2 * QEMU System Emulator block driver
4 * Copyright (c) 2003 Fabrice Bellard
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
24 #include "config-host.h"
25 #include "qemu-common.h"
27 #include "block_int.h"
29 #include "qemu-objects.h"
32 #include <sys/types.h>
34 #include <sys/ioctl.h>
35 #include <sys/queue.h>
45 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
46 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
47 BlockDriverCompletionFunc
*cb
, void *opaque
);
48 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
49 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
50 BlockDriverCompletionFunc
*cb
, void *opaque
);
51 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
52 BlockDriverCompletionFunc
*cb
, void *opaque
);
53 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
54 uint8_t *buf
, int nb_sectors
);
55 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
56 const uint8_t *buf
, int nb_sectors
);
57 static BlockDriver
*find_protocol(const char *filename
);
59 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
60 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
62 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
63 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
65 /* If non-zero, use only whitelisted block drivers */
66 static int use_bdrv_whitelist
;
68 int path_is_absolute(const char *path
)
72 /* specific case for names like: "\\.\d:" */
73 if (*path
== '/' || *path
== '\\')
76 p
= strchr(path
, ':');
82 return (*p
== '/' || *p
== '\\');
88 /* if filename is absolute, just copy it to dest. Otherwise, build a
89 path to it by considering it is relative to base_path. URL are
91 void path_combine(char *dest
, int dest_size
,
92 const char *base_path
,
100 if (path_is_absolute(filename
)) {
101 pstrcpy(dest
, dest_size
, filename
);
103 p
= strchr(base_path
, ':');
108 p1
= strrchr(base_path
, '/');
112 p2
= strrchr(base_path
, '\\');
124 if (len
> dest_size
- 1)
126 memcpy(dest
, base_path
, len
);
128 pstrcat(dest
, dest_size
, filename
);
132 void bdrv_register(BlockDriver
*bdrv
)
134 if (!bdrv
->bdrv_aio_readv
) {
135 /* add AIO emulation layer */
136 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
137 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
138 } else if (!bdrv
->bdrv_read
) {
139 /* add synchronous IO emulation layer */
140 bdrv
->bdrv_read
= bdrv_read_em
;
141 bdrv
->bdrv_write
= bdrv_write_em
;
144 if (!bdrv
->bdrv_aio_flush
)
145 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
147 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
150 /* create a new block device (by default it is empty) */
151 BlockDriverState
*bdrv_new(const char *device_name
)
153 BlockDriverState
*bs
;
155 bs
= qemu_mallocz(sizeof(BlockDriverState
));
156 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
157 if (device_name
[0] != '\0') {
158 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
163 BlockDriver
*bdrv_find_format(const char *format_name
)
166 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
167 if (!strcmp(drv1
->format_name
, format_name
)) {
174 static int bdrv_is_whitelisted(BlockDriver
*drv
)
176 static const char *whitelist
[] = {
177 CONFIG_BDRV_WHITELIST
182 return 1; /* no whitelist, anything goes */
184 for (p
= whitelist
; *p
; p
++) {
185 if (!strcmp(drv
->format_name
, *p
)) {
192 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
194 BlockDriver
*drv
= bdrv_find_format(format_name
);
195 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
198 int bdrv_create(BlockDriver
*drv
, const char* filename
,
199 QEMUOptionParameter
*options
)
201 if (!drv
->bdrv_create
)
204 return drv
->bdrv_create(filename
, options
);
207 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
211 drv
= find_protocol(filename
);
213 drv
= bdrv_find_format("file");
216 return bdrv_create(drv
, filename
, options
);
220 void get_tmp_filename(char *filename
, int size
)
222 char temp_dir
[MAX_PATH
];
224 GetTempPath(MAX_PATH
, temp_dir
);
225 GetTempFileName(temp_dir
, "qem", 0, filename
);
228 void get_tmp_filename(char *filename
, int size
)
232 /* XXX: race condition possible */
233 tmpdir
= getenv("TMPDIR");
236 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
237 fd
= mkstemp(filename
);
243 static int is_windows_drive_prefix(const char *filename
)
245 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
246 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
250 int is_windows_drive(const char *filename
)
252 if (is_windows_drive_prefix(filename
) &&
255 if (strstart(filename
, "\\\\.\\", NULL
) ||
256 strstart(filename
, "//./", NULL
))
263 * Detect host devices. By convention, /dev/cdrom[N] is always
264 * recognized as a host CDROM.
266 static BlockDriver
*find_hdev_driver(const char *filename
)
268 int score_max
= 0, score
;
269 BlockDriver
*drv
= NULL
, *d
;
271 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
272 if (d
->bdrv_probe_device
) {
273 score
= d
->bdrv_probe_device(filename
);
274 if (score
> score_max
) {
284 static BlockDriver
*find_protocol(const char *filename
)
292 if (is_windows_drive(filename
) ||
293 is_windows_drive_prefix(filename
))
294 return bdrv_find_format("file");
296 p
= strchr(filename
, ':');
298 drv1
= find_hdev_driver(filename
);
300 drv1
= bdrv_find_format("file");
305 if (len
> sizeof(protocol
) - 1)
306 len
= sizeof(protocol
) - 1;
307 memcpy(protocol
, filename
, len
);
308 protocol
[len
] = '\0';
309 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
310 if (drv1
->protocol_name
&&
311 !strcmp(drv1
->protocol_name
, protocol
)) {
318 static BlockDriver
*find_image_format(const char *filename
)
320 int ret
, score
, score_max
;
321 BlockDriver
*drv1
, *drv
;
323 BlockDriverState
*bs
;
325 drv
= find_protocol(filename
);
326 /* no need to test disk image formats for vvfat */
327 if (drv
&& strcmp(drv
->format_name
, "vvfat") == 0)
330 ret
= bdrv_file_open(&bs
, filename
, 0);
333 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
341 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
342 if (drv1
->bdrv_probe
) {
343 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
344 if (score
> score_max
) {
354 * Common part for opening disk images and files
356 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
357 int flags
, BlockDriver
*drv
)
363 bs
->is_temporary
= 0;
366 bs
->open_flags
= flags
;
367 /* buffer_alignment defaulted to 512, drivers can change this value */
368 bs
->buffer_alignment
= 512;
370 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
372 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
377 bs
->opaque
= qemu_mallocz(drv
->instance_size
);
380 * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
381 * write cache to the guest. We do need the fdatasync to flush
382 * out transactions for block allocations, and we maybe have a
383 * volatile write cache in our backing device to deal with.
