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 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
59 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
61 static BlockDriver
*first_drv
;
63 /* If non-zero, use only whitelisted block drivers */
64 static int use_bdrv_whitelist
;
66 int path_is_absolute(const char *path
)
70 /* specific case for names like: "\\.\d:" */
71 if (*path
== '/' || *path
== '\\')
74 p
= strchr(path
, ':');
80 return (*p
== '/' || *p
== '\\');
86 /* if filename is absolute, just copy it to dest. Otherwise, build a
87 path to it by considering it is relative to base_path. URL are
89 void path_combine(char *dest
, int dest_size
,
90 const char *base_path
,
98 if (path_is_absolute(filename
)) {
99 pstrcpy(dest
, dest_size
, filename
);
101 p
= strchr(base_path
, ':');
106 p1
= strrchr(base_path
, '/');
110 p2
= strrchr(base_path
, '\\');
122 if (len
> dest_size
- 1)
124 memcpy(dest
, base_path
, len
);
126 pstrcat(dest
, dest_size
, filename
);
130 void bdrv_register(BlockDriver
*bdrv
)
132 if (!bdrv
->bdrv_aio_readv
) {
133 /* add AIO emulation layer */
134 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
135 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
136 } else if (!bdrv
->bdrv_read
) {
137 /* add synchronous IO emulation layer */
138 bdrv
->bdrv_read
= bdrv_read_em
;
139 bdrv
->bdrv_write
= bdrv_write_em
;
142 if (!bdrv
->bdrv_aio_flush
)
143 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
145 bdrv
->next
= first_drv
;
149 /* create a new block device (by default it is empty) */
150 BlockDriverState
*bdrv_new(const char *device_name
)
152 BlockDriverState
*bs
;
154 bs
= qemu_mallocz(sizeof(BlockDriverState
));
155 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
156 if (device_name
[0] != '\0') {
157 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
162 BlockDriver
*bdrv_find_format(const char *format_name
)
165 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
166 if (!strcmp(drv1
->format_name
, format_name
))
172 static int bdrv_is_whitelisted(BlockDriver
*drv
)
174 static const char *whitelist
[] = {
175 CONFIG_BDRV_WHITELIST
180 return 1; /* no whitelist, anything goes */
182 for (p
= whitelist
; *p
; p
++) {
183 if (!strcmp(drv
->format_name
, *p
)) {
190 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
192 BlockDriver
*drv
= bdrv_find_format(format_name
);
193 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
196 int bdrv_create(BlockDriver
*drv
, const char* filename
,
197 QEMUOptionParameter
*options
)
199 if (!drv
->bdrv_create
)
202 return drv
->bdrv_create(filename
, options
);
206 void get_tmp_filename(char *filename
, int size
)
208 char temp_dir
[MAX_PATH
];
210 GetTempPath(MAX_PATH
, temp_dir
);
211 GetTempFileName(temp_dir
, "qem", 0, filename
);
214 void get_tmp_filename(char *filename
, int size
)
218 /* XXX: race condition possible */
219 tmpdir
= getenv("TMPDIR");
222 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
223 fd
= mkstemp(filename
);
229 static int is_windows_drive_prefix(const char *filename
)
231 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
232 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
236 int is_windows_drive(const char *filename
)
238 if (is_windows_drive_prefix(filename
) &&
241 if (strstart(filename
, "\\\\.\\", NULL
) ||
242 strstart(filename
, "//./", NULL
))
248 static BlockDriver
*find_protocol(const char *filename
)
256 if (is_windows_drive(filename
) ||
257 is_windows_drive_prefix(filename
))
258 return bdrv_find_format("raw");
260 p
= strchr(filename
, ':');
262 return bdrv_find_format("raw");
264 if (len
> sizeof(protocol
) - 1)
265 len
= sizeof(protocol
) - 1;
266 memcpy(protocol
, filename
, len
);
267 protocol
[len
] = '\0';
268 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
269 if (drv1
->protocol_name
&&
270 !strcmp(drv1
->protocol_name
, protocol
))
277 * Detect host devices. By convention, /dev/cdrom[N] is always
278 * recognized as a host CDROM.
