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
, BDRV_O_RDONLY
);
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
,
359 int ret
, open_flags
, try_rw
;
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 /* Note: for compatibility, we open disk image files as RDWR, and
448 RDONLY as fallback */
449 try_rw
= !bs
->read_only
|| bs
->is_temporary
;
450 if (!(flags
& BDRV_O_FILE
))
451 open_flags
= (try_rw
? BDRV_O_RDWR
: 0) |
452 (flags
& (BDRV_O_CACHE_MASK
|BDRV_O_NATIVE_AIO
));
454 open_flags
= flags
& ~(BDRV_O_FILE
| BDRV_O_SNAPSHOT
);
455 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
))
458 ret
= drv
->bdrv_open(bs
, filename
, open_flags
);
459 if ((ret
== -EACCES
|| ret
== -EPERM
) && !(flags
& BDRV_O_FILE
)) {
460 ret
= drv
->bdrv_open(bs
, filename
, open_flags
& ~BDRV_O_RDWR
);
464 qemu_free(bs
->opaque
);
468 if (bs
->is_temporary
)
472 if (drv
->bdrv_getlength
) {
473 bs
->total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
476 if (bs
->is_temporary
) {
480 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
481 /* if there is a backing file, use it */
482 BlockDriver
*back_drv
= NULL
;
483 bs
->backing_hd
= bdrv_new("");
484 /* pass on read_only property to the backing_hd */
485 bs
->backing_hd
->read_only
= bs
->read_only
;
486 path_combine(backing_filename
, sizeof(backing_filename
),
487 filename
, bs
->backing_file
);
488 if (bs
->backing_format
[0] != '\0')
489 back_drv
= bdrv_find_format(bs
->backing_format
);
490 ret
= bdrv_open2(bs
->backing_hd
, backing_filename
, open_flags
,
498 if (!bdrv_key_required(bs
)) {
499 /* call the change callback */
500 bs
->media_changed
= 1;
502 bs
->change_cb(bs
->change_opaque
);
507 void bdrv_close(BlockDriverState
*bs
)
511 bdrv_delete(bs
->backing_hd
);
512 bs
->drv
->bdrv_close(bs
);
513 qemu_free(bs
->opaque
);
515 if (bs
->is_temporary
) {
516 unlink(bs
->filename
);
522 /* call the change callback */
523 bs
->media_changed
= 1;
525 bs
->change_cb(bs
->change_opaque
);
529 void bdrv_delete(BlockDriverState
*bs
)
531 BlockDriverState
**pbs
;
534 while (*pbs
!= bs
&& *pbs
!= NULL
)
544 * Run consistency checks on an image
546 * Returns the number of errors or -errno when an internal error occurs
548 int bdrv_check(BlockDriverState
*bs
)
550 if (bs
->drv
->bdrv_check
== NULL
) {
554 return bs
->drv
->bdrv_check(bs
);
557 /* commit COW file into the raw image */
558 int bdrv_commit(BlockDriverState
*bs
)
560 BlockDriver
*drv
= bs
->drv
;
561 int64_t i
, total_sectors
;
563 unsigned char sector
[512];
572 if (!bs
->backing_hd
) {
576 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
577 for (i
= 0; i
< total_sectors
;) {
578 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
579 for(j
= 0; j
< n
; j
++) {
580 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
584 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
594 if (drv
->bdrv_make_empty
)
595 return 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
;
722 /* first read to align to sector start */
723 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
726 sector_num
= offset
>> BDRV_SECTOR_BITS
;
728 if (bdrv_read(bs
, sector_num
, tmp_buf
, 1) < 0)
730 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
738 /* read the sectors "in place" */
739 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
740 if (nb_sectors
> 0) {
741 if (bdrv_read(bs
, sector_num
, buf
, nb_sectors
) < 0)
743 sector_num
+= nb_sectors
;
744 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
749 /* add data from the last sector */
751 if (bdrv_read(bs
, sector_num
, tmp_buf
, 1) < 0)
753 memcpy(buf
, tmp_buf
, count
);
758 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
759 const void *buf
, int count1
)
761 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
762 int len
, nb_sectors
, count
;
766 /* first write to align to sector start */