385 if (flags
& (BDRV_O_CACHE_WB
|BDRV_O_NOCACHE
))
386 bs
->enable_write_cache
= 1;
389 * Clear flags that are internal to the block layer before opening the
392 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
395 * Snapshots should be writeable.
397 if (bs
->is_temporary
) {
398 open_flags
|= BDRV_O_RDWR
;
401 ret
= drv
->bdrv_open(bs
, filename
, open_flags
);
406 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
407 if (drv
->bdrv_getlength
) {
408 bs
->total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
411 if (bs
->is_temporary
) {
418 qemu_free(bs
->opaque
);
425 * Opens a file using a protocol (file, host_device, nbd, ...)
427 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
429 BlockDriverState
*bs
;
433 drv
= find_protocol(filename
);
439 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
450 * Opens a disk image (raw, qcow2, vmdk, ...)
452 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
457 if (flags
& BDRV_O_SNAPSHOT
) {
458 BlockDriverState
*bs1
;
461 BlockDriver
*bdrv_qcow2
;
462 QEMUOptionParameter
*options
;
463 char tmp_filename
[PATH_MAX
];
464 char backing_filename
[PATH_MAX
];
466 /* if snapshot, we create a temporary backing file and open it
467 instead of opening 'filename' directly */
469 /* if there is a backing file, use it */
471 ret
= bdrv_open(bs1
, filename
, 0, drv
);
476 total_size
= bdrv_getlength(bs1
) >> BDRV_SECTOR_BITS
;
478 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
483 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
485 /* Real path is meaningless for protocols */
487 snprintf(backing_filename
, sizeof(backing_filename
),
489 else if (!realpath(filename
, backing_filename
))
492 bdrv_qcow2
= bdrv_find_format("qcow2");
493 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
495 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
* 512);
496 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
498 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
502 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
507 filename
= tmp_filename
;
509 bs
->is_temporary
= 1;
512 /* Find the right image format driver */
514 drv
= find_image_format(filename
);
519 goto unlink_and_fail
;
523 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
525 goto unlink_and_fail
;
528 /* If there is a backing file, use it */
529 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
530 char backing_filename
[PATH_MAX
];
532 BlockDriver
*back_drv
= NULL
;
534 bs
->backing_hd
= bdrv_new("");
535 path_combine(backing_filename
, sizeof(backing_filename
),
536 filename
, bs
->backing_file
);
537 if (bs
->backing_format
[0] != '\0')
538 back_drv
= bdrv_find_format(bs
->backing_format
);
540 /* backing files always opened read-only */
542 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
544 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
549 if (bs
->is_temporary
) {
550 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
552 /* base image inherits from "parent" */
553 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
557 if (!bdrv_key_required(bs
)) {
558 /* call the change callback */
559 bs
->media_changed
= 1;
561 bs
->change_cb(bs
->change_opaque
);
567 if (bs
->is_temporary
) {
573 void bdrv_close(BlockDriverState
*bs
)
577 bdrv_delete(bs
->backing_hd
);
578 bs
->drv
->bdrv_close(bs
);
579 qemu_free(bs
->opaque
);
581 if (bs
->is_temporary
) {
582 unlink(bs
->filename
);
588 /* call the change callback */
589 bs
->media_changed
= 1;
591 bs
->change_cb(bs
->change_opaque
);
595 void bdrv_delete(BlockDriverState
*bs
)
597 /* remove from list, if necessary */
598 if (bs
->device_name
[0] != '\0') {
599 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
607 * Run consistency checks on an image
609 * Returns the number of errors or -errno when an internal error occurs
611 int bdrv_check(BlockDriverState
*bs
)
613 if (bs
->drv
->bdrv_check
== NULL
) {
617 return bs
->drv
->bdrv_check(bs
);
620 /* commit COW file into the raw image */
621 int bdrv_commit(BlockDriverState
*bs
)
623 BlockDriver
*drv
= bs
->drv
;
624 int64_t i
, total_sectors
;
625 int n
, j
, ro
, open_flags
;
626 int ret
= 0, rw_ret
= 0;
627 unsigned char sector
[512];
629 BlockDriverState
*bs_rw
, *bs_ro
;
634 if (!bs
->backing_hd
) {
638 if (bs
->backing_hd
->keep_read_only
) {
642 ro
= bs
->backing_hd
->read_only
;
643 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
644 open_flags
= bs
->backing_hd
->open_flags
;
648 bdrv_delete(bs
->backing_hd
);
649 bs
->backing_hd
= NULL
;
650 bs_rw
= bdrv_new("");
651 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
, NULL
);
654 /* try to re-open read-only */
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
;
666 bs
->backing_hd
= bs_rw
;
669 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
670 for (i
= 0; i
< total_sectors
;) {
671 if (drv
->bdrv_is_allocated(bs
, i
, 65536, &n
)) {
