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
);
442 if (flags
& BDRV_O_CACHE_WB
)
443 bs
->enable_write_cache
= 1;
446 * Clear flags that are internal to the block layer before opening the
449 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
452 * Snapshots should be writable.
454 if (bs
->is_temporary
) {
455 open_flags
|= BDRV_O_RDWR
;
458 /* Open the image, either directly or using a protocol */
459 if (drv
->bdrv_file_open
) {
460 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
462 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
464 ret
= drv
->bdrv_open(bs
, open_flags
);
472 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
474 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
480 if (bs
->is_temporary
) {
488 bdrv_delete(bs
->file
);
491 qemu_free(bs
->opaque
);
498 * Opens a file using a protocol (file, host_device, nbd, ...)
500 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
502 BlockDriverState
*bs
;
506 drv
= bdrv_find_protocol(filename
);
512 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
523 * Opens a disk image (raw, qcow2, vmdk, ...)
525 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
530 if (flags
& BDRV_O_SNAPSHOT
) {
531 BlockDriverState
*bs1
;
534 BlockDriver
*bdrv_qcow2
;
535 QEMUOptionParameter
*options
;
536 char tmp_filename
[PATH_MAX
];
537 char backing_filename
[PATH_MAX
];
539 /* if snapshot, we create a temporary backing file and open it
540 instead of opening 'filename' directly */
542 /* if there is a backing file, use it */
544 ret
= bdrv_open(bs1
, filename
, 0, drv
);
549 total_size
= bdrv_getlength(bs1
) & BDRV_SECTOR_MASK
;
551 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
556 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
558 /* Real path is meaningless for protocols */
560 snprintf(backing_filename
, sizeof(backing_filename
),
562 else if (!realpath(filename
, backing_filename
))
565 bdrv_qcow2
= bdrv_find_format("qcow2");
566 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
568 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
);
569 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
571 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
575 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
576 free_option_parameters(options
);
581 filename
= tmp_filename
;
583 bs
->is_temporary
= 1;
586 /* Find the right image format driver */
588 ret
= find_image_format(filename
, &drv
);
592 goto unlink_and_fail
;
596 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
598 goto unlink_and_fail
;
601 /* If there is a backing file, use it */
602 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
603 char backing_filename
[PATH_MAX
];
605 BlockDriver
*back_drv
= NULL
;
607 bs
->backing_hd
= bdrv_new("");
609 if (path_has_protocol(bs
->backing_file
)) {
610 pstrcpy(backing_filename
, sizeof(backing_filename
),
613 path_combine(backing_filename
, sizeof(backing_filename
),
614 filename
, bs
->backing_file
);
617 if (bs
->backing_format
[0] != '\0') {
618 back_drv
= bdrv_find_format(bs
->backing_format
);
621 /* backing files always opened read-only */
623 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
625 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
630 if (bs
->is_temporary
) {
631 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
633 /* base image inherits from "parent" */
634 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
638 if (!bdrv_key_required(bs
)) {
639 /* call the change callback */
640 bs
->media_changed
= 1;
642 bs
->change_cb(bs
->change_opaque
, CHANGE_MEDIA
);
648 if (bs
->is_temporary
) {
654 void bdrv_close(BlockDriverState
*bs
)
657 if (bs
== bs_snapshots
) {
660 if (bs
->backing_hd
) {
661 bdrv_delete(bs
->backing_hd
);
662 bs
->backing_hd
= NULL
;
664 bs
->drv
->bdrv_close(bs
);
665 qemu_free(bs
->opaque
);
667 if (bs
->is_temporary
) {
668 unlink(bs
->filename
);
674 if (bs
->file
!= NULL
) {
675 bdrv_close(bs
->file
);
678 /* call the change callback */
679 bs
->media_changed
= 1;
681 bs
->change_cb(bs
->change_opaque
, CHANGE_MEDIA
);
685 void bdrv_close_all(void)
687 BlockDriverState
*bs
;
689 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
694 /* make a BlockDriverState anonymous by removing from bdrv_state list.
695 Also, NULL terminate the device_name to prevent double remove */
696 void bdrv_make_anon(BlockDriverState
*bs
)
698 if (bs
->device_name
[0] != '\0') {
699 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
701 bs
->device_name
[0] = '\0';
704 void bdrv_delete(BlockDriverState
*bs
)
708 /* remove from list, if necessary */
712 if (bs
->file
!= NULL
) {
713 bdrv_delete(bs
->file
);
716 assert(bs
!= bs_snapshots
);
720 int bdrv_attach(BlockDriverState
*bs
, DeviceState
*qdev
)
729 void bdrv_detach(BlockDriverState
*bs
, DeviceState
*qdev
)
731 assert(bs
->peer
== qdev
);
733 bs
->change_cb
= NULL
;
734 bs
->change_opaque
= NULL
;
737 DeviceState
*bdrv_get_attached(BlockDriverState
*bs
)
743 * Run consistency checks on an image
745 * Returns 0 if the check could be completed (it doesn't mean that the image is
746 * free of errors) or -errno when an internal error occurred. The results of the
747 * check are stored in res.
