2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "config-host.h"
25 #include "qemu-common.h"
27 #include "block_int.h"
29 #include "qemu-objects.h"
32 #include <sys/types.h>
34 #include <sys/ioctl.h>
35 #include <sys/queue.h>
45 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
46 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
47 BlockDriverCompletionFunc
*cb
, void *opaque
);
48 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
49 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
50 BlockDriverCompletionFunc
*cb
, void *opaque
);
51 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
52 BlockDriverCompletionFunc
*cb
, void *opaque
);
53 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
54 uint8_t *buf
, int nb_sectors
);
55 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
56 const uint8_t *buf
, int nb_sectors
);
57 static BlockDriver
*find_protocol(const char *filename
);
59 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
60 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
62 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
63 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
65 /* If non-zero, use only whitelisted block drivers */
66 static int use_bdrv_whitelist
;
68 int path_is_absolute(const char *path
)
72 /* specific case for names like: "\\.\d:" */
73 if (*path
== '/' || *path
== '\\')
76 p
= strchr(path
, ':');
82 return (*p
== '/' || *p
== '\\');
88 /* if filename is absolute, just copy it to dest. Otherwise, build a
89 path to it by considering it is relative to base_path. URL are
91 void path_combine(char *dest
, int dest_size
,
92 const char *base_path
,
100 if (path_is_absolute(filename
)) {
101 pstrcpy(dest
, dest_size
, filename
);
103 p
= strchr(base_path
, ':');
108 p1
= strrchr(base_path
, '/');
112 p2
= strrchr(base_path
, '\\');
124 if (len
> dest_size
- 1)
126 memcpy(dest
, base_path
, len
);
128 pstrcat(dest
, dest_size
, filename
);
132 void bdrv_register(BlockDriver
*bdrv
)
134 if (!bdrv
->bdrv_aio_readv
) {
135 /* add AIO emulation layer */
136 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
137 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
138 } else if (!bdrv
->bdrv_read
) {
139 /* add synchronous IO emulation layer */
140 bdrv
->bdrv_read
= bdrv_read_em
;
141 bdrv
->bdrv_write
= bdrv_write_em
;
144 if (!bdrv
->bdrv_aio_flush
)
145 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
147 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
150 /* create a new block device (by default it is empty) */
151 BlockDriverState
*bdrv_new(const char *device_name
)
153 BlockDriverState
*bs
;
155 bs
= qemu_mallocz(sizeof(BlockDriverState
));
156 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
157 if (device_name
[0] != '\0') {
158 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
163 BlockDriver
*bdrv_find_format(const char *format_name
)
166 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
167 if (!strcmp(drv1
->format_name
, format_name
)) {
174 static int bdrv_is_whitelisted(BlockDriver
*drv
)
176 static const char *whitelist
[] = {
177 CONFIG_BDRV_WHITELIST
182 return 1; /* no whitelist, anything goes */
184 for (p
= whitelist
; *p
; p
++) {
185 if (!strcmp(drv
->format_name
, *p
)) {
192 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
194 BlockDriver
*drv
= bdrv_find_format(format_name
);
195 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
198 int bdrv_create(BlockDriver
*drv
, const char* filename
,
199 QEMUOptionParameter
*options
)
201 if (!drv
->bdrv_create
)
204 return drv
->bdrv_create(filename
, options
);
207 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
211 drv
= find_protocol(filename
);
213 drv
= bdrv_find_format("file");
216 return bdrv_create(drv
, filename
, options
);
220 void get_tmp_filename(char *filename
, int size
)
222 char temp_dir
[MAX_PATH
];
224 GetTempPath(MAX_PATH
, temp_dir
);
225 GetTempFileName(temp_dir
, "qem", 0, filename
);
228 void get_tmp_filename(char *filename
, int size
)
232 /* XXX: race condition possible */
233 tmpdir
= getenv("TMPDIR");
236 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
237 fd
= mkstemp(filename
);
243 static int is_windows_drive_prefix(const char *filename
)
245 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
246 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
250 int is_windows_drive(const char *filename
)
252 if (is_windows_drive_prefix(filename
) &&
255 if (strstart(filename
, "\\\\.\\", NULL
) ||
256 strstart(filename
, "//./", NULL
))
263 * Detect host devices. By convention, /dev/cdrom[N] is always
264 * recognized as a host CDROM.
