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
);
58 BlockDriverState
*bdrv_first
;
60 static BlockDriver
*first_drv
;
62 /* If non-zero, use only whitelisted block drivers */
63 static int use_bdrv_whitelist
;
65 int path_is_absolute(const char *path
)
69 /* specific case for names like: "\\.\d:" */
70 if (*path
== '/' || *path
== '\\')
73 p
= strchr(path
, ':');
79 return (*p
== '/' || *p
== '\\');
85 /* if filename is absolute, just copy it to dest. Otherwise, build a
86 path to it by considering it is relative to base_path. URL are
88 void path_combine(char *dest
, int dest_size
,
89 const char *base_path
,
97 if (path_is_absolute(filename
)) {
98 pstrcpy(dest
, dest_size
, filename
);
100 p
= strchr(base_path
, ':');
105 p1
= strrchr(base_path
, '/');
109 p2
= strrchr(base_path
, '\\');
121 if (len
> dest_size
- 1)
123 memcpy(dest
, base_path
, len
);
125 pstrcat(dest
, dest_size
, filename
);
129 void bdrv_register(BlockDriver
*bdrv
)
131 if (!bdrv
->bdrv_aio_readv
) {
132 /* add AIO emulation layer */
133 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
134 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
135 } else if (!bdrv
->bdrv_read
) {
136 /* add synchronous IO emulation layer */
137 bdrv
->bdrv_read
= bdrv_read_em
;
138 bdrv
->bdrv_write
= bdrv_write_em
;
141 if (!bdrv
->bdrv_aio_flush
)
142 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
144 bdrv
->next
= first_drv
;
148 /* create a new block device (by default it is empty) */
149 BlockDriverState
*bdrv_new(const char *device_name
)
151 BlockDriverState
**pbs
, *bs
;
153 bs
= qemu_mallocz(sizeof(BlockDriverState
));
154 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
155 if (device_name
[0] != '\0') {
156 /* insert at the end */
165 BlockDriver
*bdrv_find_format(const char *format_name
)
168 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
169 if (!strcmp(drv1
->format_name
, format_name
))
175 static int bdrv_is_whitelisted(BlockDriver
*drv
)
177 static const char *whitelist
[] = {
178 CONFIG_BDRV_WHITELIST
183 return 1; /* no whitelist, anything goes */
185 for (p
= whitelist
; *p
; p
++) {
186 if (!strcmp(drv
->format_name
, *p
)) {
193 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
195 BlockDriver
*drv
= bdrv_find_format(format_name
);
196 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
199 int bdrv_create(BlockDriver
*drv
, const char* filename
,
200 QEMUOptionParameter
*options
)
202 if (!drv
->bdrv_create
)
205 return drv
->bdrv_create(filename
, options
);
209 void get_tmp_filename(char *filename
, int size
)
211 char temp_dir
[MAX_PATH
];
213 GetTempPath(MAX_PATH
, temp_dir
);
214 GetTempFileName(temp_dir
, "qem", 0, filename
);
217 void get_tmp_filename(char *filename
, int size
)
221 /* XXX: race condition possible */
222 tmpdir
= getenv("TMPDIR");
225 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
226 fd
= mkstemp(filename
);
232 static int is_windows_drive_prefix(const char *filename
)
234 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
235 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
239 int is_windows_drive(const char *filename
)
241 if (is_windows_drive_prefix(filename
) &&
244 if (strstart(filename
, "\\\\.\\", NULL
) ||
245 strstart(filename
, "//./", NULL
))
251 static BlockDriver
*find_protocol(const char *filename
)
259 if (is_windows_drive(filename
) ||
260 is_windows_drive_prefix(filename
))
261 return bdrv_find_format("raw");
263 p
= strchr(filename
, ':');
265 return bdrv_find_format("raw");
267 if (len
> sizeof(protocol
) - 1)
268 len
= sizeof(protocol
) - 1;
269 memcpy(protocol
, filename
, len
);
270 protocol
[len
] = '\0';
271 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
272 if (drv1
->protocol_name
&&
273 !strcmp(drv1
->protocol_name
, protocol
))
280 * Detect host devices. By convention, /dev/cdrom[N] is always
281 * recognized as a host CDROM.