767 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
770 sector_num
= offset
>> BDRV_SECTOR_BITS
;
772 if (bdrv_read(bs
, sector_num
, tmp_buf
, 1) < 0)
774 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
775 if (bdrv_write(bs
, sector_num
, tmp_buf
, 1) < 0)
784 /* write the sectors "in place" */
785 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
786 if (nb_sectors
> 0) {
787 if (bdrv_write(bs
, sector_num
, buf
, nb_sectors
) < 0)
789 sector_num
+= nb_sectors
;
790 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
795 /* add data from the last sector */
797 if (bdrv_read(bs
, sector_num
, tmp_buf
, 1) < 0)
799 memcpy(tmp_buf
, buf
, count
);
800 if (bdrv_write(bs
, sector_num
, tmp_buf
, 1) < 0)
807 * Truncate file to 'offset' bytes (needed only for file protocols)
809 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
811 BlockDriver
*drv
= bs
->drv
;
814 if (!drv
->bdrv_truncate
)
818 return drv
->bdrv_truncate(bs
, offset
);
822 * Length of a file in bytes. Return < 0 if error or unknown.
824 int64_t bdrv_getlength(BlockDriverState
*bs
)
826 BlockDriver
*drv
= bs
->drv
;
829 if (!drv
->bdrv_getlength
) {
831 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
833 return drv
->bdrv_getlength(bs
);
836 /* return 0 as number of sectors if no device present or error */
837 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
840 length
= bdrv_getlength(bs
);
844 length
= length
>> BDRV_SECTOR_BITS
;
845 *nb_sectors_ptr
= length
;
849 uint8_t boot_ind
; /* 0x80 - active */
850 uint8_t head
; /* starting head */
851 uint8_t sector
; /* starting sector */
852 uint8_t cyl
; /* starting cylinder */
853 uint8_t sys_ind
; /* What partition type */
854 uint8_t end_head
; /* end head */
855 uint8_t end_sector
; /* end sector */
856 uint8_t end_cyl
; /* end cylinder */
857 uint32_t start_sect
; /* starting sector counting from 0 */
858 uint32_t nr_sects
; /* nr of sectors in partition */
859 } __attribute__((packed
));
861 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
862 static int guess_disk_lchs(BlockDriverState
*bs
,
863 int *pcylinders
, int *pheads
, int *psectors
)
866 int ret
, i
, heads
, sectors
, cylinders
;
871 bdrv_get_geometry(bs
, &nb_sectors
);
873 ret
= bdrv_read(bs
, 0, buf
, 1);
876 /* test msdos magic */
877 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
879 for(i
= 0; i
< 4; i
++) {
880 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
881 nr_sects
= le32_to_cpu(p
->nr_sects
);
882 if (nr_sects
&& p
->end_head
) {
883 /* We make the assumption that the partition terminates on
884 a cylinder boundary */
885 heads
= p
->end_head
+ 1;
886 sectors
= p
->end_sector
& 63;
889 cylinders
= nb_sectors
/ (heads
* sectors
);
890 if (cylinders
< 1 || cylinders
> 16383)
894 *pcylinders
= cylinders
;
896 printf("guessed geometry: LCHS=%d %d %d\n",
897 cylinders
, heads
, sectors
);
905 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
907 int translation
, lba_detected
= 0;
908 int cylinders
, heads
, secs
;
911 /* if a geometry hint is available, use it */
912 bdrv_get_geometry(bs
, &nb_sectors
);
913 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
914 translation
= bdrv_get_translation_hint(bs
);
915 if (cylinders
!= 0) {
920 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
922 /* if heads > 16, it means that a BIOS LBA
923 translation was active, so the default
924 hardware geometry is OK */
926 goto default_geometry
;
931 /* disable any translation to be in sync with
932 the logical geometry */
933 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
934 bdrv_set_translation_hint(bs
,
935 BIOS_ATA_TRANSLATION_NONE
);
940 /* if no geometry, use a standard physical disk geometry */
941 cylinders
= nb_sectors
/ (16 * 63);
943 if (cylinders
> 16383)
945 else if (cylinders
< 2)
