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"
28 #include "block_int.h"
30 #include "qemu-objects.h"
31 #include "qemu-coroutine.h"
34 #include <sys/types.h>
36 #include <sys/ioctl.h>
37 #include <sys/queue.h>
47 #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */
49 static void bdrv_dev_change_media_cb(BlockDriverState
*bs
, bool load
);
50 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
51 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
52 BlockDriverCompletionFunc
*cb
, void *opaque
);
53 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
54 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
55 BlockDriverCompletionFunc
*cb
, void *opaque
);
56 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
57 BlockDriverCompletionFunc
*cb
, void *opaque
);
58 static BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
59 BlockDriverCompletionFunc
*cb
, void *opaque
);
60 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
61 uint8_t *buf
, int nb_sectors
);
62 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
63 const uint8_t *buf
, int nb_sectors
);
64 static BlockDriverAIOCB
*bdrv_co_aio_readv_em(BlockDriverState
*bs
,
65 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
66 BlockDriverCompletionFunc
*cb
, void *opaque
);
67 static BlockDriverAIOCB
*bdrv_co_aio_writev_em(BlockDriverState
*bs
,
68 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
69 BlockDriverCompletionFunc
*cb
, void *opaque
);
70 static int coroutine_fn
bdrv_co_readv_em(BlockDriverState
*bs
,
71 int64_t sector_num
, int nb_sectors
,
73 static int coroutine_fn
bdrv_co_writev_em(BlockDriverState
*bs
,
74 int64_t sector_num
, int nb_sectors
,
76 static int coroutine_fn
bdrv_co_flush_em(BlockDriverState
*bs
);
77 static int coroutine_fn
bdrv_co_do_readv(BlockDriverState
*bs
,
78 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
79 static int coroutine_fn
bdrv_co_do_writev(BlockDriverState
*bs
,
80 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
);
82 static QTAILQ_HEAD(, BlockDriverState
) bdrv_states
=
83 QTAILQ_HEAD_INITIALIZER(bdrv_states
);
85 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
86 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
88 /* The device to use for VM snapshots */
89 static BlockDriverState
*bs_snapshots
;
91 /* If non-zero, use only whitelisted block drivers */
92 static int use_bdrv_whitelist
;
95 static int is_windows_drive_prefix(const char *filename
)
97 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
98 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
102 int is_windows_drive(const char *filename
)
104 if (is_windows_drive_prefix(filename
) &&
107 if (strstart(filename
, "\\\\.\\", NULL
) ||
108 strstart(filename
, "//./", NULL
))
114 /* check if the path starts with "<protocol>:" */
115 static int path_has_protocol(const char *path
)
118 if (is_windows_drive(path
) ||
119 is_windows_drive_prefix(path
)) {
124 return strchr(path
, ':') != NULL
;
127 int path_is_absolute(const char *path
)
131 /* specific case for names like: "\\.\d:" */
132 if (*path
== '/' || *path
== '\\')
135 p
= strchr(path
, ':');
141 return (*p
== '/' || *p
== '\\');
147 /* if filename is absolute, just copy it to dest. Otherwise, build a
148 path to it by considering it is relative to base_path. URL are
150 void path_combine(char *dest
, int dest_size
,
151 const char *base_path
,
152 const char *filename
)
159 if (path_is_absolute(filename
)) {
160 pstrcpy(dest
, dest_size
, filename
);
162 p
= strchr(base_path
, ':');
167 p1
= strrchr(base_path
, '/');
171 p2
= strrchr(base_path
, '\\');
183 if (len
> dest_size
- 1)
185 memcpy(dest
, base_path
, len
);
187 pstrcat(dest
, dest_size
, filename
);
191 void bdrv_register(BlockDriver
*bdrv
)
193 if (bdrv
->bdrv_co_readv
) {
194 /* Emulate AIO by coroutines, and sync by AIO */
195 bdrv
->bdrv_aio_readv
= bdrv_co_aio_readv_em
;
196 bdrv
->bdrv_aio_writev
= bdrv_co_aio_writev_em
;
197 bdrv
->bdrv_read
= bdrv_read_em
;
198 bdrv
->bdrv_write
= bdrv_write_em
;
200 bdrv
->bdrv_co_readv
= bdrv_co_readv_em
;
201 bdrv
->bdrv_co_writev
= bdrv_co_writev_em
;
203 if (!bdrv
->bdrv_aio_readv
) {
204 /* add AIO emulation layer */
205 bdrv
->bdrv_aio_readv
= bdrv_aio_readv_em
;
206 bdrv
->bdrv_aio_writev
= bdrv_aio_writev_em
;
207 } else if (!bdrv
->bdrv_read
) {
208 /* add synchronous IO emulation layer */
