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"
28 #include "block_int.h"
30 #include "qemu-objects.h"
33 #include <sys/types.h>
35 #include <sys/ioctl.h>
36 #include <sys/queue.h>
46 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
47 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
48 BlockDriverCompletionFunc
*cb
, void *opaque
);
49 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
50 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
51 BlockDriverCompletionFunc
*cb
, void *opaque
);
52 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
53 BlockDriverCompletionFunc
*cb
, void *opaque
);
54 static BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
55 BlockDriverCompletionFunc
*cb
, void *opaque
);
56 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
57 uint8_t *buf
, int nb_sectors
);
58 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
59 const uint8_t *buf
, int nb_sectors
);
61 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
62 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
64 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
65 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
67 /* The device to use for VM snapshots */
68 static BlockDriverState
*bs_snapshots
;
70 /* If non-zero, use only whitelisted block drivers */
71 static int use_bdrv_whitelist
;
74 static int is_windows_drive_prefix(const char *filename
)
76 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
77 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
81 int is_windows_drive(const char *filename
)
83 if (is_windows_drive_prefix(filename
) &&
86 if (strstart(filename
, "\\\\.\\", NULL
) ||
87 strstart(filename
, "//./", NULL
))
93 /* check if the path starts with "<protocol>:" */
94 static int path_has_protocol(const char *path
)
97 if (is_windows_drive(path
) ||
98 is_windows_drive_prefix(path
)) {
103 return strchr(path
, ':') != NULL
;
106 int path_is_absolute(const char *path
)
110 /* specific case for names like: "\\.\d:" */
111 if (*path
== '/' || *path
== '\\')
114 p
= strchr(path
, ':');
120 return (*p
== '/' || *p
== '\\');
126 /* if filename is absolute, just copy it to dest. Otherwise, build a
127 path to it by considering it is relative to base_path. URL are
129 void path_combine(char *dest
, int dest_size
,
130 const char *base_path
,
131 const char *filename
)
138 if (path_is_absolute(filename
)) {
139 pstrcpy(dest
, dest_size
, filename
);
141 p
= strchr(base_path
, ':');
146 p1
= strrchr(base_path
, '/');
150 p2
= strrchr(base_path
, '\\');
162 if (len
> dest_size
- 1)
164 memcpy(dest
, base_path
, len
);
166 pstrcat(dest
, dest_size
, filename
);
170 void bdrv_register(BlockDriver
*bdrv
)
172 if (!bdrv
->bdrv_aio_readv
) {
173 /* add AIO emulation layer */
174 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
175 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
176 } else if (!bdrv
->bdrv_read
) {
177 /* add synchronous IO emulation layer */
178 bdrv
->bdrv_read
= bdrv_read_em
;
179 bdrv
->bdrv_write
= bdrv_write_em
;
182 if (!bdrv
->bdrv_aio_flush
)
183 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
185 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
188 /* create a new block device (by default it is empty) */
189 BlockDriverState
*bdrv_new(const char *device_name
)
191 BlockDriverState
*bs
;
193 bs
= qemu_mallocz(sizeof(BlockDriverState
));
194 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
195 if (device_name
[0] != '\0') {
196 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
201 BlockDriver
*bdrv_find_format(const char *format_name
)
204 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
205 if (!strcmp(drv1
->format_name
, format_name
)) {
212 static int bdrv_is_whitelisted(BlockDriver
*drv
)
214 static const char *whitelist
[] = {
215 CONFIG_BDRV_WHITELIST
220 return 1; /* no whitelist, anything goes */
222 for (p
= whitelist
; *p
; p
++) {
223 if (!strcmp(drv
->format_name
, *p
)) {
230 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
232 BlockDriver
*drv
= bdrv_find_format(format_name
);
233 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
236 int bdrv_create(BlockDriver
*drv
, const char* filename
,
237 QEMUOptionParameter
*options
)
239 if (!drv
->bdrv_create
)
242 return drv
->bdrv_create(filename
, options
);
245 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
249 drv
= bdrv_find_protocol(filename
);
254 return bdrv_create(drv
, filename
, options
);
258 void get_tmp_filename(char *filename
, int size
)
260 char temp_dir
[MAX_PATH
];
262 GetTempPath(MAX_PATH
, temp_dir
);
263 GetTempFileName(temp_dir
, "qem", 0, filename
);
266 void get_tmp_filename(char *filename
, int size
)
270 /* XXX: race condition possible */
271 tmpdir
= getenv("TMPDIR");
274 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
275 fd
= mkstemp(filename
);
281 * Detect host devices. By convention, /dev/cdrom[N] is always
282 * recognized as a host CDROM.
284 static BlockDriver
*find_hdev_driver(const char *filename
)
286 int score_max
= 0, score
;
287 BlockDriver
*drv
= NULL
, *d
;
289 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
290 if (d
->bdrv_probe_device
) {
291 score
= d
->bdrv_probe_device(filename
);
292 if (score
> score_max
) {
302 BlockDriver
*bdrv_find_protocol(const char *filename
)
309 /* TODO Drivers without bdrv_file_open must be specified explicitly */
312 * XXX(hch): we really should not let host device detection
313 * override an explicit protocol specification, but moving this
314 * later breaks access to device names with colons in them.
315 * Thanks to the brain-dead persistent naming schemes on udev-
316 * based Linux systems those actually are quite common.
318 drv1
= find_hdev_driver(filename
);
323 if (!path_has_protocol(filename
)) {
324 return bdrv_find_format("file");
326 p
= strchr(filename
, ':');
329 if (len
> sizeof(protocol
) - 1)
330 len
= sizeof(protocol
) - 1;
331 memcpy(protocol
, filename
, len
);
332 protocol
[len
] = '\0';
333 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
334 if (drv1
->protocol_name
&&
335 !strcmp(drv1
->protocol_name
, protocol
)) {
342 static int find_image_format(const char *filename
, BlockDriver
**pdrv
)
344 int ret
, score
, score_max
;
345 BlockDriver
*drv1
, *drv
;
347 BlockDriverState
*bs
;
349 ret
= bdrv_file_open(&bs
, filename
, 0);
355 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
356 if (bs
->sg
|| !bdrv_is_inserted(bs
)) {
358 drv
= bdrv_find_format("raw");
366 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
375 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
376 if (drv1
->bdrv_probe
) {
377 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
378 if (score
> score_max
) {
392 * Set the current 'total_sectors' value
394 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
396 BlockDriver
*drv
= bs
->drv
;
398 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
402 /* query actual device if possible, otherwise just trust the hint */
403 if (drv
->bdrv_getlength
) {
404 int64_t length
= drv
->bdrv_getlength(bs
);
408 hint
= length
>> BDRV_SECTOR_BITS
;
411 bs
->total_sectors
= hint
;
416 * Common part for opening disk images and files
418 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
419 int flags
, BlockDriver
*drv
)
426 bs
->total_sectors
= 0;
429 bs
->open_flags
= flags
;
430 /* buffer_alignment defaulted to 512, drivers can change this value */
431 bs
->buffer_alignment
= 512;
433 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
435 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
440 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
443 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
444 * write cache to the guest. We do need the fdatasync to flush
445 * out transactions for block allocations, and we maybe have a
446 * volatile write cache in our backing device to deal with.
448 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
449 bs
->enable_write_cache
= 1;
452 * Clear flags that are internal to the block layer before opening the
455 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
458 * Snapshots should be writeable.
460 if (bs
->is_temporary
) {
461 open_flags
|= BDRV_O_RDWR
;
464 /* Open the image, either directly or using a protocol */
465 if (drv
->bdrv_file_open
) {
466 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
468 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
470 ret
= drv
->bdrv_open(bs
, open_flags
);
478 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
480 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
486 if (bs
->is_temporary
) {
494 bdrv_delete(bs
->file
);
497 qemu_free(bs
->opaque
);
504 * Opens a file using a protocol (file, host_device, nbd, ...)
