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 BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
54 BlockDriverCompletionFunc
*cb
, void *opaque
);
55 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
56 uint8_t *buf
, int nb_sectors
);
57 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
58 const uint8_t *buf
, int nb_sectors
);
60 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
61 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
63 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
64 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
66 /* If non-zero, use only whitelisted block drivers */
67 static int use_bdrv_whitelist
;
69 int path_is_absolute(const char *path
)
73 /* specific case for names like: "\\.\d:" */
74 if (*path
== '/' || *path
== '\\')
77 p
= strchr(path
, ':');
83 return (*p
== '/' || *p
== '\\');
89 /* if filename is absolute, just copy it to dest. Otherwise, build a
90 path to it by considering it is relative to base_path. URL are
92 void path_combine(char *dest
, int dest_size
,
93 const char *base_path
,
101 if (path_is_absolute(filename
)) {
102 pstrcpy(dest
, dest_size
, filename
);
104 p
= strchr(base_path
, ':');
109 p1
= strrchr(base_path
, '/');
113 p2
= strrchr(base_path
, '\\');
125 if (len
> dest_size
- 1)
127 memcpy(dest
, base_path
, len
);
129 pstrcat(dest
, dest_size
, filename
);
133 void bdrv_register(BlockDriver
*bdrv
)
135 if (!bdrv
->bdrv_aio_readv
) {
136 /* add AIO emulation layer */
137 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
138 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
139 } else if (!bdrv
->bdrv_read
) {
140 /* add synchronous IO emulation layer */
141 bdrv
->bdrv_read
= bdrv_read_em
;
142 bdrv
->bdrv_write
= bdrv_write_em
;
145 if (!bdrv
->bdrv_aio_flush
)
146 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
148 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
151 /* create a new block device (by default it is empty) */
152 BlockDriverState
*bdrv_new(const char *device_name
)
154 BlockDriverState
*bs
;
156 bs
= qemu_mallocz(sizeof(BlockDriverState
));
157 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
158 if (device_name
[0] != '\0') {
159 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
164 BlockDriver
*bdrv_find_format(const char *format_name
)
167 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
168 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
);
208 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
212 drv
= bdrv_find_protocol(filename
);
214 drv
= bdrv_find_format("file");
217 return bdrv_create(drv
, filename
, options
);
221 void get_tmp_filename(char *filename
, int size
)
223 char temp_dir
[MAX_PATH
];
225 GetTempPath(MAX_PATH
, temp_dir
);
226 GetTempFileName(temp_dir
, "qem", 0, filename
);
229 void get_tmp_filename(char *filename
, int size
)
233 /* XXX: race condition possible */
234 tmpdir
= getenv("TMPDIR");
237 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
238 fd
= mkstemp(filename
);
244 static int is_windows_drive_prefix(const char *filename
)
246 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
247 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
251 int is_windows_drive(const char *filename
)
253 if (is_windows_drive_prefix(filename
) &&
256 if (strstart(filename
, "\\\\.\\", NULL
) ||
257 strstart(filename
, "//./", NULL
))
264 * Detect host devices. By convention, /dev/cdrom[N] is always
265 * recognized as a host CDROM.
267 static BlockDriver
*find_hdev_driver(const char *filename
)
269 int score_max
= 0, score
;
270 BlockDriver
*drv
= NULL
, *d
;
272 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
273 if (d
->bdrv_probe_device
) {
274 score
= d
->bdrv_probe_device(filename
);
275 if (score
> score_max
) {
285 BlockDriver
*bdrv_find_protocol(const char *filename
)
292 /* TODO Drivers without bdrv_file_open must be specified explicitly */
295 * XXX(hch): we really should not let host device detection
296 * override an explicit protocol specification, but moving this
297 * later breaks access to device names with colons in them.
298 * Thanks to the brain-dead persistent naming schemes on udev-
299 * based Linux systems those actually are quite common.
301 drv1
= find_hdev_driver(filename
);
307 if (is_windows_drive(filename
) ||
308 is_windows_drive_prefix(filename
))
309 return bdrv_find_format("file");
312 p
= strchr(filename
, ':');
314 return bdrv_find_format("file");
317 if (len
> sizeof(protocol
) - 1)
318 len
= sizeof(protocol
) - 1;
319 memcpy(protocol
, filename
, len
);
320 protocol
[len
] = '\0';
321 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
322 if (drv1
->protocol_name
&&
323 !strcmp(drv1
->protocol_name
, protocol
)) {
330 static BlockDriver
*find_image_format(const char *filename
)
332 int ret
, score
, score_max
;
333 BlockDriver
*drv1
, *drv
;
335 BlockDriverState
*bs
;
337 ret
= bdrv_file_open(&bs
, filename
, 0);
341 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
342 if (bs
->sg
|| !bdrv_is_inserted(bs
)) {
344 return bdrv_find_format("raw");
347 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
355 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
356 if (drv1
->bdrv_probe
) {
357 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
358 if (score
> score_max
) {
368 * Set the current 'total_sectors' value
370 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
372 BlockDriver
*drv
= bs
->drv
;
374 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
378 /* query actual device if possible, otherwise just trust the hint */
379 if (drv
->bdrv_getlength
) {
380 int64_t length
= drv
->bdrv_getlength(bs
);
384 hint
= length
>> BDRV_SECTOR_BITS
;
387 bs
->total_sectors
= hint
;
392 * Common part for opening disk images and files
394 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
395 int flags
, BlockDriver
*drv
)
402 bs
->total_sectors
= 0;
405 bs
->open_flags
= flags
;
406 /* buffer_alignment defaulted to 512, drivers can change this value */
407 bs
->buffer_alignment
= 512;
409 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
411 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
416 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
419 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
420 * write cache to the guest. We do need the fdatasync to flush
421 * out transactions for block allocations, and we maybe have a
422 * volatile write cache in our backing device to deal with.