749 int bdrv_check(BlockDriverState
*bs
, BdrvCheckResult
*res
)
751 if (bs
->drv
->bdrv_check
== NULL
) {
755 memset(res
, 0, sizeof(*res
));
756 return bs
->drv
->bdrv_check(bs
, res
);
759 #define COMMIT_BUF_SECTORS 2048
761 /* commit COW file into the raw image */
762 int bdrv_commit(BlockDriverState
*bs
)
764 BlockDriver
*drv
= bs
->drv
;
765 BlockDriver
*backing_drv
;
766 int64_t sector
, total_sectors
;
767 int n
, ro
, open_flags
;
768 int ret
= 0, rw_ret
= 0;
771 BlockDriverState
*bs_rw
, *bs_ro
;
776 if (!bs
->backing_hd
) {
780 if (bs
->backing_hd
->keep_read_only
) {
784 backing_drv
= bs
->backing_hd
->drv
;
785 ro
= bs
->backing_hd
->read_only
;
786 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
787 open_flags
= bs
->backing_hd
->open_flags
;
791 bdrv_delete(bs
->backing_hd
);
792 bs
->backing_hd
= NULL
;
793 bs_rw
= bdrv_new("");
794 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
,
798 /* try to re-open read-only */
799 bs_ro
= bdrv_new("");
800 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
804 /* drive not functional anymore */
808 bs
->backing_hd
= bs_ro
;
811 bs
->backing_hd
= bs_rw
;
814 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
815 buf
= qemu_malloc(COMMIT_BUF_SECTORS
* BDRV_SECTOR_SIZE
);
817 for (sector
= 0; sector
< total_sectors
; sector
+= n
) {
818 if (drv
->bdrv_is_allocated(bs
, sector
, COMMIT_BUF_SECTORS
, &n
)) {
820 if (bdrv_read(bs
, sector
, buf
, n
) != 0) {
825 if (bdrv_write(bs
->backing_hd
, sector
, buf
, n
) != 0) {
832 if (drv
->bdrv_make_empty
) {
833 ret
= drv
->bdrv_make_empty(bs
);
838 * Make sure all data we wrote to the backing device is actually
842 bdrv_flush(bs
->backing_hd
);
849 bdrv_delete(bs
->backing_hd
);
850 bs
->backing_hd
= NULL
;
851 bs_ro
= bdrv_new("");
852 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
856 /* drive not functional anymore */
860 bs
->backing_hd
= bs_ro
;
861 bs
->backing_hd
->keep_read_only
= 0;
867 void bdrv_commit_all(void)
869 BlockDriverState
*bs
;
871 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
879 * -EINVAL - backing format specified, but no file
880 * -ENOSPC - can't update the backing file because no space is left in the
882 * -ENOTSUP - format driver doesn't support changing the backing file
884 int bdrv_change_backing_file(BlockDriverState
*bs
,
885 const char *backing_file
, const char *backing_fmt
)
887 BlockDriver
*drv
= bs
->drv
;
889 if (drv
->bdrv_change_backing_file
!= NULL
) {
890 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
896 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
901 if (!bdrv_is_inserted(bs
))
907 len
= bdrv_getlength(bs
);
912 if ((offset
> len
) || (len
- offset
< size
))
918 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
921 return bdrv_check_byte_request(bs
, sector_num
* BDRV_SECTOR_SIZE
,
922 nb_sectors
* BDRV_SECTOR_SIZE
);
925 /* return < 0 if error. See bdrv_write() for the return codes */
926 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
927 uint8_t *buf
, int nb_sectors
)
929 BlockDriver
*drv
= bs
->drv
;
933 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
936 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
939 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
940 int nb_sectors
, int dirty
)
943 unsigned long val
, idx
, bit
;
945 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
946 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
948 for (; start
<= end
; start
++) {
949 idx
= start
/ (sizeof(unsigned long) * 8);
950 bit
= start
% (sizeof(unsigned long) * 8);
951 val
= bs
->dirty_bitmap
[idx
];
953 if (!(val
& (1UL << bit
))) {
958 if (val
& (1UL << bit
)) {
960 val
&= ~(1UL << bit
);
963 bs
->dirty_bitmap
[idx
] = val
;
967 /* Return < 0 if error. Important errors are:
968 -EIO generic I/O error (may happen for all errors)
969 -ENOMEDIUM No media inserted.
970 -EINVAL Invalid sector number or nb_sectors
971 -EACCES Trying to write a read-only device
973 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
974 const uint8_t *buf
, int nb_sectors
)
976 BlockDriver
*drv
= bs
->drv
;
981 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
984 if (bs
->dirty_bitmap
) {
985 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
988 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
989 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
992 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
995 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
996 void *buf
, int count1
)
998 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
999 int len
, nb_sectors
, count
;
1004 /* first read to align to sector start */
1005 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1008 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1010 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1012 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
1020 /* read the sectors "in place" */
1021 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1022 if (nb_sectors
> 0) {
1023 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1025 sector_num
+= nb_sectors
;
1026 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1031 /* add data from the last sector */
1033 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1035 memcpy(buf
, tmp_buf
, count
);
1040 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
1041 const void *buf
, int count1
)
1043 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
1044 int len
, nb_sectors
, count
;
1049 /* first write to align to sector start */
1050 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1053 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1055 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1057 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
1058 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1067 /* write the sectors "in place" */
1068 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1069 if (nb_sectors
> 0) {
1070 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1072 sector_num
+= nb_sectors
;
1073 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1078 /* add data from the last sector */
1080 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1082 memcpy(tmp_buf
, buf
, count
);
1083 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1090 * Writes to the file and ensures that no writes are reordered across this
1091 * request (acts as a barrier)
1093 * Returns 0 on success, -errno in error cases.
1095 int bdrv_pwrite_sync(BlockDriverState
*bs
, int64_t offset
,
1096 const void *buf
, int count
)
1100 ret
= bdrv_pwrite(bs
, offset
, buf
, count
);
1105 /* No flush needed for cache=writethrough, it uses O_DSYNC */
1106 if ((bs
->open_flags
& BDRV_O_CACHE_MASK
) != 0) {
1114 * Truncate file to 'offset' bytes (needed only for file protocols)
1116 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
1118 BlockDriver
*drv
= bs
->drv
;
1122 if (!drv
->bdrv_truncate
)
1126 if (bdrv_in_use(bs
))
1128 ret
= drv
->bdrv_truncate(bs
, offset
);
1130 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1131 if (bs
->change_cb
) {
1132 bs
->change_cb(bs
->change_opaque
, CHANGE_SIZE
);
1139 * Length of a allocated file in bytes. Sparse files are counted by actual
1140 * allocated space. Return < 0 if error or unknown.
1142 int64_t bdrv_get_allocated_file_size(BlockDriverState
*bs
)
1144 BlockDriver
*drv
= bs
->drv
;
1148 if (drv
->bdrv_get_allocated_file_size
) {
1149 return drv
->bdrv_get_allocated_file_size(bs
);
1152 return bdrv_get_allocated_file_size(bs
->file
);
1158 * Length of a file in bytes. Return < 0 if error or unknown.