950 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
951 if ((*pcyls
* *pheads
) <= 131072) {
952 bdrv_set_translation_hint(bs
,
953 BIOS_ATA_TRANSLATION_LARGE
);
955 bdrv_set_translation_hint(bs
,
956 BIOS_ATA_TRANSLATION_LBA
);
960 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
964 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
965 int cyls
, int heads
, int secs
)
972 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
975 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
976 type
== BDRV_TYPE_FLOPPY
));
979 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
981 bs
->translation
= translation
;
984 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
985 int *pcyls
, int *pheads
, int *psecs
)
992 int bdrv_get_type_hint(BlockDriverState
*bs
)
997 int bdrv_get_translation_hint(BlockDriverState
*bs
)
999 return bs
->translation
;
1002 int bdrv_is_removable(BlockDriverState
*bs
)
1004 return bs
->removable
;
1007 int bdrv_is_read_only(BlockDriverState
*bs
)
1009 return bs
->read_only
;
1012 int bdrv_set_read_only(BlockDriverState
*bs
, int read_only
)
1014 int ret
= bs
->read_only
;
1015 bs
->read_only
= read_only
;
1019 int bdrv_is_sg(BlockDriverState
*bs
)
1024 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1026 return bs
->enable_write_cache
;
1029 /* XXX: no longer used */
1030 void bdrv_set_change_cb(BlockDriverState
*bs
,
1031 void (*change_cb
)(void *opaque
), void *opaque
)
1033 bs
->change_cb
= change_cb
;
1034 bs
->change_opaque
= opaque
;
1037 int bdrv_is_encrypted(BlockDriverState
*bs
)
1039 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1041 return bs
->encrypted
;
1044 int bdrv_key_required(BlockDriverState
*bs
)
1046 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1048 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1050 return (bs
->encrypted
&& !bs
->valid_key
);
1053 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1056 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1057 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1063 if (!bs
->encrypted
|| !bs
->drv
|| !bs
->drv
->bdrv_set_key
)
1065 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1068 } else if (!bs
->valid_key
) {
1070 /* call the change callback now, we skipped it on open */
1071 bs
->media_changed
= 1;
1073 bs
->change_cb(bs
->change_opaque
);
1078 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1083 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1087 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1092 for (drv
= first_drv
; drv
!= NULL
; drv
= drv
->next
) {
1093 it(opaque
, drv
->format_name
);
1097 BlockDriverState
*bdrv_find(const char *name
)
1099 BlockDriverState
*bs
;
1101 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1102 if (!strcmp(name
, bs
->device_name
))
1108 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1110 BlockDriverState
*bs
;
1112 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1117 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1119 return bs
->device_name
;
1122 void bdrv_flush(BlockDriverState
*bs
)
1124 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1125 bs
->drv
->bdrv_flush(bs
);
1128 void bdrv_flush_all(void)
1130 BlockDriverState
*bs
;
1132 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
)
1133 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1134 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
)))
1139 * Returns true iff the specified sector is present in the disk image. Drivers
1140 * not implementing the functionality are assumed to not support backing files,
1141 * hence all their sectors are reported as allocated.
1143 * 'pnum' is set to the number of sectors (including and immediately following
1144 * the specified sector) that are known to be in the same
1145 * allocated/unallocated state.
1147 * 'nb_sectors' is the max value 'pnum' should be set to.