424 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
425 bs
->enable_write_cache
= 1;
428 * Clear flags that are internal to the block layer before opening the
431 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
434 * Snapshots should be writeable.
436 if (bs
->is_temporary
) {
437 open_flags
|= BDRV_O_RDWR
;
440 /* Open the image, either directly or using a protocol */
441 if (drv
->bdrv_file_open
) {
442 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
444 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
446 ret
= drv
->bdrv_open(bs
, open_flags
);
454 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
456 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
462 if (bs
->is_temporary
) {
470 bdrv_delete(bs
->file
);
473 qemu_free(bs
->opaque
);
480 * Opens a file using a protocol (file, host_device, nbd, ...)
482 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
484 BlockDriverState
*bs
;
488 drv
= bdrv_find_protocol(filename
);
494 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
505 * Opens a disk image (raw, qcow2, vmdk, ...)
507 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
512 if (flags
& BDRV_O_SNAPSHOT
) {
513 BlockDriverState
*bs1
;
516 BlockDriver
*bdrv_qcow2
;
517 QEMUOptionParameter
*options
;
518 char tmp_filename
[PATH_MAX
];
519 char backing_filename
[PATH_MAX
];
521 /* if snapshot, we create a temporary backing file and open it
522 instead of opening 'filename' directly */
524 /* if there is a backing file, use it */
526 ret
= bdrv_open(bs1
, filename
, 0, drv
);
531 total_size
= bdrv_getlength(bs1
) & BDRV_SECTOR_MASK
;
533 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
538 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
540 /* Real path is meaningless for protocols */
542 snprintf(backing_filename
, sizeof(backing_filename
),
544 else if (!realpath(filename
, backing_filename
))
547 bdrv_qcow2
= bdrv_find_format("qcow2");
548 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
550 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
);
551 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
553 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
557 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
558 free_option_parameters(options
);
563 filename
= tmp_filename
;
565 bs
->is_temporary
= 1;
568 /* Find the right image format driver */
570 drv
= find_image_format(filename
);
575 goto unlink_and_fail
;
579 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
581 goto unlink_and_fail
;
584 /* If there is a backing file, use it */
585 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
586 char backing_filename
[PATH_MAX
];
588 BlockDriver
*back_drv
= NULL
;
590 bs
->backing_hd
= bdrv_new("");
591 path_combine(backing_filename
, sizeof(backing_filename
),
592 filename
, bs
->backing_file
);
593 if (bs
->backing_format
[0] != '\0')
594 back_drv
= bdrv_find_format(bs
->backing_format
);
596 /* backing files always opened read-only */
598 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
600 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
605 if (bs
->is_temporary
) {
606 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
608 /* base image inherits from "parent" */
609 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
613 if (!bdrv_key_required(bs
)) {
614 /* call the change callback */
615 bs
->media_changed
= 1;
617 bs
->change_cb(bs
->change_opaque
);
623 if (bs
->is_temporary
) {
629 void bdrv_close(BlockDriverState
*bs
)
632 if (bs
->backing_hd
) {
633 bdrv_delete(bs
->backing_hd
);
634 bs
->backing_hd
= NULL
;
636 bs
->drv
->bdrv_close(bs
);
637 qemu_free(bs
->opaque
);
639 if (bs
->is_temporary
) {
640 unlink(bs
->filename
);
646 if (bs
->file
!= NULL
) {
647 bdrv_close(bs
->file
);
650 /* call the change callback */
651 bs
->media_changed
= 1;
653 bs
->change_cb(bs
->change_opaque
);
657 void bdrv_close_all(void)
659 BlockDriverState
*bs
;
661 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
666 void bdrv_delete(BlockDriverState
*bs
)
668 /* remove from list, if necessary */
669 if (bs
->device_name
[0] != '\0') {
670 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
674 if (bs
->file
!= NULL
) {
675 bdrv_delete(bs
->file
);
682 * Run consistency checks on an image
684 * Returns the number of errors or -errno when an internal error occurs
686 int bdrv_check(BlockDriverState
*bs
)
688 if (bs
->drv
->bdrv_check
== NULL
) {
692 return bs
->drv
->bdrv_check(bs
);
695 /* commit COW file into the raw image */
696 int bdrv_commit(BlockDriverState
*bs
)
698 BlockDriver
*drv
= bs
->drv
;
699 int64_t i
, total_sectors
;
700 int n
, j
, ro
, open_flags
;
701 int ret
= 0, rw_ret
= 0;
702 unsigned char sector
[BDRV_SECTOR_SIZE
];
704 BlockDriverState
*bs_rw
, *bs_ro
;
709 if (!bs
->backing_hd
) {
713 if (bs
->backing_hd
->keep_read_only
) {
717 ro
= bs
->backing_hd
->read_only
;
718 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
719 open_flags
= bs
->backing_hd
->open_flags
;
723 bdrv_delete(bs
->backing_hd
);
724 bs
->backing_hd
= NULL
;
725 bs_rw
= bdrv_new("");
726 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, drv
);
729 /* try to re-open read-only */
730 bs_ro
= bdrv_new("");
731 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, drv
);
734 /* drive not functional anymore */
738 bs
->backing_hd
= bs_ro
;
741 bs
->backing_hd
= bs_rw
;
744 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
745 for (i
= 0; i
< total_sectors
;) {
746 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
747 for(j
= 0; j
< n
; j
++) {
748 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
753 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
764 if (drv
->bdrv_make_empty
) {
765 ret
= drv
->bdrv_make_empty(bs
);
770 * Make sure all data we wrote to the backing device is actually