209 bdrv
->bdrv_read
= bdrv_read_em
;
210 bdrv
->bdrv_write
= bdrv_write_em
;
214 if (!bdrv
->bdrv_aio_flush
)
215 bdrv
->bdrv_aio_flush
= bdrv_aio_flush_em
;
217 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
220 /* create a new block device (by default it is empty) */
221 BlockDriverState
*bdrv_new(const char *device_name
)
223 BlockDriverState
*bs
;
225 bs
= g_malloc0(sizeof(BlockDriverState
));
226 pstrcpy(bs
->device_name
, sizeof(bs
->device_name
), device_name
);
227 if (device_name
[0] != '\0') {
228 QTAILQ_INSERT_TAIL(&bdrv_states
, bs
, list
);
230 bdrv_iostatus_disable(bs
);
234 BlockDriver
*bdrv_find_format(const char *format_name
)
237 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
238 if (!strcmp(drv1
->format_name
, format_name
)) {
245 static int bdrv_is_whitelisted(BlockDriver
*drv
)
247 static const char *whitelist
[] = {
248 CONFIG_BDRV_WHITELIST
253 return 1; /* no whitelist, anything goes */
255 for (p
= whitelist
; *p
; p
++) {
256 if (!strcmp(drv
->format_name
, *p
)) {
263 BlockDriver
*bdrv_find_whitelisted_format(const char *format_name
)
265 BlockDriver
*drv
= bdrv_find_format(format_name
);
266 return drv
&& bdrv_is_whitelisted(drv
) ? drv
: NULL
;
269 int bdrv_create(BlockDriver
*drv
, const char* filename
,
270 QEMUOptionParameter
*options
)
272 if (!drv
->bdrv_create
)
275 return drv
->bdrv_create(filename
, options
);
278 int bdrv_create_file(const char* filename
, QEMUOptionParameter
*options
)
282 drv
= bdrv_find_protocol(filename
);
287 return bdrv_create(drv
, filename
, options
);
291 void get_tmp_filename(char *filename
, int size
)
293 char temp_dir
[MAX_PATH
];
295 GetTempPath(MAX_PATH
, temp_dir
);
296 GetTempFileName(temp_dir
, "qem", 0, filename
);
299 void get_tmp_filename(char *filename
, int size
)
303 /* XXX: race condition possible */
304 tmpdir
= getenv("TMPDIR");
307 snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
);
308 fd
= mkstemp(filename
);
314 * Detect host devices. By convention, /dev/cdrom[N] is always
315 * recognized as a host CDROM.
317 static BlockDriver
*find_hdev_driver(const char *filename
)
319 int score_max
= 0, score
;
320 BlockDriver
*drv
= NULL
, *d
;
322 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
323 if (d
->bdrv_probe_device
) {
324 score
= d
->bdrv_probe_device(filename
);
325 if (score
> score_max
) {
335 BlockDriver
*bdrv_find_protocol(const char *filename
)
342 /* TODO Drivers without bdrv_file_open must be specified explicitly */
345 * XXX(hch): we really should not let host device detection
346 * override an explicit protocol specification, but moving this
347 * later breaks access to device names with colons in them.
348 * Thanks to the brain-dead persistent naming schemes on udev-
349 * based Linux systems those actually are quite common.
351 drv1
= find_hdev_driver(filename
);
356 if (!path_has_protocol(filename
)) {
357 return bdrv_find_format("file");
359 p
= strchr(filename
, ':');
362 if (len
> sizeof(protocol
) - 1)
363 len
= sizeof(protocol
) - 1;
364 memcpy(protocol
, filename
, len
);
365 protocol
[len
] = '\0';
366 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
367 if (drv1
->protocol_name
&&
368 !strcmp(drv1
->protocol_name
, protocol
)) {
375 static int find_image_format(const char *filename
, BlockDriver
**pdrv
)
377 int ret
, score
, score_max
;
378 BlockDriver
*drv1
, *drv
;
380 BlockDriverState
*bs
;
382 ret
= bdrv_file_open(&bs
, filename
, 0);
388 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
389 if (bs
->sg
|| !bdrv_is_inserted(bs
)) {
391 drv
= bdrv_find_format("raw");
399 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
408 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
409 if (drv1
->bdrv_probe
) {
410 score
= drv1
->bdrv_probe(buf
, ret
, filename
);
411 if (score
> score_max
) {
425 * Set the current 'total_sectors' value
427 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
429 BlockDriver
*drv
= bs
->drv
;
431 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
435 /* query actual device if possible, otherwise just trust the hint */
436 if (drv
->bdrv_getlength
) {
437 int64_t length
= drv
->bdrv_getlength(bs
);
441 hint
= length
>> BDRV_SECTOR_BITS
;
444 bs
->total_sectors
= hint
;
449 * Set open flags for a given cache mode
451 * Return 0 on success, -1 if the cache mode was invalid.