672 for(j
= 0; j
< n
; j
++) {
673 if (bdrv_read(bs
, i
, sector
, 1) != 0) {
678 if (bdrv_write(bs
->backing_hd
, i
, sector
, 1) != 0) {
689 if (drv
->bdrv_make_empty
) {
690 ret
= drv
->bdrv_make_empty(bs
);
695 * Make sure all data we wrote to the backing device is actually
699 bdrv_flush(bs
->backing_hd
);
705 bdrv_delete(bs
->backing_hd
);
706 bs
->backing_hd
= NULL
;
707 bs_ro
= bdrv_new("");
708 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
, NULL
);
711 /* drive not functional anymore */
715 bs
->backing_hd
= bs_ro
;
716 bs
->backing_hd
->keep_read_only
= 0;
725 * -EINVAL - backing format specified, but no file
726 * -ENOSPC - can't update the backing file because no space is left in the
728 * -ENOTSUP - format driver doesn't support changing the backing file
730 int bdrv_change_backing_file(BlockDriverState
*bs
,
731 const char *backing_file
, const char *backing_fmt
)
733 BlockDriver
*drv
= bs
->drv
;
735 if (drv
->bdrv_change_backing_file
!= NULL
) {
736 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
742 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
747 if (!bdrv_is_inserted(bs
))
753 len
= bdrv_getlength(bs
);
758 if ((offset
> len
) || (len
- offset
< size
))
764 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
767 return bdrv_check_byte_request(bs
, sector_num
* 512, nb_sectors
* 512);
770 /* return < 0 if error. See bdrv_write() for the return codes */
771 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
772 uint8_t *buf
, int nb_sectors
)
774 BlockDriver
*drv
= bs
->drv
;
778 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
781 return drv
->bdrv_read(bs
, sector_num
, buf
, nb_sectors
);
784 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
785 int nb_sectors
, int dirty
)
788 unsigned long val
, idx
, bit
;
790 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
791 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
793 for (; start
<= end
; start
++) {
794 idx
= start
/ (sizeof(unsigned long) * 8);
795 bit
= start
% (sizeof(unsigned long) * 8);
796 val
= bs
->dirty_bitmap
[idx
];
798 if (!(val
& (1 << bit
))) {
803 if (val
& (1 << bit
)) {
808 bs
->dirty_bitmap
[idx
] = val
;
812 /* Return < 0 if error. Important errors are:
813 -EIO generic I/O error (may happen for all errors)
814 -ENOMEDIUM No media inserted.
815 -EINVAL Invalid sector number or nb_sectors
816 -EACCES Trying to write a read-only device
818 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
819 const uint8_t *buf
, int nb_sectors
)
821 BlockDriver
*drv
= bs
->drv
;
826 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
829 if (bs
->dirty_bitmap
) {
830 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
833 return drv
->bdrv_write(bs
, sector_num
, buf
, nb_sectors
);
836 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
837 void *buf
, int count1
)
839 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
840 int len
, nb_sectors
, count
;
845 /* first read to align to sector start */
846 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
849 sector_num
= offset
>> BDRV_SECTOR_BITS
;
851 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
853 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
861 /* read the sectors "in place" */
862 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
863 if (nb_sectors
> 0) {
864 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
866 sector_num
+= nb_sectors
;
867 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
872 /* add data from the last sector */
874 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
876 memcpy(buf
, tmp_buf
, count
);
881 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
882 const void *buf
, int count1
)
884 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
885 int len
, nb_sectors
, count
;
890 /* first write to align to sector start */
891 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
894 sector_num
= offset
>> BDRV_SECTOR_BITS
;
896 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
898 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
899 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
908 /* write the sectors "in place" */
909 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
910 if (nb_sectors
> 0) {
911 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
913 sector_num
+= nb_sectors
;
914 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
919 /* add data from the last sector */
921 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
923 memcpy(tmp_buf
, buf
, count
);
924 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
931 * Truncate file to 'offset' bytes (needed only for file protocols)
933 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
935 BlockDriver
*drv
= bs
->drv
;
938 if (!drv
->bdrv_truncate
)
942 return drv
->bdrv_truncate(bs
, offset
);
946 * Length of a file in bytes. Return < 0 if error or unknown.