283 static BlockDriver
*find_hdev_driver(const char *filename
)
285 int score_max
= 0, score
;
286 BlockDriver
*drv
= NULL
, *d
;
288 for (d
= first_drv
; d
; d
= d
->next
) {
289 if (d
->bdrv_probe_device
) {
290 score
= d
->bdrv_probe_device(filename
);
291 if (score
> score_max
) {
301 static BlockDriver
*find_image_format(const char *filename
)
303 int ret
, score
, score_max
;
304 BlockDriver
*drv1
, *drv
;
306 BlockDriverState
*bs
;
308 drv
= find_protocol(filename
);
309 /* no need to test disk image formats for vvfat */
310 if (drv
&& strcmp(drv
->format_name
, "vvfat") == 0)
313 ret
= bdrv_file_open(&bs
, filename
, 0);
316 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
323 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
324 if (drv1
->bdrv_probe
) {
325 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
326 if (score
> score_max
) {
335 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
337 BlockDriverState
*bs
;
341 ret
= bdrv_open2(bs
, filename
, flags
| BDRV_O_FILE
, NULL
);
351 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
)
353 return bdrv_open2(bs
, filename
, flags
, NULL
);
356 int bdrv_open2(BlockDriverState
*bs
, const char *filename
, int flags
,
360 char tmp_filename
[PATH_MAX
];
361 char backing_filename
[PATH_MAX
];
363 bs
->is_temporary
= 0;
366 /* buffer_alignment defaulted to 512, drivers can change this value */
367 bs
->buffer_alignment
= 512;
369 if (flags
& BDRV_O_SNAPSHOT
) {
370 BlockDriverState
*bs1
;
373 BlockDriver
*bdrv_qcow2
;
374 QEMUOptionParameter
*options
;
376 /* if snapshot, we create a temporary backing file and open it
377 instead of opening 'filename' directly */
379 /* if there is a backing file, use it */
381 ret
= bdrv_open2(bs1
, filename
, 0, drv
);
386 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
388 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
393 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
395 /* Real path is meaningless for protocols */
397 snprintf(backing_filename
, sizeof(backing_filename
),
399 else if (!realpath(filename
, backing_filename
))
402 bdrv_qcow2
= bdrv_find_format("qcow2");
403 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
405 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
406 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
408 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
412 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
417 filename
= tmp_filename
;
419 bs
->is_temporary
= 1;
422 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
423 if (flags
& BDRV_O_FILE
) {
424 drv
= find_protocol(filename
);
426 drv
= find_hdev_driver(filename
);
428 drv
= find_image_format(filename
);
433 goto unlink_and_fail
;
436 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
439 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
440 * write cache to the guest. We do need the fdatasync to flush
441 * out transactions for block allocations, and we maybe have a
442 * volatile write cache in our backing device to deal with.
444 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
445 bs
->enable_write_cache
= 1;
447 bs
->read_only
= (flags
& BDRV_O_RDWR
) == 0;
448 if (!(flags
& BDRV_O_FILE
)) {
449 open_flags
= (flags
& (BDRV_O_RDWR
| BDRV_O_CACHE_MASK
|BDRV_O_NATIVE_AIO
));
450 if (bs
->is_temporary
) { /* snapshot should be writeable */
451 open_flags
|= BDRV_O_RDWR
;
454 open_flags
= flags
& ~(BDRV_O_FILE
| BDRV_O_SNAPSHOT
);
456 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
459 ret
= drv
->bdrv_open(bs
, filename
, open_flags
);
462 qemu_free(bs
->opaque
);
466 if (bs
->is_temporary
)
470 if (drv
->bdrv_getlength
) {
471 bs
->total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
474 if (bs
->is_temporary
) {
478 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
479 /* if there is a backing file, use it */
480 BlockDriver
*back_drv
= NULL
;
481 bs
->backing_hd
= bdrv_new("");
482 path_combine(backing_filename
, sizeof(backing_filename
),
483 filename
, bs
->backing_file
);
484 if (bs
->backing_format
[0] != '\0')
485 back_drv
= bdrv_find_format(bs
->backing_format
);
486 ret
= bdrv_open2(bs
->backing_hd
, backing_filename
, open_flags
,
488 bs
->backing_hd
->read_only
= (open_flags
& BDRV_O_RDWR
) == 0;
495 if (!bdrv_key_required(bs
)) {
496 /* call the change callback */
497 bs
->media_changed
= 1;
499 bs
->change_cb(bs
->change_opaque
);
504 void bdrv_close(BlockDriverState
*bs
)
508 bdrv_delete(bs
->backing_hd
);
509 bs
->drv
->bdrv_close(bs
);
510 qemu_free(bs
->opaque
);
512 if (bs
->is_temporary
) {
513 unlink(bs
->filename
);
519 /* call the change callback */
520 bs
->media_changed
= 1;
522 bs
->change_cb(bs
->change_opaque
);
526 void bdrv_delete(BlockDriverState
*bs
)
528 BlockDriverState
**pbs
;
531 while (*pbs
!= bs
&& *pbs
!= NULL
)
541 * Run consistency checks on an image
543 * Returns the number of errors or -errno when an internal error occurs
545 int bdrv_check(BlockDriverState
*bs
)
547 if (bs