280 static BlockDriver
*find_hdev_driver(const char *filename
)
282 int score_max
= 0, score
;
283 BlockDriver
*drv
= NULL
, *d
;
285 for (d
= first_drv
; d
; d
= d
->next
) {
286 if (d
->bdrv_probe_device
) {
287 score
= d
->bdrv_probe_device(filename
);
288 if (score
> score_max
) {
298 static BlockDriver
*find_image_format(const char *filename
)
300 int ret
, score
, score_max
;
301 BlockDriver
*drv1
, *drv
;
303 BlockDriverState
*bs
;
305 drv
= find_protocol(filename
);
306 /* no need to test disk image formats for vvfat */
307 if (drv
&& strcmp(drv
->format_name
, "vvfat") == 0)
310 ret
= bdrv_file_open(&bs
, filename
, 0);
313 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
320 for(drv1
= first_drv
; drv1
!= NULL
; drv1
= drv1
->next
) {
321 if (drv1
->bdrv_probe
) {
322 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
323 if (score
> score_max
) {
332 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
334 BlockDriverState
*bs
;
338 drv
= find_protocol(filename
);
344 ret
= bdrv_open(bs
, filename
, flags
, drv
);
354 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
358 char tmp_filename
[PATH_MAX
];
359 char backing_filename
[PATH_MAX
];
361 bs
->is_temporary
= 0;
364 bs
->open_flags
= flags
;
365 /* buffer_alignment defaulted to 512, drivers can change this value */
366 bs
->buffer_alignment
= 512;
368 if (flags
& BDRV_O_SNAPSHOT
) {
369 BlockDriverState
*bs1
;
372 BlockDriver
*bdrv_qcow2
;
373 QEMUOptionParameter
*options
;
375 /* if snapshot, we create a temporary backing file and open it
376 instead of opening 'filename' directly */
378 /* if there is a backing file, use it */
380 ret
= bdrv_open(bs1
, filename
, 0, drv
);
385 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
387 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
392 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
394 /* Real path is meaningless for protocols */
396 snprintf(backing_filename
, sizeof(backing_filename
),
398 else if (!realpath(filename
, backing_filename
))
401 bdrv_qcow2
= bdrv_find_format("qcow2");
402 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
404 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
405 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
407 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
411 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
416 filename
= tmp_filename
;
418 bs
->is_temporary
= 1;
421 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
424 drv
= find_hdev_driver(filename
);
426 drv
= find_image_format(filename
);
432 goto unlink_and_fail
;
434 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
436 goto unlink_and_fail
;
440 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
443 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
444 * write cache to the guest. We do need the fdatasync to flush
445 * out transactions for block allocations, and we maybe have a
446 * volatile write cache in our backing device to deal with.
448 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
449 bs
->enable_write_cache
= 1;
452 * Clear flags that are internal to the block layer before opening the
455 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
458 * Snapshots should be writeable.
460 if (bs
->is_temporary
) {
461 open_flags
|= BDRV_O_RDWR
;
464 ret
= drv
->bdrv_open(bs
, filename
, open_flags
);
469 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
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
);
487 /* backing files always opened read-only */
488 open_flags
&= ~BDRV_O_RDWR
;
490 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, open_flags
, back_drv
);
495 if (bs
->is_temporary
) {
496 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
498 /* base image inherits from "parent" */
499 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
503 if (!bdrv_key_required(bs
)) {
504 /* call the change callback */
505 bs
->media_changed
= 1;
507 bs
->change_cb(bs
->change_opaque
);
512 qemu_free(bs
->opaque
);
516 if (bs
->is_temporary
)
521 void bdrv_close(BlockDriverState
*bs
)
525 bdrv_delete(bs
->backing_hd
);
526 bs
->drv
->bdrv_close(bs
);
527 qemu_free(bs
->opaque
);
529 if (bs
->is_temporary
) {
530 unlink(bs
->filename
);
536 /* call the change callback */
537 bs
->media_changed
= 1;
539 bs
->change_cb(bs
->change_opaque
);
543 void bdrv_delete(BlockDriverState
*bs
)
545 /* remove from list, if necessary */
546 if (bs
->device_name
[0] != '\0') {
547 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
555 * Run consistency checks on an image
557 * Returns the number of errors or -errno when an internal error occurs
559 int bdrv_check(BlockDriverState
*bs
)
561 if (bs
->drv
->bdrv_check
== NULL
) {
565 return bs
->drv
->bdrv_check(bs
);
568 /* commit COW file into the raw image */
569 int bdrv_commit(BlockDriverState
*bs
)
571 BlockDriver
*drv
= bs
->drv
;
572 int64_t i
, total_sectors
;
573 int n
, j
, ro
, open_flags
;
574 int ret
= 0, rw_ret
= 0;
575 unsigned char sector
[512];
577 BlockDriverState
*bs_rw
, *bs_ro
;
582 if (!bs
->backing_hd
) {
586 if (bs
->backing_hd
->keep_read_only
) {
590 ro
= bs
->backing_hd
->read_only
;
591 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
592 open_flags
= bs
->backing_hd
->open_flags
;
596 bdrv_delete(bs
->backing_hd
);
597 bs
->backing_hd
= NULL
;
