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) {
1113 int coroutine_fn
bdrv_co_readv(BlockDriverState
*bs
, int64_t sector_num
,
1114 int nb_sectors
, QEMUIOVector
*qiov
)
1116 BlockDriver
*drv
= bs
->drv
;
1118 trace_bdrv_co_readv(bs
, sector_num
, nb_sectors
);
1123 if (bdrv_check_request(bs
, sector_num
, nb_sectors
)) {
1127 return drv
->bdrv_co_readv(bs
, sector_num
, nb_sectors
, qiov
);
1130 int coroutine_fn
bdrv_co_writev(BlockDriverState
*bs
, int64_t sector_num
,
1131 int nb_sectors
, QEMUIOVector
*qiov
)
1133 BlockDriver
*drv
= bs
->drv
;
1135 trace_bdrv_co_writev(bs
, sector_num
, nb_sectors
);
1140 if (bs
->read_only
) {
1143 if (bdrv_check_request(bs
, sector_num
, nb_sectors
)) {
1147 if (bs
->dirty_bitmap
) {
1148 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1151 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
1152 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
1155 return drv
->bdrv_co_writev(bs
, sector_num
, nb_sectors
, qiov
);
1159 * Truncate file to 'offset' bytes (needed only for file protocols)
1161 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
1163 BlockDriver
*drv
= bs
->drv
;
1167 if (!drv
->bdrv_truncate
)
1171 if (bdrv_in_use(bs
))
1173 ret
= drv
->bdrv_truncate(bs
, offset
);
1175 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1176 if (bs
->change_cb
) {
1177 bs
->change_cb(bs
->change_opaque
, CHANGE_SIZE
);
1184 * Length of a allocated file in bytes. Sparse files are counted by actual
1185 * allocated space. Return < 0 if error or unknown.
1187 int64_t bdrv_get_allocated_file_size(BlockDriverState
*bs
)
1189 BlockDriver
*drv
= bs
->drv
;
1193 if (drv
->bdrv_get_allocated_file_size
) {
1194 return drv
->bdrv_get_allocated_file_size(bs
);
1197 return bdrv_get_allocated_file_size(bs
->file
);
1203 * Length of a file in bytes. Return < 0 if error or unknown.
1205 int64_t bdrv_getlength(BlockDriverState
*bs
)
1207 BlockDriver
*drv
= bs
->drv
;
1211 if (bs
->growable
|| bs
->removable
) {
1212 if (drv
->bdrv_getlength
) {
1213 return drv
->bdrv_getlength(bs
);
1216 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1219 /* return 0 as number of sectors if no device present or error */
1220 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1223 length
= bdrv_getlength(bs
);
1227 length
= length
>> BDRV_SECTOR_BITS
;
1228 *nb_sectors_ptr
= length
;
1232 uint8_t boot_ind
; /* 0x80 - active */
1233 uint8_t head
; /* starting head */
1234 uint8_t sector
; /* starting sector */
1235 uint8_t cyl
; /* starting cylinder */
1236 uint8_t sys_ind
; /* What partition type */
1237 uint8_t end_head
; /* end head */
1238 uint8_t end_sector
; /* end sector */
1239 uint8_t end_cyl
; /* end cylinder */
1240 uint32_t start_sect
; /* starting sector counting from 0 */
1241 uint32_t nr_sects
; /* nr of sectors in partition */
1242 } __attribute__((packed
));
1244 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1245 static int guess_disk_lchs(BlockDriverState
*bs
,
1246 int *pcylinders
, int *pheads
, int *psectors
)
1248 uint8_t buf
[BDRV_SECTOR_SIZE
];
1249 int ret
, i
, heads
, sectors
, cylinders
;
1250 struct partition
*p
;
1252 uint64_t nb_sectors
;
1254 bdrv_get_geometry(bs
, &nb_sectors
);
1256 ret
= bdrv_read(bs
, 0, buf
, 1);
1259 /* test msdos magic */
1260 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1262 for(i
= 0; i
< 4; i
++) {
1263 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1264 nr_sects
= le32_to_cpu(p
->nr_sects
);
1265 if (nr_sects
&& p
->end_head
) {
1266 /* We make the assumption that the partition terminates on
1267 a cylinder boundary */
1268 heads
= p
->end_head
+ 1;
1269 sectors
= p
->end_sector
& 63;
1272 cylinders
= nb_sectors
/ (heads
* sectors
);
1273 if (cylinders
< 1 || cylinders
> 16383)
1276 *psectors
= sectors
;
1277 *pcylinders
= cylinders
;
1279 printf("guessed geometry: LCHS=%d %d %d\n",
1280 cylinders
, heads
, sectors
);
1288 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1290 int translation
, lba_detected
= 0;
1291 int cylinders
, heads
, secs
;
1292 uint64_t nb_sectors
;
1294 /* if a geometry hint is available, use it */
1295 bdrv_get_geometry(bs
, &nb_sectors
);
1296 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1297 translation
= bdrv_get_translation_hint(bs
);
1298 if (cylinders
!= 0) {
1303 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1305 /* if heads > 16, it means that a BIOS LBA
1306 translation was active, so the default
1307 hardware geometry is OK */
1309 goto default_geometry
;
1314 /* disable any translation to be in sync with
1315 the logical geometry */
1316 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1317 bdrv_set_translation_hint(bs
,
1318 BIOS_ATA_TRANSLATION_NONE
);
1323 /* if no geometry, use a standard physical disk geometry */
1324 cylinders
= nb_sectors
/ (16 * 63);
1326 if (cylinders
> 16383)
1328 else if (cylinders
< 2)
1333 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1334 if ((*pcyls
* *pheads
) <= 131072) {
1335 bdrv_set_translation_hint(bs
,
1336 BIOS_ATA_TRANSLATION_LARGE
);
1338 bdrv_set_translation_hint(bs