266 static BlockDriver
*find_hdev_driver(const char *filename
)
268 int score_max
= 0, score
;
269 BlockDriver
*drv
= NULL
, *d
;
271 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
272 if (d
->bdrv_probe_device
) {
273 score
= d
->bdrv_probe_device(filename
);
274 if (score
> score_max
) {
284 static BlockDriver
*find_protocol(const char *filename
)
291 /* TODO Drivers without bdrv_file_open must be specified explicitly */
294 if (is_windows_drive(filename
) ||
295 is_windows_drive_prefix(filename
))
296 return bdrv_find_format("file");
298 p
= strchr(filename
, ':');
300 drv1
= find_hdev_driver(filename
);
302 drv1
= bdrv_find_format("file");
307 if (len
> sizeof(protocol
) - 1)
308 len
= sizeof(protocol
) - 1;
309 memcpy(protocol
, filename
, len
);
310 protocol
[len
] = '\0';
311 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
312 if (drv1
->protocol_name
&&
313 !strcmp(drv1
->protocol_name
, protocol
)) {
320 static BlockDriver
*find_image_format(const char *filename
)
322 int ret
, score
, score_max
;
323 BlockDriver
*drv1
, *drv
;
325 BlockDriverState
*bs
;
327 drv
= find_protocol(filename
);
328 /* no need to test disk image formats for vvfat */
329 if (drv
&& strcmp(drv
->format_name
, "vvfat") == 0)
332 ret
= bdrv_file_open(&bs
, filename
, 0);
335 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
343 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
344 if (drv1
->bdrv_probe
) {
345 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
346 if (score
> score_max
) {
356 * Set the current 'total_sectors' value
358 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
360 BlockDriver
*drv
= bs
->drv
;
362 /* query actual device if possible, otherwise just trust the hint */
363 if (drv
->bdrv_getlength
) {
364 int64_t length
= drv
->bdrv_getlength(bs
);
368 hint
= length
>> BDRV_SECTOR_BITS
;
371 bs
->total_sectors
= hint
;
376 * Common part for opening disk images and files
378 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
379 int flags
, BlockDriver
*drv
)
386 bs
->total_sectors
= 0;
387 bs
->is_temporary
= 0;
390 bs
->open_flags
= flags
;
391 /* buffer_alignment defaulted to 512, drivers can change this value */
392 bs
->buffer_alignment
= 512;
394 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
396 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
401 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
404 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
405 * write cache to the guest. We do need the fdatasync to flush
406 * out transactions for block allocations, and we maybe have a
407 * volatile write cache in our backing device to deal with.
409 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
410 bs
->enable_write_cache
= 1;
413 * Clear flags that are internal to the block layer before opening the
416 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
419 * Snapshots should be writeable.
421 if (bs
->is_temporary
) {
422 open_flags
|= BDRV_O_RDWR
;
425 /* Open the image, either directly or using a protocol */
426 if (drv
->bdrv_file_open
) {
427 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
429 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
431 ret
= drv
->bdrv_open(bs
, open_flags
);
439 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
441 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
447 if (bs
->is_temporary
) {
455 bdrv_delete(bs
->file
);
458 qemu_free(bs
->opaque
);
465 * Opens a file using a protocol (file, host_device, nbd, ...)
467 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
469 BlockDriverState
*bs
;
473 drv
= find_protocol(filename
);
479 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
490 * Opens a disk image (raw, qcow2, vmdk, ...)
492 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
497 if (flags
& BDRV_O_SNAPSHOT
) {
498 BlockDriverState
*bs1
;
501 BlockDriver
*bdrv_qcow2
;
502 QEMUOptionParameter
*options
;
503 char tmp_filename
[PATH_MAX
];
504 char backing_filename
[PATH_MAX
];
506 /* if snapshot, we create a temporary backing file and open it
507 instead of opening 'filename' directly */
509 /* if there is a backing file, use it */
511 ret
= bdrv_open(bs1
, filename
, 0, drv
);
516 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
518 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
523 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
525 /* Real path is meaningless for protocols */
527 snprintf(backing_filename
, sizeof(backing_filename
),
529 else if (!realpath(filename
, backing_filename
))
532 bdrv_qcow2
= bdrv_find_format("qcow2");
533 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
535 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
536 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
538 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
542 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
547 filename
= tmp_filename
;
549 bs
->is_temporary
= 1;
552 /* Find the right image format driver */
554 drv
= find_image_format(filename
);
559 goto unlink_and_fail
;
563 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
565 goto unlink_and_fail
;
568 /* If there is a backing file, use it */
569 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
570 char backing_filename
[PATH_MAX
];
572 BlockDriver
*back_drv
= NULL
;
574 bs
->backing_hd
= bdrv_new("");
575 path_combine(backing_filename
, sizeof(backing_filename
),
576 filename
, bs
->backing_file
);
577 if (bs
->backing_format
[0] != '\0')
578 back_drv
= bdrv_find_format(bs
->backing_format
);
580 /* backing files always opened read-only */
582 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
584 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
589 if (bs
->is_temporary
) {
590 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
592 /* base image inherits from "parent" */
593 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
597 if (!bdrv_key_required(bs
)) {
598 /* call the change callback */
599 bs
->media_changed
= 1;
601 bs
->change_cb(bs
->change_opaque
);
607 if (bs
->is_temporary
) {
613 void bdrv_close(BlockDriverState
*bs
)
616 if (bs
->backing_hd
) {
617 bdrv_delete(bs
->backing_hd
);
618 bs
->backing_hd
= NULL
;
620 bs
->drv
->bdrv_close(bs
);
621 qemu_free(bs
->opaque
);
623 if (bs
->is_temporary
) {
624 unlink(bs
->filename
);
630 if (bs
->file
!= NULL
) {
631 bdrv_close(bs
->file
);
634 /* call the change callback */
635 bs
->media_changed
= 1;
637 bs
->change_cb(bs