1160 int64_t bdrv_getlength(BlockDriverState
*bs
)
1162 BlockDriver
*drv
= bs
->drv
;
1166 if (bs
->growable
|| bs
->removable
) {
1167 if (drv
->bdrv_getlength
) {
1168 return drv
->bdrv_getlength(bs
);
1171 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
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_translation_hint(BlockDriverState
*bs
, int translation
)
1312 bs
->translation
= translation
;
1315 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1316 int *pcyls
, int *pheads
, int *psecs
)
1319 *pheads
= bs
->heads
;
1323 /* Recognize floppy formats */
1324 typedef struct FDFormat
{
1331 static const FDFormat fd_formats
[] = {
1332 /* First entry is default format */
1333 /* 1.44 MB 3"1/2 floppy disks */
1334 { FDRIVE_DRV_144
, 18, 80, 1, },
1335 { FDRIVE_DRV_144
, 20, 80, 1, },
1336 { FDRIVE_DRV_144
, 21, 80, 1, },
1337 { FDRIVE_DRV_144
, 21, 82, 1, },
1338 { FDRIVE_DRV_144
, 21, 83, 1, },
1339 { FDRIVE_DRV_144
, 22, 80, 1, },
1340 { FDRIVE_DRV_144
, 23, 80, 1, },
1341 { FDRIVE_DRV_144
, 24, 80, 1, },
1342 /* 2.88 MB 3"1/2 floppy disks */
1343 { FDRIVE_DRV_288
, 36, 80, 1, },
1344 { FDRIVE_DRV_288
, 39, 80, 1, },
1345 { FDRIVE_DRV_288
, 40, 80, 1, },
1346 { FDRIVE_DRV_288
, 44, 80, 1, },
1347 { FDRIVE_DRV_288
, 48, 80, 1, },
1348 /* 720 kB 3"1/2 floppy disks */
1349 { FDRIVE_DRV_144
, 9, 80, 1, },
1350 { FDRIVE_DRV_144
, 10, 80, 1, },
1351 { FDRIVE_DRV_144
, 10, 82, 1, },
1352 { FDRIVE_DRV_144
, 10, 83, 1, },
1353 { FDRIVE_DRV_144
, 13, 80, 1, },
1354 { FDRIVE_DRV_144
, 14, 80, 1, },
1355 /* 1.2 MB 5"1/4 floppy disks */
1356 { FDRIVE_DRV_120
, 15, 80, 1, },
1357 { FDRIVE_DRV_120
, 18, 80, 1, },
1358 { FDRIVE_DRV_120
, 18, 82, 1, },
1359 { FDRIVE_DRV_120
, 18, 83, 1, },
1360 { FDRIVE_DRV_120
, 20, 80, 1, },
1361 /* 720 kB 5"1/4 floppy disks */
1362 { FDRIVE_DRV_120
, 9, 80, 1, },
1363 { FDRIVE_DRV_120
, 11, 80, 1, },
1364 /* 360 kB 5"1/4 floppy disks */
1365 { FDRIVE_DRV_120
, 9, 40, 1, },
1366 { FDRIVE_DRV_120
, 9, 40, 0, },
1367 { FDRIVE_DRV_120
, 10, 41, 1, },
1368 { FDRIVE_DRV_120
, 10, 42, 1, },
1369 /* 320 kB 5"1/4 floppy disks */
1370 { FDRIVE_DRV_120
, 8, 40, 1, },
1371 { FDRIVE_DRV_120
, 8, 40, 0, },
1372 /* 360 kB must match 5"1/4 better than 3"1/2... */
1373 { FDRIVE_DRV_144
, 9, 80, 0, },
1375 { FDRIVE_DRV_NONE
, -1, -1, 0, },
1378 void bdrv_get_floppy_geometry_hint(BlockDriverState
*bs
, int *nb_heads
,
1379 int *max_track
, int *last_sect
,
1380 FDriveType drive_in
, FDriveType
*drive
)
1382 const FDFormat
*parse
;
1383 uint64_t nb_sectors
, size
;
1384 int i
, first_match
, match
;
1386 bdrv_get_geometry_hint(bs
, nb_heads
, max_track
, last_sect
);
1387 if (*nb_heads
!= 0 && *max_track
!= 0 && *last_sect
!= 0) {
1388 /* User defined disk */
1390 bdrv_get_geometry(bs
, &nb_sectors
);
1393 for (i
= 0; ; i
++) {
1394 parse
= &fd_formats
[i
];
1395 if (parse
->drive
== FDRIVE_DRV_NONE
) {
1398 if (drive_in
== parse
->drive
||
1399 drive_in
== FDRIVE_DRV_NONE
) {
1400 size
= (parse
->max_head
+ 1) * parse
->max_track
*
1402 if (nb_sectors
== size
) {
1406 if (first_match
== -1) {
1412 if (first_match
== -1) {
1415 match
= first_match
;
1417 parse
= &fd_formats
[match
];
1419 *nb_heads
= parse
->max_head
+ 1;
1420 *max_track
= parse
->max_track
;
1421 *last_sect
= parse
->last_sect
;
1422 *drive
= parse
->drive
;
1426 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1428 return bs
->translation
;
1431 void bdrv_set_on_error(BlockDriverState
*bs
, BlockErrorAction on_read_error
,
1432 BlockErrorAction on_write_error
)
1434 bs
->on_read_error
= on_read_error
;
1435 bs
->on_write_error
= on_write_error
;
1438 BlockErrorAction
bdrv_get_on_error(BlockDriverState
*bs
, int is_read
)
1440 return is_read
? bs
->on_read_error
: bs
->on_write_error
;
1443 void bdrv_set_removable(BlockDriverState
*bs
, int removable
)
1445 bs
->removable
= removable
;
1446 if (removable
&& bs
== bs_snapshots
) {
1447 bs_snapshots
= NULL
;
1451 int bdrv_is_removable(BlockDriverState
*bs
)
1453 return bs
->removable
;
1456 int bdrv_is_read_only(BlockDriverState
*bs
)
1458 return bs
->read_only
;
1461 int bdrv_is_sg(BlockDriverState
*bs
)
1466 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1468 return bs
->enable_write_cache
;
1471 /* XXX: no longer used */
1472 void bdrv_set_change_cb(BlockDriverState
*bs
,
1473 void (*change_cb
)(void *opaque
, int reason
),
1476 bs
->change_cb
= change_cb
;
1477 bs
->change_opaque
= opaque
;
1480 int bdrv_is_encrypted(BlockDriverState
*bs
)
1482 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1484 return bs
->encrypted
;
1487 int bdrv_key_required(BlockDriverState
*bs
)
1489 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1491 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1493 return (bs
->encrypted
&& !bs
->valid_key
);
1496 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1499 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1500 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1506 if (!bs
->encrypted
) {
1508 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1511 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1514 } else if (!bs
->valid_key
) {
1516 /* call the change callback now, we skipped it on open */
1517 bs
->media_changed
= 1;
1519 bs
->change_cb(bs
->change_opaque
, CHANGE_MEDIA
);
1524 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1529 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1533 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1538 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1539 it(opaque
, drv
->format_name
);
1543 BlockDriverState
*bdrv_find(const char *name
)
1545 BlockDriverState
*bs
;
1547 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1548 if (!strcmp(name
, bs
->device_name
)) {
1555 BlockDriverState
*bdrv_next(BlockDriverState
*bs
)
1558 return QTAILQ_FIRST(&bdrv_states
);
1560 return QTAILQ_NEXT(bs
, list
);
1563 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1565 BlockDriverState
*bs
;
1567 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1572 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1574 return bs
->device_name
;
1577 int bdrv_flush(BlockDriverState
*bs
)
1579 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
1583 if (bs
->drv
&& bs
->drv
->bdrv_flush
) {
1584 return bs
->drv
->bdrv_flush(bs
);
1588 * Some block drivers always operate in either writethrough or unsafe mode
1589 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1590 * the server works (because the behaviour is hardcoded or depends on
1591 * server-side configuration), so we can't ensure that everything is safe
1592 * on disk. Returning an error doesn't work because that would break guests
1593 * even if the server operates in writethrough mode.