598 bs_rw
= bdrv_new("");
599 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, NULL
);
602 /* try to re-open read-only */
603 bs_ro
= bdrv_new("");
604 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
607 /* drive not functional anymore */
611 bs
->backing_hd
= bs_ro
;
614 bs
->backing_hd
= bs_rw
;
617 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
618 for (i
= 0; i
< total_sectors
;) {
619 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
620 for(j
= 0; j
< n
; j
++) {
621 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
626 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
637 if (drv
->bdrv_make_empty
) {
638 ret
= drv
->bdrv_make_empty(bs
);
643 * Make sure all data we wrote to the backing device is actually
647 bdrv_flush(bs
->backing_hd
);
653 bdrv_delete(bs
->backing_hd
);
654 bs
->backing_hd
= NULL
;
655 bs_ro
= bdrv_new("");
656 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
659 /* drive not functional anymore */
663 bs
->backing_hd
= bs_ro
;
664 bs
->backing_hd
->keep_read_only
= 0;
673 * -EINVAL - backing format specified, but no file
674 * -ENOSPC - can't update the backing file because no space is left in the
676 * -ENOTSUP - format driver doesn't support changing the backing file
678 int bdrv_change_backing_file(BlockDriverState
*bs
,
679 const char *backing_file
, const char *backing_fmt
)
681 BlockDriver
*drv
= bs
->drv
;
683 if (drv
->bdrv_change_backing_file
!= NULL
) {
684 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
690 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
695 if (!bdrv_is_inserted(bs
))
701 len
= bdrv_getlength(bs
);
706 if ((offset
> len
) || (len
- offset
< size
))
712 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
715 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
718 /* return < 0 if error. See bdrv_write() for the return codes */
719 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
720 uint8_t *buf
, int nb_sectors
)
722 BlockDriver
*drv
= bs
->drv
;
726 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
729 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
732 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
733 int nb_sectors
, int dirty
)
736 unsigned long val
, idx
, bit
;
738 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
739 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
741 for (; start
<= end
; start
++) {
742 idx
= start
/ (sizeof(unsigned long) * 8);
743 bit
= start
% (sizeof(unsigned long) * 8);
744 val
= bs
->dirty_bitmap
[idx
];
746 if (!(val
& (1 << bit
))) {
751 if (val
& (1 << bit
)) {
756 bs
->dirty_bitmap
[idx
] = val
;
760 /* Return < 0 if error. Important errors are:
761 -EIO generic I/O error (may happen for all errors)
762 -ENOMEDIUM No media inserted.
763 -EINVAL Invalid sector number or nb_sectors
764 -EACCES Trying to write a read-only device
766 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
767 const uint8_t *buf
, int nb_sectors
)
769 BlockDriver
*drv
= bs
->drv
;
774 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
777 if (bs
->dirty_bitmap
) {
778 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
781 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
784 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
785 void *buf
, int count1
)
787 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
788 int len
, nb_sectors
, count
;
793 /* first read to align to sector start */
794 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
797 sector_num
= offset
>> BDRV_SECTOR_BITS
;
799 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
801 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
809 /* read the sectors "in place" */
810 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
811 if (nb_sectors
> 0) {
812 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
814 sector_num
+= nb_sectors
;
815 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
820 /* add data from the last sector */
822 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
824 memcpy(buf
, tmp_buf
, count
);
829 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
830 const void *buf
, int count1
)
832 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
833 int len
, nb_sectors
, count
;
838 /* first write to align to sector start */
839 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
842 sector_num
= offset
>> BDRV_SECTOR_BITS
;
844 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
846 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
847 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
856 /* write the sectors "in place" */
857 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
858 if (nb_sectors
> 0) {
859 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
861 sector_num
+= nb_sectors
;
862 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
867 /* add data from the last sector */
869 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
871 memcpy(tmp_buf
, buf
, count
);
872 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
879 * Truncate file to 'offset' bytes (needed only for file protocols)
881 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
883 BlockDriver
*drv
= bs
->drv
;
886 if (!drv
->bdrv_truncate
)
890 return drv
->bdrv_truncate(bs
, offset
);
894 * Length of a file in bytes. Return < 0 if error or unknown.
