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 * Common part for opening disk images and files
358 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
359 int flags
, BlockDriver
*drv
)
366 bs
->is_temporary
= 0;
369 bs
->open_flags
= flags
;
370 /* buffer_alignment defaulted to 512, drivers can change this value */
371 bs
->buffer_alignment
= 512;
373 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
375 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
380 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
383 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
384 * write cache to the guest. We do need the fdatasync to flush
385 * out transactions for block allocations, and we maybe have a
386 * volatile write cache in our backing device to deal with.
388 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
389 bs
->enable_write_cache
= 1;
392 * Clear flags that are internal to the block layer before opening the
395 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
398 * Snapshots should be writeable.
400 if (bs
->is_temporary
) {
401 open_flags
|= BDRV_O_RDWR
;
404 /* Open the image, either directly or using a protocol */
405 if (drv
->bdrv_file_open
) {
406 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
408 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
410 ret
= drv
->bdrv_open(bs
, open_flags
);
418 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
419 if (drv
->bdrv_getlength
) {
420 bs
->total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
423 if (bs
->is_temporary
) {
431 bdrv_delete(bs
->file
);
434 qemu_free(bs
->opaque
);
441 * Opens a file using a protocol (file, host_device, nbd, ...)
443 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
445 BlockDriverState
*bs
;
449 drv
= find_protocol(filename
);
455 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
466 * Opens a disk image (raw, qcow2, vmdk, ...)
468 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
473 if (flags
& BDRV_O_SNAPSHOT
) {
474 BlockDriverState
*bs1
;
477 BlockDriver
*bdrv_qcow2
;
478 QEMUOptionParameter
*options
;
479 char tmp_filename
[PATH_MAX
];
480 char backing_filename
[PATH_MAX
];
482 /* if snapshot, we create a temporary backing file and open it
483 instead of opening 'filename' directly */
485 /* if there is a backing file, use it */
487 ret
= bdrv_open(bs1
, filename
, 0, drv
);
492 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
494 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
499 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
501 /* Real path is meaningless for protocols */
503 snprintf(backing_filename
, sizeof(backing_filename
),
505 else if (!realpath(filename
, backing_filename
))
508 bdrv_qcow2
= bdrv_find_format("qcow2");
509 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
511 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
512 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
514 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
518 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
523 filename
= tmp_filename
;
525 bs
->is_temporary
= 1;
528 /* Find the right image format driver */
530 drv
= find_image_format(filename
);
535 goto unlink_and_fail
;
539 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
541 goto unlink_and_fail
;
544 /* If there is a backing file, use it */
545 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
546 char backing_filename
[PATH_MAX
];
548 BlockDriver
*back_drv
= NULL
;
550 bs
->backing_hd
= bdrv_new("");
551 path_combine(backing_filename
, sizeof(backing_filename
),
552 filename
, bs
->backing_file
);
553 if (bs
->backing_format
[0] != '\0')
554 back_drv
= bdrv_find_format(bs
->backing_format
);
556 /* backing files always opened read-only */
558 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
560 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
565 if (bs
->is_temporary
) {
566 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
568 /* base image inherits from "parent" */
569 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
573 if (!bdrv_key_required(bs
)) {
574 /* call the change callback */
575 bs
->media_changed
= 1;
577 bs
->change_cb(bs
->change_opaque
);
583 if (bs
->is_temporary
) {
589 void bdrv_close(BlockDriverState
*bs
)
593 bdrv_delete(bs
->backing_hd
);
594 bs
->drv
->bdrv_close(bs
);
595 qemu_free(bs
->opaque
);
597 if (bs
->is_temporary
) {
598 unlink(bs
->filename
);
604 if (bs
->file
!= NULL
) {
605 bdrv_close(bs
->file
);
608 /* call the change callback */
609 bs
->media_changed
= 1;
611 bs
->change_cb(bs
->change_opaque
);
615 void bdrv_delete(BlockDriverState
*bs
)
617 /* remove from list, if necessary */
618 if (bs
->device_name
[0] != '\0') {
619 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
623 if (bs
->file
!= NULL
) {
624 bdrv_delete(bs
->file
);
631 * Run consistency checks on an image
633 * Returns the number of errors or -errno when an internal error occurs
635 int bdrv_check(BlockDriverState
*bs
)
637 if (bs
->drv
->bdrv_check
== NULL
) {
641 return bs
->drv
->bdrv_check(bs
);
644 /* commit COW file into the raw image */
645 int bdrv_commit(BlockDriverState
*bs
)
647 BlockDriver
*drv
= bs
->drv
;
648 int64_t i
, total_sectors
;
649 int n
, j
, ro
, open_flags
;
650 int ret
= 0, rw_ret
= 0;
651 unsigned char sector
[512];
653 BlockDriverState
*bs_rw
, *bs_ro
;
658 if (!bs
->backing_hd
) {
662 if (bs
->backing_hd
->keep_read_only
) {
666 ro
= bs
->backing_hd
->read_only
;