453 int bdrv_parse_cache_flags(const char *mode
, int *flags
)
455 *flags
&= ~BDRV_O_CACHE_MASK
;
457 if (!strcmp(mode
, "off") || !strcmp(mode
, "none")) {
458 *flags
|= BDRV_O_NOCACHE
| BDRV_O_CACHE_WB
;
459 } else if (!strcmp(mode
, "directsync")) {
460 *flags
|= BDRV_O_NOCACHE
;
461 } else if (!strcmp(mode
, "writeback")) {
462 *flags
|= BDRV_O_CACHE_WB
;
463 } else if (!strcmp(mode
, "unsafe")) {
464 *flags
|= BDRV_O_CACHE_WB
;
465 *flags
|= BDRV_O_NO_FLUSH
;
466 } else if (!strcmp(mode
, "writethrough")) {
467 /* this is the default */
476 * Common part for opening disk images and files
478 static int bdrv_open_common(BlockDriverState
*bs
, const char *filename
,
479 int flags
, BlockDriver
*drv
)
485 trace_bdrv_open_common(bs
, filename
, flags
, drv
->format_name
);
488 bs
->total_sectors
= 0;
491 bs
->open_flags
= flags
;
492 bs
->buffer_alignment
= 512;
494 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
496 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
)) {
501 bs
->opaque
= g_malloc0(drv
->instance_size
);
503 if (flags
& BDRV_O_CACHE_WB
)
504 bs
->enable_write_cache
= 1;
507 * Clear flags that are internal to the block layer before opening the
510 open_flags
= flags
& ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
513 * Snapshots should be writable.
515 if (bs
->is_temporary
) {
516 open_flags
|= BDRV_O_RDWR
;
519 /* Open the image, either directly or using a protocol */
520 if (drv
->bdrv_file_open
) {
521 ret
= drv
->bdrv_file_open(bs
, filename
, open_flags
);
523 ret
= bdrv_file_open(&bs
->file
, filename
, open_flags
);
525 ret
= drv
->bdrv_open(bs
, open_flags
);
533 bs
->keep_read_only
= bs
->read_only
= !(open_flags
& BDRV_O_RDWR
);
535 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
541 if (bs
->is_temporary
) {
549 bdrv_delete(bs
->file
);
559 * Opens a file using a protocol (file, host_device, nbd, ...)
561 int bdrv_file_open(BlockDriverState
**pbs
, const char *filename
, int flags
)
563 BlockDriverState
*bs
;
567 drv
= bdrv_find_protocol(filename
);
573 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
584 * Opens a disk image (raw, qcow2, vmdk, ...)
586 int bdrv_open(BlockDriverState
*bs
, const char *filename
, int flags
,
591 if (flags
& BDRV_O_SNAPSHOT
) {
592 BlockDriverState
*bs1
;
595 BlockDriver
*bdrv_qcow2
;
596 QEMUOptionParameter
*options
;
597 char tmp_filename
[PATH_MAX
];
598 char backing_filename
[PATH_MAX
];
600 /* if snapshot, we create a temporary backing file and open it
601 instead of opening 'filename' directly */
603 /* if there is a backing file, use it */
605 ret
= bdrv_open(bs1
, filename
, 0, drv
);
610 total_size
= bdrv_getlength(bs1
) & BDRV_SECTOR_MASK
;
612 if (bs1
->drv
&& bs1
->drv
->protocol_name
)
617 get_tmp_filename(tmp_filename
, sizeof(tmp_filename
));
619 /* Real path is meaningless for protocols */
621 snprintf(backing_filename
, sizeof(backing_filename
),
623 else if (!realpath(filename
, backing_filename
))
626 bdrv_qcow2
= bdrv_find_format("qcow2");
627 options
= parse_option_parameters("", bdrv_qcow2
->create_options
, NULL
);
629 set_option_parameter_int(options
, BLOCK_OPT_SIZE
, total_size
);
630 set_option_parameter(options
, BLOCK_OPT_BACKING_FILE
, backing_filename
);
632 set_option_parameter(options
, BLOCK_OPT_BACKING_FMT
,
636 ret
= bdrv_create(bdrv_qcow2
, tmp_filename
, options
);
637 free_option_parameters(options
);
642 filename
= tmp_filename
;
644 bs
->is_temporary
= 1;
647 /* Find the right image format driver */
649 ret
= find_image_format(filename
, &drv
);
653 goto unlink_and_fail
;
657 ret
= bdrv_open_common(bs
, filename
, flags
, drv
);
659 goto unlink_and_fail
;
662 /* If there is a backing file, use it */
663 if ((flags
& BDRV_O_NO_BACKING
) == 0 && bs
->backing_file
[0] != '\0') {
664 char backing_filename
[PATH_MAX
];
666 BlockDriver
*back_drv
= NULL
;
668 bs
->backing_hd
= bdrv_new("");
670 if (path_has_protocol(bs
->backing_file
)) {
671 pstrcpy(backing_filename
, sizeof(backing_filename
),
674 path_combine(backing_filename
, sizeof(backing_filename
),
675 filename
, bs
->backing_file
);
678 if (bs
->backing_format
[0] != '\0') {
679 back_drv
= bdrv_find_format(bs
->backing_format
);
682 /* backing files always opened read-only */
684 flags
& ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
686 ret
= bdrv_open(bs
->backing_hd
, backing_filename
, back_flags
, back_drv
);
691 if (bs
->is_temporary
) {
692 bs
->backing_hd
->keep_read_only
= !(flags
& BDRV_O_RDWR
);
694 /* base image inherits from "parent" */
695 bs
->backing_hd
->keep_read_only
= bs
->keep_read_only
;
699 if (!bdrv_key_required(bs
)) {
700 bdrv_dev_change_media_cb(bs
, true);
706 if (bs
->is_temporary
) {
712 void bdrv_close(BlockDriverState
*bs
)
715 if (bs
== bs_snapshots
) {
718 if (bs
->backing_hd
) {
719 bdrv_delete(bs
->backing_hd
);
720 bs
->backing_hd
= NULL
;
722 bs
->drv
->bdrv_close(bs
);
725 if (bs
->is_temporary
) {
726 unlink(bs
->filename
);
732 if (bs
->file
!= NULL
) {
733 bdrv_close(bs
->file
);
736 bdrv_dev_change_media_cb(bs
, false);
740 void bdrv_close_all(void)
742 BlockDriverState
*bs
;
744 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
749 /* make a BlockDriverState anonymous by removing from bdrv_state list.