->change_opaque
);
641 void bdrv_delete(BlockDriverState
*bs
)
643 /* remove from list, if necessary */
644 if (bs
->device_name
[0] != '\0') {
645 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
649 if (bs
->file
!= NULL
) {
650 bdrv_delete(bs
->file
);
657 * Run consistency checks on an image
659 * Returns the number of errors or -errno when an internal error occurs
661 int bdrv_check(BlockDriverState
*bs
)
663 if (bs
->drv
->bdrv_check
== NULL
) {
667 return bs
->drv
->bdrv_check(bs
);
670 /* commit COW file into the raw image */
671 int bdrv_commit(BlockDriverState
*bs
)
673 BlockDriver
*drv
= bs
->drv
;
674 int64_t i
, total_sectors
;
675 int n
, j
, ro
, open_flags
;
676 int ret
= 0, rw_ret
= 0;
677 unsigned char sector
[512];
679 BlockDriverState
*bs_rw
, *bs_ro
;
684 if (!bs
->backing_hd
) {
688 if (bs
->backing_hd
->keep_read_only
) {
692 ro
= bs
->backing_hd
->read_only
;
693 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
694 open_flags
= bs
->backing_hd
->open_flags
;
698 bdrv_delete(bs
->backing_hd
);
699 bs
->backing_hd
= NULL
;
700 bs_rw
= bdrv_new("");
701 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, NULL
);
704 /* try to re-open read-only */
705 bs_ro
= bdrv_new("");
706 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
709 /* drive not functional anymore */
713 bs
->backing_hd
= bs_ro
;
716 bs
->backing_hd
= bs_rw
;
719 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
720 for (i
= 0; i
< total_sectors
;) {
721 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
722 for(j
= 0; j
< n
; j
++) {
723 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
728 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
739 if (drv
->bdrv_make_empty
) {
740 ret
= drv
->bdrv_make_empty(bs
);
745 * Make sure all data we wrote to the backing device is actually
749 bdrv_flush(bs
->backing_hd
);
755 bdrv_delete(bs
->backing_hd
);
756 bs
->backing_hd
= NULL
;
757 bs_ro
= bdrv_new("");
758 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
761 /* drive not functional anymore */
765 bs
->backing_hd
= bs_ro
;
766 bs
->backing_hd
->keep_read_only
= 0;
775 * -EINVAL - backing format specified, but no file
776 * -ENOSPC - can't update the backing file because no space is left in the
778 * -ENOTSUP - format driver doesn't support changing the backing file
780 int bdrv_change_backing_file(BlockDriverState
*bs
,
781 const char *backing_file
, const char *backing_fmt
)
783 BlockDriver
*drv
= bs
->drv
;
785 if (drv
->bdrv_change_backing_file
!= NULL
) {
786 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
792 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
797 if (!bdrv_is_inserted(bs
))
803 len
= bdrv_getlength(bs
);
808 if ((offset
> len
) || (len
- offset
< size
))
814 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
817 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
820 /* return < 0 if error. See bdrv_write() for the return codes */
821 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
822 uint8_t *buf
, int nb_sectors
)
824 BlockDriver
*drv
= bs
->drv
;
828 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
831 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
834 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
835 int nb_sectors
, int dirty
)
838 unsigned long val
, idx
, bit
;
840 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
841 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
843 for (; start
<= end
; start
++) {
844 idx
= start
/ (sizeof(unsigned long) * 8);
845 bit
= start
% (sizeof(unsigned long) * 8);
846 val
= bs
->dirty_bitmap
[idx
];
848 if (!(val
& (1 << bit
))) {
853 if (val
& (1 << bit
)) {
858 bs
->dirty_bitmap
[idx
] = val
;
862 /* Return < 0 if error. Important errors are:
863 -EIO generic I/O error (may happen for all errors)
864 -ENOMEDIUM No media inserted.
865 -EINVAL Invalid sector number or nb_sectors
866 -EACCES Trying to write a read-only device
868 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
869 const uint8_t *buf
, int nb_sectors
)
871 BlockDriver
*drv
= bs
->drv
;
876 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
879 if (bs
->dirty_bitmap
) {
880 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
883 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
886 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
887 void *buf
, int count1
)
889 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
890 int len
, nb_sectors
, count
;
895 /* first read to align to sector start */
896 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
899 sector_num
= offset
>> BDRV_SECTOR_BITS
;
901 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
903 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
911 /* read the sectors "in place" */
912 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
913 if (nb_sectors
> 0) {
914 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
916 sector_num
+= nb_sectors
;
917 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
922 /* add data from the last sector */
924 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
926 memcpy(buf
, tmp_buf
, count
);
931 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
932 const void *buf
, int count1
)
934 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
935 int len
, nb_sectors
, count
;
940 /* first write to align to sector start */
941 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
944 sector_num
= offset
>> BDRV_SECTOR_BITS
;
946 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
948 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
949 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
958 /* write the sectors "in place" */
959 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
960 if (nb_sectors
> 0) {
961 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
963 sector_num
+= nb_sectors
;
964 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
969 /* add data from the last sector */
971 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
973 memcpy(tmp_buf
, buf
, count
);
974 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
981 * Truncate file to 'offset' bytes (needed only for file protocols)
983 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
985 BlockDriver
*drv
= bs
->drv
;
989 if (!drv
->bdrv_truncate
)
993 ret
= drv
->bdrv_truncate(bs
, offset
);
995 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1001 * Length of a file in bytes. Return < 0 if error or unknown.