774 bdrv_flush(bs
->backing_hd
);
780 bdrv_delete(bs
->backing_hd
);
781 bs
->backing_hd
= NULL
;
782 bs_ro
= bdrv_new("");
783 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, drv
);
786 /* drive not functional anymore */
790 bs
->backing_hd
= bs_ro
;
791 bs
->backing_hd
->keep_read_only
= 0;
797 void bdrv_commit_all(void)
799 BlockDriverState
*bs
;
801 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
809 * -EINVAL - backing format specified, but no file
810 * -ENOSPC - can't update the backing file because no space is left in the
812 * -ENOTSUP - format driver doesn't support changing the backing file
814 int bdrv_change_backing_file(BlockDriverState
*bs
,
815 const char *backing_file
, const char *backing_fmt
)
817 BlockDriver
*drv
= bs
->drv
;
819 if (drv
->bdrv_change_backing_file
!= NULL
) {
820 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
826 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
831 if (!bdrv_is_inserted(bs
))
837 len
= bdrv_getlength(bs
);
842 if ((offset
> len
) || (len
- offset
< size
))
848 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
851 return bdrv_check_byte_request(bs
, sector_num
* BDRV_SECTOR_SIZE
,
852 nb_sectors
* BDRV_SECTOR_SIZE
);
855 /* return < 0 if error. See bdrv_write() for the return codes */
856 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
857 uint8_t *buf
, int nb_sectors
)
859 BlockDriver
*drv
= bs
->drv
;
863 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
866 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
869 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
870 int nb_sectors
, int dirty
)
873 unsigned long val
, idx
, bit
;
875 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
876 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
878 for (; start
<= end
; start
++) {
879 idx
= start
/ (sizeof(unsigned long) * 8);
880 bit
= start
% (sizeof(unsigned long) * 8);
881 val
= bs
->dirty_bitmap
[idx
];
883 if (!(val
& (1 << bit
))) {
888 if (val
& (1 << bit
)) {
893 bs
->dirty_bitmap
[idx
] = val
;
897 /* Return < 0 if error. Important errors are:
898 -EIO generic I/O error (may happen for all errors)
899 -ENOMEDIUM No media inserted.
900 -EINVAL Invalid sector number or nb_sectors
901 -EACCES Trying to write a read-only device
903 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
904 const uint8_t *buf
, int nb_sectors
)
906 BlockDriver
*drv
= bs
->drv
;
911 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
914 if (bs
->dirty_bitmap
) {
915 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
918 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
919 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
922 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
925 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
926 void *buf
, int count1
)
928 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
929 int len
, nb_sectors
, count
;
934 /* first read to align to sector start */
935 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
938 sector_num
= offset
>> BDRV_SECTOR_BITS
;
940 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
942 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
950 /* read the sectors "in place" */
951 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
952 if (nb_sectors
> 0) {
953 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
955 sector_num
+= nb_sectors
;
956 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
961 /* add data from the last sector */
963 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
965 memcpy(buf
, tmp_buf
, count
);
970 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
971 const void *buf
, int count1
)
973 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
974 int len
, nb_sectors
, count
;
979 /* first write to align to sector start */
980 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
983 sector_num
= offset
>> BDRV_SECTOR_BITS
;
985 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
987 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
988 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
997 /* write the sectors "in place" */
998 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
999 if (nb_sectors
> 0) {
1000 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1002 sector_num
+= nb_sectors
;
1003 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1008 /* add data from the last sector */
1010 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1012 memcpy(tmp_buf
, buf
, count
);
1013 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1020 * Writes to the file and ensures that no writes are reordered across this
1021 * request (acts as a barrier)
1023 * Returns 0 on success, -errno in error cases.
1025 int bdrv_pwrite_sync(BlockDriverState
*bs
, int64_t offset
,
1026 const void *buf
, int count
)
1030 ret
= bdrv_pwrite(bs
, offset
, buf
, count
);
1035 /* No flush needed for cache=writethrough, it uses O_DSYNC */
1036 if ((bs
->open_flags
& BDRV_O_CACHE_MASK
) != 0) {
1044 * Writes to the file and ensures that no writes are reordered across this
1045 * request (acts as a barrier)
1047 * Returns 0 on success, -errno in error cases.
1049 int bdrv_write_sync(BlockDriverState
*bs
, int64_t sector_num
,
1050 const uint8_t *buf
, int nb_sectors
)
1052 return bdrv_pwrite_sync(bs
, BDRV_SECTOR_SIZE
* sector_num
,
1053 buf
, BDRV_SECTOR_SIZE
* nb_sectors
);
1057 * Truncate file to 'offset' bytes (needed only for file protocols)
1059 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
1061 BlockDriver
*drv
= bs
->drv
;
1065 if (!drv
->bdrv_truncate
)
1069 ret
= drv
->bdrv_truncate(bs
, offset
);
1071 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1077 * Length of a file in bytes. Return < 0 if error or unknown.