750 Also, NULL terminate the device_name to prevent double remove */
751 void bdrv_make_anon(BlockDriverState
*bs
)
753 if (bs
->device_name
[0] != '\0') {
754 QTAILQ_REMOVE(&bdrv_states
, bs
, list
);
756 bs
->device_name
[0] = '\0';
759 void bdrv_delete(BlockDriverState
*bs
)
763 /* remove from list, if necessary */
767 if (bs
->file
!= NULL
) {
768 bdrv_delete(bs
->file
);
771 assert(bs
!= bs_snapshots
);
775 int bdrv_attach_dev(BlockDriverState
*bs
, void *dev
)
776 /* TODO change to DeviceState *dev when all users are qdevified */
782 bdrv_iostatus_reset(bs
);
786 /* TODO qdevified devices don't use this, remove when devices are qdevified */
787 void bdrv_attach_dev_nofail(BlockDriverState
*bs
, void *dev
)
789 if (bdrv_attach_dev(bs
, dev
) < 0) {
794 void bdrv_detach_dev(BlockDriverState
*bs
, void *dev
)
795 /* TODO change to DeviceState *dev when all users are qdevified */
797 assert(bs
->dev
== dev
);
800 bs
->dev_opaque
= NULL
;
801 bs
->buffer_alignment
= 512;
804 /* TODO change to return DeviceState * when all users are qdevified */
805 void *bdrv_get_attached_dev(BlockDriverState
*bs
)
810 void bdrv_set_dev_ops(BlockDriverState
*bs
, const BlockDevOps
*ops
,
814 bs
->dev_opaque
= opaque
;
815 if (bdrv_dev_has_removable_media(bs
) && bs
== bs_snapshots
) {
820 static void bdrv_dev_change_media_cb(BlockDriverState
*bs
, bool load
)
822 if (bs
->dev_ops
&& bs
->dev_ops
->change_media_cb
) {
823 bs
->dev_ops
->change_media_cb(bs
->dev_opaque
, load
);
827 bool bdrv_dev_has_removable_media(BlockDriverState
*bs
)
829 return !bs
->dev
|| (bs
->dev_ops
&& bs
->dev_ops
->change_media_cb
);
832 bool bdrv_dev_is_tray_open(BlockDriverState
*bs
)
834 if (bs
->dev_ops
&& bs
->dev_ops
->is_tray_open
) {
835 return bs
->dev_ops
->is_tray_open(bs
->dev_opaque
);
840 static void bdrv_dev_resize_cb(BlockDriverState
*bs
)
842 if (bs
->dev_ops
&& bs
->dev_ops
->resize_cb
) {
843 bs
->dev_ops
->resize_cb(bs
->dev_opaque
);
847 bool bdrv_dev_is_medium_locked(BlockDriverState
*bs
)
849 if (bs
->dev_ops
&& bs
->dev_ops
->is_medium_locked
) {
850 return bs
->dev_ops
->is_medium_locked(bs
->dev_opaque
);
856 * Run consistency checks on an image
858 * Returns 0 if the check could be completed (it doesn't mean that the image is
859 * free of errors) or -errno when an internal error occurred. The results of the
860 * check are stored in res.