->drv
->bdrv_check
== NULL
) {
551 return bs
->drv
->bdrv_check(bs
);
554 /* commit COW file into the raw image */
555 int bdrv_commit(BlockDriverState
*bs
)
557 BlockDriver
*drv
= bs
->drv
;
558 int64_t i
, total_sectors
;
561 unsigned char sector
[512];
570 if (!bs
->backing_hd
) {
574 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
575 for (i
= 0; i
< total_sectors
;) {
576 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
577 for(j
= 0; j
< n
; j
++) {
578 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
582 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
592 if (drv
->bdrv_make_empty
) {
593 ret
= drv
->bdrv_make_empty(bs
);
598 * Make sure all data we wrote to the backing device is actually
602 bdrv_flush(bs
->backing_hd
);
609 * -EINVAL - backing format specified, but no file
610 * -ENOSPC - can't update the backing file because no space is left in the
612 * -ENOTSUP - format driver doesn't support changing the backing file
614 int bdrv_change_backing_file(BlockDriverState
*bs
,
615 const char *backing_file
, const char *backing_fmt
)
617 BlockDriver
*drv
= bs
->drv
;
619 if (drv
->bdrv_change_backing_file
!= NULL
) {
620 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
626 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
631 if (!bdrv_is_inserted(bs
))
637 len
= bdrv_getlength(bs
);
642 if ((offset
> len
) || (len
- offset
< size
))
648 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
651 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
654 /* return < 0 if error. See bdrv_write() for the return codes */
655 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
656 uint8_t *buf
, int nb_sectors
)
658 BlockDriver
*drv
= bs
->drv
;
662 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
665 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
668 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
669 int nb_sectors
, int dirty
)
672 unsigned long val
, idx
, bit
;
674 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
675 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
677 for (; start
<= end
; start
++) {
678 idx
= start
/ (sizeof(unsigned long) * 8);
679 bit
= start
% (sizeof(unsigned long) * 8);
680 val
= bs
->dirty_bitmap
[idx
];
686 bs
->dirty_bitmap
[idx
] = val
;
690 /* Return < 0 if error. Important errors are:
691 -EIO generic I/O error (may happen for all errors)
692 -ENOMEDIUM No media inserted.
693 -EINVAL Invalid sector number or nb_sectors
694 -EACCES Trying to write a read-only device
696 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
697 const uint8_t *buf
, int nb_sectors
)
699 BlockDriver
*drv
= bs
->drv
;
704 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
707 if (bs
->dirty_bitmap
) {
708 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
711 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
714 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
715 void *buf
, int count1
)
717 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
718 int len
, nb_sectors
, count
;
723 /* first read to align to sector start */
724 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
727 sector_num
= offset
>> BDRV_SECTOR_BITS
;
729 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
731 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
739 /* read the sectors "in place" */
740 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
741 if (nb_sectors
> 0) {
742 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
744 sector_num
+= nb_sectors
;
745 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
750 /* add data from the last sector */
752 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
754 memcpy(buf
, tmp_buf
, count
);
759 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
760 const void *buf
, int count1
)
762 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
763 int len
, nb_sectors
, count
;
768 /* first write to align to sector start */
769 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
772 sector_num
= offset
>> BDRV_SECTOR_BITS
;
774 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
776 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
777 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
786 /* write the sectors "in place" */
787 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
788 if (nb_sectors
> 0) {
789 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
791 sector_num
+= nb_sectors
;
792 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
797 /* add data from the last sector */
799 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
801 memcpy(tmp_buf
, buf
, count
);
802 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
809 * Truncate file to 'offset' bytes (needed only for file protocols)
811 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
813 BlockDriver
*drv
= bs
->drv
;
816 if (!drv
->bdrv_truncate
)
820 return drv
->bdrv_truncate(bs
, offset
);
824 * Length of a file in bytes. Return < 0 if error or unknown.