1149 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1153 if (!bs
->drv
->bdrv_is_allocated
) {
1154 if (sector_num
>= bs
->total_sectors
) {
1158 n
= bs
->total_sectors
- sector_num
;
1159 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1162 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1165 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1168 Monitor
*mon
= opaque
;
1170 bs_dict
= qobject_to_qdict(obj
);
1172 monitor_printf(mon
, "%s: type=%s removable=%d",
1173 qdict_get_str(bs_dict
, "device"),
1174 qdict_get_str(bs_dict
, "type"),
1175 qdict_get_bool(bs_dict
, "removable"));
1177 if (qdict_get_bool(bs_dict
, "removable")) {
1178 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1181 if (qdict_haskey(bs_dict
, "inserted")) {
1182 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1184 monitor_printf(mon
, " file=");
1185 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1186 if (qdict_haskey(qdict
, "backing_file")) {
1187 monitor_printf(mon
, " backing_file=");
1188 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1190 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1191 qdict_get_bool(qdict
, "ro"),
1192 qdict_get_str(qdict
, "drv"),
1193 qdict_get_bool(qdict
, "encrypted"));
1195 monitor_printf(mon
, " [not inserted]");
1198 monitor_printf(mon
, "\n");
1201 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1203 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1207 * bdrv_info(): Block devices information
1209 * Each block device information is stored in a QDict and the
1210 * returned QObject is a QList of all devices.
1212 * The QDict contains the following:
1214 * - "device": device name
1215 * - "type": device type
1216 * - "removable": true if the device is removable, false otherwise
1217 * - "locked": true if the device is locked, false otherwise
1218 * - "inserted": only present if the device is inserted, it is a QDict
1219 * containing the following:
1220 * - "file": device file name
1221 * - "ro": true if read-only, false otherwise
1222 * - "drv": driver format name
1223 * - "backing_file": backing file name if one is used
1224 * - "encrypted": true if encrypted, false otherwise
1228 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1229 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1230 * { "device": "floppy0", "type": "floppy", "removable": true,
1231 * "locked": false } ]
1233 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1236 BlockDriverState
*bs
;
1238 bs_list
= qlist_new();
1240 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1242 const char *type
= "unknown";
1248 case BDRV_TYPE_CDROM
:
1251 case BDRV_TYPE_FLOPPY
:
1256 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1257 "'removable': %i, 'locked': %i }",
1258 bs
->device_name
, type
, bs
->removable
,
1260 assert(bs_obj
!= NULL
);
1264 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1266 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1267 "'encrypted': %i }",
1268 bs
->filename
, bs
->read_only
,
1269 bs
->drv
->format_name
,
1270 bdrv_is_encrypted(bs
));
1271 assert(obj
!= NULL
);
1272 if (bs
->backing_file
[0] != '\0') {
1273 QDict
*qdict
= qobject_to_qdict(obj
);
1274 qdict_put(qdict
, "backing_file",
1275 qstring_from_str(bs
->backing_file
));
1278 qdict_put_obj(bs_dict
, "inserted", obj
);
1280 qlist_append_obj(bs_list
, bs_obj
);
1283 *ret_data
= QOBJECT(bs_list
);
1286 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1289 Monitor
*mon
= opaque
;
1291 qdict
= qobject_to_qdict(data
);
1292 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1294 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1295 monitor_printf(mon
, " rd_bytes=%" PRId64
1296 " wr_bytes=%" PRId64
1297 " rd_operations=%" PRId64
1298 " wr_operations=%" PRId64
1300 qdict_get_int(qdict
, "rd_bytes"),
1301 qdict_get_int(qdict
, "wr_bytes"),
1302 qdict_get_int(qdict
, "rd_operations"),
1303 qdict_get_int(qdict
, "wr_operations"));
1306 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1308 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1312 * bdrv_info_stats(): show block device statistics
1314 * Each device statistic information is stored in a QDict and
1315 * the returned QObject is a QList of all devices.
