' is set to the number of sectors (including and immediately following
1218 * the specified sector) that are known to be in the same
1219 * allocated/unallocated state.
1221 * 'nb_sectors' is the max value 'pnum' should be set to.
1223 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1227 if (!bs
->drv
->bdrv_is_allocated
) {
1228 if (sector_num
>= bs
->total_sectors
) {
1232 n
= bs
->total_sectors
- sector_num
;
1233 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1236 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1239 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1240 BlockMonEventAction action
, int is_read
)
1243 const char *action_str
;
1246 case BDRV_ACTION_REPORT
:
1247 action_str
= "report";
1249 case BDRV_ACTION_IGNORE
:
1250 action_str
= "ignore";
1252 case BDRV_ACTION_STOP
:
1253 action_str
= "stop";
1259 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1262 is_read
? "read" : "write");
1263 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1265 qobject_decref(data
);
1268 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1271 Monitor
*mon
= opaque
;
1273 bs_dict
= qobject_to_qdict(obj
);
1275 monitor_printf(mon
, "%s: type=%s removable=%d",
1276 qdict_get_str(bs_dict
, "device"),
1277 qdict_get_str(bs_dict
, "type"),
1278 qdict_get_bool(bs_dict
, "removable"));
1280 if (qdict_get_bool(bs_dict
, "removable")) {
1281 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1284 if (qdict_haskey(bs_dict
, "inserted")) {
1285 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1287 monitor_printf(mon
, " file=");
1288 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1289 if (qdict_haskey(qdict
, "backing_file")) {
1290 monitor_printf(mon
, " backing_file=");
1291 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1293 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1294 qdict_get_bool(qdict
, "ro"),
1295 qdict_get_str(qdict
, "drv"),
1296 qdict_get_bool(qdict
, "encrypted"));
1298 monitor_printf(mon
, " [not inserted]");
1301 monitor_printf(mon
, "\n");
1304 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1306 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1310 * bdrv_info(): Block devices information
1312 * Each block device information is stored in a QDict and the
1313 * returned QObject is a QList of all devices.
1315 * The QDict contains the following:
1317 * - "device": device name
1318 * - "type": device type
1319 * - "removable": true if the device is removable, false otherwise
1320 * - "locked": true if the device is locked, false otherwise
1321 * - "inserted": only present if the device is inserted, it is a QDict
1322 * containing the following:
1323 * - "file": device file name
1324 * - "ro": true if read-only, false otherwise
1325 * - "drv": driver format name
1326 * - "backing_file": backing file name if one is used
1327 * - "encrypted": true if encrypted, false otherwise
1331 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1332 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1333 * { "device": "floppy0", "type": "floppy", "removable": true,
1334 * "locked": false } ]
1336 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1339 BlockDriverState
*bs
;
1341 bs_list
= qlist_new();
1343 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1345 const char *type
= "unknown";
1351 case BDRV_TYPE_CDROM
:
1354 case BDRV_TYPE_FLOPPY
:
1359 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1360 "'removable': %i, 'locked': %i }",
1361 bs
->device_name
, type
, bs
->removable
,
1366 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1368 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1369 "'encrypted': %i }",
1370 bs
->filename
, bs
->read_only
,
1371 bs
->drv
->format_name
,
1372 bdrv_is_encrypted(bs
));
1373 if (bs
->backing_file
[0] != '\0') {
1374 QDict
*qdict
= qobject_to_qdict(obj
);
1375 qdict_put(qdict
, "backing_file",
1376 qstring_from_str(bs
->backing_file
));
1379 qdict_put_obj(bs_dict
, "inserted", obj
);
1381 qlist_append_obj(bs_list
, bs_obj
);
1384 *ret_data
= QOBJECT(bs_list
);
1387 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1390 Monitor
*mon
= opaque
;
1392 qdict
= qobject_to_qdict(data
);
1393 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1395 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1396 monitor_printf(mon
, " rd_bytes=%" PRId64
1397 " wr_bytes=%" PRId64
1398 " rd_operations=%" PRId64
1399 " wr_operations=%" PRId64
1401 qdict_get_int(qdict
, "rd_bytes"),
1402 qdict_get_int(qdict
, "wr_bytes"),
1403 qdict_get_int(qdict
, "rd_operations"),
1404 qdict_get_int(qdict
, "wr_operations"));
1407 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1409 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1413 * bdrv_info_stats(): show block device statistics
1415 * Each device statistic information is stored in a QDict and
1416 * the returned QObject is a QList of all devices.