862 int bdrv_check(BlockDriverState
*bs
, BdrvCheckResult
*res
)
864 if (bs
->drv
->bdrv_check
== NULL
) {
868 memset(res
, 0, sizeof(*res
));
869 return bs
->drv
->bdrv_check(bs
, res
);
872 #define COMMIT_BUF_SECTORS 2048
874 /* commit COW file into the raw image */
875 int bdrv_commit(BlockDriverState
*bs
)
877 BlockDriver
*drv
= bs
->drv
;
878 BlockDriver
*backing_drv
;
879 int64_t sector
, total_sectors
;
880 int n
, ro
, open_flags
;
881 int ret
= 0, rw_ret
= 0;
884 BlockDriverState
*bs_rw
, *bs_ro
;
889 if (!bs
->backing_hd
) {
893 if (bs
->backing_hd
->keep_read_only
) {
897 backing_drv
= bs
->backing_hd
->drv
;
898 ro
= bs
->backing_hd
->read_only
;
899 strncpy(filename
, bs
->backing_hd
->filename
, sizeof(filename
));
900 open_flags
= bs
->backing_hd
->open_flags
;
904 bdrv_delete(bs
->backing_hd
);
905 bs
->backing_hd
= NULL
;
906 bs_rw
= bdrv_new("");
907 rw_ret
= bdrv_open(bs_rw
, filename
, open_flags
| BDRV_O_RDWR
,
911 /* try to re-open read-only */
912 bs_ro
= bdrv_new("");
913 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
917 /* drive not functional anymore */
921 bs
->backing_hd
= bs_ro
;
924 bs
->backing_hd
= bs_rw
;
927 total_sectors
= bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
;
928 buf
= g_malloc(COMMIT_BUF_SECTORS
* BDRV_SECTOR_SIZE
);
930 for (sector
= 0; sector
< total_sectors
; sector
+= n
) {
931 if (drv
->bdrv_is_allocated(bs
, sector
, COMMIT_BUF_SECTORS
, &n
)) {
933 if (bdrv_read(bs
, sector
, buf
, n
) != 0) {
938 if (bdrv_write(bs
->backing_hd
, sector
, buf
, n
) != 0) {
945 if (drv
->bdrv_make_empty
) {
946 ret
= drv
->bdrv_make_empty(bs
);
951 * Make sure all data we wrote to the backing device is actually
955 bdrv_flush(bs
->backing_hd
);
962 bdrv_delete(bs
->backing_hd
);
963 bs
->backing_hd
= NULL
;
964 bs_ro
= bdrv_new("");
965 ret
= bdrv_open(bs_ro
, filename
, open_flags
& ~BDRV_O_RDWR
,
969 /* drive not functional anymore */
973 bs
->backing_hd
= bs_ro
;
974 bs
->backing_hd
->keep_read_only
= 0;
980 void bdrv_commit_all(void)
982 BlockDriverState
*bs
;
984 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
992 * -EINVAL - backing format specified, but no file
993 * -ENOSPC - can't update the backing file because no space is left in the
995 * -ENOTSUP - format driver doesn't support changing the backing file
997 int bdrv_change_backing_file(BlockDriverState
*bs
,
998 const char *backing_file
, const char *backing_fmt
)
1000 BlockDriver
*drv
= bs
->drv
;
1002 if (drv
->bdrv_change_backing_file
!= NULL
) {
1003 return drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
1009 static int bdrv_check_byte_request(BlockDriverState
*bs
, int64_t offset
,
1014 if (!bdrv_is_inserted(bs
))
1020 len
= bdrv_getlength(bs
);
1025 if ((offset
> len
) || (len
- offset
< size
))
1031 static int bdrv_check_request(BlockDriverState
*bs
, int64_t sector_num
,
1034 return bdrv_check_byte_request(bs
, sector_num
* BDRV_SECTOR_SIZE
,
1035 nb_sectors
* BDRV_SECTOR_SIZE
);
1038 static inline bool bdrv_has_async_rw(BlockDriver
*drv
)
1040 return drv
->bdrv_co_readv
!= bdrv_co_readv_em
1041 || drv
->bdrv_aio_readv
!= bdrv_aio_readv_em
;
1044 static inline bool bdrv_has_async_flush(BlockDriver
*drv
)
1046 return drv
->bdrv_aio_flush
!= bdrv_aio_flush_em
;
1049 typedef struct RwCo
{
1050 BlockDriverState
*bs
;
1058 static void coroutine_fn
bdrv_rw_co_entry(void *opaque
)
1060 RwCo
*rwco
= opaque
;
1062 if (!rwco
->is_write
) {
1063 rwco
->ret
= bdrv_co_do_readv(rwco
->bs
, rwco
->sector_num
,
1064 rwco
->nb_sectors
, rwco
->qiov
);
1066 rwco
->ret
= bdrv_co_do_writev(rwco
->bs
, rwco
->sector_num
,
1067 rwco
->nb_sectors
, rwco
->qiov
);
1072 * Process a synchronous request using coroutines
1074 static int bdrv_rw_co(BlockDriverState
*bs
, int64_t sector_num
, uint8_t *buf
,
1075 int nb_sectors
, bool is_write
)
1078 struct iovec iov
= {
1079 .iov_base
= (void *)buf
,
1080 .iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
,
1085 .sector_num
= sector_num
,
1086 .nb_sectors
= nb_sectors
,
1088 .is_write
= is_write
,
1092 qemu_iovec_init_external(&qiov
, &iov
, 1);
1094 if (qemu_in_coroutine()) {
1095 /* Fast-path if already in coroutine context */
1096 bdrv_rw_co_entry(&rwco
);
1098 co
= qemu_coroutine_create(bdrv_rw_co_entry
);
1099 qemu_coroutine_enter(co
, &rwco
);
1100 while (rwco
.ret
== NOT_DONE
) {
1107 /* return < 0 if error. See bdrv_write() for the return codes */
1108 int bdrv_read(BlockDriverState
*bs
, int64_t sector_num
,
1109 uint8_t *buf
, int nb_sectors
)
1111 return bdrv_rw_co(bs
, sector_num
, buf
, nb_sectors
, false);
1114 static void set_dirty_bitmap(BlockDriverState
*bs
, int64_t sector_num
,
1115 int nb_sectors
, int dirty
)
1118 unsigned long val
, idx
, bit
;
1120 start
= sector_num
/ BDRV_SECTORS_PER_DIRTY_CHUNK
;
1121 end
= (sector_num
+ nb_sectors
- 1) / BDRV_SECTORS_PER_DIRTY_CHUNK
;
1123 for (; start
<= end
; start
++) {
1124 idx
= start
/ (sizeof(unsigned long) * 8);
1125 bit
= start
% (sizeof(unsigned long) * 8);
1126 val
= bs
->dirty_bitmap
[idx
];
1128 if (!(val
& (1UL << bit
))) {
1133 if (val
& (1UL << bit
)) {
1135 val
&= ~(1UL << bit
);
1138 bs
->dirty_bitmap
[idx
] = val
;
1142 /* Return < 0 if error. Important errors are:
1143 -EIO generic I/O error (may happen for all errors)
1144 -ENOMEDIUM No media inserted.