948 int64_t bdrv_getlength(BlockDriverState
*bs
)
950 BlockDriver
*drv
= bs
->drv
;
953 if (!drv
->bdrv_getlength
) {
955 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
957 return drv
->bdrv_getlength(bs
);
960 /* return 0 as number of sectors if no device present or error */
961 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
964 length
= bdrv_getlength(bs
);
968 length
= length
>> BDRV_SECTOR_BITS
;
969 *nb_sectors_ptr
= length
;
973 uint8_t boot_ind
; /* 0x80 - active */
974 uint8_t head
; /* starting head */
975 uint8_t sector
; /* starting sector */
976 uint8_t cyl
; /* starting cylinder */
977 uint8_t sys_ind
; /* What partition type */
978 uint8_t end_head
; /* end head */
979 uint8_t end_sector
; /* end sector */
980 uint8_t end_cyl
; /* end cylinder */
981 uint32_t start_sect
; /* starting sector counting from 0 */
982 uint32_t nr_sects
; /* nr of sectors in partition */
983 } __attribute__((packed
));
985 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
986 static int guess_disk_lchs(BlockDriverState
*bs
,
987 int *pcylinders
, int *pheads
, int *psectors
)
990 int ret
, i
, heads
, sectors
, cylinders
;
995 bdrv_get_geometry(bs
, &nb_sectors
);
997 ret
= bdrv_read(bs
, 0, buf
, 1);
1000 /* test msdos magic */
1001 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1003 for(i
= 0; i
< 4; i
++) {
1004 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1005 nr_sects
= le32_to_cpu(p
->nr_sects
);
1006 if (nr_sects
&& p
->end_head
) {
1007 /* We make the assumption that the partition terminates on
1008 a cylinder boundary */
1009 heads
= p
->end_head
+ 1;
1010 sectors
= p
->end_sector
& 63;
1013 cylinders
= nb_sectors
/ (heads
* sectors
);
1014 if (cylinders
< 1 || cylinders
> 16383)
1017 *psectors
= sectors
;
1018 *pcylinders
= cylinders
;
1020 printf("guessed geometry: LCHS=%d %d %d\n",
1021 cylinders
, heads
, sectors
);
1029 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1031 int translation
, lba_detected
= 0;
1032 int cylinders
, heads
, secs
;
1033 uint64_t nb_sectors
;
1035 /* if a geometry hint is available, use it */
1036 bdrv_get_geometry(bs
, &nb_sectors
);
1037 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1038 translation
= bdrv_get_translation_hint(bs
);
1039 if (cylinders
!= 0) {
1044 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1046 /* if heads > 16, it means that a BIOS LBA
1047 translation was active, so the default
1048 hardware geometry is OK */
1050 goto default_geometry
;
1055 /* disable any translation to be in sync with
1056 the logical geometry */
1057 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1058 bdrv_set_translation_hint(bs
,
1059 BIOS_ATA_TRANSLATION_NONE
);
1064 /* if no geometry, use a standard physical disk geometry */
1065 cylinders
= nb_sectors
/ (16 * 63);
1067 if (cylinders
> 16383)
1069 else if (cylinders
< 2)
1074 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1075 if ((*pcyls
* *pheads
) <= 131072) {
1076 bdrv_set_translation_hint(bs
,
1077 BIOS_ATA_TRANSLATION_LARGE
);
1079 bdrv_set_translation_hint(bs
,
1080 BIOS_ATA_TRANSLATION_LBA
);
1084 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1088 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1089 int cyls
, int heads
, int secs
)
1096 void bdrv_set_type_hint(BlockDriverState
*bs
, int type
)
1099 bs
->removable
= ((type
== BDRV_TYPE_CDROM
||
1100 type
== BDRV_TYPE_FLOPPY
));
1103 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1105 bs
->translation
= translation
;
1108 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1109 int *pcyls
, int *pheads
, int *psecs
)
1112 *pheads
= bs
->heads
;
1116 int bdrv_get_type_hint(BlockDriverState
*bs
)
1121 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1123 return bs
->translation
;
1126 int bdrv_is_removable(BlockDriverState
*bs
)
1128 return bs
->removable
;
1131 int bdrv_is_read_only(BlockDriverState
*bs
)
1133 return bs
->read_only
;
1136 int bdrv_is_sg(BlockDriverState
*bs
)
1141 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1143 return bs
->enable_write_cache
;
1146 /* XXX: no longer used */
1147 void bdrv_set_change_cb(BlockDriverState
*bs
,
1148 void (*change_cb
)(void *opaque
), void *opaque
)
1150 bs
->change_cb
= change_cb
;
1151 bs
->change_opaque
= opaque
;
1154 int bdrv_is_encrypted(BlockDriverState
*bs
)
1156 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1158 return bs
->encrypted
;
1161 int bdrv_key_required(BlockDriverState
*bs
)
1163 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1165 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1167 return (bs
->encrypted
&& !bs
->valid_key
);
1170 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1173 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1174 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1180 if (!bs
->encrypted
) {
1182 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1185 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1188 } else if (!bs
->valid_key
) {
1190 /* call the change callback now, we skipped it on open */
1191 bs
->media_changed
= 1;
1193 bs
->change_cb(bs
->change_opaque
);
1198 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1203 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1207 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1212 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1213 it(opaque
, drv
->format_name
);
1217 BlockDriverState
*bdrv_find(const char *name
)
1219 BlockDriverState
*bs
;
1221 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1222 if (!strcmp(name
, bs
->device_name
)) {
1229 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1231 BlockDriverState
*bs
;
1233 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1238 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1240 return bs
->device_name
;
1243 void bdrv_flush(BlockDriverState
*bs
)
1245 if (bs
->drv
&& bs
->drv
->bdrv_flush
)
1246 bs
->drv
->bdrv_flush(bs
);
1249 void bdrv_flush_all(void)
1251 BlockDriverState
*bs
;
1253 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1254 if (bs
->drv
&& !bdrv_is_read_only(bs
) &&
1255 (!bdrv_is_removable(bs
) || bdrv_is_inserted(bs
))) {
1262 * Returns true iff the specified sector is present in the disk image. Drivers
1263 * not implementing the functionality are assumed to not support backing files,
1264 * hence all their sectors are reported as allocated.
1266 * 'pnum' is set to the number of sectors (including and immediately following
1267 * the specified sector) that are known to be in the same
1268 * allocated/unallocated state.
1270 * 'nb_sectors' is the max value 'pnum' should be set to.