896 int64_t bdrv_getlength(BlockDriverState
*bs
)
898 BlockDriver
*drv
= bs
->drv
;
901 if (!drv
->bdrv_getlength
) {
903 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
905 return drv
->bdrv_getlength(bs
);
908 /* return 0 as number of sectors if no device present or error */
909 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
912 length
= bdrv_getlength(bs
);
916 length
= length
>> BDRV_SECTOR_BITS
;
917 *nb_sectors_ptr
= length
;
921 uint8_t boot_ind
; /* 0x80 - active */
922 uint8_t head
; /* starting head */
923 uint8_t sector
; /* starting sector */
924 uint8_t cyl
; /* starting cylinder */
925 uint8_t sys_ind
; /* What partition type */
926 uint8_t end_head
; /* end head */
927 uint8_t end_sector
; /* end sector */
928 uint8_t end_cyl
; /* end cylinder */
929 uint32_t start_sect
; /* starting sector counting from 0 */
930 uint32_t nr_sects
; /* nr of sectors in partition */
931 } __attribute__((packed
));
933 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
934 static int guess_disk_lchs(BlockDriverState
*bs
,
935 int *pcylinders
, int *pheads
, int *psectors
)
938 int ret
, i
, heads
, sectors
, cylinders
;
943 bdrv_get_geometry(bs
, &nb_sectors
);
945 ret
= bdrv_read(bs
, 0, buf
, 1);
948 /* test msdos magic */
949 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
951 for(i
= 0; i
< 4; i
++) {
952 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
953 nr_sects
= le32_to_cpu(p
->nr_sects
);
954 if (nr_sects
&& p
->end_head
) {
955 /* We make the assumption that the partition terminates on
956 a cylinder boundary */
957 heads
= p
->end_head
+ 1;
958 sectors
= p
->end_sector
& 63;
961 cylinders
= nb_sectors
/ (heads
* sectors
);
962 if (cylinders
< 1 || cylinders
> 16383)
966 *pcylinders
= cylinders
;
968 printf("guessed geometry: LCHS=%d %d %d\n",
969 cylinders
, heads
, sectors
);
977 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
979 int translation
, lba_detected
= 0;
980 int cylinders
, heads
, secs
;
983 /* if a geometry hint is available, use it */
984 bdrv_get_geometry(bs
, &nb_sectors
);
985 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
986 translation
= bdrv_get_translation_hint(bs
);
987 if (cylinders
!= 0) {
992 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
994 /* if heads > 16, it means that a BIOS LBA
995 translation was active, so the default
996 hardware geometry is OK */
998 goto default_geometry
;
1003 /* disable any translation to be in sync with
1004 the logical geometry */
1005 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1006 bdrv_set_translation_hint(bs
,
1007 BIOS_ATA_TRANSLATION_NONE
);
1012 /* if no geometry, use a standard physical disk geometry */
1013 cylinders
= nb_sectors
/ (16 * 63);
1015 if (cylinders
> 16383)
1017 else if (cylinders
< 2)
1022 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1023 if ((*pcyls
* *pheads
) <= 131072) {
1024 bdrv_set_translation_hint(bs
,
1025 BIOS_ATA_TRANSLATION_LARGE
);
1027 bdrv_set_translation_hint(bs
,
1028 BIOS_ATA_TRANSLATION_LBA
);
1032 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1036 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1037 int cyls
, int heads
, int secs
)
1044 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1047 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1048 type
== BDRV_TYPE_FLOPPY
));
1051 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1053 bs
->translation
= translation
;
1056 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1057 int *pcyls
, int *pheads
, int *psecs
)
1060 *pheads
= bs
->heads
;
1064 int bdrv_get_type_hint(BlockDriverState
*bs
)
1069 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1071 return bs
->translation
;
1074 int bdrv_is_removable(BlockDriverState
*bs
)
1076 return bs
->removable
;
1079 int bdrv_is_read_only(BlockDriverState
*bs
)
1081 return bs
->read_only
;
1084 int bdrv_is_sg(BlockDriverState
*bs
)
1089 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1091 return bs
->enable_write_cache
;
1094 /* XXX: no longer used */
1095 void bdrv_set_change_cb(BlockDriverState
*bs
,
1096 void (*change_cb
)(void *opaque
), void *opaque
)
1098 bs
->change_cb
= change_cb
;
1099 bs
->change_opaque
= opaque
;
1102 int bdrv_is_encrypted(BlockDriverState
*bs
)
1104 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1106 return bs
->encrypted
;
1109 int bdrv_key_required(BlockDriverState
*bs
)
1111 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1113 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1115 return (bs
->encrypted
&& !bs
->valid_key
);
1118 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1121 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1122 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1128 if (!bs
->encrypted
) {
1130 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1133 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1136 } else if (!bs
->valid_key
) {
1138 /* call the change callback now, we skipped it on open */
1139 bs
->media_changed
= 1;
1141 bs
->change_cb(bs
->change_opaque
);
1146 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1151 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1155 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1160 for (drv
= first_drv
; drv
!= NULL
; drv
= drv
->next
) {
1161 it(opaque
, drv
->format_name
);
1165 BlockDriverState
*bdrv_find(const char *name
)
1167 BlockDriverState
*bs
;
1169 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1170 if (!strcmp(name
, bs
->device_name
)) {
1177 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1179 BlockDriverState
*bs
;
1181 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1186 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1188 return bs
->device_name
;
1191 void bdrv_flush(BlockDriverState
*bs
)
1193 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1194 bs
->drv
->bdrv_flush(bs
);
1197 void bdrv_flush_all(void)
1199 BlockDriverState
*bs
;
1201 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1202 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1203 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1210 * Returns true iff the specified sector is present in the disk image. Drivers
1211 * not implementing the functionality are assumed to not support backing files,
1212 * hence all their sectors are reported as allocated.
1214 * 'pnum' is set to the number of sectors (including and immediately following
1215 * the specified sector) that are known to be in the same
1216 * allocated/unallocated state.
1218 * 'nb_sectors' is the max value 'pnum' should be set to.