1595 * Let's hope the user knows what he's doing.
1600 void bdrv_flush_all(void)
1602 BlockDriverState
*bs
;
1604 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1605 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1606 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1612 int bdrv_has_zero_init(BlockDriverState
*bs
)
1616 if (bs
->drv
->bdrv_has_zero_init
) {
1617 return bs
->drv
->bdrv_has_zero_init(bs
);
1623 int bdrv_discard(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
)
1628 if (!bs
->drv
->bdrv_discard
) {
1631 return bs
->drv
->bdrv_discard(bs
, sector_num
, nb_sectors
);
1635 * Returns true iff the specified sector is present in the disk image. Drivers
1636 * not implementing the functionality are assumed to not support backing files,
1637 * hence all their sectors are reported as allocated.
1639 * 'pnum' is set to the number of sectors (including and immediately following
1640 * the specified sector) that are known to be in the same
1641 * allocated/unallocated state.
1643 * 'nb_sectors' is the max value 'pnum' should be set to.
1645 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1649 if (!bs
->drv
->bdrv_is_allocated
) {
1650 if (sector_num
>= bs
->total_sectors
) {
1654 n
= bs
->total_sectors
- sector_num
;
1655 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1658 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1661 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1662 BlockMonEventAction action
, int is_read
)
1665 const char *action_str
;
1668 case BDRV_ACTION_REPORT
:
1669 action_str
= "report";
1671 case BDRV_ACTION_IGNORE
:
1672 action_str
= "ignore";
1674 case BDRV_ACTION_STOP
:
1675 action_str
= "stop";
1681 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1684 is_read
? "read" : "write");
1685 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1687 qobject_decref(data
);
1690 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1693 Monitor
*mon
= opaque
;
1695 bs_dict
= qobject_to_qdict(obj
);
1697 monitor_printf(mon
, "%s: removable=%d",
1698 qdict_get_str(bs_dict
, "device"),
1699 qdict_get_bool(bs_dict
, "removable"));
1701 if (qdict_get_bool(bs_dict
, "removable")) {
1702 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1705 if (qdict_haskey(bs_dict
, "inserted")) {
1706 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1708 monitor_printf(mon
, " file=");
1709 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1710 if (qdict_haskey(qdict
, "backing_file")) {
1711 monitor_printf(mon
, " backing_file=");
1712 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1714 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1715 qdict_get_bool(qdict
, "ro"),
1716 qdict_get_str(qdict
, "drv"),
1717 qdict_get_bool(qdict
, "encrypted"));
1719 monitor_printf(mon
, " [not inserted]");
1722 monitor_printf(mon
, "\n");
1725 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1727 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1730 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1733 BlockDriverState
*bs
;
1735 bs_list
= qlist_new();
1737 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1740 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
1741 "'removable': %i, 'locked': %i }",
1742 bs
->device_name
, bs
->removable
,
1747 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1749 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1750 "'encrypted': %i }",
1751 bs
->filename
, bs
->read_only
,
1752 bs
->drv
->format_name
,
1753 bdrv_is_encrypted(bs
));
1754 if (bs
->backing_file
[0] != '\0') {
1755 QDict
*qdict
= qobject_to_qdict(obj
);
1756 qdict_put(qdict
, "backing_file",
1757 qstring_from_str(bs
->backing_file
));
1760 qdict_put_obj(bs_dict
, "inserted", obj
);
1762 qlist_append_obj(bs_list
, bs_obj
);
1765 *ret_data
= QOBJECT(bs_list
);
1768 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1771 Monitor
*mon
= opaque
;
1773 qdict
= qobject_to_qdict(data
);
1774 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1776 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1777 monitor_printf(mon
, " rd_bytes=%" PRId64
1778 " wr_bytes=%" PRId64
1779 " rd_operations=%" PRId64
1780 " wr_operations=%" PRId64
1782 qdict_get_int(qdict
, "rd_bytes"),
1783 qdict_get_int(qdict
, "wr_bytes"),
1784 qdict_get_int(qdict
, "rd_operations"),
1785 qdict_get_int(qdict
, "wr_operations"));
1788 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1790 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1793 static QObject
* bdrv_info_stats_bs(BlockDriverState
*bs
)
1798 res
= qobject_from_jsonf("{ 'stats': {"
1799 "'rd_bytes': %" PRId64
","
1800 "'wr_bytes': %" PRId64
","
1801 "'rd_operations': %" PRId64
","
1802 "'wr_operations': %" PRId64
","
1803 "'wr_highest_offset': %" PRId64
1805 bs
->rd_bytes
, bs
->wr_bytes
,
1806 bs
->rd_ops
, bs
->wr_ops
,
1807 bs
->wr_highest_sector
*
1808 (uint64_t)BDRV_SECTOR_SIZE
);
1809 dict
= qobject_to_qdict(res
);
1811 if (*bs
->device_name
) {
1812 qdict_put(dict
, "device", qstring_from_str(bs
->device_name
));
1816 QObject
*parent
= bdrv_info_stats_bs(bs
->file
);
1817 qdict_put_obj(dict
, "parent", parent
);
1823 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1827 BlockDriverState
*bs
;
1829 devices
= qlist_new();
1831 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1832 obj
= bdrv_info_stats_bs(bs
);
1833 qlist_append_obj(devices
, obj
);
1836 *ret_data
= QOBJECT(devices
);
1839 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1841 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1842 return bs
->backing_file
;
1843 else if (bs
->encrypted
)
1844 return bs
->filename
;
1849 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1850 char *filename
, int filename_size
)
1852 if (!bs
->backing_file
) {
1853 pstrcpy(filename
, filename_size
, "");
1855 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1859 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1860 const uint8_t *buf
, int nb_sectors
)
1862 BlockDriver
*drv
= bs
->drv
;
1865 if (!drv
->bdrv_write_compressed
)
1867 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1870 if (bs
->dirty_bitmap