826 int64_t bdrv_getlength(BlockDriverState
*bs
)
828 BlockDriver
*drv
= bs
->drv
;
831 if (!drv
->bdrv_getlength
) {
833 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
835 return drv
->bdrv_getlength(bs
);
838 /* return 0 as number of sectors if no device present or error */
839 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
842 length
= bdrv_getlength(bs
);
846 length
= length
>> BDRV_SECTOR_BITS
;
847 *nb_sectors_ptr
= length
;
851 uint8_t boot_ind
; /* 0x80 - active */
852 uint8_t head
; /* starting head */
853 uint8_t sector
; /* starting sector */
854 uint8_t cyl
; /* starting cylinder */
855 uint8_t sys_ind
; /* What partition type */
856 uint8_t end_head
; /* end head */
857 uint8_t end_sector
; /* end sector */
858 uint8_t end_cyl
; /* end cylinder */
859 uint32_t start_sect
; /* starting sector counting from 0 */
860 uint32_t nr_sects
; /* nr of sectors in partition */
861 } __attribute__((packed
));
863 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
864 static int guess_disk_lchs(BlockDriverState
*bs
,
865 int *pcylinders
, int *pheads
, int *psectors
)
868 int ret
, i
, heads
, sectors
, cylinders
;
873 bdrv_get_geometry(bs
, &nb_sectors
);
875 ret
= bdrv_read(bs
, 0, buf
, 1);
878 /* test msdos magic */
879 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
881 for(i
= 0; i
< 4; i
++) {
882 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
883 nr_sects
= le32_to_cpu(p
->nr_sects
);
884 if (nr_sects
&& p
->end_head
) {
885 /* We make the assumption that the partition terminates on
886 a cylinder boundary */
887 heads
= p
->end_head
+ 1;
888 sectors
= p
->end_sector
& 63;
891 cylinders
= nb_sectors
/ (heads
* sectors
);
892 if (cylinders
< 1 || cylinders
> 16383)
896 *pcylinders
= cylinders
;
898 printf("guessed geometry: LCHS=%d %d %d\n",
899 cylinders
, heads
, sectors
);
907 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
909 int translation
, lba_detected
= 0;
910 int cylinders
, heads
, secs
;
913 /* if a geometry hint is available, use it */
914 bdrv_get_geometry(bs
, &nb_sectors
);
915 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
916 translation
= bdrv_get_translation_hint(bs
);
917 if (cylinders
!= 0) {
922 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
924 /* if heads > 16, it means that a BIOS LBA
925 translation was active, so the default
926 hardware geometry is OK */
928 goto default_geometry
;
933 /* disable any translation to be in sync with
934 the logical geometry */
935 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
936 bdrv_set_translation_hint(bs
,
937 BIOS_ATA_TRANSLATION_NONE
);
942 /* if no geometry, use a standard physical disk geometry */
943 cylinders
= nb_sectors
/ (16 * 63);
945 if (cylinders
> 16383)
947 else if (cylinders
< 2)
952 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
953 if ((*pcyls
* *pheads
) <= 131072) {
954 bdrv_set_translation_hint(bs
,
955 BIOS_ATA_TRANSLATION_LARGE
);
957 bdrv_set_translation_hint(bs
,
958 BIOS_ATA_TRANSLATION_LBA
);
962 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
966 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
967 int cyls
, int heads
, int secs
)
974 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
977 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
978 type
== BDRV_TYPE_FLOPPY
));
981 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
983 bs
->translation
= translation
;
986 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
987 int *pcyls
, int *pheads
, int *psecs
)
994 int bdrv_get_type_hint(BlockDriverState
*bs
)
999 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1001 return bs
->translation
;
1004 int bdrv_is_removable(BlockDriverState
*bs
)
1006 return bs
->removable
;
1009 int bdrv_is_read_only(BlockDriverState
*bs
)
1011 return bs
->read_only
;
1014 int bdrv_is_sg(BlockDriverState
*bs
)
1019 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1021 return bs
->enable_write_cache
;
1024 /* XXX: no longer used */
1025 void bdrv_set_change_cb(BlockDriverState
*bs
,
1026 void (*change_cb
)(void *opaque
), void *opaque
)
1028 bs
->change_cb
= change_cb
;
1029 bs
->change_opaque
= opaque
;
1032 int bdrv_is_encrypted(BlockDriverState
*bs
)
1034 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1036 return bs
->encrypted
;
1039 int bdrv_key_required(BlockDriverState
*bs
)
1041 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1043 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1045 return (bs
->encrypted
&& !bs
->valid_key
);
1048 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1051 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1052 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1058 if (!bs
->encrypted
|| !bs
->drv
|| !bs
->drv
->bdrv_set_key
)
1060 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1063 } else if (!bs
->valid_key
) {
1065 /* call the change callback now, we skipped it on open */
1066 bs
->media_changed
= 1;
1068 bs
->change_cb(bs
->change_opaque
);
1073 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1078 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1082 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1087 for (drv
= first_drv
; drv
!= NULL
; drv
= drv
->next
) {
1088 it(opaque
, drv
->format_name
);
1092 BlockDriverState
*bdrv_find(const char *name
)
1094 BlockDriverState
*bs
;
1096 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1097 if (!strcmp(name
, bs
->device_name
))
1103 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1105 BlockDriverState
*bs
;
1107 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1112 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1114 return bs
->device_name
;
1117 void bdrv_flush(BlockDriverState
*bs
)
1119 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1120 bs
->drv
->bdrv_flush(bs
);
1123 void bdrv_flush_all(void)
1125 BlockDriverState
*bs
;
1127 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
)
1128 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1129 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
)))
1134 * Returns true iff the specified sector is present in the disk image. Drivers
1135 * not implementing the functionality are assumed to not support backing files,
1136 * hence all their sectors are reported as allocated.
1138 * 'pnum' is set to the number of sectors (including and immediately following
1139 * the specified sector) that are known to be in the same
1140 * allocated/unallocated state.
1142 * 'nb_sectors' is the max value 'pnum' should be set to.