1418 * The QDict contains the following:
1420 * - "device": device name
1421 * - "stats": A QDict with the statistics information, it contains:
1422 * - "rd_bytes": bytes read
1423 * - "wr_bytes": bytes written
1424 * - "rd_operations": read operations
1425 * - "wr_operations": write operations
1429 * [ { "device": "ide0-hd0",
1430 * "stats": { "rd_bytes": 512,
1432 * "rd_operations": 1,
1433 * "wr_operations": 0 } },
1434 * { "device": "ide1-cd0",
1435 * "stats": { "rd_bytes": 0,
1437 * "rd_operations": 0,
1438 * "wr_operations": 0 } } ]
1440 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1444 BlockDriverState
*bs
;
1446 devices
= qlist_new();
1448 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1449 obj
= qobject_from_jsonf("{ 'device': %s, 'stats': {"
1450 "'rd_bytes': %" PRId64
","
1451 "'wr_bytes': %" PRId64
","
1452 "'rd_operations': %" PRId64
","
1453 "'wr_operations': %" PRId64
1456 bs
->rd_bytes
, bs
->wr_bytes
,
1457 bs
->rd_ops
, bs
->wr_ops
);
1458 qlist_append_obj(devices
, obj
);
1461 *ret_data
= QOBJECT(devices
);
1464 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1466 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1467 return bs
->backing_file
;
1468 else if (bs
->encrypted
)
1469 return bs
->filename
;
1474 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1475 char *filename
, int filename_size
)
1477 if (!bs
->backing_file
) {
1478 pstrcpy(filename
, filename_size
, "");
1480 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1484 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1485 const uint8_t *buf
, int nb_sectors
)
1487 BlockDriver
*drv
= bs
->drv
;
1490 if (!drv
->bdrv_write_compressed
)
1492 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1495 if (bs
->dirty_bitmap
) {
1496 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1499 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1502 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1504 BlockDriver
*drv
= bs
->drv
;
1507 if (!drv
->bdrv_get_info
)
1509 memset(bdi
, 0, sizeof(*bdi
));
1510 return drv
->bdrv_get_info(bs
, bdi
);
1513 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1514 int64_t pos
, int size
)
1516 BlockDriver
*drv
= bs
->drv
;
1519 if (!drv
->bdrv_save_vmstate
)
1521 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1524 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1525 int64_t pos
, int size
)
1527 BlockDriver
*drv
= bs
->drv
;
1530 if (!drv
->bdrv_load_vmstate
)
1532 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1535 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1537 BlockDriver
*drv
= bs
->drv
;
1539 if (!drv
|| !drv
->bdrv_debug_event
) {
1543 return drv
->bdrv_debug_event(bs
, event
);
1547 /**************************************************************/
1548 /* handling of snapshots */
1550 int bdrv_snapshot_create(BlockDriverState
*bs
,
1551 QEMUSnapshotInfo
*sn_info
)
1553 BlockDriver
*drv
= bs
->drv
;
1556 if (!drv
->bdrv_snapshot_create
)
1558 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1561 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1562 const char *snapshot_id
)
1564 BlockDriver
*drv
= bs
->drv
;
1567 if (!drv
->bdrv_snapshot_goto
)
1569 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1572 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1574 BlockDriver
*drv
= bs
->drv
;
1577 if (!drv
->bdrv_snapshot_delete
)
1579 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1582 int bdrv_snapshot_list(BlockDriverState
*bs
,
1583 QEMUSnapshotInfo
**psn_info
)
1585 BlockDriver
*drv
= bs
->drv
;
1588 if (!drv
->bdrv_snapshot_list
)
1590 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1593 #define NB_SUFFIXES 4
1595 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1597 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1602 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1605 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1606 if (size
< (10 * base
)) {
1607 snprintf(buf
, buf_size
, "%0.1f%c",
1608 (double)size
/ base
,
1611 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1612 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1613 ((size
+ (base
>> 1)) / base
),
1623 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1625 char buf1
[128], date_buf
[128], clock_buf
[128];
1635 snprintf(buf
, buf_size
,
1636 "%-10s%-20s%7s%20s%15s",
1637 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1641 ptm
= localtime(&ti
);
1642 strftime(date_buf
, sizeof(date_buf
),
1643 "%Y-%m-%d %H:%M:%S", ptm
);
1645 localtime_r(&ti
, &tm
);
1646 strftime(date_buf
, sizeof(date_buf
),
1647 "%Y-%m-%d %H:%M:%S", &tm
);
1649 secs