) {
1871 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1874 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1877 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1879 BlockDriver
*drv
= bs
->drv
;
1882 if (!drv
->bdrv_get_info
)
1884 memset(bdi
, 0, sizeof(*bdi
));
1885 return drv
->bdrv_get_info(bs
, bdi
);
1888 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1889 int64_t pos
, int size
)
1891 BlockDriver
*drv
= bs
->drv
;
1894 if (drv
->bdrv_save_vmstate
)
1895 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1897 return bdrv_save_vmstate(bs
->file
, buf
, pos
, size
);
1901 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1902 int64_t pos
, int size
)
1904 BlockDriver
*drv
= bs
->drv
;
1907 if (drv
->bdrv_load_vmstate
)
1908 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1910 return bdrv_load_vmstate(bs
->file
, buf
, pos
, size
);
1914 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1916 BlockDriver
*drv
= bs
->drv
;
1918 if (!drv
|| !drv
->bdrv_debug_event
) {
1922 return drv
->bdrv_debug_event(bs
, event
);
1926 /**************************************************************/
1927 /* handling of snapshots */
1929 int bdrv_can_snapshot(BlockDriverState
*bs
)
1931 BlockDriver
*drv
= bs
->drv
;
1932 if (!drv
|| bdrv_is_removable(bs
) || bdrv_is_read_only(bs
)) {
1936 if (!drv
->bdrv_snapshot_create
) {
1937 if (bs
->file
!= NULL
) {
1938 return bdrv_can_snapshot(bs
->file
);
1946 int bdrv_is_snapshot(BlockDriverState
*bs
)
1948 return !!(bs
->open_flags
& BDRV_O_SNAPSHOT
);
1951 BlockDriverState
*bdrv_snapshots(void)
1953 BlockDriverState
*bs
;
1956 return bs_snapshots
;
1960 while ((bs
= bdrv_next(bs
))) {
1961 if (bdrv_can_snapshot(bs
)) {
1969 int bdrv_snapshot_create(BlockDriverState
*bs
,
1970 QEMUSnapshotInfo
*sn_info
)
1972 BlockDriver
*drv
= bs
->drv
;
1975 if (drv
->bdrv_snapshot_create
)
1976 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1978 return bdrv_snapshot_create(bs
->file
, sn_info
);
1982 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1983 const char *snapshot_id
)
1985 BlockDriver
*drv
= bs
->drv
;
1990 if (drv
->bdrv_snapshot_goto
)
1991 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1994 drv
->bdrv_close(bs
);
1995 ret
= bdrv_snapshot_goto(bs
->file
, snapshot_id
);
1996 open_ret
= drv
->bdrv_open(bs
, bs
->open_flags
);
1998 bdrv_delete(bs
->file
);
2008 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
2010 BlockDriver
*drv
= bs
->drv
;
2013 if (drv
->bdrv_snapshot_delete
)
2014 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
2016 return bdrv_snapshot_delete(bs
->file
, snapshot_id
);
2020 int bdrv_snapshot_list(BlockDriverState
*bs
,
2021 QEMUSnapshotInfo
**psn_info
)
2023 BlockDriver
*drv
= bs
->drv
;
2026 if (drv
->bdrv_snapshot_list
)
2027 return drv
->bdrv_snapshot_list(bs
, psn_info
);
2029 return bdrv_snapshot_list(bs
->file
, psn_info
);
2033 int bdrv_snapshot_load_tmp(BlockDriverState
*bs
,
2034 const char *snapshot_name
)
2036 BlockDriver
*drv
= bs
->drv
;
2040 if (!bs
->read_only
) {
2043 if (drv
->bdrv_snapshot_load_tmp
) {
2044 return drv
->bdrv_snapshot_load_tmp(bs
, snapshot_name
);
2049 #define NB_SUFFIXES 4
2051 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
2053 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
2058 snprintf(buf
, buf_size
, "%" PRId64
, size
);
2061 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
2062 if (size
< (10 * base
)) {
2063 snprintf(buf
, buf_size
, "%0.1f%c",
2064 (double)size
/ base
,
2067 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
2068 snprintf(buf
, buf_size
, "%" PRId64
"%c",
2069 ((size
+ (base
>> 1)) / base
),
2079 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
2081 char buf1
[128], date_buf
[128], clock_buf
[128];
2091 snprintf(buf
, buf_size
,
2092 "%-10s%-20s%7s%20s%15s",
2093 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2097 ptm
= localtime(&ti
);
2098 strftime(date_buf
, sizeof(date_buf
),
2099 "%Y-%m-%d %H:%M:%S", ptm
);
2101 localtime_r(&ti
, &tm
);
2102 strftime(date_buf
, sizeof(date_buf
),
2103 "%Y-%m-%d %H:%M:%S", &tm
);
2105 secs
= sn
->vm_clock_nsec
/ 1000000000;
2106 snprintf(clock_buf
, sizeof(clock_buf
),
2107 "%02d:%02d:%02d.%03d",
2109 (int)((secs
/ 60) % 60),
2111 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
2112 snprintf(buf
, buf_size
,
2113 "%-10s%-20s%7s%20s%15s",
2114 sn
->id_str
, sn
->name
,
2115 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
2123 /**************************************************************/
2126 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
2127 QEMUIOVector
*qiov
, int nb_sectors
,
2128 BlockDriverCompletionFunc
*cb
, void *opaque
)
2130 BlockDriver
*drv
= bs
->drv
;
2131 BlockDriverAIOCB
*ret
;
2133 trace_bdrv_aio_readv(bs
, sector_num
, nb_sectors
, opaque
);
2137 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2140 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
2144 /* Update stats even though technically transfer has not happened. */
2145 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
2152 typedef struct BlockCompleteData
{
2153 BlockDriverCompletionFunc
*cb
;
2155 BlockDriverState
*bs
;
2158 } BlockCompleteData
;
2160 static void block_complete_cb(void *opaque
, int ret
)
2162 BlockCompleteData
*b
= opaque
;
2164 if (b
->bs
->dirty_bitmap
) {
2165 set_dirty_bitmap(b
->bs
, b
->sector_num
, b
->nb_sectors
, 1);
2167 b
->cb(b
->opaque
, ret
);
2171 static BlockCompleteData
*blk_dirty_cb_alloc(BlockDriverState
*bs
,
2174 BlockDriverCompletionFunc
*cb
,
2177 BlockCompleteData
*blkdata
= qemu_mallocz(sizeof(BlockCompleteData
));
2181 blkdata
->opaque
= opaque
;
2182 blkdata
->sector_num
= sector_num
;
2183 blkdata
->nb_sectors
= nb_sectors
;
2188 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
2189 QEMUIOVector
*qiov
, int nb_sectors
,
2190 BlockDriverCompletionFunc
*cb
, void *opaque
)
2192 BlockDriver
*drv
= bs
->drv
;
2193 BlockDriverAIOCB
*ret
;
2194 BlockCompleteData
*blk_cb_data
;
2196 trace_bdrv_aio_writev(bs
, sector_num
, nb_sectors
, opaque
);
2202 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2205 if (bs