1003 int64_t bdrv_getlength(BlockDriverState
*bs
)
1005 BlockDriver
*drv
= bs
->drv
;
1009 /* Fixed size devices use the total_sectors value for speed instead of
1010 issuing a length query (like lseek) on each call. Also, legacy block
1011 drivers don't provide a bdrv_getlength function and must use
1013 if (!bs
->growable
|| !drv
->bdrv_getlength
) {
1014 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1016 return drv
->bdrv_getlength(bs
);
1019 /* return 0 as number of sectors if no device present or error */
1020 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1023 length
= bdrv_getlength(bs
);
1027 length
= length
>> BDRV_SECTOR_BITS
;
1028 *nb_sectors_ptr
= length
;
1032 uint8_t boot_ind
; /* 0x80 - active */
1033 uint8_t head
; /* starting head */
1034 uint8_t sector
; /* starting sector */
1035 uint8_t cyl
; /* starting cylinder */
1036 uint8_t sys_ind
; /* What partition type */
1037 uint8_t end_head
; /* end head */
1038 uint8_t end_sector
; /* end sector */
1039 uint8_t end_cyl
; /* end cylinder */
1040 uint32_t start_sect
; /* starting sector counting from 0 */
1041 uint32_t nr_sects
; /* nr of sectors in partition */
1042 } __attribute__((packed
));
1044 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1045 static int guess_disk_lchs(BlockDriverState
*bs
,
1046 int *pcylinders
, int *pheads
, int *psectors
)
1049 int ret
, i
, heads
, sectors
, cylinders
;
1050 struct partition
*p
;
1052 uint64_t nb_sectors
;
1054 bdrv_get_geometry(bs
, &nb_sectors
);
1056 ret
= bdrv_read(bs
, 0, buf
, 1);
1059 /* test msdos magic */
1060 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1062 for(i
= 0; i
< 4; i
++) {
1063 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1064 nr_sects
= le32_to_cpu(p
->nr_sects
);
1065 if (nr_sects
&& p
->end_head
) {
1066 /* We make the assumption that the partition terminates on
1067 a cylinder boundary */
1068 heads
= p
->end_head
+ 1;
1069 sectors
= p
->end_sector
& 63;
1072 cylinders
= nb_sectors
/ (heads
* sectors
);
1073 if (cylinders
< 1 || cylinders
> 16383)
1076 *psectors
= sectors
;
1077 *pcylinders
= cylinders
;
1079 printf("guessed geometry: LCHS=%d %d %d\n",
1080 cylinders
, heads
, sectors
);
1088 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1090 int translation
, lba_detected
= 0;
1091 int cylinders
, heads
, secs
;
1092 uint64_t nb_sectors
;
1094 /* if a geometry hint is available, use it */
1095 bdrv_get_geometry(bs
, &nb_sectors
);
1096 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1097 translation
= bdrv_get_translation_hint(bs
);
1098 if (cylinders
!= 0) {
1103 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1105 /* if heads > 16, it means that a BIOS LBA
1106 translation was active, so the default
1107 hardware geometry is OK */
1109 goto default_geometry
;
1114 /* disable any translation to be in sync with
1115 the logical geometry */
1116 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1117 bdrv_set_translation_hint(bs
,
1118 BIOS_ATA_TRANSLATION_NONE
);
1123 /* if no geometry, use a standard physical disk geometry */
1124 cylinders
= nb_sectors
/ (16 * 63);
1126 if (cylinders
> 16383)
1128 else if (cylinders
< 2)
1133 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1134 if ((*pcyls
* *pheads
) <= 131072) {
1135 bdrv_set_translation_hint(bs
,
1136 BIOS_ATA_TRANSLATION_LARGE
);
1138 bdrv_set_translation_hint(bs
,
1139 BIOS_ATA_TRANSLATION_LBA
);
1143 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1147 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1148 int cyls
, int heads
, int secs
)
1155 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1158 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1159 type
== BDRV_TYPE_FLOPPY
));
1162 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1164 bs
->translation
= translation
;
1167 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1168 int *pcyls
, int *pheads
, int *psecs
)
1171 *pheads
= bs
->heads
;
1175 int bdrv_get_type_hint(BlockDriverState
*bs
)
1180 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1182 return bs
->translation
;
1185 int bdrv_is_removable(BlockDriverState
*bs
)
1187 return bs
->removable
;
1190 int bdrv_is_read_only(BlockDriverState
*bs
)
1192 return bs
->read_only
;
1195 int bdrv_is_sg(BlockDriverState
*bs
)
1200 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1202 return bs
->enable_write_cache
;
1205 /* XXX: no longer used */
1206 void bdrv_set_change_cb(BlockDriverState
*bs
,
1207 void (*change_cb
)(void *opaque
), void *opaque
)
1209 bs
->change_cb
= change_cb
;
1210 bs
->change_opaque
= opaque
;
1213 int bdrv_is_encrypted(BlockDriverState
*bs
)
1215 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1217 return bs
->encrypted
;
1220 int bdrv_key_required(BlockDriverState
*bs
)
1222 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1224 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1226 return (bs
->encrypted
&& !bs
->valid_key
);
1229 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1232 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1233 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1239 if (!bs
->encrypted
) {
1241 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1244 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1247 } else if (!bs
->valid_key
) {
1249 /* call the change callback now, we skipped it on open */
1250 bs
->media_changed
= 1;
1252 bs
->change_cb(bs
->change_opaque
);
1257 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1262 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1266 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1271 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1272 it(opaque
, drv
->format_name
);
1276 BlockDriverState
*bdrv_find(const char *name
)
1278 BlockDriverState
*bs
;
1280 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1281 if (!strcmp(name
, bs
->device_name
)) {
1288 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1290 BlockDriverState
*bs
;
1292 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1297 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1299 return bs
->device_name
;
1302 void bdrv_flush(BlockDriverState
*bs
)
1304 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1305 bs
->drv
->bdrv_flush(bs
);
1308 void bdrv_flush_all(void)
1310 BlockDriverState
*bs
;
1312 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1313 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1314 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1320 int bdrv_has_zero_init(BlockDriverState
*bs
)
1324 if (bs
->drv
->no_zero_init
) {
1326 } else if (bs
->file
) {
1327 return bdrv_has_zero_init(bs
->file
);
1334 * Returns true iff the specified sector is present in the disk image. Drivers
1335 * not implementing the functionality are assumed to not support backing files,
1336 * hence all their sectors are reported as allocated.