1079 int64_t bdrv_getlength(BlockDriverState
*bs
)
1081 BlockDriver
*drv
= bs
->drv
;
1085 /* Fixed size devices use the total_sectors value for speed instead of
1086 issuing a length query (like lseek) on each call. Also, legacy block
1087 drivers don't provide a bdrv_getlength function and must use
1089 if (!bs
->growable
|| !drv
->bdrv_getlength
) {
1090 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1092 return drv
->bdrv_getlength(bs
);
1095 /* return 0 as number of sectors if no device present or error */
1096 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1099 length
= bdrv_getlength(bs
);
1103 length
= length
>> BDRV_SECTOR_BITS
;
1104 *nb_sectors_ptr
= length
;
1108 uint8_t boot_ind
; /* 0x80 - active */
1109 uint8_t head
; /* starting head */
1110 uint8_t sector
; /* starting sector */
1111 uint8_t cyl
; /* starting cylinder */
1112 uint8_t sys_ind
; /* What partition type */
1113 uint8_t end_head
; /* end head */
1114 uint8_t end_sector
; /* end sector */
1115 uint8_t end_cyl
; /* end cylinder */
1116 uint32_t start_sect
; /* starting sector counting from 0 */
1117 uint32_t nr_sects
; /* nr of sectors in partition */
1118 } __attribute__((packed
));
1120 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1121 static int guess_disk_lchs(BlockDriverState
*bs
,
1122 int *pcylinders
, int *pheads
, int *psectors
)
1124 uint8_t buf
[BDRV_SECTOR_SIZE
];
1125 int ret
, i
, heads
, sectors
, cylinders
;
1126 struct partition
*p
;
1128 uint64_t nb_sectors
;
1130 bdrv_get_geometry(bs
, &nb_sectors
);
1132 ret
= bdrv_read(bs
, 0, buf
, 1);
1135 /* test msdos magic */
1136 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1138 for(i
= 0; i
< 4; i
++) {
1139 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1140 nr_sects
= le32_to_cpu(p
->nr_sects
);
1141 if (nr_sects
&& p
->end_head
) {
1142 /* We make the assumption that the partition terminates on
1143 a cylinder boundary */
1144 heads
= p
->end_head
+ 1;
1145 sectors
= p
->end_sector
& 63;
1148 cylinders
= nb_sectors
/ (heads
* sectors
);
1149 if (cylinders
< 1 || cylinders
> 16383)
1152 *psectors
= sectors
;
1153 *pcylinders
= cylinders
;
1155 printf("guessed geometry: LCHS=%d %d %d\n",
1156 cylinders
, heads
, sectors
);
1164 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1166 int translation
, lba_detected
= 0;
1167 int cylinders
, heads
, secs
;
1168 uint64_t nb_sectors
;
1170 /* if a geometry hint is available, use it */
1171 bdrv_get_geometry(bs
, &nb_sectors
);
1172 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1173 translation
= bdrv_get_translation_hint(bs
);
1174 if (cylinders
!= 0) {
1179 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1181 /* if heads > 16, it means that a BIOS LBA
1182 translation was active, so the default
1183 hardware geometry is OK */
1185 goto default_geometry
;
1190 /* disable any translation to be in sync with
1191 the logical geometry */
1192 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1193 bdrv_set_translation_hint(bs
,
1194 BIOS_ATA_TRANSLATION_NONE
);
1199 /* if no geometry, use a standard physical disk geometry */
1200 cylinders
= nb_sectors
/ (16 * 63);
1202 if (cylinders
> 16383)
1204 else if (cylinders
< 2)
1209 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1210 if ((*pcyls
* *pheads
) <= 131072) {
1211 bdrv_set_translation_hint(bs
,
1212 BIOS_ATA_TRANSLATION_LARGE
);
1214 bdrv_set_translation_hint(bs
,
1215 BIOS_ATA_TRANSLATION_LBA
);
1219 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1223 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1224 int cyls
, int heads
, int secs
)
1231 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1234 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1235 type
== BDRV_TYPE_FLOPPY
));
1238 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1240 bs
->translation
= translation
;
1243 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1244 int *pcyls
, int *pheads
, int *psecs
)
1247 *pheads
= bs
->heads
;
1251 int bdrv_get_type_hint(BlockDriverState
*bs
)
1256 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1258 return bs
->translation
;
1261 void bdrv_set_on_error(BlockDriverState
*bs
, BlockErrorAction on_read_error
,
1262 BlockErrorAction on_write_error
)
1264 bs
->on_read_error
= on_read_error
;
1265 bs
->on_write_error
= on_write_error
;
1268 BlockErrorAction
bdrv_get_on_error(BlockDriverState
*bs
, int is_read
)
1270 return is_read
? bs
->on_read_error
: bs
->on_write_error
;
1273 int bdrv_is_removable(BlockDriverState
*bs
)
1275 return bs
->removable
;
1278 int bdrv_is_read_only(BlockDriverState
*bs
)
1280 return bs
->read_only
;
1283 int bdrv_is_sg(BlockDriverState
*bs
)
1288 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1290 return bs
->enable_write_cache
;
1293 /* XXX: no longer used */
1294 void bdrv_set_change_cb(BlockDriverState
*bs
,
1295 void (*change_cb
)(void *opaque
), void *opaque
)
1297 bs
->change_cb
= change_cb
;
1298 bs
->change_opaque
= opaque
;
1301 int bdrv_is_encrypted(BlockDriverState
*bs
)
1303 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1305 return bs
->encrypted
;
1308 int bdrv_key_required(BlockDriverState
*bs
)
1310 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1312 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1314 return (bs
->encrypted
&& !bs
->valid_key
);
1317 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1320 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1321 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1327 if (!bs
->encrypted
) {
1329 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1332 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1335 } else if (!bs
->valid_key
) {
1337 /* call the change callback now, we skipped it on open */
1338 bs
->media_changed
= 1;
1340 bs
->change_cb(bs
->change_opaque
);
1345 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1350 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1354 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1359 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1360 it(opaque
, drv
->format_name
);
1364 BlockDriverState
*bdrv_find(const char *name
)
1366 BlockDriverState
*bs
;
1368 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1369 if (!strcmp(name
, bs
->device_name
)) {
1376 BlockDriverState
*bdrv_next(BlockDriverState
*bs
)
1379 return QTAILQ_FIRST(&bdrv_states
);
1381 return QTAILQ_NEXT(bs
, list
);
1384 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1386 BlockDriverState
*bs
;
1388 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1393 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1395 return bs
->device_name
;
1398 void bdrv_flush(BlockDriverState
*bs
)
1400 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
1404 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1405 bs
->drv
->bdrv_flush(bs
);
1408 void bdrv_flush_all(void)
1410 BlockDriverState
*bs
;
1412 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1413 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1414 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1420 int bdrv_has_zero_init(BlockDriverState
*bs
)
1424 if (bs
->drv
->no_zero_init
) {
1426 } else if (bs
->file
) {
1427 return bdrv_has_zero_init(bs
->file
);
1434 * Returns true iff the specified sector is present in the disk image. Drivers
1435 * not implementing the functionality are assumed to not support backing files,
1436 * hence all their sectors are reported as allocated.