506 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
508 BlockDriverState
*bs
;
512 drv
= bdrv_find_protocol(filename
);
518 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
529 * Opens a disk image (raw, qcow2, vmdk, ...)
531 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
536 if (flags
& BDRV_O_SNAPSHOT
) {
537 BlockDriverState
*bs1
;
540 BlockDriver
*bdrv_qcow2
;
541 QEMUOptionParameter
*options
;
542 char tmp_filename
[PATH_MAX
];
543 char backing_filename
[PATH_MAX
];
545 /* if snapshot, we create a temporary backing file and open it
546 instead of opening 'filename' directly */
548 /* if there is a backing file, use it */
550 ret
= bdrv_open(bs1
, filename
, 0, drv
);
555 total_size
= bdrv_getlength(bs1
) & BDRV_SECTOR_MASK
;
557 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
562 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
564 /* Real path is meaningless for protocols */
566 snprintf(backing_filename
, sizeof(backing_filename
),
568 else if (!realpath(filename
, backing_filename
))
571 bdrv_qcow2
= bdrv_find_format("qcow2");
572 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
574 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
);
575 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
577 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
581 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
582 free_option_parameters(options
);
587 filename
= tmp_filename
;
589 bs
->is_temporary
= 1;
592 /* Find the right image format driver */
594 ret
= find_image_format(filename
, &drv
);
598 goto unlink_and_fail
;
602 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
604 goto unlink_and_fail
;
607 /* If there is a backing file, use it */
608 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
609 char backing_filename
[PATH_MAX
];
611 BlockDriver
*back_drv
= NULL
;
613 bs
->backing_hd
= bdrv_new("");
615 if (path_has_protocol(bs
->backing_file
)) {
616 pstrcpy(backing_filename
, sizeof(backing_filename
),
619 path_combine(backing_filename
, sizeof(backing_filename
),
620 filename
, bs
->backing_file
);
623 if (bs
->backing_format
[0] != '\0') {
624 back_drv
= bdrv_find_format(bs
->backing_format
);
627 /* backing files always opened read-only */
629 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
631 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
636 if (bs
->is_temporary
) {
637 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
639 /* base image inherits from "parent" */
640 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
644 if (!bdrv_key_required(bs
)) {
645 /* call the change callback */
646 bs
->media_changed
= 1;
648 bs
->change_cb(bs
->change_opaque
, CHANGE_MEDIA
);
654 if (bs
->is_temporary
) {
660 void bdrv_close(BlockDriverState
*bs
)
663 if (bs
== bs_snapshots
) {
666 if (bs
->backing_hd
) {
667 bdrv_delete(bs
->backing_hd
);
668 bs
->backing_hd
= NULL
;
670 bs
->drv
->bdrv_close(bs
);
671 qemu_free(bs
->opaque
);
673 if (bs
->is_temporary
) {
674 unlink(bs
->filename
);
680 if (bs
->file
!= NULL
) {
681 bdrv_close(bs
->file
);
684 /* call the change callback */
685 bs
->media_changed
= 1;
687 bs
->change_cb(bs
->change_opaque
, CHANGE_MEDIA
);
691 void bdrv_close_all(void)
693 BlockDriverState
*bs
;
695 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
700 /* make a BlockDriverState anonymous by removing from bdrv_state list.
701 Also, NULL terminate the device_name to prevent double remove */
702 void bdrv_make_anon(BlockDriverState
*bs
)
704 if (bs
->device_name
[0] != '\0') {
705 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
707 bs
->device_name
[0] = '\0';
710 void bdrv_delete(BlockDriverState
*bs
)
714 /* remove from list, if necessary */
718 if (bs
->file
!= NULL
) {
719 bdrv_delete(bs
->file
);
722 assert(bs
!= bs_snapshots
);
726 int bdrv_attach(BlockDriverState
*bs
, DeviceState
*qdev
)
735 void bdrv_detach(BlockDriverState
*bs
, DeviceState
*qdev
)
737 assert(bs
->peer
== qdev
);
741 DeviceState
*bdrv_get_attached(BlockDriverState
*bs
)
747 * Run consistency checks on an image
749 * Returns 0 if the check could be completed (it doesn't mean that the image is
750 * free of errors) or -errno when an internal error occured. The results of the
751 * check are stored in res.
753 int bdrv_check(BlockDriverState
*bs
, BdrvCheckResult
*res
)
755 if (bs
->drv
->bdrv_check
== NULL
) {
759 memset(res
, 0, sizeof(*res
));
760 return bs
->drv
->bdrv_check(bs
, res
);
763 #define COMMIT_BUF_SECTORS 2048
765 /* commit COW file into the raw image */
766 int bdrv_commit(BlockDriverState
*bs
)
768 BlockDriver
*drv
= bs
->drv
;
769 BlockDriver
*backing_drv
;
770 int64_t sector
, total_sectors
;
771 int n
, ro
, open_flags
;
772 int ret
= 0, rw_ret
= 0;
775 BlockDriverState
*bs_rw
, *bs_ro
;
780 if (!bs
->backing_hd
) {
784 if (bs
->backing_hd
->keep_read_only
) {
788 backing_drv
= bs
->backing_hd
->drv
;
789 ro
= bs
->backing_hd
->read_only
;
790 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
791 open_flags
= bs
->backing_hd
->open_flags
;
795 bdrv_delete(bs
->backing_hd
);
796 bs
->backing_hd
= NULL
;
797 bs_rw
= bdrv_new("");
798 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
,
802 /* try to re-open read-only */
803 bs_ro
= bdrv_new("");
804 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
808 /* drive not functional anymore */
812 bs
->backing_hd
= bs_ro
;
815 bs
->backing_hd
= bs_rw
;
818 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
819 buf
= qemu_malloc(COMMIT_BUF_SECTORS
* BDRV_SECTOR_SIZE
);
821 for (sector
= 0; sector
< total_sectors
; sector
+= n
) {
822 if (drv
->bdrv_is_allocated(bs
, sector
, COMMIT_BUF_SECTORS
, &n
)) {
824 if (bdrv_read(bs
, sector
, buf
, n
) != 0) {
829 if (bdrv_write(bs
->backing_hd
, sector
, buf
, n
) != 0) {
836 if (drv
->bdrv_make_empty
) {
837 ret
= drv
->bdrv_make_empty(bs
);
842 * Make sure all data we wrote to the backing device is actually
846 bdrv_flush(bs
->backing_hd
);
853 bdrv_delete(bs
->backing_hd
);
854 bs
->backing_hd
= NULL
;
855 bs_ro
= bdrv_new("");
856 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
860 /* drive not functional anymore */
864 bs
->backing_hd
= bs_ro
;
865 bs
->backing_hd
->keep_read_only
= 0;
871 void bdrv_commit_all(void)
873 BlockDriverState
*bs
;
875 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
883 * -EINVAL - backing format specified, but no file
884 * -ENOSPC - can't update the backing file because no space is left in the
886 * -ENOTSUP - format driver doesn't support changing the backing file
888 int bdrv_change_backing_file(BlockDriverState
*bs
,
889 const char *backing_file
, const char *backing_fmt
)
891 BlockDriver
*drv
= bs
->drv
;
893 if (drv
->bdrv_change_backing_file
!= NULL
) {
894 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
900 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
905 if (!bdrv_is_inserted(bs
))
911 len
= bdrv_getlength(bs
);
916 if ((offset
> len
) || (len
- offset
< size
))
922 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
925 return bdrv_check_byte_request(bs
, sector_num
* BDRV_SECTOR_SIZE
,
926 nb_sectors
* BDRV_SECTOR_SIZE
);
929 /* return < 0 if error. See bdrv_write() for the return codes */
930 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
931 uint8_t *buf
, int nb_sectors
)
933 BlockDriver
*drv
= bs
->drv
;
937 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
940 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
943 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
944 int nb_sectors
, int dirty
)
947 unsigned long val
, idx
, bit
;
949 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
950 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
952 for (; start
<= end
; start
++) {
953 idx
= start
/ (sizeof(unsigned long) * 8);
954 bit
= start
% (sizeof(unsigned long) * 8);
955 val
= bs
->dirty_bitmap
[idx
];
957 if (!(val
& (1UL << bit
))) {
962 if (val
& (1UL << bit
)) {
964 val
&= ~(1UL << bit
);
967 bs
->dirty_bitmap
[idx
] = val
;
971 /* Return < 0 if error. Important errors are:
972 -EIO generic I/O error (may happen for all errors)
973 -ENOMEDIUM No media inserted.