1144 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1148 if (!bs
->drv
->bdrv_is_allocated
) {
1149 if (sector_num
>= bs
->total_sectors
) {
1153 n
= bs
->total_sectors
- sector_num
;
1154 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1157 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1160 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1163 Monitor
*mon
= opaque
;
1165 bs_dict
= qobject_to_qdict(obj
);
1167 monitor_printf(mon
, "%s: type=%s removable=%d",
1168 qdict_get_str(bs_dict
, "device"),
1169 qdict_get_str(bs_dict
, "type"),
1170 qdict_get_bool(bs_dict
, "removable"));
1172 if (qdict_get_bool(bs_dict
, "removable")) {
1173 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1176 if (qdict_haskey(bs_dict
, "inserted")) {
1177 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1179 monitor_printf(mon
, " file=");
1180 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1181 if (qdict_haskey(qdict
, "backing_file")) {
1182 monitor_printf(mon
, " backing_file=");
1183 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1185 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1186 qdict_get_bool(qdict
, "ro"),
1187 qdict_get_str(qdict
, "drv"),
1188 qdict_get_bool(qdict
, "encrypted"));
1190 monitor_printf(mon
, " [not inserted]");
1193 monitor_printf(mon
, "\n");
1196 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1198 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1202 * bdrv_info(): Block devices information
1204 * Each block device information is stored in a QDict and the
1205 * returned QObject is a QList of all devices.
1207 * The QDict contains the following:
1209 * - "device": device name
1210 * - "type": device type
1211 * - "removable": true if the device is removable, false otherwise
1212 * - "locked": true if the device is locked, false otherwise
1213 * - "inserted": only present if the device is inserted, it is a QDict
1214 * containing the following:
1215 * - "file": device file name
1216 * - "ro": true if read-only, false otherwise
1217 * - "drv": driver format name
1218 * - "backing_file": backing file name if one is used
1219 * - "encrypted": true if encrypted, false otherwise
1223 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1224 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1225 * { "device": "floppy0", "type": "floppy", "removable": true,
1226 * "locked": false } ]
1228 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1231 BlockDriverState
*bs
;
1233 bs_list
= qlist_new();
1235 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1237 const char *type
= "unknown";
1243 case BDRV_TYPE_CDROM
:
1246 case BDRV_TYPE_FLOPPY
:
1251 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1252 "'removable': %i, 'locked': %i }",
1253 bs
->device_name
, type
, bs
->removable
,
1255 assert(bs_obj
!= NULL
);
1259 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1261 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1262 "'encrypted': %i }",
1263 bs
->filename
, bs
->read_only
,
1264 bs
->drv
->format_name
,
1265 bdrv_is_encrypted(bs
));
1266 assert(obj
!= NULL
);
1267 if (bs
->backing_file
[0] != '\0') {
1268 QDict
*qdict
= qobject_to_qdict(obj
);
1269 qdict_put(qdict
, "backing_file",
1270 qstring_from_str(bs
->backing_file
));
1273 qdict_put_obj(bs_dict
, "inserted", obj
);
1275 qlist_append_obj(bs_list
, bs_obj
);
1278 *ret_data
= QOBJECT(bs_list
);
1281 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1284 Monitor
*mon
= opaque
;
1286 qdict
= qobject_to_qdict(data
);
1287 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1289 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1290 monitor_printf(mon
, " rd_bytes=%" PRId64
1291 " wr_bytes=%" PRId64
1292 " rd_operations=%" PRId64
1293 " wr_operations=%" PRId64
1295 qdict_get_int(qdict
, "rd_bytes"),
1296 qdict_get_int(qdict
, "wr_bytes"),
1297 qdict_get_int(qdict
, "rd_operations"),
1298 qdict_get_int(qdict
, "wr_operations"));
1301 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1303 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1307 * bdrv_info_stats(): show block device statistics
1309 * Each device statistic information is stored in a QDict and
1310 * the returned QObject is a QList of all devices.