1438 * 'pnum' is set to the number of sectors (including and immediately following
1439 * the specified sector) that are known to be in the same
1440 * allocated/unallocated state.
1442 * 'nb_sectors' is the max value 'pnum' should be set to.
1444 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1448 if (!bs
->drv
->bdrv_is_allocated
) {
1449 if (sector_num
>= bs
->total_sectors
) {
1453 n
= bs
->total_sectors
- sector_num
;
1454 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1457 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1460 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1461 BlockMonEventAction action
, int is_read
)
1464 const char *action_str
;
1467 case BDRV_ACTION_REPORT
:
1468 action_str
= "report";
1470 case BDRV_ACTION_IGNORE
:
1471 action_str
= "ignore";
1473 case BDRV_ACTION_STOP
:
1474 action_str
= "stop";
1480 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1483 is_read
? "read" : "write");
1484 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1486 qobject_decref(data
);
1489 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1492 Monitor
*mon
= opaque
;
1494 bs_dict
= qobject_to_qdict(obj
);
1496 monitor_printf(mon
, "%s: type=%s removable=%d",
1497 qdict_get_str(bs_dict
, "device"),
1498 qdict_get_str(bs_dict
, "type"),
1499 qdict_get_bool(bs_dict
, "removable"));
1501 if (qdict_get_bool(bs_dict
, "removable")) {
1502 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1505 if (qdict_haskey(bs_dict
, "inserted")) {
1506 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1508 monitor_printf(mon
, " file=");
1509 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1510 if (qdict_haskey(qdict
, "backing_file")) {
1511 monitor_printf(mon
, " backing_file=");
1512 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1514 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1515 qdict_get_bool(qdict
, "ro"),
1516 qdict_get_str(qdict
, "drv"),
1517 qdict_get_bool(qdict
, "encrypted"));
1519 monitor_printf(mon
, " [not inserted]");
1522 monitor_printf(mon
, "\n");
1525 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1527 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1530 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1533 BlockDriverState
*bs
;
1535 bs_list
= qlist_new();
1537 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1539 const char *type
= "unknown";
1545 case BDRV_TYPE_CDROM
:
1548 case BDRV_TYPE_FLOPPY
:
1553 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1554 "'removable': %i, 'locked': %i }",
1555 bs
->device_name
, type
, bs
->removable
,
1560 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1562 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1563 "'encrypted': %i }",
1564 bs
->filename
, bs
->read_only
,
1565 bs
->drv
->format_name
,
1566 bdrv_is_encrypted(bs
));
1567 if (bs
->backing_file
[0] != '\0') {
1568 QDict
*qdict
= qobject_to_qdict(obj
);
1569 qdict_put(qdict
, "backing_file",
1570 qstring_from_str(bs
->backing_file
));
1573 qdict_put_obj(bs_dict
, "inserted", obj
);
1575 qlist_append_obj(bs_list
, bs_obj
);
1578 *ret_data
= QOBJECT(bs_list
);
1581 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1584 Monitor
*mon
= opaque
;
1586 qdict
= qobject_to_qdict(data
);
1587 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1589 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1590 monitor_printf(mon
, " rd_bytes=%" PRId64
1591 " wr_bytes=%" PRId64
1592 " rd_operations=%" PRId64
1593 " wr_operations=%" PRId64
1595 qdict_get_int(qdict
, "rd_bytes"),
1596 qdict_get_int(qdict
, "wr_bytes"),
1597 qdict_get_int(qdict
, "rd_operations"),
1598 qdict_get_int(qdict
, "wr_operations"));
1601 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1603 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1606 static QObject
* bdrv_info_stats_bs(BlockDriverState
*bs
)
1611 res
= qobject_from_jsonf("{ 'stats': {"
1612 "'rd_bytes': %" PRId64
","
1613 "'wr_bytes': %" PRId64
","
1614 "'rd_operations': %" PRId64
","
1615 "'wr_operations': %" PRId64
","
1616 "'wr_highest_offset': %" PRId64
1618 bs
->rd_bytes
, bs
->wr_bytes
,
1619 bs
->rd_ops
, bs
->wr_ops
,
1620 bs
->wr_highest_sector
*
1621 (uint64_t)BDRV_SECTOR_SIZE
);
1622 dict
= qobject_to_qdict(res
);
1624 if (*bs
->device_name
) {
1625 qdict_put(dict
, "device", qstring_from_str(bs
->device_name
));
1629 QObject
*parent
= bdrv_info_stats_bs(bs
->file
);
1630 qdict_put_obj(dict
, "parent", parent
);
1636 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1640 BlockDriverState
*bs
;
1642 devices
= qlist_new();
1644 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1645 obj
= bdrv_info_stats_bs(bs
);
1646 qlist_append_obj(devices
, obj
);
1649 *ret_data
= QOBJECT(devices
);
1652 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1654 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1655 return bs
->backing_file
;
1656 else if (bs
->encrypted
)
1657 return bs
->filename
;
1662 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1663 char *filename
, int filename_size
)
1665 if (!bs
->backing_file
) {
1666 pstrcpy(filename
, filename_size
, "");
1668 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1672 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1673 const uint8_t *buf
, int nb_sectors
)
1675 BlockDriver
*drv
= bs
->drv
;
1678 if (!drv
->bdrv_write_compressed
)
1680 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1683 if (bs
->dirty_bitmap
) {
1684 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1687 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1690 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1692 BlockDriver
*drv
= bs
->drv
;
1695 if (!drv
->bdrv_get_info
)
1697 memset(bdi
, 0, sizeof(*bdi
));
1698 return drv
->bdrv_get_info(bs
, bdi
);
1701 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1702 int64_t pos