667 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
668 open_flags
= bs
->backing_hd
->open_flags
;
672 bdrv_delete(bs
->backing_hd
);
673 bs
->backing_hd
= NULL
;
674 bs_rw
= bdrv_new("");
675 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, NULL
);
678 /* try to re-open read-only */
679 bs_ro
= bdrv_new("");
680 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
683 /* drive not functional anymore */
687 bs
->backing_hd
= bs_ro
;
690 bs
->backing_hd
= bs_rw
;
693 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
694 for (i
= 0; i
< total_sectors
;) {
695 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
696 for(j
= 0; j
< n
; j
++) {
697 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
702 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
713 if (drv
->bdrv_make_empty
) {
714 ret
= drv
->bdrv_make_empty(bs
);
719 * Make sure all data we wrote to the backing device is actually
723 bdrv_flush(bs
->backing_hd
);
729 bdrv_delete(bs
->backing_hd
);
730 bs
->backing_hd
= NULL
;
731 bs_ro
= bdrv_new("");
732 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
735 /* drive not functional anymore */
739 bs
->backing_hd
= bs_ro
;
740 bs
->backing_hd
->keep_read_only
= 0;
749 * -EINVAL - backing format specified, but no file
750 * -ENOSPC - can't update the backing file because no space is left in the
752 * -ENOTSUP - format driver doesn't support changing the backing file
754 int bdrv_change_backing_file(BlockDriverState
*bs
,
755 const char *backing_file
, const char *backing_fmt
)
757 BlockDriver
*drv
= bs
->drv
;
759 if (drv
->bdrv_change_backing_file
!= NULL
) {
760 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
766 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
771 if (!bdrv_is_inserted(bs
))
777 len
= bdrv_getlength(bs
);
782 if ((offset
> len
) || (len
- offset
< size
))
788 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
791 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
794 /* return < 0 if error. See bdrv_write() for the return codes */
795 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
796 uint8_t *buf
, int nb_sectors
)
798 BlockDriver
*drv
= bs
->drv
;
802 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
805 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
808 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
809 int nb_sectors
, int dirty
)
812 unsigned long val
, idx
, bit
;
814 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
815 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
817 for (; start
<= end
; start
++) {
818 idx
= start
/ (sizeof(unsigned long) * 8);
819 bit
= start
% (sizeof(unsigned long) * 8);
820 val
= bs
->dirty_bitmap
[idx
];
822 if (!(val
& (1 << bit
))) {
827 if (val
& (1 << bit
)) {
832 bs
->dirty_bitmap
[idx
] = val
;
836 /* Return < 0 if error. Important errors are:
837 -EIO generic I/O error (may happen for all errors)
838 -ENOMEDIUM No media inserted.
839 -EINVAL Invalid sector number or nb_sectors
840 -EACCES Trying to write a read-only device
842 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
843 const uint8_t *buf
, int nb_sectors
)
845 BlockDriver
*drv
= bs
->drv
;
850 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
853 if (bs
->dirty_bitmap
) {
854 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
857 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
860 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
861 void *buf
, int count1
)
863 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
864 int len
, nb_sectors
, count
;
869 /* first read to align to sector start */
870 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
873 sector_num
= offset
>> BDRV_SECTOR_BITS
;
875 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
877 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
885 /* read the sectors "in place" */
886 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
887 if (nb_sectors
> 0) {
888 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
890 sector_num
+= nb_sectors
;
891 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
896 /* add data from the last sector */
898 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
900 memcpy(buf
, tmp_buf
, count
);
905 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
906 const void *buf
, int count1
)
908 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
909 int len
, nb_sectors
, count
;
914 /* first write to align to sector start */
915 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
918 sector_num
= offset
>> BDRV_SECTOR_BITS
;
920 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
922 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
923 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
932 /* write the sectors "in place" */
933 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
934 if (nb_sectors
> 0) {
935 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
937 sector_num
+= nb_sectors
;
938 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
943 /* add data from the last sector */
945 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
947 memcpy(tmp_buf
, buf
, count
);
948 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
955 * Truncate file to 'offset' bytes (needed only for file protocols)
957 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
959 BlockDriver
*drv
= bs
->drv
;
962 if (!drv
->bdrv_truncate
)
966 return drv
->bdrv_truncate(bs
, offset
);
970 * Length of a file in bytes. Return < 0 if error or unknown.