,
1339 BIOS_ATA_TRANSLATION_LBA
);
1343 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1347 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1348 int cyls
, int heads
, int secs
)
1355 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1357 bs
->translation
= translation
;
1360 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1361 int *pcyls
, int *pheads
, int *psecs
)
1364 *pheads
= bs
->heads
;
1368 /* Recognize floppy formats */
1369 typedef struct FDFormat
{
1376 static const FDFormat fd_formats
[] = {
1377 /* First entry is default format */
1378 /* 1.44 MB 3"1/2 floppy disks */
1379 { FDRIVE_DRV_144
, 18, 80, 1, },
1380 { FDRIVE_DRV_144
, 20, 80, 1, },
1381 { FDRIVE_DRV_144
, 21, 80, 1, },
1382 { FDRIVE_DRV_144
, 21, 82, 1, },
1383 { FDRIVE_DRV_144
, 21, 83, 1, },
1384 { FDRIVE_DRV_144
, 22, 80, 1, },
1385 { FDRIVE_DRV_144
, 23, 80, 1, },
1386 { FDRIVE_DRV_144
, 24, 80, 1, },
1387 /* 2.88 MB 3"1/2 floppy disks */
1388 { FDRIVE_DRV_288
, 36, 80, 1, },
1389 { FDRIVE_DRV_288
, 39, 80, 1, },
1390 { FDRIVE_DRV_288
, 40, 80, 1, },
1391 { FDRIVE_DRV_288
, 44, 80, 1, },
1392 { FDRIVE_DRV_288
, 48, 80, 1, },
1393 /* 720 kB 3"1/2 floppy disks */
1394 { FDRIVE_DRV_144
, 9, 80, 1, },
1395 { FDRIVE_DRV_144
, 10, 80, 1, },
1396 { FDRIVE_DRV_144
, 10, 82, 1, },
1397 { FDRIVE_DRV_144
, 10, 83, 1, },
1398 { FDRIVE_DRV_144
, 13, 80, 1, },
1399 { FDRIVE_DRV_144
, 14, 80, 1, },
1400 /* 1.2 MB 5"1/4 floppy disks */
1401 { FDRIVE_DRV_120
, 15, 80, 1, },
1402 { FDRIVE_DRV_120
, 18, 80, 1, },
1403 { FDRIVE_DRV_120
, 18, 82, 1, },
1404 { FDRIVE_DRV_120
, 18, 83, 1, },
1405 { FDRIVE_DRV_120
, 20, 80, 1, },
1406 /* 720 kB 5"1/4 floppy disks */
1407 { FDRIVE_DRV_120
, 9, 80, 1, },
1408 { FDRIVE_DRV_120
, 11, 80, 1, },
1409 /* 360 kB 5"1/4 floppy disks */
1410 { FDRIVE_DRV_120
, 9, 40, 1, },
1411 { FDRIVE_DRV_120
, 9, 40, 0, },
1412 { FDRIVE_DRV_120
, 10, 41, 1, },
1413 { FDRIVE_DRV_120
, 10, 42, 1, },
1414 /* 320 kB 5"1/4 floppy disks */
1415 { FDRIVE_DRV_120
, 8, 40, 1, },
1416 { FDRIVE_DRV_120
, 8, 40, 0, },
1417 /* 360 kB must match 5"1/4 better than 3"1/2... */
1418 { FDRIVE_DRV_144
, 9, 80, 0, },
1420 { FDRIVE_DRV_NONE
, -1, -1, 0, },
1423 void bdrv_get_floppy_geometry_hint(BlockDriverState
*bs
, int *nb_heads
,
1424 int *max_track
, int *last_sect
,
1425 FDriveType drive_in
, FDriveType
*drive
)
1427 const FDFormat
*parse
;
1428 uint64_t nb_sectors
, size
;
1429 int i
, first_match
, match
;
1431 bdrv_get_geometry_hint(bs
, nb_heads
, max_track
, last_sect
);
1432 if (*nb_heads
!= 0 && *max_track
!= 0 && *last_sect
!= 0) {
1433 /* User defined disk */
1435 bdrv_get_geometry(bs
, &nb_sectors
);
1438 for (i
= 0; ; i
++) {
1439 parse
= &fd_formats
[i
];
1440 if (parse
->drive
== FDRIVE_DRV_NONE
) {
1443 if (drive_in
== parse
->drive
||
1444 drive_in
== FDRIVE_DRV_NONE
) {
1445 size
= (parse
->max_head
+ 1) * parse
->max_track
*
1447 if (nb_sectors
== size
) {
1451 if (first_match
== -1) {
1457 if (first_match
== -1) {
1460 match
= first_match
;
1462 parse
= &fd_formats
[match
];
1464 *nb_heads
= parse
->max_head
+ 1;
1465 *max_track
= parse
->max_track
;
1466 *last_sect
= parse
->last_sect
;
1467 *drive
= parse
->drive
;
1471 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1473 return bs
->translation
;
1476 void bdrv_set_on_error(BlockDriverState
*bs
, BlockErrorAction on_read_error
,
1477 BlockErrorAction on_write_error
)
1479 bs
->on_read_error
= on_read_error
;
1480 bs
->on_write_error
= on_write_error
;
1483 BlockErrorAction
bdrv_get_on_error(BlockDriverState
*bs
, int is_read
)
1485 return is_read
? bs
->on_read_error
: bs
->on_write_error
;
1488 void bdrv_set_removable(BlockDriverState
*bs
, int removable
)
1490 bs
->removable
= removable
;
1491 if (removable
&& bs
== bs_snapshots
) {
1492 bs_snapshots
= NULL
;
1496 int bdrv_is_removable(BlockDriverState
*bs
)
1498 return bs
->removable
;
1501 int bdrv_is_read_only(BlockDriverState
*bs
)
1503 return bs
->read_only
;
1506 int bdrv_is_sg(BlockDriverState
*bs
)
1511 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1513 return bs
->enable_write_cache
;
1516 /* XXX: no longer used */
1517 void bdrv_set_change_cb(BlockDriverState
*bs
,
1518 void (*change_cb
)(void *opaque
, int reason
),
1521 bs
->change_cb
= change_cb
;
1522 bs
->change_opaque
= opaque
;
1525 int bdrv_is_encrypted(BlockDriverState
*bs
)
1527 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1529 return bs
->encrypted
;
1532 int bdrv_key_required(BlockDriverState
*bs
)
1534 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1536 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1538 return (bs
->encrypted
&& !bs
->valid_key
);
1541 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1544 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1545 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1551 if (!bs
->encrypted
) {
1553 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1556 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1559 } else if (!bs
->valid_key
) {
1561 /* call the change callback now, we skipped it on open */