1317 * The QDict contains the following:
1319 * - "device": device name
1320 * - "stats": A QDict with the statistics information, it contains:
1321 * - "rd_bytes": bytes read
1322 * - "wr_bytes": bytes written
1323 * - "rd_operations": read operations
1324 * - "wr_operations": write operations
1328 * [ { "device": "ide0-hd0",
1329 * "stats": { "rd_bytes": 512,
1331 * "rd_operations": 1,
1332 * "wr_operations": 0 } },
1333 * { "device": "ide1-cd0",
1334 * "stats": { "rd_bytes": 0,
1336 * "rd_operations": 0,
1337 * "wr_operations": 0 } } ]
1339 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1343 BlockDriverState
*bs
;
1345 devices
= qlist_new();
1347 for (bs
= bdrv_first
; bs
!= NULL
; bs
= bs
->next
) {
1348 obj
= qobject_from_jsonf("{ 'device': %s, 'stats': {"
1349 "'rd_bytes': %" PRId64
","
1350 "'wr_bytes': %" PRId64
","
1351 "'rd_operations': %" PRId64
","
1352 "'wr_operations': %" PRId64
1355 bs
->rd_bytes
, bs
->wr_bytes
,
1356 bs
->rd_ops
, bs
->wr_ops
);
1357 assert(obj
!= NULL
);
1358 qlist_append_obj(devices
, obj
);
1361 *ret_data
= QOBJECT(devices
);
1364 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1366 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1367 return bs
->backing_file
;
1368 else if (bs
->encrypted
)
1369 return bs
->filename
;
1374 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1375 char *filename
, int filename_size
)
1377 if (!bs
->backing_file
) {
1378 pstrcpy(filename
, filename_size
, "");
1380 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1384 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1385 const uint8_t *buf
, int nb_sectors
)
1387 BlockDriver
*drv
= bs
->drv
;
1390 if (!drv
->bdrv_write_compressed
)
1392 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1395 if (bs
->dirty_bitmap
) {
1396 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1399 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1402 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1404 BlockDriver
*drv
= bs
->drv
;
1407 if (!drv
->bdrv_get_info
)
1409 memset(bdi
, 0, sizeof(*bdi
));
1410 return drv
->bdrv_get_info(bs
, bdi
);
1413 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1414 int64_t pos
, int size
)
1416 BlockDriver
*drv
= bs
->drv
;
1419 if (!drv
->bdrv_save_vmstate
)
1421 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1424 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1425 int64_t pos
, int size
)
1427 BlockDriver
*drv
= bs
->drv
;
1430 if (!drv
->bdrv_load_vmstate
)
1432 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1435 /**************************************************************/
1436 /* handling of snapshots */
1438 int bdrv_snapshot_create(BlockDriverState
*bs
,
1439 QEMUSnapshotInfo
*sn_info
)
1441 BlockDriver
*drv
= bs
->drv
;
1444 if (!drv
->bdrv_snapshot_create
)
1446 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1449 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1450 const char *snapshot_id
)
1452 BlockDriver
*drv
= bs
->drv
;
1455 if (!drv
->bdrv_snapshot_goto
)
1457 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1460 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1462 BlockDriver
*drv
= bs
->drv
;
1465 if (!drv
->bdrv_snapshot_delete
)
1467 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1470 int bdrv_snapshot_list(BlockDriverState
*bs
,
1471 QEMUSnapshotInfo
**psn_info
)
1473 BlockDriver
*drv
= bs
->drv
;
1476 if (!drv
->bdrv_snapshot_list
)
1478 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1481 #define NB_SUFFIXES 4
1483 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1485 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1490 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1493 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1494 if (size
< (10 * base
)) {
1495 snprintf(buf
, buf_size
, "%0.1f%c",
1496 (double)size
/ base
,
1499 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1500 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1501 ((size
+ (base
>> 1)) / base
),
1511 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1513 char buf1
[128], date_buf
[128], clock_buf
[128];
1523 snprintf(buf
, buf_size
,
1524 "%-10s%-20s%7s%20s%15s",
1525 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1529 ptm
= localtime(&ti
);
1530 strftime(date_buf
, sizeof(date_buf
),
1531 "%Y-%m-%d %H:%M:%S", ptm
);
1533 localtime_r(&ti
, &tm
);
1534 strftime(date_buf
, sizeof(date_buf
),
1535 "%Y-%m-%d %H:%M:%S", &tm
);
1537 secs
= sn
->vm_clock_nsec