1220 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1224 if (!bs
->drv
->bdrv_is_allocated
) {
1225 if (sector_num
>= bs
->total_sectors
) {
1229 n
= bs
->total_sectors
- sector_num
;
1230 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1233 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1236 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1237 BlockMonEventAction action
, int is_read
)
1240 const char *action_str
;
1243 case BDRV_ACTION_REPORT
:
1244 action_str
= "report";
1246 case BDRV_ACTION_IGNORE
:
1247 action_str
= "ignore";
1249 case BDRV_ACTION_STOP
:
1250 action_str
= "stop";
1256 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1259 is_read
? "read" : "write");
1260 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1262 qobject_decref(data
);
1265 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1268 Monitor
*mon
= opaque
;
1270 bs_dict
= qobject_to_qdict(obj
);
1272 monitor_printf(mon
, "%s: type=%s removable=%d",
1273 qdict_get_str(bs_dict
, "device"),
1274 qdict_get_str(bs_dict
, "type"),
1275 qdict_get_bool(bs_dict
, "removable"));
1277 if (qdict_get_bool(bs_dict
, "removable")) {
1278 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1281 if (qdict_haskey(bs_dict
, "inserted")) {
1282 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1284 monitor_printf(mon
, " file=");
1285 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1286 if (qdict_haskey(qdict
, "backing_file")) {
1287 monitor_printf(mon
, " backing_file=");
1288 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1290 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1291 qdict_get_bool(qdict
, "ro"),
1292 qdict_get_str(qdict
, "drv"),
1293 qdict_get_bool(qdict
, "encrypted"));
1295 monitor_printf(mon
, " [not inserted]");
1298 monitor_printf(mon
, "\n");
1301 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1303 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1307 * bdrv_info(): Block devices information
1309 * Each block device information is stored in a QDict and the
1310 * returned QObject is a QList of all devices.
1312 * The QDict contains the following:
1314 * - "device": device name
1315 * - "type": device type
1316 * - "removable": true if the device is removable, false otherwise
1317 * - "locked": true if the device is locked, false otherwise
1318 * - "inserted": only present if the device is inserted, it is a QDict
1319 * containing the following:
1320 * - "file": device file name
1321 * - "ro": true if read-only, false otherwise
1322 * - "drv": driver format name
1323 * - "backing_file": backing file name if one is used
1324 * - "encrypted": true if encrypted, false otherwise
1328 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1329 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1330 * { "device": "floppy0", "type": "floppy", "removable": true,
1331 * "locked": false } ]
1333 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1336 BlockDriverState
*bs
;
1338 bs_list
= qlist_new();
1340 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1342 const char *type
= "unknown";
1348 case BDRV_TYPE_CDROM
:
1351 case BDRV_TYPE_FLOPPY
:
1356 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1357 "'removable': %i, 'locked': %i }",
1358 bs
->device_name
, type
, bs
->removable
,
1363 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1365 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1366 "'encrypted': %i }",
1367 bs
->filename
, bs
->read_only
,
1368 bs
->drv
->format_name
,
1369 bdrv_is_encrypted(bs
));
1370 if (bs
->backing_file
[0] != '\0') {
1371 QDict
*qdict
= qobject_to_qdict(obj
);
1372 qdict_put(qdict
, "backing_file",
1373 qstring_from_str(bs
->backing_file
));
1376 qdict_put_obj(bs_dict
, "inserted", obj
);
1378 qlist_append_obj(bs_list
, bs_obj
);
1381 *ret_data
= QOBJECT(bs_list
);
1384 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1387 Monitor
*mon
= opaque
;
1389 qdict
= qobject_to_qdict(data
);
1390 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1392 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1393 monitor_printf(mon
, " rd_bytes=%" PRId64
1394 " wr_bytes=%" PRId64
1395 " rd_operations=%" PRId64
1396 " wr_operations=%" PRId64
1398 qdict_get_int(qdict
, "rd_bytes"),
1399 qdict_get_int(qdict
, "wr_bytes"),
1400 qdict_get_int(qdict
, "rd_operations"),
1401 qdict_get_int(qdict
, "wr_operations"));
1404 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1406 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1410 * bdrv_info_stats(): show block device statistics
1412 * Each device statistic information is stored in a QDict and
1413 * the returned QObject is a QList of all devices.