1145 -EINVAL Invalid sector number or nb_sectors
1146 -EACCES Trying to write a read-only device
1148 int bdrv_write(BlockDriverState
*bs
, int64_t sector_num
,
1149 const uint8_t *buf
, int nb_sectors
)
1151 return bdrv_rw_co(bs
, sector_num
, (uint8_t *)buf
, nb_sectors
, true);
1154 int bdrv_pread(BlockDriverState
*bs
, int64_t offset
,
1155 void *buf
, int count1
)
1157 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
1158 int len
, nb_sectors
, count
;
1163 /* first read to align to sector start */
1164 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1167 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1169 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1171 memcpy(buf
, tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), len
);
1179 /* read the sectors "in place" */
1180 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1181 if (nb_sectors
> 0) {
1182 if ((ret
= bdrv_read(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1184 sector_num
+= nb_sectors
;
1185 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1190 /* add data from the last sector */
1192 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1194 memcpy(buf
, tmp_buf
, count
);
1199 int bdrv_pwrite(BlockDriverState
*bs
, int64_t offset
,
1200 const void *buf
, int count1
)
1202 uint8_t tmp_buf
[BDRV_SECTOR_SIZE
];
1203 int len
, nb_sectors
, count
;
1208 /* first write to align to sector start */
1209 len
= (BDRV_SECTOR_SIZE
- offset
) & (BDRV_SECTOR_SIZE
- 1);
1212 sector_num
= offset
>> BDRV_SECTOR_BITS
;
1214 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1216 memcpy(tmp_buf
+ (offset
& (BDRV_SECTOR_SIZE
- 1)), buf
, len
);
1217 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1226 /* write the sectors "in place" */
1227 nb_sectors
= count
>> BDRV_SECTOR_BITS
;
1228 if (nb_sectors
> 0) {
1229 if ((ret
= bdrv_write(bs
, sector_num
, buf
, nb_sectors
)) < 0)
1231 sector_num
+= nb_sectors
;
1232 len
= nb_sectors
<< BDRV_SECTOR_BITS
;
1237 /* add data from the last sector */
1239 if ((ret
= bdrv_read(bs
, sector_num
, tmp_buf
, 1)) < 0)
1241 memcpy(tmp_buf
, buf
, count
);
1242 if ((ret
= bdrv_write(bs
, sector_num
, tmp_buf
, 1)) < 0)
1249 * Writes to the file and ensures that no writes are reordered across this
1250 * request (acts as a barrier)
1252 * Returns 0 on success, -errno in error cases.
1254 int bdrv_pwrite_sync(BlockDriverState
*bs
, int64_t offset
,
1255 const void *buf
, int count
)
1259 ret
= bdrv_pwrite(bs
, offset
, buf
, count
);
1264 /* No flush needed for cache modes that use O_DSYNC */
1265 if ((bs
->open_flags
& BDRV_O_CACHE_WB
) != 0) {
1273 * Handle a read request in coroutine context
1275 static int coroutine_fn
bdrv_co_do_readv(BlockDriverState
*bs
,
1276 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
1278 BlockDriver
*drv
= bs
->drv
;
1283 if (bdrv_check_request(bs
, sector_num
, nb_sectors
)) {
1287 return drv
->bdrv_co_readv(bs
, sector_num
, nb_sectors
, qiov
);
1290 int coroutine_fn
bdrv_co_readv(BlockDriverState
*bs
, int64_t sector_num
,
1291 int nb_sectors
, QEMUIOVector
*qiov
)
1293 trace_bdrv_co_readv(bs
, sector_num
, nb_sectors
);
1295 return bdrv_co_do_readv(bs
, sector_num
, nb_sectors
, qiov
);
1299 * Handle a write request in coroutine context
1301 static int coroutine_fn
bdrv_co_do_writev(BlockDriverState
*bs
,
1302 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
1304 BlockDriver
*drv
= bs
->drv
;
1309 if (bs
->read_only
) {
1312 if (bdrv_check_request(bs
, sector_num
, nb_sectors
)) {
1316 if (bs
->dirty_bitmap
) {
1317 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
1320 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
1321 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
1324 return drv
->bdrv_co_writev(bs
, sector_num
, nb_sectors
, qiov
);
1327 int coroutine_fn
bdrv_co_writev(BlockDriverState
*bs
, int64_t sector_num
,
1328 int nb_sectors
, QEMUIOVector
*qiov
)
1330 trace_bdrv_co_writev(bs
, sector_num
, nb_sectors
);
1332 return bdrv_co_do_writev(bs
, sector_num
, nb_sectors
, qiov
);
1336 * Truncate file to 'offset' bytes (needed only for file protocols)
1338 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
1340 BlockDriver
*drv
= bs
->drv
;
1344 if (!drv
->bdrv_truncate
)
1348 if (bdrv_in_use(bs
))
1350 ret
= drv
->bdrv_truncate(bs
, offset
);
1352 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
1353 bdrv_dev_resize_cb(bs
);
1359 * Length of a allocated file in bytes. Sparse files are counted by actual
1360 * allocated space. Return < 0 if error or unknown.