974 -EINVAL Invalid sector number or nb_sectors
975 -EACCES Trying to write a read-only device
977 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
978 const uint8_t *buf
, int nb_sectors
)
980 BlockDriver
*drv
= bs
->drv
;
985 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
988 if (bs
->dirty_bitmap
) {
989 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
992 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
993 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
996 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
999 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
1000 void *buf
, int count1
)
1002 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
1003 int len
, nb_sectors
, count
;
1008 /* first read to align to sector start */
1009 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1012 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1014 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1016 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
1024 /* read the sectors "in place" */
1025 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1026 if (nb_sectors
> 0) {
1027 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1029 sector_num
+= nb_sectors
;
1030 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1035 /* add data from the last sector */
1037 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1039 memcpy(buf
, tmp_buf
, count
);
1044 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
1045 const void *buf
, int count1
)
1047 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
1048 int len
, nb_sectors
, count
;
1053 /* first write to align to sector start */
1054 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1057 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1059 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1061 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
1062 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1071 /* write the sectors "in place" */
1072 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1073 if (nb_sectors
> 0) {
1074 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1076 sector_num
+= nb_sectors
;
1077 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1082 /* add data from the last sector */
1084 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1086 memcpy(tmp_buf
, buf
, count
);
1087 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1094 * Writes to the file and ensures that no writes are reordered across this
1095 * request (acts as a barrier)
1097 * Returns 0 on success, -errno in error cases.
1099 int bdrv_pwrite_sync(BlockDriverState
*bs
, int64_t offset
,
1100 const void *buf
, int count
)
1104 ret
= bdrv_pwrite(bs
, offset
, buf
, count
);
1109 /* No flush needed for cache=writethrough, it uses O_DSYNC */
1110 if ((bs
->open_flags
& BDRV_O_CACHE_MASK
) != 0) {
1118 * Writes to the file and ensures that no writes are reordered across this
1119 * request (acts as a barrier)
1121 * Returns 0 on success, -errno in error cases.
1123 int bdrv_write_sync(BlockDriverState
*bs
, int64_t sector_num
,
1124 const uint8_t *buf
, int nb_sectors
)
1126 return bdrv_pwrite_sync(bs
, BDRV_SECTOR_SIZE
* sector_num
,
1127 buf
, BDRV_SECTOR_SIZE
* nb_sectors
);
1131 * Truncate file to 'offset' bytes (needed only for file protocols)
1133 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
1135 BlockDriver
*drv
= bs
->drv
;
1139 if (!drv
->bdrv_truncate
)
1143 if (bdrv_in_use(bs
))
1145 ret
= drv
->bdrv_truncate(bs
, offset
);
1147 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1148 if (bs
->change_cb
) {
1149 bs
->change_cb(bs
->change_opaque
, CHANGE_SIZE
);
1156 * Length of a file in bytes. Return < 0 if error or unknown.
1158 int64_t bdrv_getlength(BlockDriverState
*bs
)
1160 BlockDriver
*drv
= bs
->drv
;
1164 /* Fixed size devices use the total_sectors value for speed instead of
1165 issuing a length query (like lseek) on each call. Also, legacy block
1166 drivers don't provide a bdrv_getlength function and must use
1168 if (!bs
->growable
|| !drv
->bdrv_getlength
) {
1169 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1171 return drv
->bdrv_getlength(bs
);
1174 /* return 0 as number of sectors if no device present or error */
1175 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1178 length
= bdrv_getlength(bs
);
1182 length
= length
>> BDRV_SECTOR_BITS
;
1183 *nb_sectors_ptr
= length
;
1187 uint8_t boot_ind
; /* 0x80 - active */
1188 uint8_t head
; /* starting head */
1189 uint8_t sector
; /* starting sector */
1190 uint8_t cyl
; /* starting cylinder */
1191 uint8_t sys_ind
; /* What partition type */
1192 uint8_t end_head
; /* end head */
1193 uint8_t end_sector
; /* end sector */
1194 uint8_t end_cyl
; /* end cylinder */
1195 uint32_t start_sect
; /* starting sector counting from 0 */
1196 uint32_t nr_sects
; /* nr of sectors in partition */
1197 } __attribute__((packed
));
1199 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1200 static int guess_disk_lchs(BlockDriverState
*bs
,
1201 int *pcylinders
, int *pheads
, int *psectors
)
1203 uint8_t buf
[BDRV_SECTOR_SIZE
];
1204 int ret
, i
, heads
, sectors
, cylinders
;
1205 struct partition
*p
;
1207 uint64_t nb_sectors
;
1209 bdrv_get_geometry(bs
, &nb_sectors
);
1211 ret
= bdrv_read(bs
, 0, buf
, 1);
1214 /* test msdos magic */
1215 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1217 for(i
= 0; i
< 4; i
++) {
1218 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1219 nr_sects
= le32_to_cpu(p
->nr_sects
);
1220 if (nr_sects
&& p
->end_head
) {
1221 /* We make the assumption that the partition terminates on
1222 a cylinder boundary */
1223 heads
= p
->end_head
+ 1;
1224 sectors
= p
->end_sector
& 63;
1227 cylinders
= nb_sectors
/ (heads
* sectors
);
1228 if (cylinders
< 1 || cylinders
> 16383)
1231 *psectors
= sectors
;
1232 *pcylinders
= cylinders
;
1234 printf("guessed geometry: LCHS=%d %d %d\n",
1235 cylinders
, heads
, sectors
);
1243 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1245 int translation
, lba_detected
= 0;
1246 int cylinders
, heads
, secs
;
1247 uint64_t nb_sectors
;
1249 /* if a geometry hint is available, use it */
1250 bdrv_get_geometry(bs
, &nb_sectors
);
1251 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1252 translation
= bdrv_get_translation_hint(bs
);
1253 if (cylinders
!= 0) {
1258 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1260 /* if heads > 16, it means that a BIOS LBA
1261 translation was active, so the default
1262 hardware geometry is OK */
1264 goto default_geometry
;
1269 /* disable any translation to be in sync with
1270 the logical geometry */
1271 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1272 bdrv_set_translation_hint(bs
,
1273 BIOS_ATA_TRANSLATION_NONE
);
1278 /* if no geometry, use a standard physical disk geometry */
1279 cylinders
= nb_sectors
/ (16 * 63);
1281 if (cylinders
> 16383)
1283 else if (cylinders
< 2)