1272 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1276 if (!bs
->drv
->bdrv_is_allocated
) {
1277 if (sector_num
>= bs
->total_sectors
) {
1281 n
= bs
->total_sectors
- sector_num
;
1282 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1285 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1288 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1289 BlockMonEventAction action
, int is_read
)
1292 const char *action_str
;
1295 case BDRV_ACTION_REPORT
:
1296 action_str
= "report";
1298 case BDRV_ACTION_IGNORE
:
1299 action_str
= "ignore";
1301 case BDRV_ACTION_STOP
:
1302 action_str
= "stop";
1308 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1311 is_read
? "read" : "write");
1312 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1314 qobject_decref(data
);
1317 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1320 Monitor
*mon
= opaque
;
1322 bs_dict
= qobject_to_qdict(obj
);
1324 monitor_printf(mon
, "%s: type=%s removable=%d",
1325 qdict_get_str(bs_dict
, "device"),
1326 qdict_get_str(bs_dict
, "type"),
1327 qdict_get_bool(bs_dict
, "removable"));
1329 if (qdict_get_bool(bs_dict
, "removable")) {
1330 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1333 if (qdict_haskey(bs_dict
, "inserted")) {
1334 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1336 monitor_printf(mon
, " file=");
1337 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1338 if (qdict_haskey(qdict
, "backing_file")) {
1339 monitor_printf(mon
, " backing_file=");
1340 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1342 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1343 qdict_get_bool(qdict
, "ro"),
1344 qdict_get_str(qdict
, "drv"),
1345 qdict_get_bool(qdict
, "encrypted"));
1347 monitor_printf(mon
, " [not inserted]");
1350 monitor_printf(mon
, "\n");
1353 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1355 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1359 * bdrv_info(): Block devices information
1361 * Each block device information is stored in a QDict and the
1362 * returned QObject is a QList of all devices.
1364 * The QDict contains the following:
1366 * - "device": device name
1367 * - "type": device type
1368 * - "removable": true if the device is removable, false otherwise
1369 * - "locked": true if the device is locked, false otherwise
1370 * - "inserted": only present if the device is inserted, it is a QDict
1371 * containing the following:
1372 * - "file": device file name
1373 * - "ro": true if read-only, false otherwise
1374 * - "drv": driver format name
1375 * - "backing_file": backing file name if one is used
1376 * - "encrypted": true if encrypted, false otherwise
1380 * [ { "device": "ide0-hd0", "type": "hd", "removable": false, "locked": false,
1381 * "inserted": { "file": "/tmp/foobar", "ro": false, "drv": "qcow2" } },
1382 * { "device": "floppy0", "type": "floppy", "removable": true,
1383 * "locked": false } ]
1385 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1388 BlockDriverState
*bs
;
1390 bs_list
= qlist_new();
1392 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1394 const char *type
= "unknown";
1400 case BDRV_TYPE_CDROM
:
1403 case BDRV_TYPE_FLOPPY
:
1408 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': %s, "
1409 "'removable': %i, 'locked': %i }",
1410 bs
->device_name
, type
, bs
->removable
,
1415 QDict
*bs_dict
= qobject_to_qdict(bs_obj
);
1417 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1418 "'encrypted': %i }",
1419 bs
->filename
, bs
->read_only
,
1420 bs
->drv
->format_name
,
1421 bdrv_is_encrypted(bs
));
1422 if (bs
->backing_file
[0] != '\0') {
1423 QDict
*qdict
= qobject_to_qdict(obj
);
1424 qdict_put(qdict
, "backing_file",
1425 qstring_from_str(bs
->backing_file
));
1428 qdict_put_obj(bs_dict
, "inserted", obj
);
1430 qlist_append_obj(bs_list
, bs_obj
);
1433 *ret_data
= QOBJECT(bs_list
);
1436 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1439 Monitor
*mon
= opaque
;
1441 qdict
= qobject_to_qdict(data
);
1442 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1444 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
1445 monitor_printf(mon
, " rd_bytes=%" PRId64
1446 " wr_bytes=%" PRId64
1447 " rd_operations=%" PRId64
1448 " wr_operations=%" PRId64
1450 qdict_get_int(qdict
, "rd_bytes"),
1451 qdict_get_int(qdict
, "wr_bytes"),
1452 qdict_get_int(qdict
, "rd_operations"),
1453 qdict_get_int(qdict
, "wr_operations"));
1456 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
1458 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
1462 * bdrv_info_stats(): show block device statistics
1464 * Each device statistic information is stored in a QDict and
1465 * the returned QObject is a QList of all devices.