1362 int64_t bdrv_get_allocated_file_size(BlockDriverState
*bs
)
1364 BlockDriver
*drv
= bs
->drv
;
1368 if (drv
->bdrv_get_allocated_file_size
) {
1369 return drv
->bdrv_get_allocated_file_size(bs
);
1372 return bdrv_get_allocated_file_size(bs
->file
);
1378 * Length of a file in bytes. Return < 0 if error or unknown.
1380 int64_t bdrv_getlength(BlockDriverState
*bs
)
1382 BlockDriver
*drv
= bs
->drv
;
1386 if (bs
->growable
|| bdrv_dev_has_removable_media(bs
)) {
1387 if (drv
->bdrv_getlength
) {
1388 return drv
->bdrv_getlength(bs
);
1391 return bs
->total_sectors
* BDRV_SECTOR_SIZE
;
1394 /* return 0 as number of sectors if no device present or error */
1395 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
1398 length
= bdrv_getlength(bs
);
1402 length
= length
>> BDRV_SECTOR_BITS
;
1403 *nb_sectors_ptr
= length
;
1407 uint8_t boot_ind
; /* 0x80 - active */
1408 uint8_t head
; /* starting head */
1409 uint8_t sector
; /* starting sector */
1410 uint8_t cyl
; /* starting cylinder */
1411 uint8_t sys_ind
; /* What partition type */
1412 uint8_t end_head
; /* end head */
1413 uint8_t end_sector
; /* end sector */
1414 uint8_t end_cyl
; /* end cylinder */
1415 uint32_t start_sect
; /* starting sector counting from 0 */
1416 uint32_t nr_sects
; /* nr of sectors in partition */
1419 /* try to guess the disk logical geometry from the MSDOS partition table. Return 0 if OK, -1 if could not guess */
1420 static int guess_disk_lchs(BlockDriverState
*bs
,
1421 int *pcylinders
, int *pheads
, int *psectors
)
1423 uint8_t buf
[BDRV_SECTOR_SIZE
];
1424 int ret
, i
, heads
, sectors
, cylinders
;
1425 struct partition
*p
;
1427 uint64_t nb_sectors
;
1429 bdrv_get_geometry(bs
, &nb_sectors
);
1431 ret
= bdrv_read(bs
, 0, buf
, 1);
1434 /* test msdos magic */
1435 if (buf
[510] != 0x55 || buf
[511] != 0xaa)
1437 for(i
= 0; i
< 4; i
++) {
1438 p
= ((struct partition
*)(buf
+ 0x1be)) + i
;
1439 nr_sects
= le32_to_cpu(p
->nr_sects
);
1440 if (nr_sects
&& p
->end_head
) {
1441 /* We make the assumption that the partition terminates on
1442 a cylinder boundary */
1443 heads
= p
->end_head
+ 1;
1444 sectors
= p
->end_sector
& 63;
1447 cylinders
= nb_sectors
/ (heads
* sectors
);
1448 if (cylinders
< 1 || cylinders
> 16383)
1451 *psectors
= sectors
;
1452 *pcylinders
= cylinders
;
1454 printf("guessed geometry: LCHS=%d %d %d\n",
1455 cylinders
, heads
, sectors
);
1463 void bdrv_guess_geometry(BlockDriverState
*bs
, int *pcyls
, int *pheads
, int *psecs
)
1465 int translation
, lba_detected
= 0;
1466 int cylinders
, heads
, secs
;
1467 uint64_t nb_sectors
;
1469 /* if a geometry hint is available, use it */
1470 bdrv_get_geometry(bs
, &nb_sectors
);
1471 bdrv_get_geometry_hint(bs
, &cylinders
, &heads
, &secs
);
1472 translation
= bdrv_get_translation_hint(bs
);
1473 if (cylinders
!= 0) {
1478 if (guess_disk_lchs(bs
, &cylinders
, &heads
, &secs
) == 0) {
1480 /* if heads > 16, it means that a BIOS LBA
1481 translation was active, so the default
1482 hardware geometry is OK */
1484 goto default_geometry
;
1489 /* disable any translation to be in sync with
1490 the logical geometry */
1491 if (translation
== BIOS_ATA_TRANSLATION_AUTO