1288 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1289 if ((*pcyls
* *pheads
) <= 131072) {
1290 bdrv_set_translation_hint(bs
,
1291 BIOS_ATA_TRANSLATION_LARGE
);
1293 bdrv_set_translation_hint(bs
,
1294 BIOS_ATA_TRANSLATION_LBA
);
1298 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1302 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1303 int cyls
, int heads
, int secs
)
1310 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1313 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1314 type
== BDRV_TYPE_FLOPPY
));
1317 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1319 bs
->translation
= translation
;
1322 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1323 int *pcyls
, int *pheads
, int *psecs
)
1326 *pheads
= bs
->heads
;
1330 /* Recognize floppy formats */
1331 typedef struct FDFormat
{
1338 static const FDFormat fd_formats
[] = {
1339 /* First entry is default format */
1340 /* 1.44 MB 3"1/2 floppy disks */
1341 { FDRIVE_DRV_144
, 18, 80, 1, },
1342 { FDRIVE_DRV_144
, 20, 80, 1, },
1343 { FDRIVE_DRV_144
, 21, 80, 1, },
1344 { FDRIVE_DRV_144
, 21, 82, 1, },
1345 { FDRIVE_DRV_144
, 21, 83, 1, },
1346 { FDRIVE_DRV_144
, 22, 80, 1, },
1347 { FDRIVE_DRV_144
, 23, 80, 1, },
1348 { FDRIVE_DRV_144
, 24, 80, 1, },
1349 /* 2.88 MB 3"1/2 floppy disks */
1350 { FDRIVE_DRV_288
, 36, 80, 1, },
1351 { FDRIVE_DRV_288
, 39, 80, 1, },
1352 { FDRIVE_DRV_288
, 40, 80, 1, },
1353 { FDRIVE_DRV_288
, 44, 80, 1, },
1354 { FDRIVE_DRV_288
, 48, 80, 1, },
1355 /* 720 kB 3"1/2 floppy disks */
1356 { FDRIVE_DRV_144
, 9, 80, 1, },
1357 { FDRIVE_DRV_144
, 10, 80, 1, },
1358 { FDRIVE_DRV_144
, 10, 82, 1, },
1359 { FDRIVE_DRV_144
, 10, 83, 1, },
1360 { FDRIVE_DRV_144
, 13, 80, 1, },
1361 { FDRIVE_DRV_144
, 14, 80, 1, },
1362 /* 1.2 MB 5"1/4 floppy disks */
1363 { FDRIVE_DRV_120
, 15, 80, 1, },
1364 { FDRIVE_DRV_120
, 18, 80, 1, },
1365 { FDRIVE_DRV_120
, 18, 82, 1, },
1366 { FDRIVE_DRV_120
, 18, 83, 1, },
1367 { FDRIVE_DRV_120
, 20, 80, 1, },
1368 /* 720 kB 5"1/4 floppy disks */
1369 { FDRIVE_DRV_120
, 9, 80, 1, },
1370 { FDRIVE_DRV_120
, 11, 80, 1, },
1371 /* 360 kB 5"1/4 floppy disks */
1372 { FDRIVE_DRV_120
, 9, 40, 1, },
1373 { FDRIVE_DRV_120
, 9, 40, 0, },
1374 { FDRIVE_DRV_120
, 10, 41, 1, },
1375 { FDRIVE_DRV_120
, 10, 42, 1, },
1376 /* 320 kB 5"1/4 floppy disks */
1377 { FDRIVE_DRV_120
, 8, 40, 1, },
1378 { FDRIVE_DRV_120
, 8, 40, 0, },
1379 /* 360 kB must match 5"1/4 better than 3"1/2... */
1380 { FDRIVE_DRV_144
, 9, 80, 0, },
1382 { FDRIVE_DRV_NONE
, -1, -1, 0, },
1385 void bdrv_get_floppy_geometry_hint(BlockDriverState
*bs
, int *nb_heads
,
1386 int *max_track
, int *last_sect
,
1387 FDriveType drive_in
, FDriveType
*drive
)
1389 const FDFormat
*parse
;
1390 uint64_t nb_sectors
, size
;
1391 int i
, first_match
, match
;
1393 bdrv_get_geometry_hint(bs
, nb_heads
, max_track
, last_sect
);
1394 if (*nb_heads
!= 0 && *max_track
!= 0 && *last_sect
!= 0) {
1395 /* User defined disk */
1397 bdrv_get_geometry(bs
, &nb_sectors
);
1400 for (i
= 0; ; i
++) {
1401 parse
= &fd_formats
[i
];
1402 if (parse
->drive
== FDRIVE_DRV_NONE
) {
1405 if (drive_in
== parse
->drive
||
1406 drive_in
== FDRIVE_DRV_NONE
) {
1407 size
= (parse
->max_head
+ 1) * parse
->max_track
*
1409 if (nb_sectors
== size
) {
1413 if (first_match
== -1) {
1419 if (first_match
== -1) {
1422 match
= first_match
;
1424 parse
= &fd_formats
[match
];
1426 *nb_heads
= parse
->max_head
+ 1;
1427 *max_track
= parse
->max_track
;
1428 *last_sect
= parse
->last_sect
;
1429 *drive
= parse
->drive
;
1433 int bdrv_get_type_hint(BlockDriverState
*bs
)
1438 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1440 return bs
->translation
;
1443 void bdrv_set_on_error(BlockDriverState
*bs
, BlockErrorAction on_read_error
,
1444 BlockErrorAction on_write_error
)
1446 bs
->on_read_error
= on_read_error
;
1447 bs
->on_write_error
= on_write_error
;
1450 BlockErrorAction
bdrv_get_on_error(BlockDriverState
*bs
, int is_read
)
1452 return is_read
? bs
->on_read_error
: bs
->on_write_error
;
1455 void bdrv_set_removable(BlockDriverState
*bs
, int removable
)
1457 bs
->removable
= removable
;
1458 if (removable
&& bs
== bs_snapshots
) {
1459 bs_snapshots
= NULL
;
1463 int bdrv_is_removable(BlockDriverState
*bs
)
1465 return bs
->removable
;
1468 int bdrv_is_read_only(BlockDriverState
*bs
)
1470 return bs
->read_only
;
1473 int bdrv_is_sg(BlockDriverState
*bs
)
1478 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1480 return bs
->enable_write_cache
;
1483 /* XXX: no longer used */
1484 void bdrv_set_change_cb(BlockDriverState
*bs
,
1485 void (*change_cb
)(void *opaque
, int reason
),
1488 bs
->change_cb
= change_cb
;
1489 bs
->change_opaque
= opaque
;
1492 int bdrv_is_encrypted(BlockDriverState
*bs
)
1494 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1496 return bs
->encrypted
;
1499 int bdrv_key_required(BlockDriverState
*bs
)
1501 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1503 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1505 return (bs
->encrypted
&& !bs
->valid_key
);
1508 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1511 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1512 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1518 if (!bs
->encrypted
) {
1520 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1523 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1526 } else if (!bs
->valid_key
) {
1528 /* call the change callback now, we skipped it on open */
1529 bs
->media_changed
= 1;
1531 bs
->change_cb(bs
->change_opaque
, CHANGE_MEDIA
);
1536 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1541 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1545 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1550 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1551 it(opaque
, drv
->format_name
);
1555 BlockDriverState
*bdrv_find(const char *name
)
1557 BlockDriverState
*bs
;
1559 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1560 if (!strcmp(name
, bs
->device_name
)) {
1567 BlockDriverState
*bdrv_next(BlockDriverState
*bs
)
1570 return QTAILQ_FIRST(&bdrv_states
);
1572 return QTAILQ_NEXT(bs
, list
);
1575 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1577 BlockDriverState
*bs
;
1579 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1584 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1586 return bs
->device_name
;
1589 int bdrv_flush(BlockDriverState
*bs
)
1591 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
1595 if (bs
->drv
&& bs
->drv
->bdrv_flush
) {
1596 return bs
->drv
->bdrv_flush(bs
);
1600 * Some block drivers always operate in either writethrough or unsafe mode
1601 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1602 * the server works (because the behaviour is hardcoded or depends on
1603 * server-side configuration), so we can't ensure that everything is safe
1604 * on disk. Returning an error doesn't work because that would break guests
1605 * even if the server operates in writethrough mode.