= sn
->vm_clock_nsec
/ 1000000000;
1650 snprintf(clock_buf
, sizeof(clock_buf
),
1651 "%02d:%02d:%02d.%03d",
1653 (int)((secs
/ 60) % 60),
1655 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1656 snprintf(buf
, buf_size
,
1657 "%-10s%-20s%7s%20s%15s",
1658 sn
->id_str
, sn
->name
,
1659 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1667 /**************************************************************/
1670 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1671 QEMUIOVector
*qiov
, int nb_sectors
,
1672 BlockDriverCompletionFunc
*cb
, void *opaque
)
1674 BlockDriver
*drv
= bs
->drv
;
1675 BlockDriverAIOCB
*ret
;
1679 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1682 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1686 /* Update stats even though technically transfer has not happened. */
1687 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1694 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1695 QEMUIOVector
*qiov
, int nb_sectors
,
1696 BlockDriverCompletionFunc
*cb
, void *opaque
)
1698 BlockDriver
*drv
= bs
->drv
;
1699 BlockDriverAIOCB
*ret
;
1705 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1708 if (bs
->dirty_bitmap
) {
1709 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1712 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1716 /* Update stats even though technically transfer has not happened. */
1717 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1725 typedef struct MultiwriteCB
{
1730 BlockDriverCompletionFunc
*cb
;
1732 QEMUIOVector
*free_qiov
;
1737 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1741 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1742 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1743 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1744 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1748 static void multiwrite_cb(void *opaque
, int ret
)
1750 MultiwriteCB
*mcb
= opaque
;
1752 if (ret
< 0 && !mcb
->error
) {
1754 multiwrite_user_cb(mcb
);
1757 mcb
->num_requests
--;
1758 if (mcb
->num_requests
== 0) {
1759 if (mcb
->error
== 0) {
1760 multiwrite_user_cb(mcb
);
1766 static int multiwrite_req_compare(const void *a
, const void *b
)
1768 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1772 * Takes a bunch of requests and tries to merge them. Returns the number of
1773 * requests that remain after merging.
1775 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1776 int num_reqs
, MultiwriteCB
*mcb
)
1780 // Sort requests by start sector
1781 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1783 // Check if adjacent requests touch the same clusters. If so, combine them,
1784 // filling up gaps with zero sectors.
1786 for (i
= 1; i
< num_reqs
; i
++) {
1788 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1790 // This handles the cases that are valid for all block drivers, namely
1791 // exactly sequential writes and overlapping writes.
1792 if (reqs
[i
].sector
<= oldreq_last
) {
1796 // The block driver may decide that it makes sense to combine requests
1797 // even if there is a gap of some sectors between them. In this case,
1798 // the gap is filled with zeros (therefore only applicable for yet
1799 // unused space in format like qcow2).
1800 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1801 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1804 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
1810 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1811 qemu_iovec_init(qiov
,
1812 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1814 // Add the first request to the merged one. If the requests are
1815 // overlapping, drop the last sectors of the first request.
1816 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1817 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1819 // We might need to add some zeros between the two requests
1820 if (reqs
[i
].sector
> oldreq_last
) {
1821 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1822 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1823 memset(buf
, 0, zero_bytes
);
1824 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1825 mcb
->callbacks
[i
].free_buf
= buf
;
1828 // Add the second request
1829 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
1831 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
1832 reqs
[outidx
].qiov
= qiov
;
1834 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
1837 reqs
[outidx
].sector
= reqs
[i
].sector
;
1838 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
1839 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
1847 * Submit multiple AIO write requests at once.