1312 * The QDict contains the following:
1314 * - "device": device name
1315 * - "stats": A QDict with the statistics information, it contains:
1316 * - "rd_bytes": bytes read
1317 * - "wr_bytes": bytes written
1318 * - "rd_operations": read operations
1319 * - "wr_operations": write operations
1323 * [ { "device": "ide0-hd0",
1324 * "stats": { "rd_bytes": 512,
1326 * "rd_operations": 1,
1327 * "wr_operations": 0 } },
1328 * { "device": "ide1-cd0",
1329 * "stats": { "rd_bytes": 0,
1331 * "rd_operations": 0,
1332 * "wr_operations": 0 } } ]
1334 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1338 BlockDriverState
*bs
;
1340 devices
= qlist_new();
1342 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1343 obj
= qobject_from_jsonf("{ 'device': %s, 'stats': {"
1344 "'rd_bytes': %" PRId64
","
1345 "'wr_bytes': %" PRId64
","
1346 "'rd_operations': %" PRId64
","
1347 "'wr_operations': %" PRId64
1350 bs
->rd_bytes
, bs
->wr_bytes
,
1351 bs
->rd_ops
, bs
->wr_ops
);
1352 assert(obj
!= NULL
);
1353 qlist_append_obj(devices
, obj
);
1356 *ret_data
= QOBJECT(devices
);
1359 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1361 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1362 return bs
->backing_file
;
1363 else if (bs
->encrypted
)
1364 return bs
->filename
;
1369 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1370 char *filename
, int filename_size
)
1372 if (!bs
->backing_file
) {
1373 pstrcpy(filename
, filename_size
, "");
1375 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1379 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1380 const uint8_t *buf
, int nb_sectors
)
1382 BlockDriver
*drv
= bs
->drv
;
1385 if (!drv
->bdrv_write_compressed
)
1387 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1390 if (bs
->dirty_bitmap
) {
1391 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1394 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1397 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1399 BlockDriver
*drv
= bs
->drv
;
1402 if (!drv
->bdrv_get_info
)
1404 memset(bdi
, 0, sizeof(*bdi
));
1405 return drv
->bdrv_get_info(bs
, bdi
);
1408 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1409 int64_t pos
, int size
)
1411 BlockDriver
*drv
= bs
->drv
;
1414 if (!drv
->bdrv_save_vmstate
)
1416 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1419 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1420 int64_t pos
, int size
)
1422 BlockDriver
*drv
= bs
->drv
;
1425 if (!drv
->bdrv_load_vmstate
)
1427 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1430 /**************************************************************/
1431 /* handling of snapshots */
1433 int bdrv_snapshot_create(BlockDriverState
*bs
,
1434 QEMUSnapshotInfo
*sn_info
)
1436 BlockDriver
*drv
= bs
->drv
;
1439 if (!drv
->bdrv_snapshot_create
)
1441 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1444 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1445 const char *snapshot_id
)
1447 BlockDriver
*drv
= bs
->drv
;
1450 if (!drv
->bdrv_snapshot_goto
)
1452 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1455 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1457 BlockDriver
*drv
= bs
->drv
;
1460 if (!drv
->bdrv_snapshot_delete
)
1462 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1465 int bdrv_snapshot_list(BlockDriverState
*bs
,
1466 QEMUSnapshotInfo
**psn_info
)
1468 BlockDriver
*drv
= bs
->drv
;
1471 if (!drv
->bdrv_snapshot_list
)
1473 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1476 #define NB_SUFFIXES 4
1478 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1480 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1485 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1488 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1489 if (size
< (10 * base
)) {
1490 snprintf(buf
, buf_size
, "%0.1f%c",
1491 (double)size
/ base
,
1494 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1495 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1496 ((size
+ (base
>> 1)) / base
),
1506 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1508 char buf1
[128], date_buf
[128], clock_buf
[128];
1518 snprintf(buf
, buf_size
,
1519 "%-10s%-20s%7s%20s%15s",
1520 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1524 ptm
= localtime(&ti
);
1525 strftime(date_buf
, sizeof(date_buf
),
1526 "%Y-%m-%d %H:%M:%S", ptm
);
1528 localtime_r(&ti
, &tm
);
1529 strftime(date_buf
, sizeof(date_buf
),
1530 "%Y-%m-%d %H:%M:%S", &tm
);
1532 secs
= sn
->vm_clock_nsec
/ 1000000000;
1533 snprintf(clock_buf
, sizeof(clock_buf
),
1534 "%02d:%02d:%02d.%03d",
1536 (int)((secs
/ 60) % 60),
1538 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1539 snprintf(buf
, buf_size
,
1540 "%-10s%-20s%7s%20s%15s",
1541 sn
->id_str
, sn
->name
,
1542 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1550 /**************************************************************/
1553 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1554 QEMUIOVector
*qiov
, int nb_sectors
,
1555 BlockDriverCompletionFunc
*cb
, void *opaque
)
1557 BlockDriver
*drv
= bs
->drv
;
1558 BlockDriverAIOCB
*ret
;
1562 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1565 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1569 /* Update stats even though technically transfer has not happened. */
1570 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1577 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1578 QEMUIOVector
*qiov
, int nb_sectors
,
1579 BlockDriverCompletionFunc
*cb
, void *opaque
)
1581 BlockDriver
*drv
= bs
->drv
;
1582 BlockDriverAIOCB
*ret
;
1588 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1591 if (bs
->dirty_bitmap
) {
1592 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1595 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1599 /* Update stats even though technically transfer has not happened. */
1600 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1608 typedef struct MultiwriteCB
{
1613 BlockDriverCompletionFunc
*cb
;
1615 QEMUIOVector
*free_qiov
;
1620 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1624 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1625 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1626 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1627 qemu_free(mcb
->callbacks
[i
].free_buf
);
1631 static void multiwrite_cb(void *opaque
, int ret
)
1633 MultiwriteCB
*mcb
= opaque
;
1637 multiwrite_user_cb(mcb
);
1640 mcb
->num_requests
--;
1641 if (mcb
->num_requests
== 0) {
1642 if (mcb
->error
== 0) {
1643 multiwrite_user_cb(mcb
);
1649 static int multiwrite_req_compare(const void *a
, const void *b
)
1651 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1655 * Takes a bunch of requests and tries to merge them. Returns the number of
1656 * requests that remain after merging.