->dirty_bitmap
) {
2206 blk_cb_data
= blk_dirty_cb_alloc(bs
, sector_num
, nb_sectors
, cb
,
2208 cb
= &block_complete_cb
;
2209 opaque
= blk_cb_data
;
2212 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
2216 /* Update stats even though technically transfer has not happened. */
2217 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
2219 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
2220 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
2228 typedef struct MultiwriteCB
{
2233 BlockDriverCompletionFunc
*cb
;
2235 QEMUIOVector
*free_qiov
;
2240 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
2244 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2245 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
2246 if (mcb
->callbacks
[i
].free_qiov
) {
2247 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
2249 qemu_free(mcb
->callbacks
[i
].free_qiov
);
2250 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
2254 static void multiwrite_cb(void *opaque
, int ret
)
2256 MultiwriteCB
*mcb
= opaque
;
2258 trace_multiwrite_cb(mcb
, ret
);
2260 if (ret
< 0 && !mcb
->error
) {
2264 mcb
->num_requests
--;
2265 if (mcb
->num_requests
== 0) {
2266 multiwrite_user_cb(mcb
);
2271 static int multiwrite_req_compare(const void *a
, const void *b
)
2273 const BlockRequest
*req1
= a
, *req2
= b
;
2276 * Note that we can't simply subtract req2->sector from req1->sector
2277 * here as that could overflow the return value.
2279 if (req1
->sector
> req2
->sector
) {
2281 } else if (req1
->sector
< req2
->sector
) {
2289 * Takes a bunch of requests and tries to merge them. Returns the number of
2290 * requests that remain after merging.
2292 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
2293 int num_reqs
, MultiwriteCB
*mcb
)
2297 // Sort requests by start sector
2298 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
2300 // Check if adjacent requests touch the same clusters. If so, combine them,
2301 // filling up gaps with zero sectors.
2303 for (i
= 1; i
< num_reqs
; i
++) {
2305 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
2307 // This handles the cases that are valid for all block drivers, namely
2308 // exactly sequential writes and overlapping writes.
2309 if (reqs
[i
].sector
<= oldreq_last
) {
2313 // The block driver may decide that it makes sense to combine requests
2314 // even if there is a gap of some sectors between them. In this case,
2315 // the gap is filled with zeros (therefore only applicable for yet
2316 // unused space in format like qcow2).
2317 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
2318 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
2321 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
2327 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
2328 qemu_iovec_init(qiov
,
2329 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
2331 // Add the first request to the merged one. If the requests are
2332 // overlapping, drop the last sectors of the first request.
2333 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
2334 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
2336 // We might need to add some zeros between the two requests
2337 if (reqs
[i
].sector
> oldreq_last
) {
2338 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
2339 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
2340 memset(buf
, 0, zero_bytes
);
2341 qemu_iovec_add(qiov
, buf
, zero_bytes
);
2342 mcb
->callbacks
[i
].free_buf
= buf
;
2345 // Add the second request
2346 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2348 reqs
[outidx
].nb_sectors
= qiov
->size
>> 9;
2349 reqs
[outidx
].qiov
= qiov
;
2351 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2354 reqs
[outidx
].sector
= reqs
[i
].sector
;
2355 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2356 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2364 * Submit multiple AIO write requests at once.
2366 * On success, the function returns 0 and all requests in the reqs array have
2367 * been submitted. In error case this function returns -1, and any of the
2368 * requests may or may not be submitted yet. In particular, this means that the
2369 * callback will be called for some of the requests, for others it won't. The
2370 * caller must check the error field of the BlockRequest to wait for the right
2371 * callbacks (if error != 0, no callback will be called).
2373 * The implementation may modify the contents of the reqs array, e.g. to merge
2374 * requests. However, the fields opaque and error are left unmodified as they
2375 * are used to signal failure for a single request to the caller.
2377 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2379 BlockDriverAIOCB
*acb
;
2383 /* don't submit writes if we don't have a medium */
2384 if (bs
->drv
== NULL
) {
2385 for (i
= 0; i
< num_reqs
; i
++) {
2386 reqs
[i
].error
= -ENOMEDIUM
;
2391 if (num_reqs
== 0) {
2395 // Create MultiwriteCB structure
2396 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2397 mcb
->num_requests
= 0;
2398 mcb
->num_callbacks
= num_reqs
;
2400 for (i
= 0; i
< num_reqs
; i
++) {
2401 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2402 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2405 // Check for mergable requests
2406 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2408 trace_bdrv_aio_multiwrite(mcb
, mcb
->num_callbacks
, num_reqs
);
2411 * Run the aio requests. As soon as one request can't be submitted
2412 * successfully, fail all requests that are not yet submitted (we must
2413 * return failure for all requests anyway)
2415 * num_requests cannot be set to the right value immediately: If
2416 * bdrv_aio_writev fails for some request, num_requests would be too high
2417 * and therefore multiwrite_cb() would never recognize the multiwrite
2418 * request as completed. We also cannot use the loop variable i to set it
2419 * when the first request fails because the callback may already have been
2420 * called for previously submitted requests. Thus, num_requests must be
2421 * incremented for each request that is submitted.