1338 * 'pnum' is set to the number of sectors (including and immediately following
1339 * the specified sector) that are known to be in the same
1340 * allocated/unallocated state.
1342 * 'nb_sectors' is the max value 'pnum' should be set to.
1344 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1348 if (!bs
->drv
->bdrv_is_allocated
) {
1349 if (sector_num
>= bs
->total_sectors
) {
1353 n
= bs
->total_sectors
- sector_num
;
1354 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1357 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1360 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1361 BlockMonEventAction action
, int is_read
)
1364 const char *action_str
;
1367 case BDRV_ACTION_REPORT
:
1368 action_str
= "report";
1370 case BDRV_ACTION_IGNORE
:
1371 action_str
= "ignore";
1373 case BDRV_ACTION_STOP
:
1374 action_str
= "stop";
1380 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1383 is_read
? "read" : "write");
1384 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1386 qobject_decref(data
);
1389 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1392 Monitor
*mon
= opaque
;
1394 bs_dict
= qobject_to_qdict(obj
);
1396 monitor_printf(mon
, "%s: type=%s removable=%d",
1397 qdict_get_str(bs_dict
, "device"),
1398 qdict_get_str(bs_dict
, "type"),
1399 qdict_get_bool(bs_dict
, "removable"));
1401 if (qdict_get_bool(bs_dict
, "removable")) {
1402 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1405 if (qdict_haskey(bs_dict
, "inserted")) {
1406 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1408 monitor_printf(mon
, " file=");
1409 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1410 if (qdict_haskey(qdict
, "backing_file")) {
1411 monitor_printf(mon
, " backing_file=");
1412 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1414 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1415 qdict_get_bool(qdict
, "ro"),
1416 qdict_get_str(qdict
, "drv"),
1417 qdict_get_bool(qdict
, "encrypted"));
1419 monitor_printf(mon
, " [not inserted]");
1422 monitor_printf(mon
, "\n");
1425 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1427 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1431 * bdrv_info(): Block devices information
1433 * Each block device information is stored in a QDict and the
1434 * returned QObject is a QList of all devices.
1436 * The QDict contains the following:
1438 * - "device": device name
1439 * - "type": device type
1440 * - "removable": true if the device is removable, false otherwise
1441 * - "locked": true if the device is locked, false otherwise
1442 * - "inserted": only present if the device is inserted, it is a QDict
1443 * containing the following:
1444 * - "file": device file name
1445 * - "ro": true if read-only, false otherwise
1446 * - "drv": driver format name
1447 * - "backing_file": backing file name if one is used
1448 * - "encrypted": true if encrypted, false otherwise
1452 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1453 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1454 * { "device": "floppy0", "type": "floppy", "removable": true,
1455 * "locked": false } ]
1457 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1460 BlockDriverState
*bs
;
1462 bs_list
= qlist_new();
1464 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1466 const char *type
= "unknown";
1472 case BDRV_TYPE_CDROM
:
1475 case BDRV_TYPE_FLOPPY
:
1480 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1481 "'removable': %i, 'locked': %i }",
1482 bs
->device_name
, type
, bs
->removable
,
1487 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1489 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1490 "'encrypted': %i }",
1491 bs
->filename
, bs
->read_only
,
1492 bs
->drv
->format_name
,
1493 bdrv_is_encrypted(bs
));
1494 if (bs
->backing_file
[0] != '\0') {
1495 QDict
*qdict
= qobject_to_qdict(obj
);
1496 qdict_put(qdict
, "backing_file",
1497 qstring_from_str(bs
->backing_file
));
1500 qdict_put_obj(bs_dict
, "inserted", obj
);
1502 qlist_append_obj(bs_list
, bs_obj
);
1505 *ret_data
= QOBJECT(bs_list
);
1508 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1511 Monitor
*mon
= opaque
;
1513 qdict
= qobject_to_qdict(data
);
1514 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1516 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1517 monitor_printf(mon
, " rd_bytes=%" PRId64
1518 " wr_bytes=%" PRId64
1519 " rd_operations=%" PRId64
1520 " wr_operations=%" PRId64
1522 qdict_get_int(qdict
, "rd_bytes"),
1523 qdict_get_int(qdict
, "wr_bytes"),
1524 qdict_get_int(qdict
, "rd_operations"),
1525 qdict_get_int(qdict
, "wr_operations"));
1528 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1530 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1534 * bdrv_info_stats(): show block device statistics
1536 * Each device statistic information is stored in a QDict and
1537 * the returned QObject is a QList of all devices.