1607 * Let's hope the user knows what he's doing.
1612 void bdrv_flush_all(void)
1614 BlockDriverState
*bs
;
1616 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1617 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1618 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1624 int bdrv_has_zero_init(BlockDriverState
*bs
)
1628 if (bs
->drv
->bdrv_has_zero_init
) {
1629 return bs
->drv
->bdrv_has_zero_init(bs
);
1635 int bdrv_discard(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
)
1640 if (!bs
->drv
->bdrv_discard
) {
1643 return bs
->drv
->bdrv_discard(bs
, sector_num
, nb_sectors
);
1647 * Returns true iff the specified sector is present in the disk image. Drivers
1648 * not implementing the functionality are assumed to not support backing files,
1649 * hence all their sectors are reported as allocated.
1651 * 'pnum' is set to the number of sectors (including and immediately following
1652 * the specified sector) that are known to be in the same
1653 * allocated/unallocated state.
1655 * 'nb_sectors' is the max value 'pnum' should be set to.
1657 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1661 if (!bs
->drv
->bdrv_is_allocated
) {
1662 if (sector_num
>= bs
->total_sectors
) {
1666 n
= bs
->total_sectors
- sector_num
;
1667 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1670 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1673 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1674 BlockMonEventAction action
, int is_read
)
1677 const char *action_str
;
1680 case BDRV_ACTION_REPORT
:
1681 action_str
= "report";
1683 case BDRV_ACTION_IGNORE
:
1684 action_str
= "ignore";
1686 case BDRV_ACTION_STOP
:
1687 action_str
= "stop";
1693 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1696 is_read
? "read" : "write");
1697 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1699 qobject_decref(data
);
1702 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1705 Monitor
*mon
= opaque
;
1707 bs_dict
= qobject_to_qdict(obj
);
1709 monitor_printf(mon
, "%s: type=%s removable=%d",
1710 qdict_get_str(bs_dict
, "device"),
1711 qdict_get_str(bs_dict
, "type"),
1712 qdict_get_bool(bs_dict
, "removable"));
1714 if (qdict_get_bool(bs_dict
, "removable")) {
1715 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1718 if (qdict_haskey(bs_dict
, "inserted")) {
1719 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1721 monitor_printf(mon
, " file=");
1722 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1723 if (qdict_haskey(qdict
, "backing_file")) {
1724 monitor_printf(mon
, " backing_file=");
1725 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1727 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1728 qdict_get_bool(qdict
, "ro"),
1729 qdict_get_str(qdict
, "drv"),
1730 qdict_get_bool(qdict
, "encrypted"));
1732 monitor_printf(mon
, " [not inserted]");
1735 monitor_printf(mon
, "\n");
1738 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1740 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1743 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1746 BlockDriverState
*bs
;
1748 bs_list
= qlist_new();
1750 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1752 const char *type
= "unknown";
1758 case BDRV_TYPE_CDROM
:
1761 case BDRV_TYPE_FLOPPY
:
1766 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1767 "'removable': %i, 'locked': %i }",
1768 bs
->device_name
, type
, bs
->removable
,
1773 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1775 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1776 "'encrypted': %i }",
1777 bs
->filename
, bs
->read_only
,
1778 bs
->drv
->format_name
,
1779 bdrv_is_encrypted(bs
));
1780 if (bs
->backing_file
[0] != '\0') {
1781 QDict
*qdict
= qobject_to_qdict(obj
);
1782 qdict_put(qdict
, "backing_file",
1783 qstring_from_str(bs
->backing_file
));
1786 qdict_put_obj(bs_dict
, "inserted", obj
);
1788 qlist_append_obj(bs_list
, bs_obj
);
1791 *ret_data
= QOBJECT(bs_list
);
1794 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1797 Monitor
*mon
= opaque
;
1799 qdict
= qobject_to_qdict(data
);
1800 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1802 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1803 monitor_printf(mon
, " rd_bytes=%" PRId64
1804 " wr_bytes=%" PRId64
1805 " rd_operations=%" PRId64
1806 " wr_operations=%" PRId64
1808 qdict_get_int(qdict
, "rd_bytes"),
1809 qdict_get_int(qdict
, "wr_bytes"),
1810 qdict_get_int(qdict
, "rd_operations"),
1811 qdict_get_int(qdict
, "wr_operations"));
1814 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1816 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1819 static QObject
* bdrv_info_stats_bs(BlockDriverState
*bs
)
1824 res
= qobject_from_jsonf("{ 'stats': {"
1825 "'rd_bytes': %" PRId64
","
1826 "'wr_bytes': %" PRId64
","
1827 "'rd_operations': %" PRId64
","
1828 "'wr_operations': %" PRId64
","
1829 "'wr_highest_offset': %" PRId64
1831 bs
->rd_bytes
, bs
->wr_bytes
,
1832 bs
->rd_ops
, bs
->wr_ops
,
1833 bs
->wr_highest_sector
*
1834 (uint64_t)BDRV_SECTOR_SIZE
);
1835 dict
= qobject_to_qdict(res
);
1837 if (*bs
->device_name
) {
1838 qdict_put(dict
, "device", qstring_from_str(bs
->device_name
));
1842 QObject
*parent
= bdrv_info_stats_bs(bs
->file
);
1843 qdict_put_obj(dict
, "parent", parent
);
1849 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1853 BlockDriverState
*bs
;
1855 devices
= qlist_new();
1857 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1858 obj
= bdrv_info_stats_bs(bs
);
1859 qlist_append_obj(devices
, obj
);
1862 *ret_data
= QOBJECT(devices
);
1865 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1867 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1868 return bs
->backing_file
;
1869 else if (bs
->encrypted
)
1870 return bs
->filename
;
1875 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1876 char *filename
, int filename_size
)
1878 if (!bs
->backing_file
) {
1879 pstrcpy(filename
, filename_size
, "");
1881 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1885 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1886 const uint8_t *buf
, int nb_sectors
)
1888 BlockDriver
*drv
= bs
->drv
;
1891 if (!drv
->bdrv_write_compressed
)
1893 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1896 if (bs
->dirty_bitmap
) {
1897 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1900 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1903 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1905 BlockDriver
*drv
= bs
->drv
;
1908 if (!drv
->bdrv_get_info
)
1910 memset(bdi
, 0, sizeof(*bdi
));
1911 return drv
->bdrv_get_info(bs
, bdi
);
1914 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1915 int64_t pos
, int size
)
1917 BlockDriver
*drv
= bs
->drv
;
1920 if (drv
->bdrv_save_vmstate
)
1921 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1923 return bdrv_save_vmstate(bs
->file
, buf
, pos
, size
);
1927 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1928 int64_t pos
, int size
)
1930 BlockDriver
*drv
= bs
->drv
;
1933 if (drv
->bdrv_load_vmstate
)