2423 * The problem that callbacks may be called early also means that we need
2424 * to take care that num_requests doesn't become 0 before all requests are
2425 * submitted - multiwrite_cb() would consider the multiwrite request
2426 * completed. A dummy request that is "completed" by a manual call to
2427 * multiwrite_cb() takes care of this.
2429 mcb
->num_requests
= 1;
2431 // Run the aio requests
2432 for (i
= 0; i
< num_reqs
; i
++) {
2433 mcb
->num_requests
++;
2434 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2435 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2438 // We can only fail the whole thing if no request has been
2439 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2440 // complete and report the error in the callback.
2442 trace_bdrv_aio_multiwrite_earlyfail(mcb
);
2445 trace_bdrv_aio_multiwrite_latefail(mcb
, i
);
2446 multiwrite_cb(mcb
, -EIO
);
2452 /* Complete the dummy request */
2453 multiwrite_cb(mcb
, 0);
2458 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2459 reqs
[i
].error
= -EIO
;
2465 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2466 BlockDriverCompletionFunc
*cb
, void *opaque
)
2468 BlockDriver
*drv
= bs
->drv
;
2470 trace_bdrv_aio_flush(bs
, opaque
);
2472 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
2473 return bdrv_aio_noop_em(bs
, cb
, opaque
);
2478 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2481 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2483 acb
->pool
->cancel(acb
);
2487 /**************************************************************/
2488 /* async block device emulation */
2490 typedef struct BlockDriverAIOCBSync
{
2491 BlockDriverAIOCB common
;
2494 /* vector translation state */
2498 } BlockDriverAIOCBSync
;
2500 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2502 BlockDriverAIOCBSync
*acb
=
2503 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2504 qemu_bh_delete(acb
->bh
);
2506 qemu_aio_release(acb
);
2509 static AIOPool bdrv_em_aio_pool
= {
2510 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2511 .cancel
= bdrv_aio_cancel_em
,
2514 static void bdrv_aio_bh_cb(void *opaque
)
2516 BlockDriverAIOCBSync
*acb
= opaque
;
2519 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2520 qemu_vfree(acb
->bounce
);
2521 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2522 qemu_bh_delete(acb
->bh
);
2524 qemu_aio_release(acb
);
2527 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2531 BlockDriverCompletionFunc
*cb
,
2536 BlockDriverAIOCBSync
*acb
;
2538 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2539 acb
->is_write
= is_write
;
2541 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2544 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2547 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2548 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2550 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2553 qemu_bh_schedule(acb
->bh
);
2555 return &acb
->common
;
2558 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2559 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2560 BlockDriverCompletionFunc
*cb
, void *opaque
)
2562 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2565 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2566 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2567 BlockDriverCompletionFunc
*cb
, void *opaque
)
2569 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2572 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2573 BlockDriverCompletionFunc
*cb
, void *opaque
)
2575 BlockDriverAIOCBSync
*acb
;
2577 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2578 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2584 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2587 qemu_bh_schedule(acb
->bh
);
2588 return &acb
->common
;
2591 static BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
2592 BlockDriverCompletionFunc
*cb
, void *opaque
)
2594 BlockDriverAIOCBSync
*acb
;
2596 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2597 acb
->is_write
= 1; /* don't bounce in the completion handler */
2603 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2606 qemu_bh_schedule(acb
->bh
);
2607 return &acb
->common
;
2610 /**************************************************************/
2611 /* sync block device emulation */
2613 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2615 *(int *)opaque
= ret
;
2618 #define NOT_DONE 0x7fffffff
2620 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2621 uint8_t *buf
, int nb_sectors
)
2624 BlockDriverAIOCB
*acb
;
2628 async_context_push();
2630 async_ret
= NOT_DONE
;
2631 iov
.iov_base
= (void *)buf
;
2632 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2633 qemu_iovec_init_external(&qiov
, &iov
, 1);
2634 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2635 bdrv_rw_em_cb
, &async_ret
);
2641 while (async_ret
== NOT_DONE
) {
2647 async_context_pop();
2651 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2652 const uint8_t *buf
, int nb_sectors
)
2655 BlockDriverAIOCB
*acb
;
2659 async_context_push();
2661 async_ret
= NOT_DONE
;
2662 iov
.iov_base
= (void *)buf
;
2663 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2664 qemu_iovec_init_external(&qiov
, &iov
, 1);
2665 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2666 bdrv_rw_em_cb
, &async_ret
);
2671 while (async_ret
== NOT_DONE
) {
2676 async_context_pop();
2680 void bdrv_init(void)
2682 module_call_init(MODULE_INIT_BLOCK
);
2685 void bdrv_init_with_whitelist(void)
2687 use_bdrv_whitelist
= 1;
2691 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2692 BlockDriverCompletionFunc
*cb
, void *opaque
)
2694 BlockDriverAIOCB
*acb
;
2696 if (pool
->free_aiocb
) {
2697 acb
= pool
->free_aiocb
;
2698 pool
->free_aiocb
= acb
->next
;
2700 acb
= qemu_mallocz(pool
->aiocb_size
);
2705 acb
->opaque
= opaque
;
2709 void qemu_aio_release(void *p
)
2711 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2712 AIOPool
*pool
= acb
->pool
;
2713 acb
->next
= pool
->free_aiocb
;
2714 pool
->free_aiocb
= acb
;
2717 /**************************************************************/
2718 /* removable device support */
2721 * Return TRUE if the media is present
2723 int bdrv_is_inserted(BlockDriverState
*bs
)
2725 BlockDriver
*drv
= bs
->drv
;
2729 if (!drv
->bdrv_is_inserted
)
2730 return !bs
->tray_open
;
2731 ret
= drv
->bdrv_is_inserted(bs
);
2736 * Return TRUE if the media changed since the last call to this
2737 * function. It is currently only used for floppy disks
2739 int bdrv_media_changed(BlockDriverState
*bs
)
2741 BlockDriver
*drv
= bs
->drv
;
2744 if (!drv
|| !drv
->bdrv_media_changed
)
2747 ret
= drv
->bdrv_media_changed(bs
);
2748 if (ret
== -ENOTSUP
)
2749 ret
= bs
->media_changed
;
2750 bs
->media_changed
= 0;
2755 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2757 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2759 BlockDriver
*drv
= bs
->drv
;
2761 if (eject_flag
&& bs
->locked
) {
2765 if (drv
&& drv
->bdrv_eject
) {
2766 drv
->bdrv_eject(bs
, eject_flag
);
2768 bs
->tray_open
= eject_flag
;
2772 int bdrv_is_locked(BlockDriverState
*bs
)
2778 * Lock or unlock the media (if it is locked, the user won't be able
2779 * to eject it manually).