1658 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1659 int num_reqs
, MultiwriteCB
*mcb
)
1663 // Sort requests by start sector
1664 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1666 // Check if adjacent requests touch the same clusters. If so, combine them,
1667 // filling up gaps with zero sectors.
1669 for (i
= 1; i
< num_reqs
; i
++) {
1671 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1673 // This handles the cases that are valid for all block drivers, namely
1674 // exactly sequential writes and overlapping writes.
1675 if (reqs
[i
].sector
<= oldreq_last
) {
1679 // The block driver may decide that it makes sense to combine requests
1680 // even if there is a gap of some sectors between them. In this case,
1681 // the gap is filled with zeros (therefore only applicable for yet
1682 // unused space in format like qcow2).
1683 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1684 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1689 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1690 qemu_iovec_init(qiov
,
1691 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1693 // Add the first request to the merged one. If the requests are
1694 // overlapping, drop the last sectors of the first request.
1695 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1696 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1698 // We might need to add some zeros between the two requests
1699 if (reqs
[i
].sector
> oldreq_last
) {
1700 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1701 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1702 memset(buf
, 0, zero_bytes
);
1703 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1704 mcb
->callbacks
[i
].free_buf
= buf
;
1707 // Add the second request
1708 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
1710 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
1711 reqs
[outidx
].qiov
= qiov
;
1713 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
1716 reqs
[outidx
].sector
= reqs
[i
].sector
;
1717 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
1718 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
1726 * Submit multiple AIO write requests at once.
1728 * On success, the function returns 0 and all requests in the reqs array have
1729 * been submitted. In error case this function returns -1, and any of the
1730 * requests may or may not be submitted yet. In particular, this means that the
1731 * callback will be called for some of the requests, for others it won't. The
1732 * caller must check the error field of the BlockRequest to wait for the right
1733 * callbacks (if error != 0, no callback will be called).
1735 * The implementation may modify the contents of the reqs array, e.g. to merge
1736 * requests. However, the fields opaque and error are left unmodified as they
1737 * are used to signal failure for a single request to the caller.
1739 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
1741 BlockDriverAIOCB
*acb
;
1745 if (num_reqs
== 0) {
1749 // Create MultiwriteCB structure
1750 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
1751 mcb
->num_requests
= 0;
1752 mcb
->num_callbacks
= num_reqs
;
1754 for (i
= 0; i
< num_reqs
; i
++) {
1755 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
1756 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
1759 // Check for mergable requests
1760 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
1762 // Run the aio requests
1763 for (i
= 0; i
< num_reqs
; i
++) {
1764 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
1765 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
1768 // We can only fail the whole thing if no request has been
1769 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1770 // complete and report the error in the callback.
1771 if (mcb
->num_requests
== 0) {
1772 reqs
[i
].error
= EIO
;
1779 mcb
->num_requests
++;
1790 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
1791 BlockDriverCompletionFunc
*cb
, void *opaque
)
1793 BlockDriver
*drv
= bs
->drv
;
1797 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
1800 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
1802 acb
->pool
->cancel(acb
);
1806 /**************************************************************/
1807 /* async block device emulation */
1809 typedef struct BlockDriverAIOCBSync
{
1810 BlockDriverAIOCB common
;
1813 /* vector translation state */
1817 } BlockDriverAIOCBSync
;
1819 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
1821 BlockDriverAIOCBSync
*acb
= (BlockDriverAIOCBSync
*)blockacb
;
1822 qemu_bh_delete(acb
->bh
);
1824 qemu_aio_release(acb
);
1827 static AIOPool bdrv_em_aio_pool
= {
1828 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
1829 .cancel
= bdrv_aio_cancel_em
,
1832 static void bdrv_aio_bh_cb(void *opaque
)
1834 BlockDriverAIOCBSync
*acb
= opaque
;
1837 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
1838 qemu_vfree(acb
->bounce
);
1839 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
1840 qemu_bh_delete(acb
->bh
);
1842 qemu_aio_release(acb
);
1845 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
1849 BlockDriverCompletionFunc
*cb
,
1854 BlockDriverAIOCBSync
*acb
;
1856 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
1857 acb
->is_write
= is_write
;
1859 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
1862 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
1865 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
1866 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1868 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1871 qemu_bh_schedule(acb