/ 1000000000;
1538 snprintf(clock_buf
, sizeof(clock_buf
),
1539 "%02d:%02d:%02d.%03d",
1541 (int)((secs
/ 60) % 60),
1543 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1544 snprintf(buf
, buf_size
,
1545 "%-10s%-20s%7s%20s%15s",
1546 sn
->id_str
, sn
->name
,
1547 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1555 /**************************************************************/
1558 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1559 QEMUIOVector
*qiov
, int nb_sectors
,
1560 BlockDriverCompletionFunc
*cb
, void *opaque
)
1562 BlockDriver
*drv
= bs
->drv
;
1563 BlockDriverAIOCB
*ret
;
1567 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1570 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1574 /* Update stats even though technically transfer has not happened. */
1575 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1582 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1583 QEMUIOVector
*qiov
, int nb_sectors
,
1584 BlockDriverCompletionFunc
*cb
, void *opaque
)
1586 BlockDriver
*drv
= bs
->drv
;
1587 BlockDriverAIOCB
*ret
;
1593 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1596 if (bs
->dirty_bitmap
) {
1597 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1600 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1604 /* Update stats even though technically transfer has not happened. */
1605 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1613 typedef struct MultiwriteCB
{
1618 BlockDriverCompletionFunc
*cb
;
1620 QEMUIOVector
*free_qiov
;
1625 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1629 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1630 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1631 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1632 qemu_free(mcb
->callbacks
[i
].free_buf
);
1636 static void multiwrite_cb(void *opaque
, int ret
)
1638 MultiwriteCB
*mcb
= opaque
;
1642 multiwrite_user_cb(mcb
);
1645 mcb
->num_requests
--;
1646 if (mcb
->num_requests
== 0) {
1647 if (mcb
->error
== 0) {
1648 multiwrite_user_cb(mcb
);
1654 static int multiwrite_req_compare(const void *a
, const void *b
)
1656 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1660 * Takes a bunch of requests and tries to merge them. Returns the number of
1661 * requests that remain after merging.
1663 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1664 int num_reqs
, MultiwriteCB
*mcb
)
1668 // Sort requests by start sector
1669 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1671 // Check if adjacent requests touch the same clusters. If so, combine them,
1672 // filling up gaps with zero sectors.
1674 for (i
= 1; i
< num_reqs
; i
++) {
1676 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1678 // This handles the cases that are valid for all block drivers, namely
1679 // exactly sequential writes and overlapping writes.
1680 if (reqs
[i
].sector
<= oldreq_last
) {
1684 // The block driver may decide that it makes sense to combine requests
1685 // even if there is a gap of some sectors between them. In this case,
1686 // the gap is filled with zeros (therefore only applicable for yet
1687 // unused space in format like qcow2).
1688 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1689 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1694 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1695 qemu_iovec_init(qiov
,
1696 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1698 // Add the first request to the merged one. If the requests are
1699 // overlapping, drop the last sectors of the first request.
1700 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1701 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1703 // We might need to add some zeros between the two requests
1704 if (reqs
[i
].sector
> oldreq_last
) {
1705 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1706 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1707 memset(buf
, 0, zero_bytes
);
1708 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1709 mcb
->callbacks
[i
].free_buf
= buf
;
1712 // Add the second request
1713 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
1715 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
1716 reqs
[outidx
].qiov
= qiov
;
1718 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
1721 reqs
[outidx
].sector
= reqs
[i
].sector
;
1722 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
1723 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
1731 * Submit multiple AIO write requests at once.