972 int64_t bdrv_getlength(BlockDriverState
*bs
)
974 BlockDriver
*drv
= bs
->drv
;
977 if (!drv
->bdrv_getlength
) {
979 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
981 return drv
->bdrv_getlength(bs
);
984 /* return 0 as number of sectors if no device present or error */
985 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
988 length
= bdrv_getlength(bs
);
992 length
= length
>> BDRV_SECTOR_BITS
;
993 *nb_sectors_ptr
= length
;
997 uint8_t boot_ind
; /* 0x80 - active */
998 uint8_t head
; /* starting head */
999 uint8_t sector
; /* starting sector */
1000 uint8_t cyl
; /* starting cylinder */
1001 uint8_t sys_ind
; /* What partition type */
1002 uint8_t end_head
; /* end head */
1003 uint8_t end_sector
; /* end sector */
1004 uint8_t end_cyl
; /* end cylinder */
1005 uint32_t start_sect
; /* starting sector counting from 0 */
1006 uint32_t nr_sects
; /* nr of sectors in partition */
1007 } __attribute__((packed
));
1009 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1010 static int guess_disk_lchs(BlockDriverState
*bs
,
1011 int *pcylinders
, int *pheads
, int *psectors
)
1014 int ret
, i
, heads
, sectors
, cylinders
;
1015 struct partition
*p
;
1017 uint64_t nb_sectors
;
1019 bdrv_get_geometry(bs
, &nb_sectors
);
1021 ret
= bdrv_read(bs
, 0, buf
, 1);
1024 /* test msdos magic */
1025 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1027 for(i
= 0; i
< 4; i
++) {
1028 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1029 nr_sects
= le32_to_cpu(p
->nr_sects
);
1030 if (nr_sects
&& p
->end_head
) {
1031 /* We make the assumption that the partition terminates on
1032 a cylinder boundary */
1033 heads
= p
->end_head
+ 1;
1034 sectors
= p
->end_sector
& 63;
1037 cylinders
= nb_sectors
/ (heads
* sectors
);
1038 if (cylinders
< 1 || cylinders
> 16383)
1041 *psectors
= sectors
;
1042 *pcylinders
= cylinders
;
1044 printf("guessed geometry: LCHS=%d %d %d\n",
1045 cylinders
, heads
, sectors
);
1053 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1055 int translation
, lba_detected
= 0;
1056 int cylinders
, heads
, secs
;
1057 uint64_t nb_sectors
;
1059 /* if a geometry hint is available, use it */
1060 bdrv_get_geometry(bs
, &nb_sectors
);
1061 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1062 translation
= bdrv_get_translation_hint(bs
);
1063 if (cylinders
!= 0) {
1068 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1070 /* if heads > 16, it means that a BIOS LBA
1071 translation was active, so the default
1072 hardware geometry is OK */
1074 goto default_geometry
;
1079 /* disable any translation to be in sync with
1080 the logical geometry */
1081 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1082 bdrv_set_translation_hint(bs
,
1083 BIOS_ATA_TRANSLATION_NONE
);
1088 /* if no geometry, use a standard physical disk geometry */
1089 cylinders
= nb_sectors
/ (16 * 63);
1091 if (cylinders
> 16383)
1093 else if (cylinders
< 2)
1098 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1099 if ((*pcyls
* *pheads
) <= 131072) {
1100 bdrv_set_translation_hint(bs
,
1101 BIOS_ATA_TRANSLATION_LARGE
);
1103 bdrv_set_translation_hint(bs
,
1104 BIOS_ATA_TRANSLATION_LBA
);
1108 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1112 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1113 int cyls
, int heads
, int secs
)
1120 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1123 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1124 type
== BDRV_TYPE_FLOPPY
));
1127 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1129 bs
->translation
= translation
;
1132 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1133 int *pcyls
, int *pheads
, int *psecs
)
1136 *pheads
= bs
->heads
;
1140 int bdrv_get_type_hint(BlockDriverState
*bs
)
1145 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1147 return bs
->translation
;
1150 int bdrv_is_removable(BlockDriverState
*bs
)
1152 return bs
->removable
;
1155 int bdrv_is_read_only(BlockDriverState
*bs
)
1157 return bs
->read_only
;
1160 int bdrv_is_sg(BlockDriverState
*bs
)
1165 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1167 return bs
->enable_write_cache
;
1170 /* XXX: no longer used */
1171 void bdrv_set_change_cb(BlockDriverState
*bs
,
1172 void (*change_cb
)(void *opaque
), void *opaque
)
1174 bs
->change_cb
= change_cb
;
1175 bs
->change_opaque
= opaque
;
1178 int bdrv_is_encrypted(BlockDriverState
*bs
)
1180 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1182 return bs
->encrypted