1849 * On success, the function returns 0 and all requests in the reqs array have
1850 * been submitted. In error case this function returns -1, and any of the
1851 * requests may or may not be submitted yet. In particular, this means that the
1852 * callback will be called for some of the requests, for others it won't. The
1853 * caller must check the error field of the BlockRequest to wait for the right
1854 * callbacks (if error != 0, no callback will be called).
1856 * The implementation may modify the contents of the reqs array, e.g. to merge
1857 * requests. However, the fields opaque and error are left unmodified as they
1858 * are used to signal failure for a single request to the caller.
1860 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
1862 BlockDriverAIOCB
*acb
;
1866 if (num_reqs
== 0) {
1870 // Create MultiwriteCB structure
1871 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
1872 mcb
->num_requests
= 0;
1873 mcb
->num_callbacks
= num_reqs
;
1875 for (i
= 0; i
< num_reqs
; i
++) {
1876 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
1877 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
1880 // Check for mergable requests
1881 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
1883 // Run the aio requests
1884 for (i
= 0; i
< num_reqs
; i
++) {
1885 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
1886 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
1889 // We can only fail the whole thing if no request has been
1890 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1891 // complete and report the error in the callback.
1892 if (mcb
->num_requests
== 0) {
1893 reqs
[i
].error
= -EIO
;
1896 mcb
->num_requests
++;
1897 multiwrite_cb(mcb
, -EIO
);
1901 mcb
->num_requests
++;
1912 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
1913 BlockDriverCompletionFunc
*cb
, void *opaque
)
1915 BlockDriver
*drv
= bs
->drv
;
1919 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
1922 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
1924 acb
->pool
->cancel(acb
);
1928 /**************************************************************/
1929 /* async block device emulation */
1931 typedef struct BlockDriverAIOCBSync
{
1932 BlockDriverAIOCB common
;
1935 /* vector translation state */
1939 } BlockDriverAIOCBSync
;
1941 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
1943 BlockDriverAIOCBSync
*acb
= (BlockDriverAIOCBSync
*)blockacb
;
1944 qemu_bh_delete(acb
->bh
);
1946 qemu_aio_release(acb
);
1949 static AIOPool bdrv_em_aio_pool
= {
1950 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
1951 .cancel
= bdrv_aio_cancel_em
,
1954 static void bdrv_aio_bh_cb(void *opaque
)
1956 BlockDriverAIOCBSync
*acb
= opaque
;
1959 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
1960 qemu_vfree(acb
->bounce
);
1961 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
1962 qemu_bh_delete(acb
->bh
);
1964 qemu_aio_release(acb
);
1967 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
1971 BlockDriverCompletionFunc
*cb
,
1976 BlockDriverAIOCBSync
*acb
;
1978 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
1979 acb
->is_write
= is_write
;
1981 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
1984 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
1987 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
1988 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1990 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1993 qemu_bh_schedule(acb
->bh
);
1995 return &acb
->common
;
1998 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
1999 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2000 BlockDriverCompletionFunc
*cb
, void *opaque
)
2002 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2005 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2006 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2007 BlockDriverCompletionFunc
*cb
, void *opaque
)
2009 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2012 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2013 BlockDriverCompletionFunc
*cb
, void *opaque
)
2015 BlockDriverAIOCBSync
*acb
;
2017 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2018 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2024 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2027 qemu_bh_schedule(acb
->bh
);
2028 return &acb
->common
;
2031 /**************************************************************/
2032 /* sync block device emulation */
2034 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2036 *(int *)opaque
= ret
;