1415 * The QDict contains the following:
1417 * - "device": device name
1418 * - "stats": A QDict with the statistics information, it contains:
1419 * - "rd_bytes": bytes read
1420 * - "wr_bytes": bytes written
1421 * - "rd_operations": read operations
1422 * - "wr_operations": write operations
1426 * [ { "device": "ide0-hd0",
1427 * "stats": { "rd_bytes": 512,
1429 * "rd_operations": 1,
1430 * "wr_operations": 0 } },
1431 * { "device": "ide1-cd0",
1432 * "stats": { "rd_bytes": 0,
1434 * "rd_operations": 0,
1435 * "wr_operations": 0 } } ]
1437 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1441 BlockDriverState
*bs
;
1443 devices
= qlist_new();
1445 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1446 obj
= qobject_from_jsonf("{ 'device': %s, 'stats': {"
1447 "'rd_bytes': %" PRId64
","
1448 "'wr_bytes': %" PRId64
","
1449 "'rd_operations': %" PRId64
","
1450 "'wr_operations': %" PRId64
1453 bs
->rd_bytes
, bs
->wr_bytes
,
1454 bs
->rd_ops
, bs
->wr_ops
);
1455 qlist_append_obj(devices
, obj
);
1458 *ret_data
= QOBJECT(devices
);
1461 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1463 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1464 return bs
->backing_file
;
1465 else if (bs
->encrypted
)
1466 return bs
->filename
;
1471 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1472 char *filename
, int filename_size
)
1474 if (!bs
->backing_file
) {
1475 pstrcpy(filename
, filename_size
, "");
1477 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1481 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1482 const uint8_t *buf
, int nb_sectors
)
1484 BlockDriver
*drv
= bs
->drv
;
1487 if (!drv
->bdrv_write_compressed
)
1489 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1492 if (bs
->dirty_bitmap
) {
1493 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1496 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1499 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1501 BlockDriver
*drv
= bs
->drv
;
1504 if (!drv
->bdrv_get_info
)
1506 memset(bdi
, 0, sizeof(*bdi
));
1507 return drv
->bdrv_get_info(bs
, bdi
);
1510 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1511 int64_t pos
, int size
)
1513 BlockDriver
*drv
= bs
->drv
;
1516 if (!drv
->bdrv_save_vmstate
)
1518 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1521 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1522 int64_t pos
, int size
)
1524 BlockDriver
*drv
= bs
->drv
;
1527 if (!drv
->bdrv_load_vmstate
)
1529 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1532 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1534 BlockDriver
*drv
= bs
->drv
;
1536 if (!drv
|| !drv
->bdrv_debug_event
) {
1540 return drv
->bdrv_debug_event(bs
, event
);
1544 /**************************************************************/
1545 /* handling of snapshots */
1547 int bdrv_snapshot_create(BlockDriverState
*bs
,
1548 QEMUSnapshotInfo
*sn_info
)
1550 BlockDriver
*drv
= bs
->drv
;
1553 if (!drv
->bdrv_snapshot_create
)
1555 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1558 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1559 const char *snapshot_id
)
1561 BlockDriver
*drv
= bs
->drv
;
1564 if (!drv
->bdrv_snapshot_goto
)
1566 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1569 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1571 BlockDriver
*drv
= bs
->drv
;
1574 if (!drv
->bdrv_snapshot_delete
)
1576 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1579 int bdrv_snapshot_list(BlockDriverState
*bs
,
1580 QEMUSnapshotInfo
**psn_info
)
1582 BlockDriver
*drv
= bs
->drv
;
1585 if (!drv
->bdrv_snapshot_list
)
1587 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1590 #define NB_SUFFIXES 4
1592 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1594 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1599 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1602 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1603 if (size
< (10 * base
)) {
1604 snprintf(buf
, buf_size
, "%0.1f%c",
1605 (double)size
/ base
,
1608 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1609 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1610 ((size
+ (base
>> 1)) / base
),
1620 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1622 char buf1
[128], date_buf
[128], clock_buf
[128];
1632 snprintf(buf
, buf_size
,
1633 "%-10s%-20s%7s%20s%15s",
1634 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1638 ptm
= localtime(&ti
);
1639 strftime(date_buf
, sizeof(date_buf
),
1640 "%Y-%m-%d %H:%M:%S", ptm
);
1642 localtime_r(&ti
, &tm
);
1643 strftime(date_buf
, sizeof(date_buf
),
1644 "%Y-%m-%d %H:%M:%S", &tm
);
1646 secs
= sn
->vm_clock_nsec
/ 1000000000;
1647 snprintf(clock_buf
, sizeof(clock_buf
),
1648 "%02d:%02d:%02d.%03d",
1650 (int)((secs
/ 60) % 60),
1652 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1653 snprintf(buf
, buf_size
,
1654 "%-10s%-20s%7s%20s%15s",
1655 sn
->id_str
, sn
->name
,
1656 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1664 /**************************************************************/
1667 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1668 QEMUIOVector
*qiov
, int nb_sectors
,
1669 BlockDriverCompletionFunc
*cb
, void *opaque
)
1671 BlockDriver
*drv
= bs
->drv
;
1672 BlockDriverAIOCB
*ret
;
1676 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1679 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1683 /* Update stats even though technically transfer has not happened. */
1684 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1691 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1692 QEMUIOVector
*qiov
, int nb_sectors
,
1693 BlockDriverCompletionFunc
*cb
, void *opaque
)
1695 BlockDriver
*drv
= bs
->drv
;
1696 BlockDriverAIOCB
*ret
;
1702 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1705 if (bs
->dirty_bitmap
) {
1706 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1709 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1713 /* Update stats even though technically transfer has not happened. */
1714 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1722 typedef struct MultiwriteCB
{
1727 BlockDriverCompletionFunc
*cb
;
1729 QEMUIOVector
*free_qiov
;
1734 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1738 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1739 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1740 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1741 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1745 static void multiwrite_cb(void *opaque
, int ret
)
1747 MultiwriteCB
*mcb
= opaque
;
1749 if (ret
< 0 && !mcb
->error
) {
1751 multiwrite_user_cb(mcb
);
1754 mcb
->num_requests
--;
1755 if (mcb
->num_requests
== 0) {
1756 if (mcb
->error
== 0) {
1757 multiwrite_user_cb(mcb
);
1763 static int multiwrite_req_compare(const void *a
, const void *b
)
1765 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1769 * Takes a bunch of requests and tries to merge them. Returns the number of
1770 * requests that remain after merging.