;
1185 int bdrv_key_required(BlockDriverState
*bs
)
1187 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1189 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1191 return (bs
->encrypted
&& !bs
->valid_key
);
1194 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1197 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1198 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1204 if (!bs
->encrypted
) {
1206 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1209 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1212 } else if (!bs
->valid_key
) {
1214 /* call the change callback now, we skipped it on open */
1215 bs
->media_changed
= 1;
1217 bs
->change_cb(bs
->change_opaque
);
1222 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1227 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1231 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1236 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1237 it(opaque
, drv
->format_name
);
1241 BlockDriverState
*bdrv_find(const char *name
)
1243 BlockDriverState
*bs
;
1245 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1246 if (!strcmp(name
, bs
->device_name
)) {
1253 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1255 BlockDriverState
*bs
;
1257 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1262 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1264 return bs
->device_name
;
1267 void bdrv_flush(BlockDriverState
*bs
)
1269 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1270 bs
->drv
->bdrv_flush(bs
);
1273 void bdrv_flush_all(void)
1275 BlockDriverState
*bs
;
1277 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1278 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1279 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1285 int bdrv_has_zero_init(BlockDriverState
*bs
)
1289 if (bs
->drv
->no_zero_init
) {
1291 } else if (bs
->file
) {
1292 return bdrv_has_zero_init(bs
->file
);
1299 * Returns true iff the specified sector is present in the disk image. Drivers
1300 * not implementing the functionality are assumed to not support backing files,
1301 * hence all their sectors are reported as allocated.
1303 * 'pnum' is set to the number of sectors (including and immediately following
1304 * the specified sector) that are known to be in the same
1305 * allocated/unallocated state.
1307 * 'nb_sectors' is the max value 'pnum' should be set to.
1309 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1313 if (!bs
->drv
->bdrv_is_allocated
) {
1314 if (sector_num
>= bs
->total_sectors
) {
1318 n
= bs
->total_sectors
- sector_num
;
1319 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1322 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1325 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1326 BlockMonEventAction action
, int is_read
)
1329 const char *action_str
;
1332 case BDRV_ACTION_REPORT
:
1333 action_str
= "report";
1335 case BDRV_ACTION_IGNORE
:
1336 action_str
= "ignore";
1338 case BDRV_ACTION_STOP
:
1339 action_str
= "stop";
1345 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1348 is_read
? "read" : "write");
1349 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1351 qobject_decref(data
);
1354 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1357 Monitor
*mon
= opaque
;
1359 bs_dict
= qobject_to_qdict(obj
);
1361 monitor_printf(mon
, "%s: type=%s removable=%d",
1362 qdict_get_str(bs_dict
, "device"),
1363 qdict_get_str(bs_dict
, "type"),
1364 qdict_get_bool(bs_dict
, "removable"));
1366 if (qdict_get_bool(bs_dict
, "removable")) {
1367 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1370 if (qdict_haskey(bs_dict
, "inserted")) {
1371 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1373 monitor_printf(mon
, " file=");
1374 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1375 if (qdict_haskey(qdict
, "backing_file")) {
1376 monitor_printf(mon
, " backing_file=");
1377 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1379 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1380 qdict_get_bool(qdict
, "ro"),
1381 qdict_get_str(qdict
, "drv"),
1382 qdict_get_bool(qdict
, "encrypted"));
1384 monitor_printf(mon
, " [not inserted]");
1387 monitor_printf(mon
, "\n");
1390 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1392 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1396 * bdrv_info(): Block devices information
1398 * Each block device information is stored in a QDict and the
1399 * returned QObject is a QList of all devices.