1562 bs
->media_changed
= 1;
1564 bs
->change_cb(bs
->change_opaque
, CHANGE_MEDIA
);
1569 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1574 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1578 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1583 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1584 it(opaque
, drv
->format_name
);
1588 BlockDriverState
*bdrv_find(const char *name
)
1590 BlockDriverState
*bs
;
1592 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1593 if (!strcmp(name
, bs
->device_name
)) {
1600 BlockDriverState
*bdrv_next(BlockDriverState
*bs
)
1603 return QTAILQ_FIRST(&bdrv_states
);
1605 return QTAILQ_NEXT(bs
, list
);
1608 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1610 BlockDriverState
*bs
;
1612 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1617 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1619 return bs
->device_name
;
1622 int bdrv_flush(BlockDriverState
*bs
)
1624 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
1628 if (bs
->drv
&& bs
->drv
->bdrv_flush
) {
1629 return bs
->drv
->bdrv_flush(bs
);
1633 * Some block drivers always operate in either writethrough or unsafe mode
1634 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1635 * the server works (because the behaviour is hardcoded or depends on
1636 * server-side configuration), so we can't ensure that everything is safe
1637 * on disk. Returning an error doesn't work because that would break guests
1638 * even if the server operates in writethrough mode.
1640 * Let's hope the user knows what he's doing.
1645 void bdrv_flush_all(void)
1647 BlockDriverState
*bs
;
1649 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1650 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1651 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1657 int bdrv_has_zero_init(BlockDriverState
*bs
)
1661 if (bs
->drv
->bdrv_has_zero_init
) {
1662 return bs
->drv
->bdrv_has_zero_init(bs
);
1668 int bdrv_discard(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
)
1673 if (!bs
->drv
->bdrv_discard
) {
1676 return bs
->drv
->bdrv_discard(bs
, sector_num
, nb_sectors
);
1680 * Returns true iff the specified sector is present in the disk image. Drivers
1681 * not implementing the functionality are assumed to not support backing files,
1682 * hence all their sectors are reported as allocated.
1684 * 'pnum' is set to the number of sectors (including and immediately following
1685 * the specified sector) that are known to be in the same
1686 * allocated/unallocated state.
1688 * 'nb_sectors' is the max value 'pnum' should be set to.
1690 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1694 if (!bs
->drv
->bdrv_is_allocated
) {
1695 if (sector_num
>= bs
->total_sectors
) {
1699 n
= bs
->total_sectors
- sector_num
;
1700 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1703 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1706 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1707 BlockMonEventAction action
, int is_read
)
1710 const char *action_str
;
1713 case BDRV_ACTION_REPORT
:
1714 action_str
= "report";
1716 case BDRV_ACTION_IGNORE
:
1717 action_str
= "ignore";
1719 case BDRV_ACTION_STOP
:
1720 action_str
= "stop";
1726 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1729 is_read
? "read" : "write");
1730 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1732 qobject_decref(data
);
1735 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1738 Monitor
*mon
= opaque
;
1740 bs_dict
= qobject_to_qdict(obj
);
1742 monitor_printf(mon
, "%s: removable=%d",
1743 qdict_get_str(bs_dict
, "device"),
1744 qdict_get_bool(bs_dict
, "removable"));
1746 if (qdict_get_bool(bs_dict
, "removable")) {
1747 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1750 if (qdict_haskey(bs_dict
, "inserted")) {
1751 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1753 monitor_printf(mon
, " file=");
1754 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1755 if (qdict_haskey(qdict
, "backing_file")) {
1756 monitor_printf(mon
, " backing_file=");
1757 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1759 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1760 qdict_get_bool(qdict
, "ro"),
1761 qdict_get_str(qdict
, "drv"),
1762 qdict_get_bool(qdict
, "encrypted"));
1764 monitor_printf(mon
, " [not inserted]");
1767 monitor_printf(mon
, "\n");
1770 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1772 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1775 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1778 BlockDriverState
*bs
;
1780 bs_list
= qlist_new();
1782 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1785 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
1786 "'removable': %i, 'locked': %i }",
1787 bs
->device_name
, bs
->removable
,
1792 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1794 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1795 "'encrypted': %i }",
1796 bs
->filename
, bs
->read_only
,
1797 bs
->drv
->format_name
,
1798 bdrv_is_encrypted(bs
));
1799 if (bs
->backing_file