1539 * The QDict contains the following:
1541 * - "device": device name
1542 * - "stats": A QDict with the statistics information, it contains:
1543 * - "rd_bytes": bytes read
1544 * - "wr_bytes": bytes written
1545 * - "rd_operations": read operations
1546 * - "wr_operations": write operations
1550 * [ { "device": "ide0-hd0",
1551 * "stats": { "rd_bytes": 512,
1553 * "rd_operations": 1,
1554 * "wr_operations": 0 } },
1555 * { "device": "ide1-cd0",
1556 * "stats": { "rd_bytes": 0,
1558 * "rd_operations": 0,
1559 * "wr_operations": 0 } } ]
1561 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1565 BlockDriverState
*bs
;
1567 devices
= qlist_new();
1569 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1570 obj
= qobject_from_jsonf("{ 'device': %s, 'stats': {"
1571 "'rd_bytes': %" PRId64
","
1572 "'wr_bytes': %" PRId64
","
1573 "'rd_operations': %" PRId64
","
1574 "'wr_operations': %" PRId64
1577 bs
->rd_bytes
, bs
->wr_bytes
,
1578 bs
->rd_ops
, bs
->wr_ops
);
1579 qlist_append_obj(devices
, obj
);
1582 *ret_data
= QOBJECT(devices
);
1585 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1587 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1588 return bs
->backing_file
;
1589 else if (bs
->encrypted
)
1590 return bs
->filename
;
1595 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1596 char *filename
, int filename_size
)
1598 if (!bs
->backing_file
) {
1599 pstrcpy(filename
, filename_size
, "");
1601 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1605 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1606 const uint8_t *buf
, int nb_sectors
)
1608 BlockDriver
*drv
= bs
->drv
;
1611 if (!drv
->bdrv_write_compressed
)
1613 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1616 if (bs
->dirty_bitmap
) {
1617 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1620 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1623 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1625 BlockDriver
*drv
= bs
->drv
;
1628 if (!drv
->bdrv_get_info
)
1630 memset(bdi
, 0, sizeof(*bdi
));
1631 return drv
->bdrv_get_info(bs
, bdi
);
1634 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1635 int64_t pos
, int size
)
1637 BlockDriver
*drv
= bs
->drv
;
1640 if (!drv
->bdrv_save_vmstate
)
1642 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1645 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1646 int64_t pos
, int size
)
1648 BlockDriver
*drv
= bs
->drv
;
1651 if (!drv
->bdrv_load_vmstate
)
1653 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1656 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1658 BlockDriver
*drv
= bs
->drv
;
1660 if (!drv
|| !drv
->bdrv_debug_event
) {
1664 return drv
->bdrv_debug_event(bs
, event
);
1668 /**************************************************************/
1669 /* handling of snapshots */
1671 int bdrv_snapshot_create(BlockDriverState
*bs
,
1672 QEMUSnapshotInfo
*sn_info
)
1674 BlockDriver
*drv
= bs
->drv
;
1677 if (!drv
->bdrv_snapshot_create
)
1679 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1682 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1683 const char *snapshot_id
)
1685 BlockDriver
*drv
= bs
->drv
;
1688 if (!drv
->bdrv_snapshot_goto
)
1690 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1693 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1695 BlockDriver
*drv
= bs
->drv
;
1698 if (!drv
->bdrv_snapshot_delete
)
1700 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1703 int bdrv_snapshot_list(BlockDriverState
*bs
,
1704 QEMUSnapshotInfo
**psn_info
)
1706 BlockDriver
*drv
= bs
->drv
;
1709 if (!drv
->bdrv_snapshot_list
)
1711 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1714 #define NB_SUFFIXES 4
1716 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1718 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1723 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1726 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1727 if (size
< (10 * base
)) {
1728 snprintf(buf
, buf_size
, "%0.1f%c",
1729 (double)size
/ base
,
1732 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1733 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1734 ((size
+ (base
>> 1)) / base
),
1744 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1746 char buf1
[128], date_buf
[128], clock_buf
[128];
1756 snprintf(buf
, buf_size
,
1757 "%-10s%-20s%7s%20s%15s",
1758 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1762 ptm
= localtime(&ti
);
1763 strftime(date_buf
, sizeof(date_buf
),
1764 "%Y-%m-%d %H:%M:%S", ptm
);
1766 localtime_r(&ti
, &tm
);
1767 strftime(date_buf
, sizeof(date_buf
),
1768 "%Y-%m-%d %H:%M:%S", &tm
);
1770 secs
= sn
->vm_clock_nsec
/ 1000000000;
1771 snprintf(clock_buf
, sizeof(clock_buf
),
1772 "%02d:%02d:%02d.%03d",
1774 (int)((secs
/ 60) % 60),
1776 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1777 snprintf(buf
, buf_size
,
1778 "%-10s%-20s%7s%20s%15s",
1779 sn
->id_str
, sn
->name
,
1780 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1788 /**************************************************************/
1791 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1792 QEMUIOVector
*qiov
, int nb_sectors
,
1793 BlockDriverCompletionFunc
*cb
, void *opaque
)
1795 BlockDriver
*drv
= bs
->drv
;
1796 BlockDriverAIOCB
*ret
;
1800 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1803 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1807 /* Update stats even though technically transfer has not happened. */
1808 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1815 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1816 QEMUIOVector
*qiov
, int nb_sectors
,
1817 BlockDriverCompletionFunc
*cb
, void *opaque
)
1819 BlockDriver
*drv
= bs
->drv
;
1820 BlockDriverAIOCB
*ret
;
1826 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1829 if (bs
->dirty_bitmap
) {
1830 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1833 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1837 /* Update stats even though technically transfer has not happened. */
1838 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1846 typedef struct MultiwriteCB
{
1851 BlockDriverCompletionFunc
*cb
;
1853 QEMUIOVector
*free_qiov
;
1858 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1862 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1863 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1864 if (mcb
->callbacks
[i
].free_qiov
) {
1865 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
1867 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1868 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1872 static void multiwrite_cb(void *opaque
, int ret
)
1874 MultiwriteCB
*mcb
= opaque
;
1876 if (ret
< 0 && !mcb
->error
) {
1878 multiwrite_user_cb(mcb
);
1881 mcb
->num_requests
--;
1882 if (mcb
->num_requests
== 0) {
1883 if (mcb
->error
== 0) {
1884 multiwrite_user_cb(mcb
);
1890 static int multiwrite_req_compare(const void *a
, const void *b
)
1892 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1896 * Takes a bunch of requests and tries to merge them. Returns the number of
1897 * requests that remain after merging.