1401 * The QDict contains the following:
1403 * - "device": device name
1404 * - "type": device type
1405 * - "removable": true if the device is removable, false otherwise
1406 * - "locked": true if the device is locked, false otherwise
1407 * - "inserted": only present if the device is inserted, it is a QDict
1408 * containing the following:
1409 * - "file": device file name
1410 * - "ro": true if read-only, false otherwise
1411 * - "drv": driver format name
1412 * - "backing_file": backing file name if one is used
1413 * - "encrypted": true if encrypted, false otherwise
1417 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1418 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1419 * { "device": "floppy0", "type": "floppy", "removable": true,
1420 * "locked": false } ]
1422 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1425 BlockDriverState
*bs
;
1427 bs_list
= qlist_new();
1429 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1431 const char *type
= "unknown";
1437 case BDRV_TYPE_CDROM
:
1440 case BDRV_TYPE_FLOPPY
:
1445 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1446 "'removable': %i, 'locked': %i }",
1447 bs
->device_name
, type
, bs
->removable
,
1452 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1454 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1455 "'encrypted': %i }",
1456 bs
->filename
, bs
->read_only
,
1457 bs
->drv
->format_name
,
1458 bdrv_is_encrypted(bs
));
1459 if (bs
->backing_file
[0] != '\0') {
1460 QDict
*qdict
= qobject_to_qdict(obj
);
1461 qdict_put(qdict
, "backing_file",
1462 qstring_from_str(bs
->backing_file
));
1465 qdict_put_obj(bs_dict
, "inserted", obj
);
1467 qlist_append_obj(bs_list
, bs_obj
);
1470 *ret_data
= QOBJECT(bs_list
);
1473 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1476 Monitor
*mon
= opaque
;
1478 qdict
= qobject_to_qdict(data
);
1479 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1481 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1482 monitor_printf(mon
, " rd_bytes=%" PRId64
1483 " wr_bytes=%" PRId64
1484 " rd_operations=%" PRId64
1485 " wr_operations=%" PRId64
1487 qdict_get_int(qdict
, "rd_bytes"),
1488 qdict_get_int(qdict
, "wr_bytes"),
1489 qdict_get_int(qdict
, "rd_operations"),
1490 qdict_get_int(qdict
, "wr_operations"));
1493 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1495 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1499 * bdrv_info_stats(): show block device statistics
1501 * Each device statistic information is stored in a QDict and
1502 * the returned QObject is a QList of all devices.
1504 * The QDict contains the following:
1506 * - "device": device name
1507 * - "stats": A QDict with the statistics information, it contains:
1508 * - "rd_bytes": bytes read
1509 * - "wr_bytes": bytes written
1510 * - "rd_operations": read operations
1511 * - "wr_operations": write operations
1515 * [ { "device": "ide0-hd0",
1516 * "stats": { "rd_bytes": 512,
1518 * "rd_operations": 1,
1519 * "wr_operations": 0 } },
1520 * { "device": "ide1-cd0",
1521 * "stats": { "rd_bytes": 0,
1523 * "rd_operations": 0,
1524 * "wr_operations": 0 } } ]
1526 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1530 BlockDriverState
*bs
;
1532 devices
= qlist_new();
1534 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1535 obj
= qobject_from_jsonf("{ 'device': %s, 'stats': {"
1536 "'rd_bytes': %" PRId64
","
1537 "'wr_bytes': %" PRId64
","
1538 "'rd_operations': %" PRId64
","
1539 "'wr_operations': %" PRId64
1542 bs
->rd_bytes
, bs
->wr_bytes
,
1543 bs
->rd_ops
, bs
->wr_ops
);
1544 qlist_append_obj(devices
, obj
);
1547 *ret_data
= QOBJECT(devices
);
1550 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1552 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1553 return bs
->backing_file
;
1554 else if (bs
->encrypted
)
1555 return bs
->filename
;
1560 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1561 char *filename
, int filename_size
)
1563 if (!bs
->backing_file
) {
1564 pstrcpy(filename
, filename_size
, "");
1566 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1570 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1571 const uint8_t *buf
, int nb_sectors
)
1573 BlockDriver
*drv
= bs
->drv
;
1576 if (!drv
->bdrv_write_compressed
)
1578 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1581 if (bs
->dirty_bitmap
) {
1582 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1585 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1588 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1590 BlockDriver
*drv
= bs
->drv
;
1593 if (!drv
->bdrv_get_info
)
1595 memset(bdi
, 0, sizeof(*bdi
));
1596 return drv
->bdrv_get_info(bs
, bdi
);
1599 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1600 int64_t pos
, int size
)
1602 BlockDriver
*drv
= bs
->drv
;
1605 if (!drv
->bdrv_save_vmstate
)
1607 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1610 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1611 int64_t pos
, int size
)
1613 BlockDriver
*drv
= bs
->drv
;
1616 if (!drv
->bdrv_load_vmstate
)
1618 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1621 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1623 BlockDriver