[0] != '\0') {
1800 QDict
*qdict
= qobject_to_qdict(obj
);
1801 qdict_put(qdict
, "backing_file",
1802 qstring_from_str(bs
->backing_file
));
1805 qdict_put_obj(bs_dict
, "inserted", obj
);
1807 qlist_append_obj(bs_list
, bs_obj
);
1810 *ret_data
= QOBJECT(bs_list
);
1813 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1816 Monitor
*mon
= opaque
;
1818 qdict
= qobject_to_qdict(data
);
1819 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1821 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1822 monitor_printf(mon
, " rd_bytes=%" PRId64
1823 " wr_bytes=%" PRId64
1824 " rd_operations=%" PRId64
1825 " wr_operations=%" PRId64
1827 qdict_get_int(qdict
, "rd_bytes"),
1828 qdict_get_int(qdict
, "wr_bytes"),
1829 qdict_get_int(qdict
, "rd_operations"),
1830 qdict_get_int(qdict
, "wr_operations"));
1833 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1835 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1838 static QObject
* bdrv_info_stats_bs(BlockDriverState
*bs
)
1843 res
= qobject_from_jsonf("{ 'stats': {"
1844 "'rd_bytes': %" PRId64
","
1845 "'wr_bytes': %" PRId64
","
1846 "'rd_operations': %" PRId64
","
1847 "'wr_operations': %" PRId64
","
1848 "'wr_highest_offset': %" PRId64
1850 bs
->rd_bytes
, bs
->wr_bytes
,
1851 bs
->rd_ops
, bs
->wr_ops
,
1852 bs
->wr_highest_sector
*
1853 (uint64_t)BDRV_SECTOR_SIZE
);
1854 dict
= qobject_to_qdict(res
);
1856 if (*bs
->device_name
) {
1857 qdict_put(dict
, "device", qstring_from_str(bs
->device_name
));
1861 QObject
*parent
= bdrv_info_stats_bs(bs
->file
);
1862 qdict_put_obj(dict
, "parent", parent
);
1868 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1872 BlockDriverState
*bs
;
1874 devices
= qlist_new();
1876 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1877 obj
= bdrv_info_stats_bs(bs
);
1878 qlist_append_obj(devices
, obj
);
1881 *ret_data
= QOBJECT(devices
);
1884 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1886 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1887 return bs
->backing_file
;
1888 else if (bs
->encrypted
)
1889 return bs
->filename
;
1894 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1895 char *filename
, int filename_size
)
1897 if (!bs
->backing_file
) {
1898 pstrcpy(filename
, filename_size
, "");
1900 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1904 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1905 const uint8_t *buf
, int nb_sectors
)
1907 BlockDriver
*drv
= bs
->drv
;
1910 if (!drv
->bdrv_write_compressed
)
1912 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1915 if (bs
->dirty_bitmap
) {
1916 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1919 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1922 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1924 BlockDriver
*drv
= bs
->drv
;
1927 if (!drv
->bdrv_get_info
)
1929 memset(bdi
, 0, sizeof(*bdi
));
1930 return drv
->bdrv_get_info(bs
, bdi
);
1933 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1934 int64_t pos
, int size
)
1936 BlockDriver
*drv
= bs
->drv
;
1939 if (drv
->bdrv_save_vmstate
)
1940 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1942 return bdrv_save_vmstate(bs
->file
, buf
, pos
, size
);
1946 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1947 int64_t pos
, int size
)
1949 BlockDriver
*drv
= bs
->drv
;
1952 if (drv
->bdrv_load_vmstate
)
1953 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1955 return bdrv_load_vmstate(bs
->file
, buf
, pos
, size
);
1959 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1961 BlockDriver
*drv
= bs
->drv
;
1963 if (!drv
|| !drv
->bdrv_debug_event
) {
1967 return drv
->bdrv_debug_event(bs
, event
);
1971 /**************************************************************/
1972 /* handling of snapshots */
1974 int bdrv_can_snapshot(BlockDriverState
*bs
)
1976 BlockDriver
*drv
= bs
->drv
;
1977 if (!drv
|| bdrv_is_removable(bs
) || bdrv_is_read_only(bs
)) {
1981 if (!drv
->bdrv_snapshot_create
) {
1982 if (bs
->file
!= NULL
) {
1983 return bdrv_can_snapshot(bs
->file
);
1991 int bdrv_is_snapshot(BlockDriverState
*bs
)
1993 return !!(bs
->open_flags
& BDRV_O_SNAPSHOT
);
1996 BlockDriverState
*bdrv_snapshots(void)
1998 BlockDriverState
*bs
;
2001 return bs_snapshots
;
2005 while ((bs
= bdrv_next(bs
))) {
2006 if (bdrv_can_snapshot(bs
)) {
2014 int bdrv_snapshot_create(BlockDriverState
*bs
,
2015 QEMUSnapshotInfo
*sn_info
)
2017 BlockDriver
*drv
= bs
->drv
;
2020 if (drv
->bdrv_snapshot_create
)
2021 return drv
->bdrv_snapshot_create(bs
, sn_info
);
2023 return bdrv_snapshot_create(bs
->file
, sn_info
);
2027 int bdrv_snapshot_goto(BlockDriverState
*bs
,
2028 const char *snapshot_id
)
2030 BlockDriver
*drv
= bs
->drv
;
2035 if (drv
->bdrv_snapshot_goto
)
2036 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
2039 drv
->bdrv_close(bs
);
2040 ret
= bdrv_snapshot_goto(bs
->file
, snapshot_id
);
2041 open_ret
= drv
->bdrv_open(bs
, bs
->open_flags
);
2043 bdrv_delete(bs
->file
);
2053 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
2055 BlockDriver
*drv
= bs
->drv
;
2058 if (drv
->bdrv_snapshot_delete
)