, int size
)
1704 BlockDriver
*drv
= bs
->drv
;
1707 if (drv
->bdrv_save_vmstate
)
1708 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1710 return bdrv_save_vmstate(bs
->file
, buf
, pos
, size
);
1714 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1715 int64_t pos
, int size
)
1717 BlockDriver
*drv
= bs
->drv
;
1720 if (drv
->bdrv_load_vmstate
)
1721 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1723 return bdrv_load_vmstate(bs
->file
, buf
, pos
, size
);
1727 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1729 BlockDriver
*drv
= bs
->drv
;
1731 if (!drv
|| !drv
->bdrv_debug_event
) {
1735 return drv
->bdrv_debug_event(bs
, event
);
1739 /**************************************************************/
1740 /* handling of snapshots */
1742 int bdrv_can_snapshot(BlockDriverState
*bs
)
1744 BlockDriver
*drv
= bs
->drv
;
1745 if (!drv
|| bdrv_is_removable(bs
) || bdrv_is_read_only(bs
)) {
1749 if (!drv
->bdrv_snapshot_create
) {
1750 if (bs
->file
!= NULL
) {
1751 return bdrv_can_snapshot(bs
->file
);
1759 int bdrv_snapshot_create(BlockDriverState
*bs
,
1760 QEMUSnapshotInfo
*sn_info
)
1762 BlockDriver
*drv
= bs
->drv
;
1765 if (drv
->bdrv_snapshot_create
)
1766 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1768 return bdrv_snapshot_create(bs
->file
, sn_info
);
1772 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1773 const char *snapshot_id
)
1775 BlockDriver
*drv
= bs
->drv
;
1780 if (drv
->bdrv_snapshot_goto
)
1781 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1784 drv
->bdrv_close(bs
);
1785 ret
= bdrv_snapshot_goto(bs
->file
, snapshot_id
);
1786 open_ret
= drv
->bdrv_open(bs
, bs
->open_flags
);
1788 bdrv_delete(bs
->file
);
1798 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1800 BlockDriver
*drv
= bs
->drv
;
1803 if (drv
->bdrv_snapshot_delete
)
1804 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1806 return bdrv_snapshot_delete(bs
->file
, snapshot_id
);
1810 int bdrv_snapshot_list(BlockDriverState
*bs
,
1811 QEMUSnapshotInfo
**psn_info
)
1813 BlockDriver
*drv
= bs
->drv
;
1816 if (drv
->bdrv_snapshot_list
)
1817 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1819 return bdrv_snapshot_list(bs
->file
, psn_info
);
1823 #define NB_SUFFIXES 4
1825 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1827 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1832 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1835 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1836 if (size
< (10 * base
)) {
1837 snprintf(buf
, buf_size
, "%0.1f%c",
1838 (double)size
/ base
,
1841 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1842 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1843 ((size
+ (base
>> 1)) / base
),
1853 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1855 char buf1
[128], date_buf
[128], clock_buf
[128];
1865 snprintf(buf
, buf_size
,
1866 "%-10s%-20s%7s%20s%15s",
1867 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1871 ptm
= localtime(&ti
);
1872 strftime(date_buf
, sizeof(date_buf
),
1873 "%Y-%m-%d %H:%M:%S", ptm
);
1875 localtime_r(&ti
, &tm
);
1876 strftime(date_buf
, sizeof(date_buf
),
1877 "%Y-%m-%d %H:%M:%S", &tm
);
1879 secs
= sn
->vm_clock_nsec
/ 1000000000;
1880 snprintf(clock_buf
, sizeof(clock_buf
),
1881 "%02d:%02d:%02d.%03d",
1883 (int)((secs
/ 60) % 60),
1885 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1886 snprintf(buf
, buf_size
,
1887 "%-10s%-20s%7s%20s%15s",
1888 sn
->id_str
, sn
->name
,
1889 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1897 /**************************************************************/
1900 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1901 QEMUIOVector
*qiov
, int nb_sectors
,
1902 BlockDriverCompletionFunc
*cb
, void *opaque
)
1904 BlockDriver
*drv
= bs
->drv
;
1905 BlockDriverAIOCB
*ret
;
1909 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1912 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1916 /* Update stats even though technically transfer has not happened. */
1917 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1924 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1925 QEMUIOVector
*qiov
, int nb_sectors
,
1926 BlockDriverCompletionFunc
*cb
, void *opaque
)
1928 BlockDriver
*drv
= bs
->drv
;
1929 BlockDriverAIOCB
*ret
;
1935 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1938 if (bs
->dirty_bitmap
) {
1939 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1942 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1946 /* Update stats even though technically transfer has not happened. */
1947 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1949 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
1950 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
1958 typedef struct MultiwriteCB
{
1963 BlockDriverCompletionFunc
*cb
;
1965 QEMUIOVector
*free_qiov
;
1970 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1974 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1975 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1976 if (mcb
->callbacks
[i
].free_qiov
) {
1977 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
1979 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1980 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1984 static void multiwrite_cb(void *opaque
, int ret
)
1986 MultiwriteCB
*mcb
= opaque
;
1988 if (ret
< 0 && !mcb
->error
) {
1990 multiwrite_user_cb(mcb
);
1993 mcb
->num_requests
--;
1994 if (mcb
->num_requests
== 0) {
1995 if (mcb
->error
== 0) {
1996 multiwrite_user_cb(mcb
);
2002 static int multiwrite_req_compare(const void *a
, const void *b
)
2004 const BlockRequest
*req1
= a
, *req2
= b
;
2007 * Note that we can't simply subtract req2->sector from req1->sector
2008 * here as that could overflow the return value.