*drv
= bs
->drv
;
1625 if (!drv
|| !drv
->bdrv_debug_event
) {
1629 return drv
->bdrv_debug_event(bs
, event
);
1633 /**************************************************************/
1634 /* handling of snapshots */
1636 int bdrv_snapshot_create(BlockDriverState
*bs
,
1637 QEMUSnapshotInfo
*sn_info
)
1639 BlockDriver
*drv
= bs
->drv
;
1642 if (!drv
->bdrv_snapshot_create
)
1644 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1647 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1648 const char *snapshot_id
)
1650 BlockDriver
*drv
= bs
->drv
;
1653 if (!drv
->bdrv_snapshot_goto
)
1655 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1658 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1660 BlockDriver
*drv
= bs
->drv
;
1663 if (!drv
->bdrv_snapshot_delete
)
1665 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1668 int bdrv_snapshot_list(BlockDriverState
*bs
,
1669 QEMUSnapshotInfo
**psn_info
)
1671 BlockDriver
*drv
= bs
->drv
;
1674 if (!drv
->bdrv_snapshot_list
)
1676 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1679 #define NB_SUFFIXES 4
1681 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1683 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1688 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1691 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1692 if (size
< (10 * base
)) {
1693 snprintf(buf
, buf_size
, "%0.1f%c",
1694 (double)size
/ base
,
1697 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1698 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1699 ((size
+ (base
>> 1)) / base
),
1709 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1711 char buf1
[128], date_buf
[128], clock_buf
[128];
1721 snprintf(buf
, buf_size
,
1722 "%-10s%-20s%7s%20s%15s",
1723 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1727 ptm
= localtime(&ti
);
1728 strftime(date_buf
, sizeof(date_buf
),
1729 "%Y-%m-%d %H:%M:%S", ptm
);
1731 localtime_r(&ti
, &tm
);
1732 strftime(date_buf
, sizeof(date_buf
),
1733 "%Y-%m-%d %H:%M:%S", &tm
);
1735 secs
= sn
->vm_clock_nsec
/ 1000000000;
1736 snprintf(clock_buf
, sizeof(clock_buf
),
1737 "%02d:%02d:%02d.%03d",
1739 (int)((secs
/ 60) % 60),
1741 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1742 snprintf(buf
, buf_size
,
1743 "%-10s%-20s%7s%20s%15s",
1744 sn
->id_str
, sn
->name
,
1745 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1753 /**************************************************************/
1756 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1757 QEMUIOVector
*qiov
, int nb_sectors
,
1758 BlockDriverCompletionFunc
*cb
, void *opaque
)
1760 BlockDriver
*drv
= bs
->drv
;
1761 BlockDriverAIOCB
*ret
;
1765 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1768 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1772 /* Update stats even though technically transfer has not happened. */
1773 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1780 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1781 QEMUIOVector
*qiov
, int nb_sectors
,
1782 BlockDriverCompletionFunc
*cb
, void *opaque
)
1784 BlockDriver
*drv
= bs
->drv
;
1785 BlockDriverAIOCB
*ret
;
1791 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1794 if (bs
->dirty_bitmap
) {
1795 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1798 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1802 /* Update stats even though technically transfer has not happened. */
1803 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1811 typedef struct MultiwriteCB
{
1816 BlockDriverCompletionFunc
*cb
;
1818 QEMUIOVector
*free_qiov
;
1823 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1827 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1828 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1829 if (mcb
->callbacks
[i
].free_qiov
) {
1830 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
1832 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1833 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1837 static void multiwrite_cb(void *opaque
, int ret
)
1839 MultiwriteCB
*mcb
= opaque
;
1841 if (ret
< 0 && !mcb
->error
) {
1843 multiwrite_user_cb(mcb
);
1846 mcb
->num_requests
--;
1847 if (mcb
->num_requests
== 0) {
1848 if (mcb
->error
== 0) {
1849 multiwrite_user_cb(mcb
);
1855 static int multiwrite_req_compare(const void *a
, const void *b
)
1857 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1861 * Takes a bunch of requests and tries to merge them. Returns the number of
1862 * requests that remain after merging.
1864 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1865 int num_reqs
, MultiwriteCB
*mcb
)
1869 // Sort requests by start sector
1870 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1872 // Check if adjacent requests touch the same clusters. If so, combine them,
1873 // filling up gaps with zero sectors.
1875 for (i
= 1; i
< num_reqs
; i
++) {
1877 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1879 // This handles the cases that are valid for all block drivers, namely
1880 // exactly sequential writes and overlapping writes.