2059 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
2061 return bdrv_snapshot_delete(bs
->file
, snapshot_id
);
2065 int bdrv_snapshot_list(BlockDriverState
*bs
,
2066 QEMUSnapshotInfo
**psn_info
)
2068 BlockDriver
*drv
= bs
->drv
;
2071 if (drv
->bdrv_snapshot_list
)
2072 return drv
->bdrv_snapshot_list(bs
, psn_info
);
2074 return bdrv_snapshot_list(bs
->file
, psn_info
);
2078 int bdrv_snapshot_load_tmp(BlockDriverState
*bs
,
2079 const char *snapshot_name
)
2081 BlockDriver
*drv
= bs
->drv
;
2085 if (!bs
->read_only
) {
2088 if (drv
->bdrv_snapshot_load_tmp
) {
2089 return drv
->bdrv_snapshot_load_tmp(bs
, snapshot_name
);
2094 #define NB_SUFFIXES 4
2096 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
2098 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
2103 snprintf(buf
, buf_size
, "%" PRId64
, size
);
2106 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
2107 if (size
< (10 * base
)) {
2108 snprintf(buf
, buf_size
, "%0.1f%c",
2109 (double)size
/ base
,
2112 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
2113 snprintf(buf
, buf_size
, "%" PRId64
"%c",
2114 ((size
+ (base
>> 1)) / base
),
2124 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
2126 char buf1
[128], date_buf
[128], clock_buf
[128];
2136 snprintf(buf
, buf_size
,
2137 "%-10s%-20s%7s%20s%15s",
2138 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2142 ptm
= localtime(&ti
);
2143 strftime(date_buf
, sizeof(date_buf
),
2144 "%Y-%m-%d %H:%M:%S", ptm
);
2146 localtime_r(&ti
, &tm
);
2147 strftime(date_buf
, sizeof(date_buf
),
2148 "%Y-%m-%d %H:%M:%S", &tm
);
2150 secs
= sn
->vm_clock_nsec
/ 1000000000;
2151 snprintf(clock_buf
, sizeof(clock_buf
),
2152 "%02d:%02d:%02d.%03d",
2154 (int)((secs
/ 60) % 60),
2156 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
2157 snprintf(buf
, buf_size
,
2158 "%-10s%-20s%7s%20s%15s",
2159 sn
->id_str
, sn
->name
,
2160 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
2168 /**************************************************************/
2171 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
2172 QEMUIOVector
*qiov
, int nb_sectors
,
2173 BlockDriverCompletionFunc
*cb
, void *opaque
)
2175 BlockDriver
*drv
= bs
->drv
;
2176 BlockDriverAIOCB
*ret
;
2178 trace_bdrv_aio_readv(bs
, sector_num
, nb_sectors
, opaque
);
2182 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2185 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
2189 /* Update stats even though technically transfer has not happened. */
2190 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
2197 typedef struct BlockCompleteData
{
2198 BlockDriverCompletionFunc
*cb
;
2200 BlockDriverState
*bs
;
2203 } BlockCompleteData
;
2205 static void block_complete_cb(void *opaque
, int ret
)
2207 BlockCompleteData
*b
= opaque
;
2209 if (b
->bs
->dirty_bitmap
) {
2210 set_dirty_bitmap(b
->bs
, b
->sector_num
, b
->nb_sectors
, 1);
2212 b
->cb(b
->opaque
, ret
);
2216 static BlockCompleteData
*blk_dirty_cb_alloc(BlockDriverState
*bs
,
2219 BlockDriverCompletionFunc
*cb
,
2222 BlockCompleteData
*blkdata
= qemu_mallocz(sizeof(BlockCompleteData
));
2226 blkdata
->opaque
= opaque
;
2227 blkdata
->sector_num
= sector_num
;
2228 blkdata
->nb_sectors
= nb_sectors
;
2233 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
2234 QEMUIOVector
*qiov
, int nb_sectors
,
2235 BlockDriverCompletionFunc
*cb
, void *opaque
)
2237 BlockDriver
*drv
= bs
->drv
;
2238 BlockDriverAIOCB
*ret
;
2239 BlockCompleteData
*blk_cb_data
;
2241 trace_bdrv_aio_writev(bs
, sector_num
, nb_sectors
, opaque
);
2247 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2250 if (bs
->dirty_bitmap
) {
2251 blk_cb_data
= blk_dirty_cb_alloc(bs
, sector_num
, nb_sectors
, cb
,
2253 cb
= &block_complete_cb
;
2254 opaque
= blk_cb_data
;
2257 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
2261 /* Update stats even though technically transfer has not happened. */
2262 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
2264 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
2265 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
2273 typedef struct MultiwriteCB
{
2278 BlockDriverCompletionFunc
*cb
;
2280 QEMUIOVector
*free_qiov
;
2285 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
2289 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2290 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
2291 if (mcb
->callbacks
[i
].free_qiov
) {
2292 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
2294 qemu_free(mcb
->callbacks
[i
].free_qiov
);
2295 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
2299 static void multiwrite_cb(void *opaque
, int ret
)
2301 MultiwriteCB
*mcb
= opaque
;
2303 trace_multiwrite_cb(mcb
, ret
);
2305 if (ret
< 0 && !mcb
->error
) {
2309 mcb
->num_requests
--;
2310 if (mcb
->num_requests
== 0) {
2311 multiwrite_user_cb(mcb
);
2316 static int multiwrite_req_compare(const void *a
, const void *b
)
2318 const BlockRequest
*req1
= a
, *req2
= b
;
2321 * Note that we can't simply subtract req2->sector from req1->sector
2322 * here as that could overflow the return value.