1899 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1900 int num_reqs
, MultiwriteCB
*mcb
)
1904 // Sort requests by start sector
1905 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1907 // Check if adjacent requests touch the same clusters. If so, combine them,
1908 // filling up gaps with zero sectors.
1910 for (i
= 1; i
< num_reqs
; i
++) {
1912 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1914 // This handles the cases that are valid for all block drivers, namely
1915 // exactly sequential writes and overlapping writes.
1916 if (reqs
[i
].sector
<= oldreq_last
) {
1920 // The block driver may decide that it makes sense to combine requests
1921 // even if there is a gap of some sectors between them. In this case,
1922 // the gap is filled with zeros (therefore only applicable for yet
1923 // unused space in format like qcow2).
1924 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1925 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1928 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
1934 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1935 qemu_iovec_init(qiov
,
1936 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1938 // Add the first request to the merged one. If the requests are
1939 // overlapping, drop the last sectors of the first request.
1940 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1941 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1943 // We might need to add some zeros between the two requests
1944 if (reqs
[i
].sector
> oldreq_last
) {
1945 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1946 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1947 memset(buf
, 0, zero_bytes
);
1948 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1949 mcb
->callbacks
[i
].free_buf
= buf
;
1952 // Add the second request
1953 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
1955 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
1956 reqs
[outidx
].qiov
= qiov
;
1958 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
1961 reqs
[outidx
].sector
= reqs
[i
].sector
;
1962 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
1963 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
1971 * Submit multiple AIO write requests at once.
1973 * On success, the function returns 0 and all requests in the reqs array have
1974 * been submitted. In error case this function returns -1, and any of the
1975 * requests may or may not be submitted yet. In particular, this means that the
1976 * callback will be called for some of the requests, for others it won't. The
1977 * caller must check the error field of the BlockRequest to wait for the right
1978 * callbacks (if error != 0, no callback will be called).
1980 * The implementation may modify the contents of the reqs array, e.g. to merge
1981 * requests. However, the fields opaque and error are left unmodified as they
1982 * are used to signal failure for a single request to the caller.
1984 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
1986 BlockDriverAIOCB
*acb
;
1990 if (num_reqs
== 0) {
1994 // Create MultiwriteCB structure
1995 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
1996 mcb
->num_requests
= 0;
1997 mcb
->num_callbacks
= num_reqs
;
1999 for (i
= 0; i
< num_reqs
; i
++) {
2000 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2001 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2004 // Check for mergable requests
2005 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2007 // Run the aio requests
2008 for (i
= 0; i
< num_reqs
; i
++) {
2009 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2010 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2013 // We can only fail the whole thing if no request has been
2014 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2015 // complete and report the error in the callback.