->bh
);
1873 return &acb
->common
;
1876 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
1877 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1878 BlockDriverCompletionFunc
*cb
, void *opaque
)
1880 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
1883 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
1884 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1885 BlockDriverCompletionFunc
*cb
, void *opaque
)
1887 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
1890 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
1891 BlockDriverCompletionFunc
*cb
, void *opaque
)
1893 BlockDriverAIOCBSync
*acb
;
1895 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
1896 acb
->is_write
= 1; /* don't bounce in the completion hadler */
1902 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
1905 qemu_bh_schedule(acb
->bh
);
1906 return &acb
->common
;
1909 /**************************************************************/
1910 /* sync block device emulation */
1912 static void bdrv_rw_em_cb(void *opaque
, int ret
)
1914 *(int *)opaque
= ret
;
1917 #define NOT_DONE 0x7fffffff
1919 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
1920 uint8_t *buf
, int nb_sectors
)
1923 BlockDriverAIOCB
*acb
;
1927 async_context_push();
1929 async_ret
= NOT_DONE
;
1930 iov
.iov_base
= (void *)buf
;
1931 iov
.iov_len
= nb_sectors
* 512;
1932 qemu_iovec_init_external(&qiov
, &iov
, 1);
1933 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
1934 bdrv_rw_em_cb
, &async_ret
);
1940 while (async_ret
== NOT_DONE
) {
1946 async_context_pop();
1950 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
1951 const uint8_t *buf
, int nb_sectors
)
1954 BlockDriverAIOCB
*acb
;
1958 async_context_push();
1960 async_ret
= NOT_DONE
;
1961 iov
.iov_base
= (void *)buf
;
1962 iov
.iov_len
= nb_sectors
* 512;
1963 qemu_iovec_init_external(&qiov
, &iov
, 1);
1964 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
1965 bdrv_rw_em_cb
, &async_ret
);
1970 while (async_ret
== NOT_DONE
) {
1975 async_context_pop();
1979 void bdrv_init(void)
1981 module_call_init(MODULE_INIT_BLOCK
);
1984 void bdrv_init_with_whitelist(void)
1986 use_bdrv_whitelist
= 1;
1990 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
1991 BlockDriverCompletionFunc
*cb
, void *opaque
)
1993 BlockDriverAIOCB
*acb
;
1995 if (pool
->free_aiocb
) {
1996 acb
= pool
->free_aiocb
;
1997 pool
->free_aiocb
= acb
->next
;
1999 acb
= qemu_mallocz(pool
->aiocb_size
);
2004 acb
->opaque
= opaque
;
2008 void qemu_aio_release(void *p
)
2010 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2011 AIOPool
*pool
= acb
->pool
;
2012 acb
->next
= pool
->free_aiocb
;
2013 pool
->free_aiocb
= acb
;
2016 /**************************************************************/
2017 /* removable device support */
2020 * Return TRUE if the media is present
2022 int bdrv_is_inserted(BlockDriverState
*bs
)
2024 BlockDriver
*drv
= bs
->drv
;
2028 if (!drv
->bdrv_is_inserted
)
2030 ret
= drv
->bdrv_is_inserted(bs
);
2035 * Return TRUE if the media changed since the last call to this
2036 * function. It is currently only used for floppy disks
2038 int bdrv_media_changed(BlockDriverState
*bs
)
2040 BlockDriver
*drv
= bs
->drv
;
2043 if (!drv
|| !drv
->bdrv_media_changed
)
2046 ret
= drv
->bdrv_media_changed(bs
);
2047 if (ret
== -ENOTSUP
)
2048 ret
= bs
->media_changed
;
2049 bs
->media_changed
= 0;
2054 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2056 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2058 BlockDriver
*drv
= bs
->drv
;
2065 if (!drv
|| !drv
->bdrv_eject
) {
2068 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2070 if (ret
== -ENOTSUP
) {
2079 int bdrv_is_locked(BlockDriverState
*bs
)
2085 * Lock or unlock the media (if it is locked, the user won't be able
2086 * to eject it manually).
2088 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2090 BlockDriver
*drv
= bs
->drv
;
2092 bs
->locked
= locked
;
2093 if (drv
&& drv
->bdrv_set_locked
) {
2094 drv
->bdrv_set_locked(bs
, locked
);
2098 /* needed for generic scsi interface */
2100 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2102 BlockDriver
*drv
= bs
->drv
;
2104 if (drv
&& drv
->bdrv_ioctl
)
2105 return drv
->bdrv_ioctl(bs
, req
, buf
);
2109 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2110 unsigned long int req
, void *buf
,
2111 BlockDriverCompletionFunc
*cb
, void *opaque
)
2113 BlockDriver
*drv
= bs
->drv
;
2115 if (drv
&& drv
->bdrv_aio_ioctl
)
2116 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2122 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2124 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2127 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2129 int64_t bitmap_size
;
2132 if (!bs
->dirty_bitmap
) {
2133 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2134 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2135 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2137 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2140 if (bs
->dirty_bitmap
) {
2141 qemu_free(bs
->dirty_bitmap
);
2142 bs
->dirty_bitmap
= NULL
;
2147 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2149 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2151 if (bs
->dirty_bitmap
&&
2152 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2153 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2154 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2160 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2163 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);