1733 * On success, the function returns 0 and all requests in the reqs array have
1734 * been submitted. In error case this function returns -1, and any of the
1735 * requests may or may not be submitted yet. In particular, this means that the
1736 * callback will be called for some of the requests, for others it won't. The
1737 * caller must check the error field of the BlockRequest to wait for the right
1738 * callbacks (if error != 0, no callback will be called).
1740 * The implementation may modify the contents of the reqs array, e.g. to merge
1741 * requests. However, the fields opaque and error are left unmodified as they
1742 * are used to signal failure for a single request to the caller.
1744 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
1746 BlockDriverAIOCB
*acb
;
1750 if (num_reqs
== 0) {
1754 // Create MultiwriteCB structure
1755 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
1756 mcb
->num_requests
= 0;
1757 mcb
->num_callbacks
= num_reqs
;
1759 for (i
= 0; i
< num_reqs
; i
++) {
1760 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
1761 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
1764 // Check for mergable requests
1765 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
1767 // Run the aio requests
1768 for (i
= 0; i
< num_reqs
; i
++) {
1769 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
1770 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
1773 // We can only fail the whole thing if no request has been
1774 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1775 // complete and report the error in the callback.
1776 if (mcb
->num_requests
== 0) {
1777 reqs
[i
].error
= EIO
;
1784 mcb
->num_requests
++;
1795 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
1796 BlockDriverCompletionFunc
*cb
, void *opaque
)
1798 BlockDriver
*drv
= bs
->drv
;
1802 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
1805 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
1807 acb
->pool
->cancel(acb
);
1811 /**************************************************************/
1812 /* async block device emulation */
1814 typedef struct BlockDriverAIOCBSync
{
1815 BlockDriverAIOCB common
;
1818 /* vector translation state */
1822 } BlockDriverAIOCBSync
;
1824 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
1826 BlockDriverAIOCBSync
*acb
= (BlockDriverAIOCBSync
*)blockacb
;
1827 qemu_bh_delete(acb
->bh
);
1829 qemu_aio_release(acb
);
1832 static AIOPool bdrv_em_aio_pool
= {
1833 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
1834 .cancel
= bdrv_aio_cancel_em
,
1837 static void bdrv_aio_bh_cb(void *opaque
)
1839 BlockDriverAIOCBSync
*acb
= opaque
;
1842 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
1843 qemu_vfree(acb
->bounce
);
1844 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
1845 qemu_bh_delete(acb
->bh
);
1847 qemu_aio_release(acb
);
1850 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
1854 BlockDriverCompletionFunc
*cb
,
1859 BlockDriverAIOCBSync
*acb
;
1861 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
1862 acb
->is_write
= is_write
;
1864 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
1867 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
1870 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
1871 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1873 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1876 qemu_bh_schedule(acb
->bh
);
1878 return &acb
->common
;
1881 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
1882 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1883 BlockDriverCompletionFunc
*cb
, void *opaque
)
1885 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
1888 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
1889 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1890 BlockDriverCompletionFunc
*cb
, void *opaque
)
1892 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
1895 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
1896 BlockDriverCompletionFunc
*cb
, void *opaque
)
1898 BlockDriverAIOCBSync
*acb
;
1900 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
1901 acb
->is_write
= 1; /* don't bounce in the completion hadler */
1907 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
1910 qemu_bh_schedule(acb
->bh
);
1911 return &acb
->common
;
1914 /**************************************************************/
1915 /* sync block device emulation */
1917 static void bdrv_rw_em_cb(void *opaque
, int ret
)