1772 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1773 int num_reqs
, MultiwriteCB
*mcb
)
1777 // Sort requests by start sector
1778 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1780 // Check if adjacent requests touch the same clusters. If so, combine them,
1781 // filling up gaps with zero sectors.
1783 for (i
= 1; i
< num_reqs
; i
++) {
1785 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1787 // This handles the cases that are valid for all block drivers, namely
1788 // exactly sequential writes and overlapping writes.
1789 if (reqs
[i
].sector
<= oldreq_last
) {
1793 // The block driver may decide that it makes sense to combine requests
1794 // even if there is a gap of some sectors between them. In this case,
1795 // the gap is filled with zeros (therefore only applicable for yet
1796 // unused space in format like qcow2).
1797 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1798 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1801 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
1807 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1808 qemu_iovec_init(qiov
,
1809 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1811 // Add the first request to the merged one. If the requests are
1812 // overlapping, drop the last sectors of the first request.
1813 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1814 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1816 // We might need to add some zeros between the two requests
1817 if (reqs
[i
].sector
> oldreq_last
) {
1818 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1819 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1820 memset(buf
, 0, zero_bytes
);
1821 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1822 mcb
->callbacks
[i
].free_buf
= buf
;
1825 // Add the second request
1826 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
1828 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
1829 reqs
[outidx
].qiov
= qiov
;
1831 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
1834 reqs
[outidx
].sector
= reqs
[i
].sector
;
1835 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
1836 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
1844 * Submit multiple AIO write requests at once.
1846 * On success, the function returns 0 and all requests in the reqs array have
1847 * been submitted. In error case this function returns -1, and any of the
1848 * requests may or may not be submitted yet. In particular, this means that the
1849 * callback will be called for some of the requests, for others it won't. The
1850 * caller must check the error field of the BlockRequest to wait for the right
1851 * callbacks (if error != 0, no callback will be called).
1853 * The implementation may modify the contents of the reqs array, e.g. to merge
1854 * requests. However, the fields opaque and error are left unmodified as they
1855 * are used to signal failure for a single request to the caller.
1857 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
1859 BlockDriverAIOCB
*acb
;
1863 if (num_reqs
== 0) {
1867 // Create MultiwriteCB structure
1868 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
1869 mcb
->num_requests
= 0;
1870 mcb
->num_callbacks
= num_reqs
;
1872 for (i
= 0; i
< num_reqs
; i
++) {
1873 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
1874 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
1877 // Check for mergable requests
1878 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
1880 // Run the aio requests
1881 for (i
= 0; i
< num_reqs
; i
++) {
1882 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
1883 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
1886 // We can only fail the whole thing if no request has been
1887 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1888 // complete and report the error in the callback.
1889 if (mcb
->num_requests
== 0) {
1890 reqs
[i
].error
= -EIO
;
1893 mcb
->num_requests
++;
1894 multiwrite_cb(mcb
, -EIO
);
1898 mcb
->num_requests
++;
1909 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
1910 BlockDriverCompletionFunc
*cb
, void *opaque
)
1912 BlockDriver
*drv
= bs
->drv
;
1916 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
1919 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
1921 acb
->pool
->cancel(acb
);
1925 /**************************************************************/
1926 /* async block device emulation */
1928 typedef struct BlockDriverAIOCBSync
{
1929 BlockDriverAIOCB common
;
1932 /* vector translation state */
1936 } BlockDriverAIOCBSync
;
1938 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
1940 BlockDriverAIOCBSync
*acb
= (BlockDriverAIOCBSync
*)blockacb
;
1941 qemu_bh_delete(acb
->bh
);
1943 qemu_aio_release(acb
);
1946 static AIOPool bdrv_em_aio_pool
= {
1947 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
1948 .cancel
= bdrv_aio_cancel_em
,
1951 static void bdrv_aio_bh_cb(void *opaque
)
1953 BlockDriverAIOCBSync
*acb
= opaque
;
1956 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
1957 qemu_vfree(acb
->bounce
);
1958 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
1959 qemu_bh_delete(acb
->bh
);
1961 qemu_aio_release(acb
);
1964 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
1968 BlockDriverCompletionFunc
*cb
,
1973 BlockDriverAIOCBSync
*acb
;
1975 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
1976 acb
->is_write
= is_write
;
1978 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
1981 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
1984 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
1985 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1987 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