2781 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2783 BlockDriver
*drv
= bs
->drv
;
2785 trace_bdrv_set_locked(bs
, locked
);
2787 bs
->locked
= locked
;
2788 if (drv
&& drv
->bdrv_set_locked
) {
2789 drv
->bdrv_set_locked(bs
, locked
);
2793 /* needed for generic scsi interface */
2795 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2797 BlockDriver
*drv
= bs
->drv
;
2799 if (drv
&& drv
->bdrv_ioctl
)
2800 return drv
->bdrv_ioctl(bs
, req
, buf
);
2804 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2805 unsigned long int req
, void *buf
,
2806 BlockDriverCompletionFunc
*cb
, void *opaque
)
2808 BlockDriver
*drv
= bs
->drv
;
2810 if (drv
&& drv
->bdrv_aio_ioctl
)
2811 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2817 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2819 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2822 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2824 int64_t bitmap_size
;
2826 bs
->dirty_count
= 0;
2828 if (!bs
->dirty_bitmap
) {
2829 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2830 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2831 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2833 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2836 if (bs
->dirty_bitmap
) {
2837 qemu_free(bs
->dirty_bitmap
);
2838 bs
->dirty_bitmap
= NULL
;
2843 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2845 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2847 if (bs
->dirty_bitmap
&&
2848 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2849 return !!(bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2850 (1UL << (chunk
% (sizeof(unsigned long) * 8))));
2856 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2859 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2862 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2864 return bs
->dirty_count
;
2867 void bdrv_set_in_use(BlockDriverState
*bs
, int in_use
)
2869 assert(bs
->in_use
!= in_use
);
2870 bs
->in_use
= in_use
;
2873 int bdrv_in_use(BlockDriverState
*bs
)
2878 int bdrv_img_create(const char *filename
, const char *fmt
,
2879 const char *base_filename
, const char *base_fmt
,
2880 char *options
, uint64_t img_size
, int flags
)
2882 QEMUOptionParameter
*param
= NULL
, *create_options
= NULL
;
2883 QEMUOptionParameter
*backing_fmt
, *backing_file
, *size
;
2884 BlockDriverState
*bs
= NULL
;
2885 BlockDriver
*drv
, *proto_drv
;
2886 BlockDriver
*backing_drv
= NULL
;
2889 /* Find driver and parse its options */
2890 drv
= bdrv_find_format(fmt
);
2892 error_report("Unknown file format '%s'", fmt
);
2897 proto_drv
= bdrv_find_protocol(filename
);
2899 error_report("Unknown protocol '%s'", filename
);
2904 create_options
= append_option_parameters(create_options
,
2905 drv
->create_options
);
2906 create_options
= append_option_parameters(create_options
,
2907 proto_drv
->create_options
);
2909 /* Create parameter list with default values */
2910 param
= parse_option_parameters("", create_options
, param
);
2912 set_option_parameter_int(param
, BLOCK_OPT_SIZE
, img_size
);
2914 /* Parse -o options */
2916 param
= parse_option_parameters(options
, create_options
, param
);
2917 if (param
== NULL
) {
2918 error_report("Invalid options for file format '%s'.", fmt
);
2924 if (base_filename
) {
2925 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FILE
,
2927 error_report("Backing file not supported for file format '%s'",
2935 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FMT
, base_fmt
)) {
2936 error_report("Backing file format not supported for file "
2937 "format '%s'", fmt
);
2943 backing_file
= get_option_parameter(param
, BLOCK_OPT_BACKING_FILE
);
2944 if (backing_file
&& backing_file
->value
.s
) {
2945 if (!strcmp(filename
, backing_file
->value
.s
)) {
2946 error_report("Error: Trying to create an image with the "
2947 "same filename as the backing file");
2953 backing_fmt
= get_option_parameter(param
, BLOCK_OPT_BACKING_FMT
);
2954 if (backing_fmt
&& backing_fmt
->value
.s
) {
2955 backing_drv
= bdrv_find_format(backing_fmt
->value
.s
);
2957 error_report("Unknown backing file format '%s'",
2958 backing_fmt
->value
.s
);
2964 // The size for the image must always be specified, with one exception:
2965 // If we are using a backing file, we can obtain the size from there
2966 size
= get_option_parameter(param
, BLOCK_OPT_SIZE
);
2967 if (size
&& size
->value
.n
== -1) {
2968 if (backing_file
&& backing_file
->value
.s
) {
2974 ret
= bdrv_open(bs
, backing_file
->value
.s
, flags
, backing_drv
);
2976 error_report("Could not open '%s'", backing_file
->value
.s
);
2979 bdrv_get_geometry(bs
, &size
);
2982 snprintf(buf
, sizeof(buf
), "%" PRId64
, size
);
2983 set_option_parameter(param
, BLOCK_OPT_SIZE
, buf
);
2985 error_report("Image creation needs a size parameter");
2991 printf("Formatting '%s', fmt=%s ", filename
, fmt
);
2992 print_option_parameters(param
);
2995 ret
= bdrv_create(drv
, filename
, param
);
2998 if (ret
== -ENOTSUP
) {
2999 error_report("Formatting or formatting option not supported for "
3000 "file format '%s'", fmt
);
3001 } else if (ret
== -EFBIG
) {
3002 error_report("The image size is too large for file format '%s'",
3005 error_report("%s: error while creating %s: %s", filename
, fmt
,
3011 free_option_parameters(create_options
);
3012 free_option_parameters(param
);