1881 if (reqs
[i
].sector
<= oldreq_last
) {
1885 // The block driver may decide that it makes sense to combine requests
1886 // even if there is a gap of some sectors between them. In this case,
1887 // the gap is filled with zeros (therefore only applicable for yet
1888 // unused space in format like qcow2).
1889 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1890 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1893 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
1899 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1900 qemu_iovec_init(qiov
,
1901 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1903 // Add the first request to the merged one. If the requests are
1904 // overlapping, drop the last sectors of the first request.
1905 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1906 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1908 // We might need to add some zeros between the two requests
1909 if (reqs
[i
].sector
> oldreq_last
) {
1910 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1911 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1912 memset(buf
, 0, zero_bytes
);
1913 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1914 mcb
->callbacks
[i
].free_buf
= buf
;
1917 // Add the second request
1918 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
1920 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
1921 reqs
[outidx
].qiov
= qiov
;
1923 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
1926 reqs
[outidx
].sector
= reqs
[i
].sector
;
1927 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
1928 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
1936 * Submit multiple AIO write requests at once.
1938 * On success, the function returns 0 and all requests in the reqs array have
1939 * been submitted. In error case this function returns -1, and any of the
1940 * requests may or may not be submitted yet. In particular, this means that the
1941 * callback will be called for some of the requests, for others it won't. The
1942 * caller must check the error field of the BlockRequest to wait for the right
1943 * callbacks (if error != 0, no callback will be called).
1945 * The implementation may modify the contents of the reqs array, e.g. to merge
1946 * requests. However, the fields opaque and error are left unmodified as they
1947 * are used to signal failure for a single request to the caller.
1949 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
1951 BlockDriverAIOCB
*acb
;
1955 if (num_reqs
== 0) {
1959 // Create MultiwriteCB structure
1960 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
1961 mcb
->num_requests
= 0;
1962 mcb
->num_callbacks
= num_reqs
;
1964 for (i
= 0; i
< num_reqs
; i
++) {
1965 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
1966 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
1969 // Check for mergable requests
1970 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
1972 // Run the aio requests
1973 for (i
= 0; i
< num_reqs
; i
++) {
1974 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
1975 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
1978 // We can only fail the whole thing if no request has been
1979 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1980 // complete and report the error in the callback.
1981 if (mcb
->num_requests
== 0) {
1982 reqs
[i
].error
= -EIO
;
1985 mcb
->num_requests
++;
1986 multiwrite_cb(mcb
, -EIO
);
1990 mcb
->num_requests
++;
2001 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2002 BlockDriverCompletionFunc
*cb
, void *opaque
)
2004 BlockDriver
*drv
= bs
->drv
;
2008 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2011 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2013 acb
->pool
->cancel(acb
);
2017 /**************************************************************/
2018 /* async block device emulation */
2020 typedef struct BlockDriverAIOCBSync
{
2021 BlockDriverAIOCB common
;
2024 /* vector translation state */
2028 } BlockDriverAIOCBSync
;
2030 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2032 BlockDriverAIOCBSync
*acb
= (BlockDriverAIOCBSync
*)blockacb
;
2033 qemu_bh_delete(acb
->bh
);
2035 qemu_aio_release(acb
);
2038 static AIOPool bdrv_em_aio_pool
= {
2039 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2040 .cancel
= bdrv_aio_cancel_em
,
2043 static void bdrv_aio_bh_cb(void *opaque
)
2045 BlockDriverAIOCBSync
*acb
= opaque
;
2048 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2049 qemu_vfree(acb
->bounce
);
2050 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2051 qemu_bh_delete(acb
->bh
);
2053 qemu_aio_release(acb
);
2056 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2060 BlockDriverCompletionFunc
*cb
,
2065 BlockDriverAIOCBSync
*acb
;
2067 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2068 acb
->is_write
= is_write
;
2070 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2073 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2076 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2077 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2079 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2082 qemu_bh_schedule(acb
->bh
);
2084 return &acb
->common
;
2087 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2088 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2089 BlockDriverCompletionFunc
*cb
, void *opaque
)