2324 if (req1
->sector
> req2
->sector
) {
2326 } else if (req1
->sector
< req2
->sector
) {
2334 * Takes a bunch of requests and tries to merge them. Returns the number of
2335 * requests that remain after merging.
2337 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
2338 int num_reqs
, MultiwriteCB
*mcb
)
2342 // Sort requests by start sector
2343 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
2345 // Check if adjacent requests touch the same clusters. If so, combine them,
2346 // filling up gaps with zero sectors.
2348 for (i
= 1; i
< num_reqs
; i
++) {
2350 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
2352 // This handles the cases that are valid for all block drivers, namely
2353 // exactly sequential writes and overlapping writes.
2354 if (reqs
[i
].sector
<= oldreq_last
) {
2358 // The block driver may decide that it makes sense to combine requests
2359 // even if there is a gap of some sectors between them. In this case,
2360 // the gap is filled with zeros (therefore only applicable for yet
2361 // unused space in format like qcow2).
2362 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
2363 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
2366 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
2372 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
2373 qemu_iovec_init(qiov
,
2374 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
2376 // Add the first request to the merged one. If the requests are
2377 // overlapping, drop the last sectors of the first request.
2378 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
2379 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
2381 // We might need to add some zeros between the two requests
2382 if (reqs
[i
].sector
> oldreq_last
) {
2383 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
2384 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
2385 memset(buf
, 0, zero_bytes
);
2386 qemu_iovec_add(qiov
, buf
, zero_bytes
);
2387 mcb
->callbacks
[i
].free_buf
= buf
;
2390 // Add the second request
2391 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2393 reqs
[outidx
].nb_sectors
= qiov
->size
>> 9;
2394 reqs
[outidx
].qiov
= qiov
;
2396 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2399 reqs
[outidx
].sector
= reqs
[i
].sector
;
2400 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2401 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2409 * Submit multiple AIO write requests at once.
2411 * On success, the function returns 0 and all requests in the reqs array have
2412 * been submitted. In error case this function returns -1, and any of the
2413 * requests may or may not be submitted yet. In particular, this means that the
2414 * callback will be called for some of the requests, for others it won't. The
2415 * caller must check the error field of the BlockRequest to wait for the right
2416 * callbacks (if error != 0, no callback will be called).
2418 * The implementation may modify the contents of the reqs array, e.g. to merge
2419 * requests. However, the fields opaque and error are left unmodified as they
2420 * are used to signal failure for a single request to the caller.
2422 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2424 BlockDriverAIOCB
*acb
;
2428 /* don't submit writes if we don't have a medium */
2429 if (bs
->drv
== NULL
) {
2430 for (i
= 0; i
< num_reqs
; i
++) {
2431 reqs
[i
].error
= -ENOMEDIUM
;
2436 if (num_reqs
== 0) {
2440 // Create MultiwriteCB structure
2441 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2442 mcb
->num_requests
= 0;
2443 mcb
->num_callbacks
= num_reqs
;
2445 for (i
= 0; i
< num_reqs
; i
++) {
2446 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2447 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2450 // Check for mergable requests
2451 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2453 trace_bdrv_aio_multiwrite(mcb
, mcb
->num_callbacks
, num_reqs
);
2456 * Run the aio requests. As soon as one request can't be submitted
2457 * successfully, fail all requests that are not yet submitted (we must
2458 * return failure for all requests anyway)
2460 * num_requests cannot be set to the right value immediately: If
2461 * bdrv_aio_writev fails for some request, num_requests would be too high
2462 * and therefore multiwrite_cb() would never recognize the multiwrite
2463 * request as completed. We also cannot use the loop variable i to set it
2464 * when the first request fails because the callback may already have been
2465 * called for previously submitted requests. Thus, num_requests must be
2466 * incremented for each request that is submitted.
2468 * The problem that callbacks may be called early also means that we need
2469 * to take care that num_requests doesn't become 0 before all requests are
2470 * submitted - multiwrite_cb() would consider the multiwrite request
2471 * completed. A dummy request that is "completed" by a manual call to
2472 * multiwrite_cb() takes care of this.
2474 mcb
->num_requests
= 1;
2476 // Run the aio requests
2477 for (i
= 0; i
< num_reqs
; i
++) {
2478 mcb
->num_requests
++;
2479 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2480 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2483 // We can only fail the whole thing if no request has been
2484 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2485 // complete and report the error in the callback.