1467 * The QDict contains the following:
1469 * - "device": device name
1470 * - "stats": A QDict with the statistics information, it contains:
1471 * - "rd_bytes": bytes read
1472 * - "wr_bytes": bytes written
1473 * - "rd_operations": read operations
1474 * - "wr_operations": write operations
1478 * [ { "device": "ide0-hd0",
1479 * "stats": { "rd_bytes": 512,
1481 * "rd_operations": 1,
1482 * "wr_operations": 0 } },
1483 * { "device": "ide1-cd0",
1484 * "stats": { "rd_bytes": 0,
1486 * "rd_operations": 0,
1487 * "wr_operations": 0 } } ]
1489 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
1493 BlockDriverState
*bs
;
1495 devices
= qlist_new();
1497 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1498 obj
= qobject_from_jsonf("{ 'device': %s, 'stats': {"
1499 "'rd_bytes': %" PRId64
","
1500 "'wr_bytes': %" PRId64
","
1501 "'rd_operations': %" PRId64
","
1502 "'wr_operations': %" PRId64
1505 bs
->rd_bytes
, bs
->wr_bytes
,
1506 bs
->rd_ops
, bs
->wr_ops
);
1507 qlist_append_obj(devices
, obj
);
1510 *ret_data
= QOBJECT(devices
);
1513 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
1515 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1516 return bs
->backing_file
;
1517 else if (bs
->encrypted
)
1518 return bs
->filename
;
1523 void bdrv_get_backing_filename(BlockDriverState
*bs
,
1524 char *filename
, int filename_size
)
1526 if (!bs
->backing_file
) {
1527 pstrcpy(filename
, filename_size
, "");
1529 pstrcpy(filename
, filename_size
, bs
->backing_file
);
1533 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
1534 const uint8_t *buf
, int nb_sectors
)
1536 BlockDriver
*drv
= bs
->drv
;
1539 if (!drv
->bdrv_write_compressed
)
1541 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1544 if (bs
->dirty_bitmap
) {
1545 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1548 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
1551 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
1553 BlockDriver
*drv
= bs
->drv
;
1556 if (!drv
->bdrv_get_info
)
1558 memset(bdi
, 0, sizeof(*bdi
));
1559 return drv
->bdrv_get_info(bs
, bdi
);
1562 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
1563 int64_t pos
, int size
)
1565 BlockDriver
*drv
= bs
->drv
;
1568 if (!drv
->bdrv_save_vmstate
)
1570 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
1573 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
1574 int64_t pos
, int size
)
1576 BlockDriver
*drv
= bs
->drv
;
1579 if (!drv
->bdrv_load_vmstate
)
1581 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
1584 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
1586 BlockDriver
*drv
= bs
->drv
;
1588 if (!drv
|| !drv
->bdrv_debug_event
) {
1592 return drv
->bdrv_debug_event(bs
, event
);
1596 /**************************************************************/
1597 /* handling of snapshots */
1599 int bdrv_snapshot_create(BlockDriverState
*bs
,
1600 QEMUSnapshotInfo
*sn_info
)
1602 BlockDriver
*drv
= bs
->drv
;
1605 if (!drv
->bdrv_snapshot_create
)
1607 return drv
->bdrv_snapshot_create(bs
, sn_info
);
1610 int bdrv_snapshot_goto(BlockDriverState
*bs
,
1611 const char *snapshot_id
)
1613 BlockDriver
*drv
= bs
->drv
;
1616 if (!drv
->bdrv_snapshot_goto
)
1618 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
1621 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
1623 BlockDriver
*drv
= bs
->drv
;
1626 if (!drv
->bdrv_snapshot_delete
)
1628 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
1631 int bdrv_snapshot_list(BlockDriverState
*bs
,
1632 QEMUSnapshotInfo
**psn_info
)
1634 BlockDriver
*drv
= bs
->drv
;
1637 if (!drv
->bdrv_snapshot_list
)
1639 return drv
->bdrv_snapshot_list(bs
, psn_info
);
1642 #define NB_SUFFIXES 4
1644 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
1646 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
1651 snprintf(buf
, buf_size
, "%" PRId64
, size
);
1654 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
1655 if (size
< (10 * base
)) {
1656 snprintf(buf
, buf_size
, "%0.1f%c",
1657 (double)size
/ base
,
1660 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
1661 snprintf(buf
, buf_size
, "%" PRId64
"%c",
1662 ((size
+ (base
>> 1)) / base
),
1672 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
1674 char buf1
[128], date_buf
[128], clock_buf
[128];
1684 snprintf(buf
, buf_size
,
1685 "%-10s%-20s%7s%20s%15s",
1686 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
1690 ptm
= localtime(&ti
);
1691 strftime(date_buf
, sizeof(date_buf
),
1692 "%Y-%m-%d %H:%M:%S", ptm
);
1694 localtime_r(&ti
, &tm
);
1695 strftime(date_buf
, sizeof(date_buf
),
1696 "%Y-%m-%d %H:%M:%S", &tm
);
1698 secs
= sn
->vm_clock_nsec
/ 1000000000;
1699 snprintf(clock_buf
, sizeof(clock_buf
),
1700 "%02d:%02d:%02d.%03d",
1702 (int)((secs
/ 60) % 60),
1704 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
1705 snprintf(buf
, buf_size
,
1706 "%-10s%-20s%7s%20s%15s",
1707 sn
->id_str
, sn
->name
,
1708 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
1716 /**************************************************************/
1719 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
1720 QEMUIOVector
*qiov
, int nb_sectors
,
1721 BlockDriverCompletionFunc
*cb
, void *opaque
)
1723 BlockDriver
*drv
= bs
->drv
;
1724 BlockDriverAIOCB
*ret
;
1728 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1731 ret
= drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
1735 /* Update stats even though technically transfer has not happened. */
1736 bs
->rd_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1743 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
1744 QEMUIOVector
*qiov
, int nb_sectors
,
1745 BlockDriverCompletionFunc
*cb
, void *opaque
)
1747 BlockDriver
*drv
= bs
->drv
;
1748 BlockDriverAIOCB
*ret
;
1754 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
1757 if (bs
->dirty_bitmap
) {
1758 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1761 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
1765 /* Update stats even though technically transfer has not happened. */
1766 bs
->wr_bytes
+= (unsigned) nb_sectors
* BDRV_SECTOR_SIZE
;
1774 typedef struct MultiwriteCB
{
1779 BlockDriverCompletionFunc
*cb
;
1781 QEMUIOVector
*free_qiov
;
1786 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
1790 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
1791 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
1792 if (mcb
->callbacks
[i
].free_qiov
) {
1793 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
1795 qemu_free(mcb
->callbacks
[i
].free_qiov
);
1796 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
1800 static void multiwrite_cb(void *opaque
, int ret
)
1802 MultiwriteCB
*mcb
= opaque
;
1804 if (ret
< 0 && !mcb
->error
) {
1806 multiwrite_user_cb(mcb
);
1809 mcb
->num_requests
--;
1810 if (mcb
->num_requests
== 0) {
1811 if (mcb
->error
== 0) {
1812 multiwrite_user_cb(mcb
);
1818 static int multiwrite_req_compare(const void *a
, const void *b
)
1820 return (((BlockRequest
*) a
)->sector
- ((BlockRequest
*) b
)->sector
);
1824 * Takes a bunch of requests and tries to merge them. Returns the number of
1825 * requests that remain after merging.