2091 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2094 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2095 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2096 BlockDriverCompletionFunc
*cb
, void *opaque
)
2098 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2101 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2102 BlockDriverCompletionFunc
*cb
, void *opaque
)
2104 BlockDriverAIOCBSync
*acb
;
2106 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2107 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2113 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2116 qemu_bh_schedule(acb
->bh
);
2117 return &acb
->common
;
2120 /**************************************************************/
2121 /* sync block device emulation */
2123 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2125 *(int *)opaque
= ret
;
2128 #define NOT_DONE 0x7fffffff
2130 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2131 uint8_t *buf
, int nb_sectors
)
2134 BlockDriverAIOCB
*acb
;
2138 async_context_push();
2140 async_ret
= NOT_DONE
;
2141 iov
.iov_base
= (void *)buf
;
2142 iov
.iov_len
= nb_sectors
* 512;
2143 qemu_iovec_init_external(&qiov
, &iov
, 1);
2144 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2145 bdrv_rw_em_cb
, &async_ret
);
2151 while (async_ret
== NOT_DONE
) {
2157 async_context_pop();
2161 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2162 const uint8_t *buf
, int nb_sectors
)
2165 BlockDriverAIOCB
*acb
;
2169 async_context_push();
2171 async_ret
= NOT_DONE
;
2172 iov
.iov_base
= (void *)buf
;
2173 iov
.iov_len
= nb_sectors
* 512;
2174 qemu_iovec_init_external(&qiov
, &iov
, 1);
2175 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2176 bdrv_rw_em_cb
, &async_ret
);
2181 while (async_ret
== NOT_DONE
) {
2186 async_context_pop();
2190 void bdrv_init(void)
2192 module_call_init(MODULE_INIT_BLOCK
);
2195 void bdrv_init_with_whitelist(void)
2197 use_bdrv_whitelist
= 1;
2201 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2202 BlockDriverCompletionFunc
*cb
, void *opaque
)
2204 BlockDriverAIOCB
*acb
;
2206 if (pool
->free_aiocb
) {
2207 acb
= pool
->free_aiocb
;
2208 pool
->free_aiocb
= acb
->next
;
2210 acb
= qemu_mallocz(pool
->aiocb_size
);
2215 acb
->opaque
= opaque
;
2219 void qemu_aio_release(void *p
)
2221 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2222 AIOPool
*pool
= acb
->pool
;
2223 acb
->next
= pool
->free_aiocb
;
2224 pool
->free_aiocb
= acb
;
2227 /**************************************************************/
2228 /* removable device support */
2231 * Return TRUE if the media is present
2233 int bdrv_is_inserted(BlockDriverState
*bs
)
2235 BlockDriver
*drv
= bs
->drv
;
2239 if (!drv
->bdrv_is_inserted
)
2241 ret
= drv
->bdrv_is_inserted(bs
);
2246 * Return TRUE if the media changed since the last call to this
2247 * function. It is currently only used for floppy disks
2249 int bdrv_media_changed(BlockDriverState
*bs
)
2251 BlockDriver
*drv
= bs
->drv
;
2254 if (!drv
|| !drv
->bdrv_media_changed
)
2257 ret
= drv
->bdrv_media_changed(bs
);
2258 if (ret
== -ENOTSUP
)
2259 ret
= bs
->media_changed
;
2260 bs
->media_changed
= 0;
2265 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2267 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2269 BlockDriver
*drv
= bs
->drv
;
2276 if (!drv
|| !drv
->bdrv_eject
) {
2279 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2281 if (ret
== -ENOTSUP
) {
2290 int bdrv_is_locked(BlockDriverState
*bs
)
2296 * Lock or unlock the media (if it is locked, the user won't be able
2297 * to eject it manually).
2299 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2301 BlockDriver
*drv
= bs
->drv
;
2303 bs
->locked
= locked
;
2304 if (drv
&& drv
->bdrv_set_locked
) {
2305 drv
->bdrv_set_locked(bs
, locked
);
2309 /* needed for generic scsi interface */
2311 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2313 BlockDriver
*drv
= bs
->drv
;
2315 if (drv
&& drv
->bdrv_ioctl
)
2316 return drv
->bdrv_ioctl(bs
, req
, buf
);
2320 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2321 unsigned long int req
, void *buf
,
2322 BlockDriverCompletionFunc
*cb
, void *opaque
)
2324 BlockDriver
*drv
= bs
->drv
;
2326 if (drv
&& drv
->bdrv_aio_ioctl
)
2327 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2333 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2335 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2338 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2340 int64_t bitmap_size
;
2342 bs
->dirty_count
= 0;
2344 if (!bs
->dirty_bitmap
) {
2345 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2346 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2347 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2349 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2352 if (bs
->dirty_bitmap
) {
2353 qemu_free(bs
->dirty_bitmap
);
2354 bs
->dirty_bitmap
= NULL
;
2359 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2361 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2363 if (bs
->dirty_bitmap
&&
2364 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2365 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2366 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2372 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2375 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2378 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2380 return bs
->dirty_count
;