1934 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1936 return bdrv_load_vmstate(bs
->file
, buf
, pos
, size
);
1940 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1942 BlockDriver
*drv
= bs
->drv
;
1944 if (!drv
|| !drv
->bdrv_debug_event
) {
1948 return drv
->bdrv_debug_event(bs
, event
);
1952 /**************************************************************/
1953 /* handling of snapshots */
1955 int bdrv_can_snapshot(BlockDriverState
*bs
)
1957 BlockDriver
*drv
= bs
->drv
;
1958 if (!drv
|| bdrv_is_removable(bs
) || bdrv_is_read_only(bs
)) {
1962 if (!drv
->bdrv_snapshot_create
) {
1963 if (bs
->file
!= NULL
) {
1964 return bdrv_can_snapshot(bs
->file
);
1972 int bdrv_is_snapshot(BlockDriverState
*bs
)
1974 return !!(bs
->open_flags
& BDRV_O_SNAPSHOT
);
1977 BlockDriverState
*bdrv_snapshots(void)
1979 BlockDriverState
*bs
;
1982 return bs_snapshots
;
1986 while ((bs
= bdrv_next(bs
))) {
1987 if (bdrv_can_snapshot(bs
)) {
1995 int bdrv_snapshot_create(BlockDriverState
*bs
,
1996 QEMUSnapshotInfo
*sn_info
)
1998 BlockDriver
*drv
= bs
->drv
;
2001 if (drv
->bdrv_snapshot_create
)
2002 return drv
->bdrv_snapshot_create(bs
, sn_info
);
2004 return bdrv_snapshot_create(bs
->file
, sn_info
);
2008 int bdrv_snapshot_goto(BlockDriverState
*bs
,
2009 const char *snapshot_id
)
2011 BlockDriver
*drv
= bs
->drv
;
2016 if (drv
->bdrv_snapshot_goto
)
2017 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
2020 drv
->bdrv_close(bs
);
2021 ret
= bdrv_snapshot_goto(bs
->file
, snapshot_id
);
2022 open_ret
= drv
->bdrv_open(bs
, bs
->open_flags
);
2024 bdrv_delete(bs
->file
);
2034 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
2036 BlockDriver
*drv
= bs
->drv
;
2039 if (drv
->bdrv_snapshot_delete
)
2040 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
2042 return bdrv_snapshot_delete(bs
->file
, snapshot_id
);
2046 int bdrv_snapshot_list(BlockDriverState
*bs
,
2047 QEMUSnapshotInfo
**psn_info
)
2049 BlockDriver
*drv
= bs
->drv
;
2052 if (drv
->bdrv_snapshot_list
)
2053 return drv
->bdrv_snapshot_list(bs
, psn_info
);
2055 return bdrv_snapshot_list(bs
->file
, psn_info
);
2059 int bdrv_snapshot_load_tmp(BlockDriverState
*bs
,
2060 const char *snapshot_name
)
2062 BlockDriver
*drv
= bs
->drv
;
2066 if (!bs
->read_only
) {
2069 if (drv
->bdrv_snapshot_load_tmp
) {
2070 return drv
->bdrv_snapshot_load_tmp(bs
, snapshot_name
);
2075 #define NB_SUFFIXES 4
2077 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
2079 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
2084 snprintf(buf
, buf_size
, "%" PRId64
, size
);
2087 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
2088 if (size
< (10 * base
)) {
2089 snprintf(buf
, buf_size
, "%0.1f%c",
2090 (double)size
/ base
,
2093 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
2094 snprintf(buf
, buf_size
, "%" PRId64
"%c",
2095 ((size
+ (base
>> 1)) / base
),
2105 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
2107 char buf1
[128], date_buf
[128], clock_buf
[128];
2117 snprintf(buf
, buf_size
,
2118 "%-10s%-20s%7s%20s%15s",
2119 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2123 ptm
= localtime(&ti
);
2124 strftime(date_buf
, sizeof(date_buf
),
2125 "%Y-%m-%d %H:%M:%S", ptm
);
2127 localtime_r(&ti
, &tm
);
2128 strftime(date_buf
, sizeof(date_buf
),
2129 "%Y-%m-%d %H:%M:%S", &tm
);
2131 secs
= sn
->vm_clock_nsec
/ 1000000000;
2132 snprintf(clock_buf
, sizeof(clock_buf
),
2133 "%02d:%02d:%02d.%03d",
2135 (int)((secs
/ 60) % 60),
2137 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
2138 snprintf(buf
, buf_size
,
2139 "%-10s%-20s%7s%20s%15s",
2140 sn
->id_str
, sn
->name
,
2141 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
2149 /**************************************************************/
2152 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
2153 QEMUIOVector
*qiov
, int nb_sectors
,
2154 BlockDriverCompletionFunc
*cb
, void *opaque
)
2156 BlockDriver
*drv
= bs
->drv
;
2157 BlockDriverAIOCB
*ret
;
2159 trace_bdrv_aio_readv(bs
, sector_num
, nb_sectors
, opaque
);
2163 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2166 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
2170 /* Update stats even though technically transfer has not happened. */
2171 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
2178 typedef struct BlockCompleteData
{
2179 BlockDriverCompletionFunc
*cb
;
2181 BlockDriverState
*bs
;
2184 } BlockCompleteData
;
2186 static void block_complete_cb(void *opaque
, int ret
)
2188 BlockCompleteData
*b
= opaque
;
2190 if (b
->bs
->dirty_bitmap
) {
2191 set_dirty_bitmap(b
->bs
, b
->sector_num
, b
->nb_sectors
, 1);
2193 b
->cb(b
->opaque
, ret
);
2197 static BlockCompleteData
*blk_dirty_cb_alloc(BlockDriverState
*bs
,
2200 BlockDriverCompletionFunc
*cb
,
2203 BlockCompleteData
*blkdata
= qemu_mallocz(sizeof(BlockCompleteData
));
2207 blkdata
->opaque
= opaque
;
2208 blkdata
->sector_num
= sector_num
;
2209 blkdata
->nb_sectors
= nb_sectors
;
2214 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
2215 QEMUIOVector
*qiov
, int nb_sectors
,
2216 BlockDriverCompletionFunc
*cb
, void *opaque
)
2218 BlockDriver
*drv
= bs
->drv
;
2219 BlockDriverAIOCB
*ret
;
2220 BlockCompleteData
*blk_cb_data
;
2222 trace_bdrv_aio_writev(bs
, sector_num
, nb_sectors
, opaque
);
2228 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2231 if (bs
->dirty_bitmap
) {
2232 blk_cb_data
= blk_dirty_cb_alloc(bs
, sector_num
, nb_sectors
, cb
,
2234 cb
= &block_complete_cb
;
2235 opaque
= blk_cb_data
;
2238 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
2242 /* Update stats even though technically transfer has not happened. */
2243 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
2245 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
2246 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
2254 typedef struct MultiwriteCB
{
2259 BlockDriverCompletionFunc
*cb
;
2261 QEMUIOVector
*free_qiov
;
2266 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
2270 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2271 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
2272 if (mcb
->callbacks
[i
].free_qiov
) {
2273 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
2275 qemu_free(mcb
->callbacks
[i
].free_qiov
);
2276 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
2280 static void multiwrite_cb(void *opaque
, int ret
)
2282 MultiwriteCB
*mcb
= opaque
;
2284 trace_multiwrite_cb(mcb
, ret
);
2286 if (ret
< 0 && !mcb
->error
) {
2290 mcb
->num_requests
--;
2291 if (mcb
->num_requests
== 0) {
2292 multiwrite_user_cb(mcb
);
2297 static int multiwrite_req_compare(const void *a
, const void *b
)
2299 const BlockRequest
*req1
= a
, *req2
= b
;
2302 * Note that we can't simply subtract req2->sector from req1->sector
2303 * here as that could overflow the return value.
2305 if (req1
->sector
> req2
->sector
) {
2307 } else if (req1
->sector
< req2
->sector
) {
2315 * Takes a bunch of requests and tries to merge them. Returns the number of
2316 * requests that remain after merging.