1919 *(int *)opaque
= ret
;
1922 #define NOT_DONE 0x7fffffff
1924 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
1925 uint8_t *buf
, int nb_sectors
)
1928 BlockDriverAIOCB
*acb
;
1932 async_context_push();
1934 async_ret
= NOT_DONE
;
1935 iov
.iov_base
= (void *)buf
;
1936 iov
.iov_len
= nb_sectors
* 512;
1937 qemu_iovec_init_external(&qiov
, &iov
, 1);
1938 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
1939 bdrv_rw_em_cb
, &async_ret
);
1945 while (async_ret
== NOT_DONE
) {
1951 async_context_pop();
1955 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
1956 const uint8_t *buf
, int nb_sectors
)
1959 BlockDriverAIOCB
*acb
;
1963 async_context_push();
1965 async_ret
= NOT_DONE
;
1966 iov
.iov_base
= (void *)buf
;
1967 iov
.iov_len
= nb_sectors
* 512;
1968 qemu_iovec_init_external(&qiov
, &iov
, 1);
1969 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
1970 bdrv_rw_em_cb
, &async_ret
);
1975 while (async_ret
== NOT_DONE
) {
1980 async_context_pop();
1984 void bdrv_init(void)
1986 module_call_init(MODULE_INIT_BLOCK
);
1989 void bdrv_init_with_whitelist(void)
1991 use_bdrv_whitelist
= 1;
1995 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
1996 BlockDriverCompletionFunc
*cb
, void *opaque
)
1998 BlockDriverAIOCB
*acb
;
2000 if (pool
->free_aiocb
) {
2001 acb
= pool
->free_aiocb
;
2002 pool
->free_aiocb
= acb
->next
;
2004 acb
= qemu_mallocz(pool
->aiocb_size
);
2009 acb
->opaque
= opaque
;
2013 void qemu_aio_release(void *p
)
2015 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2016 AIOPool
*pool
= acb
->pool
;
2017 acb
->next
= pool
->free_aiocb
;
2018 pool
->free_aiocb
= acb
;
2021 /**************************************************************/
2022 /* removable device support */
2025 * Return TRUE if the media is present
2027 int bdrv_is_inserted(BlockDriverState
*bs
)
2029 BlockDriver
*drv
= bs
->drv
;
2033 if (!drv
->bdrv_is_inserted
)
2035 ret
= drv
->bdrv_is_inserted(bs
);
2040 * Return TRUE if the media changed since the last call to this
2041 * function. It is currently only used for floppy disks
2043 int bdrv_media_changed(BlockDriverState
*bs
)
2045 BlockDriver
*drv
= bs
->drv
;
2048 if (!drv
|| !drv
->bdrv_media_changed
)
2051 ret
= drv
->bdrv_media_changed(bs
);
2052 if (ret
== -ENOTSUP
)
2053 ret
= bs
->media_changed
;
2054 bs
->media_changed
= 0;
2059 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2061 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2063 BlockDriver
*drv
= bs
->drv
;
2070 if (!drv
|| !drv
->bdrv_eject
) {
2073 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2075 if (ret
== -ENOTSUP
) {
2084 int bdrv_is_locked(BlockDriverState
*bs
)
2090 * Lock or unlock the media (if it is locked, the user won't be able
2091 * to eject it manually).
2093 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2095 BlockDriver
*drv
= bs
->drv
;
2097 bs
->locked
= locked
;
2098 if (drv
&& drv
->bdrv_set_locked
) {
2099 drv
->bdrv_set_locked(bs
, locked
);
2103 /* needed for generic scsi interface */
2105 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2107 BlockDriver
*drv
= bs
->drv
;
2109 if (drv
&& drv
->bdrv_ioctl
)
2110 return drv
->bdrv_ioctl(bs
, req
, buf
);
2114 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2115 unsigned long int req
, void *buf
,
2116 BlockDriverCompletionFunc
*cb
, void *opaque
)
2118 BlockDriver
*drv
= bs
->drv
;
2120 if (drv
&& drv
->bdrv_aio_ioctl
)
2121 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2127 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2129 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2132 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2134 int64_t bitmap_size
;
2137 if (!bs
->dirty_bitmap
) {
2138 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2139 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2140 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2142 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2145 if (bs
->dirty_bitmap
) {
2146 qemu_free(bs
->dirty_bitmap
);
2147 bs
->dirty_bitmap
= NULL
;
2152 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2154 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2156 if (bs
->dirty_bitmap
&&
2157 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2158 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2159 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2165 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2168 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);