2039 #define NOT_DONE 0x7fffffff
2041 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2042 uint8_t *buf
, int nb_sectors
)
2045 BlockDriverAIOCB
*acb
;
2049 async_context_push();
2051 async_ret
= NOT_DONE
;
2052 iov
.iov_base
= (void *)buf
;
2053 iov
.iov_len
= nb_sectors
* 512;
2054 qemu_iovec_init_external(&qiov
, &iov
, 1);
2055 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2056 bdrv_rw_em_cb
, &async_ret
);
2062 while (async_ret
== NOT_DONE
) {
2068 async_context_pop();
2072 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2073 const uint8_t *buf
, int nb_sectors
)
2076 BlockDriverAIOCB
*acb
;
2080 async_context_push();
2082 async_ret
= NOT_DONE
;
2083 iov
.iov_base
= (void *)buf
;
2084 iov
.iov_len
= nb_sectors
* 512;
2085 qemu_iovec_init_external(&qiov
, &iov
, 1);
2086 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2087 bdrv_rw_em_cb
, &async_ret
);
2092 while (async_ret
== NOT_DONE
) {
2097 async_context_pop();
2101 void bdrv_init(void)
2103 module_call_init(MODULE_INIT_BLOCK
);
2106 void bdrv_init_with_whitelist(void)
2108 use_bdrv_whitelist
= 1;
2112 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2113 BlockDriverCompletionFunc
*cb
, void *opaque
)
2115 BlockDriverAIOCB
*acb
;
2117 if (pool
->free_aiocb
) {
2118 acb
= pool
->free_aiocb
;
2119 pool
->free_aiocb
= acb
->next
;
2121 acb
= qemu_mallocz(pool
->aiocb_size
);
2126 acb
->opaque
= opaque
;
2130 void qemu_aio_release(void *p
)
2132 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2133 AIOPool
*pool
= acb
->pool
;
2134 acb
->next
= pool
->free_aiocb
;
2135 pool
->free_aiocb
= acb
;
2138 /**************************************************************/
2139 /* removable device support */
2142 * Return TRUE if the media is present
2144 int bdrv_is_inserted(BlockDriverState
*bs
)
2146 BlockDriver
*drv
= bs
->drv
;
2150 if (!drv
->bdrv_is_inserted
)
2152 ret
= drv
->bdrv_is_inserted(bs
);
2157 * Return TRUE if the media changed since the last call to this
2158 * function. It is currently only used for floppy disks
2160 int bdrv_media_changed(BlockDriverState
*bs
)
2162 BlockDriver
*drv
= bs
->drv
;
2165 if (!drv
|| !drv
->bdrv_media_changed
)
2168 ret
= drv
->bdrv_media_changed(bs
);
2169 if (ret
== -ENOTSUP
)
2170 ret
= bs
->media_changed
;
2171 bs
->media_changed
= 0;
2176 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2178 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2180 BlockDriver
*drv
= bs
->drv
;
2187 if (!drv
|| !drv
->bdrv_eject
) {
2190 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2192 if (ret
== -ENOTSUP
) {
2201 int bdrv_is_locked(BlockDriverState
*bs
)
2207 * Lock or unlock the media (if it is locked, the user won't be able
2208 * to eject it manually).
2210 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2212 BlockDriver
*drv
= bs
->drv
;
2214 bs
->locked
= locked
;
2215 if (drv
&& drv
->bdrv_set_locked
) {
2216 drv
->bdrv_set_locked(bs
, locked
);
2220 /* needed for generic scsi interface */
2222 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2224 BlockDriver
*drv
= bs
->drv
;
2226 if (drv
&& drv
->bdrv_ioctl
)
2227 return drv
->bdrv_ioctl(bs
, req
, buf
);
2231 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2232 unsigned long int req
, void *buf
,
2233 BlockDriverCompletionFunc
*cb
, void *opaque
)
2235 BlockDriver
*drv
= bs
->drv
;
2237 if (drv
&& drv
->bdrv_aio_ioctl
)
2238 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2244 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2246 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2249 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2251 int64_t bitmap_size
;
2253 bs
->dirty_count
= 0;
2255 if (!bs
->dirty_bitmap
) {
2256 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2257 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2258 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2260 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2263 if (bs
->dirty_bitmap
) {
2264 qemu_free(bs
->dirty_bitmap
);
2265 bs
->dirty_bitmap
= NULL
;
2270 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2272 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2274 if (bs
->dirty_bitmap
&&
2275 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2276 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2277 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2283 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2286 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2289 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2291 return bs
->dirty_count
;