2318 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
2319 int num_reqs
, MultiwriteCB
*mcb
)
2323 // Sort requests by start sector
2324 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
2326 // Check if adjacent requests touch the same clusters. If so, combine them,
2327 // filling up gaps with zero sectors.
2329 for (i
= 1; i
< num_reqs
; i
++) {
2331 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
2333 // This handles the cases that are valid for all block drivers, namely
2334 // exactly sequential writes and overlapping writes.
2335 if (reqs
[i
].sector
<= oldreq_last
) {
2339 // The block driver may decide that it makes sense to combine requests
2340 // even if there is a gap of some sectors between them. In this case,
2341 // the gap is filled with zeros (therefore only applicable for yet
2342 // unused space in format like qcow2).
2343 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
2344 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
2347 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
2353 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
2354 qemu_iovec_init(qiov
,
2355 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
2357 // Add the first request to the merged one. If the requests are
2358 // overlapping, drop the last sectors of the first request.
2359 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
2360 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
2362 // We might need to add some zeros between the two requests
2363 if (reqs
[i
].sector
> oldreq_last
) {
2364 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
2365 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
2366 memset(buf
, 0, zero_bytes
);
2367 qemu_iovec_add(qiov
, buf
, zero_bytes
);
2368 mcb
->callbacks
[i
].free_buf
= buf
;
2371 // Add the second request
2372 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2374 reqs
[outidx
].nb_sectors
= qiov
->size
>> 9;
2375 reqs
[outidx
].qiov
= qiov
;
2377 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2380 reqs
[outidx
].sector
= reqs
[i
].sector
;
2381 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2382 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2390 * Submit multiple AIO write requests at once.
2392 * On success, the function returns 0 and all requests in the reqs array have
2393 * been submitted. In error case this function returns -1, and any of the
2394 * requests may or may not be submitted yet. In particular, this means that the
2395 * callback will be called for some of the requests, for others it won't. The
2396 * caller must check the error field of the BlockRequest to wait for the right
2397 * callbacks (if error != 0, no callback will be called).
2399 * The implementation may modify the contents of the reqs array, e.g. to merge
2400 * requests. However, the fields opaque and error are left unmodified as they
2401 * are used to signal failure for a single request to the caller.
2403 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2405 BlockDriverAIOCB
*acb
;
2409 /* don't submit writes if we don't have a medium */
2410 if (bs
->drv
== NULL
) {
2411 for (i
= 0; i
< num_reqs
; i
++) {
2412 reqs
[i
].error
= -ENOMEDIUM
;
2417 if (num_reqs
== 0) {
2421 // Create MultiwriteCB structure
2422 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2423 mcb
->num_requests
= 0;
2424 mcb
->num_callbacks
= num_reqs
;
2426 for (i
= 0; i
< num_reqs
; i
++) {
2427 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2428 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2431 // Check for mergable requests
2432 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2434 trace_bdrv_aio_multiwrite(mcb
, mcb
->num_callbacks
, num_reqs
);
2437 * Run the aio requests. As soon as one request can't be submitted
2438 * successfully, fail all requests that are not yet submitted (we must
2439 * return failure for all requests anyway)
2441 * num_requests cannot be set to the right value immediately: If
2442 * bdrv_aio_writev fails for some request, num_requests would be too high
2443 * and therefore multiwrite_cb() would never recognize the multiwrite
2444 * request as completed. We also cannot use the loop variable i to set it
2445 * when the first request fails because the callback may already have been
2446 * called for previously submitted requests. Thus, num_requests must be
2447 * incremented for each request that is submitted.
2449 * The problem that callbacks may be called early also means that we need
2450 * to take care that num_requests doesn't become 0 before all requests are
2451 * submitted - multiwrite_cb() would consider the multiwrite request
2452 * completed. A dummy request that is "completed" by a manual call to
2453 * multiwrite_cb() takes care of this.
2455 mcb
->num_requests
= 1;
2457 // Run the aio requests
2458 for (i
= 0; i
< num_reqs
; i
++) {
2459 mcb
->num_requests
++;
2460 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2461 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2464 // We can only fail the whole thing if no request has been
2465 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2466 // complete and report the error in the callback.
2468 trace_bdrv_aio_multiwrite_earlyfail(mcb
);
2471 trace_bdrv_aio_multiwrite_latefail(mcb
, i
);
2472 multiwrite_cb(mcb
, -EIO
);
2478 /* Complete the dummy request */
2479 multiwrite_cb(mcb
, 0);
2484 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2485 reqs
[i
].error
= -EIO
;
2491 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2492 BlockDriverCompletionFunc
*cb
, void *opaque
)
2494 BlockDriver
*drv
= bs
->drv
;
2496 trace_bdrv_aio_flush(bs
, opaque
);
2498 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
2499 return bdrv_aio_noop_em(bs
, cb
, opaque
);
2504 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2507 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2509 acb
->pool
->cancel(acb
);
2513 /**************************************************************/
2514 /* async block device emulation */
2516 typedef struct BlockDriverAIOCBSync
{
2517 BlockDriverAIOCB common
;
2520 /* vector translation state */
2524 } BlockDriverAIOCBSync
;
2526 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2528 BlockDriverAIOCBSync
*acb
=
2529 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2530 qemu_bh_delete(acb
->bh
);
2532 qemu_aio_release(acb
);
2535 static AIOPool bdrv_em_aio_pool
= {
2536 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2537 .cancel
= bdrv_aio_cancel_em
,
2540 static void bdrv_aio_bh_cb(void *opaque
)
2542 BlockDriverAIOCBSync
*acb
= opaque
;
2545 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2546 qemu_vfree(acb
->bounce
);
2547 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2548 qemu_bh_delete(acb
->bh
);
2550 qemu_aio_release(acb
);
2553 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2557 BlockDriverCompletionFunc
*cb
,
2562 BlockDriverAIOCBSync
*acb
;
2564 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2565 acb
->is_write
= is_write
;
2567 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2570 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2573 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2574 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2576 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2579 qemu_bh_schedule(acb
->bh
);
2581 return &acb
->common
;
2584 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2585 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2586 BlockDriverCompletionFunc
*cb
, void *opaque
)
2588 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2591 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2592 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2593 BlockDriverCompletionFunc
*cb
, void *opaque
)
2595 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2598 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2599 BlockDriverCompletionFunc
*cb
, void *opaque
)
2601 BlockDriverAIOCBSync
*acb
;
2603 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2604 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2610 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2613 qemu_bh_schedule(acb
->bh
);
2614 return &acb
->common
;
2617 static BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
2618 BlockDriverCompletionFunc
*cb
, void *opaque
)
2620 BlockDriverAIOCBSync
*acb
;
2622 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2623 acb
->is_write