2016 if (mcb
->num_requests
== 0) {
2017 reqs
[i
].error
= -EIO
;
2020 mcb
->num_requests
++;
2021 multiwrite_cb(mcb
, -EIO
);
2025 mcb
->num_requests
++;
2036 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2037 BlockDriverCompletionFunc
*cb
, void *opaque
)
2039 BlockDriver
*drv
= bs
->drv
;
2043 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2046 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2048 acb
->pool
->cancel(acb
);
2052 /**************************************************************/
2053 /* async block device emulation */
2055 typedef struct BlockDriverAIOCBSync
{
2056 BlockDriverAIOCB common
;
2059 /* vector translation state */
2063 } BlockDriverAIOCBSync
;
2065 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2067 BlockDriverAIOCBSync
*acb
= (BlockDriverAIOCBSync
*)blockacb
;
2068 qemu_bh_delete(acb
->bh
);
2070 qemu_aio_release(acb
);
2073 static AIOPool bdrv_em_aio_pool
= {
2074 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2075 .cancel
= bdrv_aio_cancel_em
,
2078 static void bdrv_aio_bh_cb(void *opaque
)
2080 BlockDriverAIOCBSync
*acb
= opaque
;
2083 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2084 qemu_vfree(acb
->bounce
);
2085 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2086 qemu_bh_delete(acb
->bh
);
2088 qemu_aio_release(acb
);
2091 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2095 BlockDriverCompletionFunc
*cb
,
2100 BlockDriverAIOCBSync
*acb
;
2102 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2103 acb
->is_write
= is_write
;
2105 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2108 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2111 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2112 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2114 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2117 qemu_bh_schedule(acb
->bh
);
2119 return &acb
->common
;
2122 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2123 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2124 BlockDriverCompletionFunc
*cb
, void *opaque
)
2126 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2129 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2130 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2131 BlockDriverCompletionFunc
*cb
, void *opaque
)
2133 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2136 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2137 BlockDriverCompletionFunc
*cb
, void *opaque
)
2139 BlockDriverAIOCBSync
*acb
;
2141 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2142 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2148 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2151 qemu_bh_schedule(acb
->bh
);
2152 return &acb
->common
;
2155 /**************************************************************/
2156 /* sync block device emulation */
2158 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2160 *(int *)opaque
= ret
;
2163 #define NOT_DONE 0x7fffffff
2165 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2166 uint8_t *buf
, int nb_sectors
)
2169 BlockDriverAIOCB
*acb
;
2173 async_context_push();
2175 async_ret
= NOT_DONE
;
2176 iov
.iov_base
= (void *)buf
;
2177 iov
.iov_len
= nb_sectors
* 512;
2178 qemu_iovec_init_external(&qiov
, &iov
, 1);
2179 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2180 bdrv_rw_em_cb
, &async_ret
);
2186 while (async_ret
== NOT_DONE
) {
2192 async_context_pop();
2196 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2197 const uint8_t *buf
, int nb_sectors
)
2200 BlockDriverAIOCB
*acb
;
2204 async_context_push();
2206 async_ret
= NOT_DONE
;
2207 iov
.iov_base
= (void *)buf
;
2208 iov
.iov_len
= nb_sectors
* 512;
2209 qemu_iovec_init_external(&qiov
, &iov
, 1);
2210 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2211 bdrv_rw_em_cb
, &async_ret
);
2216 while (async_ret
== NOT_DONE
) {
2221 async_context_pop();
2225 void bdrv_init(void)
2227 module_call_init(MODULE_INIT_BLOCK
);
2230 void bdrv_init_with_whitelist(void)
2232 use_bdrv_whitelist
= 1;
2236 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2237 BlockDriverCompletionFunc
*cb
, void *opaque
)
2239 BlockDriverAIOCB
*acb
;
2241 if (pool
->free_aiocb
) {
2242 acb
= pool
->free_aiocb
;
2243 pool
->free_aiocb
= acb
->next
;
2245 acb
= qemu_mallocz(pool
->aiocb_size
);
2250 acb
->opaque
= opaque
;
2254 void qemu_aio_release(void *p
)
2256 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2257 AIOPool
*pool
= acb
->pool
;
2258 acb
->next
= pool
->free_aiocb
;
2259 pool
->free_aiocb
= acb
;
2262 /**************************************************************/
2263 /* removable device support */
2266 * Return TRUE if the media is present
2268 int bdrv_is_inserted(BlockDriverState
*bs
)
2270 BlockDriver
*drv
= bs
->drv
;
2274 if (!drv
->bdrv_is_inserted
)
2276 ret
= drv
->bdrv_is_inserted(bs
);
2281 * Return TRUE if the media changed since the last call to this
2282 * function. It is currently only used for floppy disks
2284 int bdrv_media_changed(BlockDriverState
*bs
)
2286 BlockDriver
*drv
= bs
->drv
;
2289 if (!drv
|| !drv
->bdrv_media_changed
)
2292 ret
= drv
->bdrv_media_changed(bs
);
2293 if (ret
== -ENOTSUP
)
2294 ret
= bs
->media_changed
;
2295 bs
->media_changed
= 0;
2300 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2302 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2304 BlockDriver
*drv
= bs
->drv
;
2311 if (!drv
|| !drv
->bdrv_eject
) {
2314 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2316 if (ret
== -ENOTSUP
) {
2325 int bdrv_is_locked(BlockDriverState
*bs
)
2331 * Lock or unlock the media (if it is locked, the user won't be able
2332 * to eject it manually).
2334 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2336 BlockDriver
*drv
= bs
->drv
;
2338 bs
->locked
= locked
;
2339 if (drv
&& drv
->bdrv_set_locked
) {
2340 drv
->bdrv_set_locked(bs
, locked
);
2344 /* needed for generic scsi interface */
2346 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2348 BlockDriver
*drv
= bs
->drv
;
2350 if (drv
&& drv
->bdrv_ioctl
)
2351 return drv
->bdrv_ioctl(bs
, req
, buf
);
2355 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2356 unsigned long int req
, void *buf
,
2357 BlockDriverCompletionFunc
*cb
, void *opaque
)
2359 BlockDriver
*drv
= bs
->drv
;
2361 if (drv
&& drv
->bdrv_aio_ioctl
)
2362 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2368 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2370 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2373 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2375 int64_t bitmap_size
;
2377 bs
->dirty_count
= 0;
2379 if (!bs
->dirty_bitmap
) {
2380 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2381 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2382 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2384 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2387 if (bs
->dirty_bitmap
) {
2388 qemu_free(bs
->dirty_bitmap
);
2389 bs
->dirty_bitmap
= NULL
;
2394 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2396 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2398 if (bs
->dirty_bitmap
&&
2399 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2400 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2401 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2407 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2410 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2413 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2415 return bs
->dirty_count
;