1990 qemu_bh_schedule(acb
->bh
);
1992 return &acb
->common
;
1995 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
1996 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
1997 BlockDriverCompletionFunc
*cb
, void *opaque
)
1999 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2002 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2003 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2004 BlockDriverCompletionFunc
*cb
, void *opaque
)
2006 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2009 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2010 BlockDriverCompletionFunc
*cb
, void *opaque
)
2012 BlockDriverAIOCBSync
*acb
;
2014 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2015 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2021 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2024 qemu_bh_schedule(acb
->bh
);
2025 return &acb
->common
;
2028 /**************************************************************/
2029 /* sync block device emulation */
2031 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2033 *(int *)opaque
= ret
;
2036 #define NOT_DONE 0x7fffffff
2038 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2039 uint8_t *buf
, int nb_sectors
)
2042 BlockDriverAIOCB
*acb
;
2046 async_context_push();
2048 async_ret
= NOT_DONE
;
2049 iov
.iov_base
= (void *)buf
;
2050 iov
.iov_len
= nb_sectors
* 512;
2051 qemu_iovec_init_external(&qiov
, &iov
, 1);
2052 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2053 bdrv_rw_em_cb
, &async_ret
);
2059 while (async_ret
== NOT_DONE
) {
2065 async_context_pop();
2069 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2070 const uint8_t *buf
, int nb_sectors
)
2073 BlockDriverAIOCB
*acb
;
2077 async_context_push();
2079 async_ret
= NOT_DONE
;
2080 iov
.iov_base
= (void *)buf
;
2081 iov
.iov_len
= nb_sectors
* 512;
2082 qemu_iovec_init_external(&qiov
, &iov
, 1);
2083 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2084 bdrv_rw_em_cb
, &async_ret
);
2089 while (async_ret
== NOT_DONE
) {
2094 async_context_pop();
2098 void bdrv_init(void)
2100 module_call_init(MODULE_INIT_BLOCK
);
2103 void bdrv_init_with_whitelist(void)
2105 use_bdrv_whitelist
= 1;
2109 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2110 BlockDriverCompletionFunc
*cb
, void *opaque
)
2112 BlockDriverAIOCB
*acb
;
2114 if (pool
->free_aiocb
) {
2115 acb
= pool
->free_aiocb
;
2116 pool
->free_aiocb
= acb
->next
;
2118 acb
= qemu_mallocz(pool
->aiocb_size
);
2123 acb
->opaque
= opaque
;
2127 void qemu_aio_release(void *p
)
2129 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2130 AIOPool
*pool
= acb
->pool
;
2131 acb
->next
= pool
->free_aiocb
;
2132 pool
->free_aiocb
= acb
;
2135 /**************************************************************/
2136 /* removable device support */
2139 * Return TRUE if the media is present
2141 int bdrv_is_inserted(BlockDriverState
*bs
)
2143 BlockDriver
*drv
= bs
->drv
;
2147 if (!drv
->bdrv_is_inserted
)
2149 ret
= drv
->bdrv_is_inserted(bs
);
2154 * Return TRUE if the media changed since the last call to this
2155 * function. It is currently only used for floppy disks
2157 int bdrv_media_changed(BlockDriverState
*bs
)
2159 BlockDriver
*drv
= bs
->drv
;
2162 if (!drv
|| !drv
->bdrv_media_changed
)
2165 ret
= drv
->bdrv_media_changed(bs
);
2166 if (ret
== -ENOTSUP
)
2167 ret
= bs
->media_changed
;
2168 bs
->media_changed
= 0;
2173 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2175 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2177 BlockDriver
*drv
= bs
->drv
;
2184 if (!drv
|| !drv
->bdrv_eject
) {
2187 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2189 if (ret
== -ENOTSUP
) {
2198 int bdrv_is_locked(BlockDriverState
*bs
)
2204 * Lock or unlock the media (if it is locked, the user won't be able
2205 * to eject it manually).
2207 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2209 BlockDriver
*drv
= bs
->drv
;
2211 bs
->locked
= locked
;
2212 if (drv
&& drv
->bdrv_set_locked
) {
2213 drv
->bdrv_set_locked(bs
, locked
);
2217 /* needed for generic scsi interface */
2219 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2221 BlockDriver
*drv
= bs
->drv
;
2223 if (drv
&& drv
->bdrv_ioctl
)
2224 return drv
->bdrv_ioctl(bs
, req
, buf
);
2228 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2229 unsigned long int req
, void *buf
,
2230 BlockDriverCompletionFunc
*cb
, void *opaque
)
2232 BlockDriver
*drv
= bs
->drv
;
2234 if (drv
&& drv
->bdrv_aio_ioctl
)
2235 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2241 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2243 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2246 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2248 int64_t bitmap_size
;
2250 bs
->dirty_count
= 0;
2252 if (!bs
->dirty_bitmap
) {
2253 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2254 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2255 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2257 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2260 if (bs
->dirty_bitmap
) {
2261 qemu_free(bs
->dirty_bitmap
);
2262 bs
->dirty_bitmap
= NULL
;
2267 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2269 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2271 if (bs
->dirty_bitmap
&&
2272 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2273 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2274 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2280 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2283 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2286 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2288 return bs
->dirty_count
;