2487 trace_bdrv_aio_multiwrite_earlyfail(mcb
);
2490 trace_bdrv_aio_multiwrite_latefail(mcb
, i
);
2491 multiwrite_cb(mcb
, -EIO
);
2497 /* Complete the dummy request */
2498 multiwrite_cb(mcb
, 0);
2503 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2504 reqs
[i
].error
= -EIO
;
2510 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2511 BlockDriverCompletionFunc
*cb
, void *opaque
)
2513 BlockDriver
*drv
= bs
->drv
;
2515 trace_bdrv_aio_flush(bs
, opaque
);
2517 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
2518 return bdrv_aio_noop_em(bs
, cb
, opaque
);
2523 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2526 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2528 acb
->pool
->cancel(acb
);
2532 /**************************************************************/
2533 /* async block device emulation */
2535 typedef struct BlockDriverAIOCBSync
{
2536 BlockDriverAIOCB common
;
2539 /* vector translation state */
2543 } BlockDriverAIOCBSync
;
2545 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2547 BlockDriverAIOCBSync
*acb
=
2548 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2549 qemu_bh_delete(acb
->bh
);
2551 qemu_aio_release(acb
);
2554 static AIOPool bdrv_em_aio_pool
= {
2555 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2556 .cancel
= bdrv_aio_cancel_em
,
2559 static void bdrv_aio_bh_cb(void *opaque
)
2561 BlockDriverAIOCBSync
*acb
= opaque
;
2564 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2565 qemu_vfree(acb
->bounce
);
2566 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2567 qemu_bh_delete(acb
->bh
);
2569 qemu_aio_release(acb
);
2572 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2576 BlockDriverCompletionFunc
*cb
,
2581 BlockDriverAIOCBSync
*acb
;
2583 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2584 acb
->is_write
= is_write
;
2586 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2589 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2592 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2593 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2595 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2598 qemu_bh_schedule(acb
->bh
);
2600 return &acb
->common
;
2603 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2604 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2605 BlockDriverCompletionFunc
*cb
, void *opaque
)
2607 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2610 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2611 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2612 BlockDriverCompletionFunc
*cb
, void *opaque
)
2614 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2617 static BlockDriverAIOCB
*bdrv_aio_flush_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 hadler */
2629 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2632 qemu_bh_schedule(acb
->bh
);
2633 return &acb
->common
;
2636 static BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
2637 BlockDriverCompletionFunc
*cb
, void *opaque
)
2639 BlockDriverAIOCBSync
*acb
;
2641 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2642 acb
->is_write
= 1; /* don't bounce in the completion handler */
2648 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2651 qemu_bh_schedule(acb
->bh
);
2652 return &acb
->common
;
2655 /**************************************************************/
2656 /* sync block device emulation */
2658 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2660 *(int *)opaque
= ret
;
2663 #define NOT_DONE 0x7fffffff
2665 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2666 uint8_t *buf
, int nb_sectors
)
2669 BlockDriverAIOCB
*acb
;
2673 async_context_push();
2675 async_ret
= NOT_DONE
;
2676 iov
.iov_base
= (void *)buf
;
2677 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2678 qemu_iovec_init_external(&qiov
, &iov
, 1);
2679 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2680 bdrv_rw_em_cb
, &async_ret
);
2686 while (async_ret
== NOT_DONE
) {
2692 async_context_pop();
2696 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2697 const uint8_t *buf
, int nb_sectors
)
2700 BlockDriverAIOCB
*acb
;
2704 async_context_push();
2706 async_ret
= NOT_DONE
;
2707 iov
.iov_base
= (void *)buf
;
2708 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2709 qemu_iovec_init_external(&qiov
, &iov
, 1);
2710 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2711 bdrv_rw_em_cb
, &async_ret
);
2716 while (async_ret
== NOT_DONE
) {
2721 async_context_pop();
2725 void bdrv_init(void)
2727 module_call_init(MODULE_INIT_BLOCK
);
2730 void bdrv_init_with_whitelist(void)
2732 use_bdrv_whitelist
= 1;
2736 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2737 BlockDriverCompletionFunc
*cb
, void *opaque
)
2739 BlockDriverAIOCB
*acb
;
2741 if (pool
->free_aiocb
) {
2742 acb
= pool
->free_aiocb
;
2743 pool
->free_aiocb
= acb
->next
;
2745 acb
= qemu_mallocz(pool
->aiocb_size
);
2750 acb
->opaque
= opaque
;
2754 void qemu_aio_release(void *p
)
2756 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2757 AIOPool
*pool
= acb
->pool
;
2758 acb
->next
= pool
->free_aiocb
;
2759 pool
->free_aiocb
= acb
;
2762 /**************************************************************/
2763 /* removable device support */
2766 * Return TRUE if the media is present
2768 int bdrv_is_inserted(BlockDriverState
*bs
)
2770 BlockDriver
*drv
= bs
->drv
;
2774 if (!drv
->bdrv_is_inserted
)
2775 return !bs
->tray_open
;
2776 ret
= drv
->bdrv_is_inserted(bs
);
2781 * Return TRUE if the media changed since the last call to this
2782 * function. It is currently only used for floppy disks
2784 int bdrv_media_changed(BlockDriverState
*bs
)
2786 BlockDriver
*drv
= bs
->drv
;
2789 if (!drv
|| !drv
->bdrv_media_changed
)
2792 ret
= drv
->bdrv_media_changed(bs
);
2793 if (ret
== -ENOTSUP
)
2794 ret
= bs
->media_changed
;
2795 bs
->media_changed
= 0;
2800 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2802 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2804 BlockDriver
*drv
= bs
->drv
;
2806 if (eject_flag
&& bs
->locked
) {
2810 if (drv
&& drv
->bdrv_eject
) {
2811 drv
->bdrv_eject(bs
, eject_flag
);
2813 bs
->tray_open
= eject_flag
;
2817 int bdrv_is_locked(BlockDriverState
*bs
)
2823 * Lock or unlock the media (if it is locked, the user won't be able
2824 * to eject it manually).