2010 if (req1
->sector
> req2
->sector
) {
2012 } else if (req1
->sector
< req2
->sector
) {
2020 * Takes a bunch of requests and tries to merge them. Returns the number of
2021 * requests that remain after merging.
2023 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
2024 int num_reqs
, MultiwriteCB
*mcb
)
2028 // Sort requests by start sector
2029 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
2031 // Check if adjacent requests touch the same clusters. If so, combine them,
2032 // filling up gaps with zero sectors.
2034 for (i
= 1; i
< num_reqs
; i
++) {
2036 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
2038 // This handles the cases that are valid for all block drivers, namely
2039 // exactly sequential writes and overlapping writes.
2040 if (reqs
[i
].sector
<= oldreq_last
) {
2044 // The block driver may decide that it makes sense to combine requests
2045 // even if there is a gap of some sectors between them. In this case,
2046 // the gap is filled with zeros (therefore only applicable for yet
2047 // unused space in format like qcow2).
2048 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
2049 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
2052 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
2058 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
2059 qemu_iovec_init(qiov
,
2060 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
2062 // Add the first request to the merged one. If the requests are
2063 // overlapping, drop the last sectors of the first request.
2064 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
2065 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
2067 // We might need to add some zeros between the two requests
2068 if (reqs
[i
].sector
> oldreq_last
) {
2069 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
2070 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
2071 memset(buf
, 0, zero_bytes
);
2072 qemu_iovec_add(qiov
, buf
, zero_bytes
);
2073 mcb
->callbacks
[i
].free_buf
= buf
;
2076 // Add the second request
2077 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2079 reqs
[outidx
].nb_sectors
= qiov
->size
>> 9;
2080 reqs
[outidx
].qiov
= qiov
;
2082 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2085 reqs
[outidx
].sector
= reqs
[i
].sector
;
2086 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2087 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2095 * Submit multiple AIO write requests at once.
2097 * On success, the function returns 0 and all requests in the reqs array have
2098 * been submitted. In error case this function returns -1, and any of the
2099 * requests may or may not be submitted yet. In particular, this means that the
2100 * callback will be called for some of the requests, for others it won't. The
2101 * caller must check the error field of the BlockRequest to wait for the right
2102 * callbacks (if error != 0, no callback will be called).
2104 * The implementation may modify the contents of the reqs array, e.g. to merge
2105 * requests. However, the fields opaque and error are left unmodified as they
2106 * are used to signal failure for a single request to the caller.
2108 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2110 BlockDriverAIOCB
*acb
;
2114 if (num_reqs
== 0) {
2118 // Create MultiwriteCB structure
2119 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2120 mcb
->num_requests
= 0;
2121 mcb
->num_callbacks
= num_reqs
;
2123 for (i
= 0; i
< num_reqs
; i
++) {
2124 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2125 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2128 // Check for mergable requests
2129 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2131 // Run the aio requests
2132 for (i
= 0; i
< num_reqs
; i
++) {
2133 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2134 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2137 // We can only fail the whole thing if no request has been
2138 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2139 // complete and report the error in the callback.