= 1; /* don't bounce in the completion handler */
2629 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2632 qemu_bh_schedule(acb
->bh
);
2633 return &acb
->common
;
2636 /**************************************************************/
2637 /* sync block device emulation */
2639 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2641 *(int *)opaque
= ret
;
2644 #define NOT_DONE 0x7fffffff
2646 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2647 uint8_t *buf
, int nb_sectors
)
2650 BlockDriverAIOCB
*acb
;
2654 async_context_push();
2656 async_ret
= NOT_DONE
;
2657 iov
.iov_base
= (void *)buf
;
2658 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2659 qemu_iovec_init_external(&qiov
, &iov
, 1);
2660 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2661 bdrv_rw_em_cb
, &async_ret
);
2667 while (async_ret
== NOT_DONE
) {
2673 async_context_pop();
2677 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2678 const uint8_t *buf
, int nb_sectors
)
2681 BlockDriverAIOCB
*acb
;
2685 async_context_push();
2687 async_ret
= NOT_DONE
;
2688 iov
.iov_base
= (void *)buf
;
2689 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2690 qemu_iovec_init_external(&qiov
, &iov
, 1);
2691 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2692 bdrv_rw_em_cb
, &async_ret
);
2697 while (async_ret
== NOT_DONE
) {
2702 async_context_pop();
2706 void bdrv_init(void)
2708 module_call_init(MODULE_INIT_BLOCK
);
2711 void bdrv_init_with_whitelist(void)
2713 use_bdrv_whitelist
= 1;
2717 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2718 BlockDriverCompletionFunc
*cb
, void *opaque
)
2720 BlockDriverAIOCB
*acb
;
2722 if (pool
->free_aiocb
) {
2723 acb
= pool
->free_aiocb
;
2724 pool
->free_aiocb
= acb
->next
;
2726 acb
= qemu_mallocz(pool
->aiocb_size
);
2731 acb
->opaque
= opaque
;
2735 void qemu_aio_release(void *p
)
2737 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2738 AIOPool
*pool
= acb
->pool
;
2739 acb
->next
= pool
->free_aiocb
;
2740 pool
->free_aiocb
= acb
;
2743 /**************************************************************/
2744 /* removable device support */
2747 * Return TRUE if the media is present
2749 int bdrv_is_inserted(BlockDriverState
*bs
)
2751 BlockDriver
*drv
= bs
->drv
;
2755 if (!drv
->bdrv_is_inserted
)
2756 return !bs
->tray_open
;
2757 ret
= drv
->bdrv_is_inserted(bs
);
2762 * Return TRUE if the media changed since the last call to this
2763 * function. It is currently only used for floppy disks
2765 int bdrv_media_changed(BlockDriverState
*bs
)
2767 BlockDriver
*drv
= bs
->drv
;
2770 if (!drv
|| !drv
->bdrv_media_changed
)
2773 ret
= drv
->bdrv_media_changed(bs
);
2774 if (ret
== -ENOTSUP
)
2775 ret
= bs
->media_changed
;
2776 bs
->media_changed
= 0;
2781 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2783 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2785 BlockDriver
*drv
= bs
->drv
;
2792 if (!drv
|| !drv
->bdrv_eject
) {
2795 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2797 if (ret
== -ENOTSUP
) {
2801 bs
->tray_open
= eject_flag
;
2807 int bdrv_is_locked(BlockDriverState
*bs
)
2813 * Lock or unlock the media (if it is locked, the user won't be able
2814 * to eject it manually).
2816 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2818 BlockDriver
*drv
= bs
->drv
;
2820 bs
->locked
= locked
;
2821 if (drv
&& drv
->bdrv_set_locked
) {
2822 drv
->bdrv_set_locked(bs
, locked
);
2826 /* needed for generic scsi interface */
2828 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2830 BlockDriver
*drv
= bs
->drv
;
2832 if (drv
&& drv
->bdrv_ioctl
)
2833 return drv
->bdrv_ioctl(bs
, req
, buf
);
2837 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2838 unsigned long int req
, void *buf
,
2839 BlockDriverCompletionFunc
*cb
, void *opaque
)
2841 BlockDriver
*drv
= bs
->drv
;
2843 if (drv
&& drv
->bdrv_aio_ioctl
)
2844 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2850 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2852 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2855 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2857 int64_t bitmap_size
;
2859 bs
->dirty_count
= 0;
2861 if (!bs
->dirty_bitmap
) {
2862 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2863 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2864 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2866 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2869 if (bs
->dirty_bitmap
) {
2870 qemu_free(bs
->dirty_bitmap
);
2871 bs
->dirty_bitmap
= NULL
;
2876 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2878 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2880 if (bs
->dirty_bitmap
&&
2881 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2882 return !!(bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2883 (1UL << (chunk
% (sizeof(unsigned long) * 8))));
2889 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2892 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2895 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2897 return bs
->dirty_count
;
2900 void bdrv_set_in_use(BlockDriverState
*bs
, int in_use
)
2902 assert(bs
->in_use
!= in_use
);
2903 bs
->in_use
= in_use
;
2906 int bdrv_in_use(BlockDriverState
*bs
)
2911 int bdrv_img_create(const char *filename
, const char *fmt
,
2912 const char *base_filename
, const char *base_fmt
,
2913 char *options
, uint64_t img_size
, int flags
)
2915 QEMUOptionParameter
*param
= NULL
, *create_options
= NULL
;
2916 QEMUOptionParameter
*backing_fmt
, *backing_file
;
2917 BlockDriverState
*bs
= NULL
;
2918 BlockDriver
*drv
, *proto_drv
;
2919 BlockDriver
*backing_drv
= NULL
;
2922 /* Find driver and parse its options */
2923 drv
= bdrv_find_format(fmt
);
2925 error_report("Unknown file format '%s'", fmt
);
2930 proto_drv
= bdrv_find_protocol(filename
);
2932 error_report("Unknown protocol '%s'", filename
);
2937 create_options
= append_option_parameters(create_options
,
2938 drv
->create_options
);
2939 create_options
= append_option_parameters(create_options
,
2940 proto_drv
->create_options
);
2942 /* Create parameter list with default values */
2943 param
= parse_option_parameters("", create_options
, param
);
2945 set_option_parameter_int(param
, BLOCK_OPT_SIZE
, img_size
);
2947 /* Parse -o options */
2949 param
= parse_option_parameters(options
, create_options
, param
);
2950 if (param
== NULL
) {
2951 error_report("Invalid options for file format '%s'.", fmt
);
2957 if (base_filename
) {
2958 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FILE
,
2960 error_report("Backing file not supported for file format '%s'",
2968 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FMT
, base_fmt
)) {
2969 error_report("Backing file format not supported for file "
2970 "format '%s'", fmt
);
2976 backing_file
= get_option_parameter(param
, BLOCK_OPT_BACKING_FILE
);
2977 if (backing_file
&& backing_file
->value
.s
) {
2978 if (!strcmp(filename
, backing_file
->value
.s
)) {
2979 error_report("Error: Trying to create an image with the "
2980 "same filename as the backing file");
2986 backing_fmt
= get_option_parameter(param
, BLOCK_OPT_BACKING_FMT
);
2987 if (backing_fmt
&& backing_fmt
->value
.s
) {
2988 backing_drv
= bdrv_find_format(backing_fmt
->value
.s
);
2990 error_report("Unknown backing file format '%s'",
2991 backing_fmt
->value
.s
);
2997 // The size for the image must always be specified, with one exception:
2998 // If we are using a backing file, we can obtain the size from there
2999 if (get_option_parameter(param
, BLOCK_OPT_SIZE
)->value
.n
== -1) {
3000 if (backing_file
&& backing_file
->value
.s
) {
3006 ret
= bdrv_open(bs
, backing_file
->value
.s
, flags
, backing_drv
);
3008 error_report("Could not open '%s'", backing_file
->value
.s
);
3011 bdrv_get_geometry(bs
, &size
);
3014 snprintf(buf
, sizeof(buf
), "%" PRId64
, size
);
3015 set_option_parameter(param
, BLOCK_OPT_SIZE
, buf
);
3017 error_report("Image creation needs a size parameter");
3023 printf("Formatting '%s', fmt=%s ", filename
, fmt
);
3024 print_option_parameters(param
);
3027 ret
= bdrv_create(drv
, filename
, param
);
3030 if (ret
== -ENOTSUP
) {
3031 error_report("Formatting or formatting option not supported for "
3032 "file format '%s'", fmt
);
3033 } else if (ret
== -EFBIG
) {
3034 error_report("The image size is too large for file format '%s'",
3037 error_report("%s: error while creating %s: %s", filename
, fmt
,
3043 free_option_parameters(create_options
);
3044 free_option_parameters(param
);