2826 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2828 BlockDriver
*drv
= bs
->drv
;
2830 trace_bdrv_set_locked(bs
, locked
);
2832 bs
->locked
= locked
;
2833 if (drv
&& drv
->bdrv_set_locked
) {
2834 drv
->bdrv_set_locked(bs
, locked
);
2838 /* needed for generic scsi interface */
2840 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2842 BlockDriver
*drv
= bs
->drv
;
2844 if (drv
&& drv
->bdrv_ioctl
)
2845 return drv
->bdrv_ioctl(bs
, req
, buf
);
2849 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2850 unsigned long int req
, void *buf
,
2851 BlockDriverCompletionFunc
*cb
, void *opaque
)
2853 BlockDriver
*drv
= bs
->drv
;
2855 if (drv
&& drv
->bdrv_aio_ioctl
)
2856 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2862 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2864 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2867 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2869 int64_t bitmap_size
;
2871 bs
->dirty_count
= 0;
2873 if (!bs
->dirty_bitmap
) {
2874 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2875 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2876 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2878 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2881 if (bs
->dirty_bitmap
) {
2882 qemu_free(bs
->dirty_bitmap
);
2883 bs
->dirty_bitmap
= NULL
;
2888 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2890 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2892 if (bs
->dirty_bitmap
&&
2893 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2894 return !!(bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2895 (1UL << (chunk
% (sizeof(unsigned long) * 8))));
2901 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2904 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2907 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2909 return bs
->dirty_count
;
2912 void bdrv_set_in_use(BlockDriverState
*bs
, int in_use
)
2914 assert(bs
->in_use
!= in_use
);
2915 bs
->in_use
= in_use
;
2918 int bdrv_in_use(BlockDriverState
*bs
)
2923 int bdrv_img_create(const char *filename
, const char *fmt
,
2924 const char *base_filename
, const char *base_fmt
,
2925 char *options
, uint64_t img_size
, int flags
)
2927 QEMUOptionParameter
*param
= NULL
, *create_options
= NULL
;
2928 QEMUOptionParameter
*backing_fmt
, *backing_file
, *size
;
2929 BlockDriverState
*bs
= NULL
;
2930 BlockDriver
*drv
, *proto_drv
;
2931 BlockDriver
*backing_drv
= NULL
;
2934 /* Find driver and parse its options */
2935 drv
= bdrv_find_format(fmt
);
2937 error_report("Unknown file format '%s'", fmt
);
2942 proto_drv
= bdrv_find_protocol(filename
);
2944 error_report("Unknown protocol '%s'", filename
);
2949 create_options
= append_option_parameters(create_options
,
2950 drv
->create_options
);
2951 create_options
= append_option_parameters(create_options
,
2952 proto_drv
->create_options
);
2954 /* Create parameter list with default values */
2955 param
= parse_option_parameters("", create_options
, param
);
2957 set_option_parameter_int(param
, BLOCK_OPT_SIZE
, img_size
);
2959 /* Parse -o options */
2961 param
= parse_option_parameters(options
, create_options
, param
);
2962 if (param
== NULL
) {
2963 error_report("Invalid options for file format '%s'.", fmt
);
2969 if (base_filename
) {
2970 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FILE
,
2972 error_report("Backing file not supported for file format '%s'",
2980 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FMT
, base_fmt
)) {
2981 error_report("Backing file format not supported for file "
2982 "format '%s'", fmt
);
2988 backing_file
= get_option_parameter(param
, BLOCK_OPT_BACKING_FILE
);
2989 if (backing_file
&& backing_file
->value
.s
) {
2990 if (!strcmp(filename
, backing_file
->value
.s
)) {
2991 error_report("Error: Trying to create an image with the "
2992 "same filename as the backing file");
2998 backing_fmt
= get_option_parameter(param
, BLOCK_OPT_BACKING_FMT
);
2999 if (backing_fmt
&& backing_fmt
->value
.s
) {
3000 backing_drv
= bdrv_find_format(backing_fmt
->value
.s
);
3002 error_report("Unknown backing file format '%s'",
3003 backing_fmt
->value
.s
);
3009 // The size for the image must always be specified, with one exception:
3010 // If we are using a backing file, we can obtain the size from there
3011 size
= get_option_parameter(param
, BLOCK_OPT_SIZE
);
3012 if (size
&& size
->value
.n
== -1) {
3013 if (backing_file
&& backing_file
->value
.s
) {
3019 ret
= bdrv_open(bs
, backing_file
->value
.s
, flags
, backing_drv
);
3021 error_report("Could not open '%s'", backing_file
->value
.s
);
3024 bdrv_get_geometry(bs
, &size
);
3027 snprintf(buf
, sizeof(buf
), "%" PRId64
, size
);
3028 set_option_parameter(param
, BLOCK_OPT_SIZE
, buf
);
3030 error_report("Image creation needs a size parameter");
3036 printf("Formatting '%s', fmt=%s ", filename
, fmt
);
3037 print_option_parameters(param
);
3040 ret
= bdrv_create(drv
, filename
, param
);
3043 if (ret
== -ENOTSUP
) {
3044 error_report("Formatting or formatting option not supported for "
3045 "file format '%s'", fmt
);
3046 } else if (ret
== -EFBIG
) {
3047 error_report("The image size is too large for file format '%s'",
3050 error_report("%s: error while creating %s: %s", filename
, fmt
,
3056 free_option_parameters(create_options
);
3057 free_option_parameters(param
);