2140 if (mcb
->num_requests
== 0) {
2141 reqs
[i
].error
= -EIO
;
2144 mcb
->num_requests
++;
2145 multiwrite_cb(mcb
, -EIO
);
2149 mcb
->num_requests
++;
2160 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2161 BlockDriverCompletionFunc
*cb
, void *opaque
)
2163 BlockDriver
*drv
= bs
->drv
;
2165 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
2166 return bdrv_aio_noop_em(bs
, cb
, opaque
);
2171 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2174 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2176 acb
->pool
->cancel(acb
);
2180 /**************************************************************/
2181 /* async block device emulation */
2183 typedef struct BlockDriverAIOCBSync
{
2184 BlockDriverAIOCB common
;
2187 /* vector translation state */
2191 } BlockDriverAIOCBSync
;
2193 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2195 BlockDriverAIOCBSync
*acb
=
2196 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2197 qemu_bh_delete(acb
->bh
);
2199 qemu_aio_release(acb
);
2202 static AIOPool bdrv_em_aio_pool
= {
2203 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2204 .cancel
= bdrv_aio_cancel_em
,
2207 static void bdrv_aio_bh_cb(void *opaque
)
2209 BlockDriverAIOCBSync
*acb
= opaque
;
2212 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2213 qemu_vfree(acb
->bounce
);
2214 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2215 qemu_bh_delete(acb
->bh
);
2217 qemu_aio_release(acb
);
2220 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2224 BlockDriverCompletionFunc
*cb
,
2229 BlockDriverAIOCBSync
*acb
;
2231 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2232 acb
->is_write
= is_write
;
2234 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2237 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2240 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2241 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2243 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2246 qemu_bh_schedule(acb
->bh
);
2248 return &acb
->common
;
2251 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2252 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2253 BlockDriverCompletionFunc
*cb
, void *opaque
)
2255 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2258 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2259 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2260 BlockDriverCompletionFunc
*cb
, void *opaque
)
2262 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2265 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2266 BlockDriverCompletionFunc
*cb
, void *opaque
)
2268 BlockDriverAIOCBSync
*acb
;
2270 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2271 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2277 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2280 qemu_bh_schedule(acb
->bh
);
2281 return &acb
->common
;
2284 static BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
2285 BlockDriverCompletionFunc
*cb
, void *opaque
)
2287 BlockDriverAIOCBSync
*acb
;
2289 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2290 acb
->is_write
= 1; /* don't bounce in the completion handler */
2296 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2299 qemu_bh_schedule(acb
->bh
);
2300 return &acb
->common
;
2303 /**************************************************************/
2304 /* sync block device emulation */
2306 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2308 *(int *)opaque
= ret
;
2311 #define NOT_DONE 0x7fffffff
2313 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2314 uint8_t *buf
, int nb_sectors
)
2317 BlockDriverAIOCB
*acb
;
2321 async_context_push();
2323 async_ret
= NOT_DONE
;
2324 iov
.iov_base
= (void *)buf
;
2325 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2326 qemu_iovec_init_external(&qiov
, &iov
, 1);
2327 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2328 bdrv_rw_em_cb
, &async_ret
);
2334 while (async_ret
== NOT_DONE
) {
2340 async_context_pop();
2344 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2345 const uint8_t *buf
, int nb_sectors
)
2348 BlockDriverAIOCB
*acb
;
2352 async_context_push();
2354 async_ret
= NOT_DONE
;
2355 iov
.iov_base
= (void *)buf
;
2356 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2357 qemu_iovec_init_external(&qiov
, &iov
, 1);
2358 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2359 bdrv_rw_em_cb
, &async_ret
);
2364 while (async_ret
== NOT_DONE
) {
2369 async_context_pop();
2373 void bdrv_init(void)
2375 module_call_init(MODULE_INIT_BLOCK
);
2378 void bdrv_init_with_whitelist(void)
2380 use_bdrv_whitelist
= 1;
2384 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2385 BlockDriverCompletionFunc
*cb
, void *opaque
)
2387 BlockDriverAIOCB
*acb
;
2389 if (pool
->free_aiocb
) {
2390 acb
= pool
->free_aiocb
;
2391 pool
->free_aiocb
= acb
->next
;
2393 acb
= qemu_mallocz(pool
->aiocb_size
);
2398 acb
->opaque
= opaque
;
2402 void qemu_aio_release(void *p
)
2404 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2405 AIOPool
*pool
= acb
->pool
;
2406 acb
->next
= pool
->free_aiocb
;
2407 pool
->free_aiocb
= acb
;
2410 /**************************************************************/
2411 /* removable device support */
2414 * Return TRUE if the media is present
2416 int bdrv_is_inserted(BlockDriverState
*bs
)
2418 BlockDriver
*drv
= bs
->drv
;
2422 if (!drv
->bdrv_is_inserted
)
2424 ret
= drv
->bdrv_is_inserted(bs
);
2429 * Return TRUE if the media changed since the last call to this
2430 * function. It is currently only used for floppy disks
2432 int bdrv_media_changed(BlockDriverState
*bs
)
2434 BlockDriver
*drv
= bs
->drv
;
2437 if (!drv
|| !drv
->bdrv_media_changed
)
2440 ret
= drv
->bdrv_media_changed(bs
);
2441 if (ret
== -ENOTSUP
)
2442 ret
= bs
->media_changed
;
2443 bs
->media_changed
= 0;
2448 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2450 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2452 BlockDriver
*drv
= bs
->drv
;
2459 if (!drv
|| !drv
->bdrv_eject
) {
2462 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2464 if (ret
== -ENOTSUP
) {
2473 int bdrv_is_locked(BlockDriverState
*bs
)
2479 * Lock or unlock the media (if it is locked, the user won't be able
2480 * to eject it manually).
2482 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2484 BlockDriver
*drv
= bs
->drv
;
2486 bs
->locked
= locked
;
2487 if (drv
&& drv
->bdrv_set_locked
) {
2488 drv
->bdrv_set_locked(bs
, locked
);
2492 /* needed for generic scsi interface */
2494 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2496 BlockDriver
*drv
= bs
->drv
;
2498 if (drv
&& drv
->bdrv_ioctl
)
2499 return drv
->bdrv_ioctl(bs
, req
, buf
);
2503 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2504 unsigned long int req
, void *buf
,
2505 BlockDriverCompletionFunc
*cb
, void *opaque
)
2507 BlockDriver
*drv
= bs
->drv
;
2509 if (drv
&& drv
->bdrv_aio_ioctl
)
2510 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2516 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2518 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2521 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2523 int64_t bitmap_size
;
2525 bs
->dirty_count
= 0;
2527 if (!bs
->dirty_bitmap
) {
2528 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2529 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2530 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2532 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2535 if (bs
->dirty_bitmap
) {
2536 qemu_free(bs
->dirty_bitmap
);
2537 bs
->dirty_bitmap
= NULL
;
2542 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2544 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2546 if (bs
->dirty_bitmap
&&
2547 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2548 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2549 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2555 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2558 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2561 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2563 return bs
->dirty_count
;