1827 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
1828 int num_reqs
, MultiwriteCB
*mcb
)
1832 // Sort requests by start sector
1833 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
1835 // Check if adjacent requests touch the same clusters. If so, combine them,
1836 // filling up gaps with zero sectors.
1838 for (i
= 1; i
< num_reqs
; i
++) {
1840 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
1842 // This handles the cases that are valid for all block drivers, namely
1843 // exactly sequential writes and overlapping writes.
1844 if (reqs
[i
].sector
<= oldreq_last
) {
1848 // The block driver may decide that it makes sense to combine requests
1849 // even if there is a gap of some sectors between them. In this case,
1850 // the gap is filled with zeros (therefore only applicable for yet
1851 // unused space in format like qcow2).
1852 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
1853 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
1856 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
1862 QEMUIOVector
*qiov
= qemu_mallocz(sizeof(*qiov
));
1863 qemu_iovec_init(qiov
,
1864 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
1866 // Add the first request to the merged one. If the requests are
1867 // overlapping, drop the last sectors of the first request.
1868 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
1869 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
1871 // We might need to add some zeros between the two requests
1872 if (reqs
[i
].sector
> oldreq_last
) {
1873 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
1874 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
1875 memset(buf
, 0, zero_bytes
);
1876 qemu_iovec_add(qiov
, buf
, zero_bytes
);
1877 mcb
->callbacks
[i
].free_buf
= buf
;
1880 // Add the second request
1881 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
1883 reqs
[outidx
].nb_sectors
+= reqs
[i
].nb_sectors
;
1884 reqs
[outidx
].qiov
= qiov
;
1886 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
1889 reqs
[outidx
].sector
= reqs
[i
].sector
;
1890 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
1891 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
1899 * Submit multiple AIO write requests at once.
1901 * On success, the function returns 0 and all requests in the reqs array have
1902 * been submitted. In error case this function returns -1, and any of the
1903 * requests may or may not be submitted yet. In particular, this means that the
1904 * callback will be called for some of the requests, for others it won't. The
1905 * caller must check the error field of the BlockRequest to wait for the right
1906 * callbacks (if error != 0, no callback will be called).
1908 * The implementation may modify the contents of the reqs array, e.g. to merge
1909 * requests. However, the fields opaque and error are left unmodified as they
1910 * are used to signal failure for a single request to the caller.
1912 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
1914 BlockDriverAIOCB
*acb
;
1918 if (num_reqs
== 0) {
1922 // Create MultiwriteCB structure
1923 mcb
= qemu_mallocz(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
1924 mcb
->num_requests
= 0;
1925 mcb
->num_callbacks
= num_reqs
;
1927 for (i
= 0; i
< num_reqs
; i
++) {
1928 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
1929 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
1932 // Check for mergable requests
1933 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
1935 // Run the aio requests
1936 for (i
= 0; i
< num_reqs
; i
++) {
1937 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
1938 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
1941 // We can only fail the whole thing if no request has been
1942 // submitted yet. Otherwise we'll wait for the submitted AIOs to
1943 // complete and report the error in the callback.
1944 if (mcb
->num_requests
== 0) {
1945 reqs
[i
].error
= -EIO
;
1948 mcb
->num_requests
++;
1949 multiwrite_cb(mcb
, -EIO
);
1953 mcb
->num_requests
++;
1964 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
1965 BlockDriverCompletionFunc
*cb
, void *opaque
)
1967 BlockDriver
*drv
= bs
->drv
;
1971 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
1974 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
1976 acb
->pool
->cancel(acb
);
1980 /**************************************************************/
1981 /* async block device emulation */
1983 typedef struct BlockDriverAIOCBSync
{
1984 BlockDriverAIOCB common
;
1987 /* vector translation state */
1991 } BlockDriverAIOCBSync
;
1993 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
1995 BlockDriverAIOCBSync
*acb
= (BlockDriverAIOCBSync
*)blockacb
;
1996 qemu_bh_delete(acb
->bh
);
1998 qemu_aio_release(acb
);
2001 static AIOPool bdrv_em_aio_pool
= {
2002 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2003 .cancel
= bdrv_aio_cancel_em
,
2006 static void bdrv_aio_bh_cb(void *opaque
)
2008 BlockDriverAIOCBSync
*acb
= opaque
;
2011 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2012 qemu_vfree(acb
->bounce
);
2013 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2014 qemu_bh_delete(acb
->bh
);
2016 qemu_aio_release(acb
);
2019 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2023 BlockDriverCompletionFunc
*cb
,
2028 BlockDriverAIOCBSync
*acb
;
2030 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2031 acb
->is_write
= is_write
;
2033 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2036 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2039 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2040 acb
->ret
= bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2042 acb
->ret
= bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2045 qemu_bh_schedule(acb
->bh
);
2047 return &acb
->common
;
2050 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2051 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2052 BlockDriverCompletionFunc
*cb
, void *opaque
)
2054 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2057 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2058 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2059 BlockDriverCompletionFunc
*cb
, void *opaque
)
2061 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2064 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2065 BlockDriverCompletionFunc
*cb
, void *opaque
)
2067 BlockDriverAIOCBSync
*acb
;
2069 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2070 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2076 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2079 qemu_bh_schedule(acb
->bh
);
2080 return &acb
->common
;
2083 /**************************************************************/
2084 /* sync block device emulation */
2086 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2088 *(int *)opaque
= ret
;
2091 #define NOT_DONE 0x7fffffff
2093 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2094 uint8_t *buf
, int nb_sectors
)
2097 BlockDriverAIOCB
*acb
;
2101 async_context_push();
2103 async_ret
= NOT_DONE
;
2104 iov
.iov_base
= (void *)buf
;
2105 iov
.iov_len
= nb_sectors
* 512;
2106 qemu_iovec_init_external(&qiov
, &iov
, 1);
2107 acb
= bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2108 bdrv_rw_em_cb
, &async_ret
);
2114 while (async_ret
== NOT_DONE
) {
2120 async_context_pop();
2124 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2125 const uint8_t *buf
, int nb_sectors
)
2128 BlockDriverAIOCB
*acb
;
2132 async_context_push();
2134 async_ret
= NOT_DONE
;
2135 iov
.iov_base
= (void *)buf
;
2136 iov
.iov_len
= nb_sectors
* 512;
2137 qemu_iovec_init_external(&qiov
, &iov
, 1);
2138 acb
= bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2139 bdrv_rw_em_cb
, &async_ret
);
2144 while (async_ret
== NOT_DONE
) {
2149 async_context_pop();
2153 void bdrv_init(void)
2155 module_call_init(MODULE_INIT_BLOCK
);
2158 void bdrv_init_with_whitelist(void)
2160 use_bdrv_whitelist
= 1;
2164 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2165 BlockDriverCompletionFunc
*cb
, void *opaque
)
2167 BlockDriverAIOCB
*acb
;
2169 if (pool
->free_aiocb
) {
2170 acb
= pool
->free_aiocb
;
2171 pool
->free_aiocb
= acb
->next
;
2173 acb
= qemu_mallocz(pool
->aiocb_size
);
2178 acb
->opaque
= opaque
;
2182 void qemu_aio_release(void *p
)
2184 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
2185 AIOPool
*pool
= acb
->pool
;
2186 acb
->next
= pool
->free_aiocb
;
2187 pool
->free_aiocb
= acb
;
2190 /**************************************************************/
2191 /* removable device support */
2194 * Return TRUE if the media is present
2196 int bdrv_is_inserted(BlockDriverState
*bs
)
2198 BlockDriver
*drv
= bs
->drv
;
2202 if (!drv
->bdrv_is_inserted
)
2204 ret
= drv
->bdrv_is_inserted(bs
);
2209 * Return TRUE if the media changed since the last call to this
2210 * function. It is currently only used for floppy disks
2212 int bdrv_media_changed(BlockDriverState
*bs
)
2214 BlockDriver
*drv
= bs
->drv
;
2217 if (!drv
|| !drv
->bdrv_media_changed
)
2220 ret
= drv
->bdrv_media_changed(bs
);
2221 if (ret
== -ENOTSUP
)
2222 ret
= bs
->media_changed
;
2223 bs
->media_changed
= 0;
2228 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
2230 int bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
2232 BlockDriver
*drv
= bs
->drv
;
2239 if (!drv
|| !drv
->bdrv_eject
) {
2242 ret
= drv
->bdrv_eject(bs
, eject_flag
);
2244 if (ret
== -ENOTSUP
) {
2253 int bdrv_is_locked(BlockDriverState
*bs
)
2259 * Lock or unlock the media (if it is locked, the user won't be able
2260 * to eject it manually).
2262 void bdrv_set_locked(BlockDriverState
*bs
, int locked
)
2264 BlockDriver
*drv
= bs
->drv
;
2266 bs
->locked
= locked
;
2267 if (drv
&& drv
->bdrv_set_locked
) {
2268 drv
->bdrv_set_locked(bs
, locked
);
2272 /* needed for generic scsi interface */
2274 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
2276 BlockDriver
*drv
= bs
->drv
;
2278 if (drv
&& drv
->bdrv_ioctl
)
2279 return drv
->bdrv_ioctl(bs
, req
, buf
);
2283 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
2284 unsigned long int req
, void *buf
,
2285 BlockDriverCompletionFunc
*cb
, void *opaque
)
2287 BlockDriver
*drv
= bs
->drv
;
2289 if (drv
&& drv
->bdrv_aio_ioctl
)
2290 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
2296 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
2298 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
2301 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
2303 int64_t bitmap_size
;
2305 bs
->dirty_count
= 0;
2307 if (!bs
->dirty_bitmap
) {
2308 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
2309 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
2310 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
2312 bs
->dirty_bitmap
= qemu_mallocz(bitmap_size
);
2315 if (bs
->dirty_bitmap
) {
2316 qemu_free(bs
->dirty_bitmap
);
2317 bs
->dirty_bitmap
= NULL
;
2322 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
2324 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
2326 if (bs
->dirty_bitmap
&&
2327 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
2328 return bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
2329 (1 << (chunk
% (sizeof(unsigned long) * 8)));
2335 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
2338 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
2341 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
2343 return bs
->dirty_count
;