) {
1492 bdrv_set_translation_hint(bs
,
1493 BIOS_ATA_TRANSLATION_NONE
);
1498 /* if no geometry, use a standard physical disk geometry */
1499 cylinders
= nb_sectors
/ (16 * 63);
1501 if (cylinders
> 16383)
1503 else if (cylinders
< 2)
1508 if ((lba_detected
== 1) && (translation
== BIOS_ATA_TRANSLATION_AUTO
)) {
1509 if ((*pcyls
* *pheads
) <= 131072) {
1510 bdrv_set_translation_hint(bs
,
1511 BIOS_ATA_TRANSLATION_LARGE
);
1513 bdrv_set_translation_hint(bs
,
1514 BIOS_ATA_TRANSLATION_LBA
);
1518 bdrv_set_geometry_hint(bs
, *pcyls
, *pheads
, *psecs
);
1522 void bdrv_set_geometry_hint(BlockDriverState
*bs
,
1523 int cyls
, int heads
, int secs
)
1530 void bdrv_set_translation_hint(BlockDriverState
*bs
, int translation
)
1532 bs
->translation
= translation
;
1535 void bdrv_get_geometry_hint(BlockDriverState
*bs
,
1536 int *pcyls
, int *pheads
, int *psecs
)
1539 *pheads
= bs
->heads
;
1543 /* Recognize floppy formats */
1544 typedef struct FDFormat
{
1551 static const FDFormat fd_formats
[] = {
1552 /* First entry is default format */
1553 /* 1.44 MB 3"1/2 floppy disks */
1554 { FDRIVE_DRV_144
, 18, 80, 1, },
1555 { FDRIVE_DRV_144
, 20, 80, 1, },
1556 { FDRIVE_DRV_144
, 21, 80, 1, },
1557 { FDRIVE_DRV_144
, 21, 82, 1, },
1558 { FDRIVE_DRV_144
, 21, 83, 1, },
1559 { FDRIVE_DRV_144
, 22, 80, 1, },
1560 { FDRIVE_DRV_144
, 23, 80, 1, },
1561 { FDRIVE_DRV_144
, 24, 80, 1, },
1562 /* 2.88 MB 3"1/2 floppy disks */
1563 { FDRIVE_DRV_288
, 36, 80, 1, },
1564 { FDRIVE_DRV_288
, 39, 80, 1, },
1565 { FDRIVE_DRV_288
, 40, 80, 1, },
1566 { FDRIVE_DRV_288
, 44, 80, 1, },
1567 { FDRIVE_DRV_288
, 48, 80, 1, },
1568 /* 720 kB 3"1/2 floppy disks */
1569 { FDRIVE_DRV_144
, 9, 80, 1, },
1570 { FDRIVE_DRV_144
, 10, 80, 1, },
1571 { FDRIVE_DRV_144
, 10, 82, 1, },
1572 { FDRIVE_DRV_144
, 10, 83, 1, },
1573 { FDRIVE_DRV_144
, 13, 80, 1, },
1574 { FDRIVE_DRV_144
, 14, 80, 1, },
1575 /* 1.2 MB 5"1/4 floppy disks */
1576 { FDRIVE_DRV_120
, 15, 80, 1, },
1577 { FDRIVE_DRV_120
, 18, 80, 1, },
1578 { FDRIVE_DRV_120
, 18, 82, 1, },
1579 { FDRIVE_DRV_120
, 18, 83, 1, },
1580 { FDRIVE_DRV_120
, 20, 80, 1, },
1581 /* 720 kB 5"1/4 floppy disks */
1582 { FDRIVE_DRV_120
, 9, 80, 1, },
1583 { FDRIVE_DRV_120
, 11, 80, 1, },
1584 /* 360 kB 5"1/4 floppy disks */
1585 { FDRIVE_DRV_120
, 9, 40, 1, },
1586 { FDRIVE_DRV_120
, 9, 40, 0, },
1587 { FDRIVE_DRV_120
, 10, 41, 1, },
1588 { FDRIVE_DRV_120
, 10, 42, 1, },
1589 /* 320 kB 5"1/4 floppy disks */
1590 { FDRIVE_DRV_120
, 8, 40, 1, },
1591 { FDRIVE_DRV_120
, 8, 40, 0, },
1592 /* 360 kB must match 5"1/4 better than 3"1/2... */
1593 { FDRIVE_DRV_144
, 9, 80, 0, },
1595 { FDRIVE_DRV_NONE
, -1, -1, 0, },
1598 void bdrv_get_floppy_geometry_hint(BlockDriverState
*bs
, int *nb_heads
,
1599 int *max_track
, int *last_sect
,
1600 FDriveType drive_in
, FDriveType
*drive
)
1602 const FDFormat
*parse
;
1603 uint64_t nb_sectors
, size
;
1604 int i
, first_match
, match
;
1606 bdrv_get_geometry_hint(bs
, nb_heads
, max_track
, last_sect
);
1607 if (*nb_heads
!= 0 && *max_track
!= 0 && *last_sect
!= 0) {
1608 /* User defined disk */
1610 bdrv_get_geometry(bs
, &nb_sectors
);
1613 for (i
= 0; ; i
++) {
1614 parse
= &fd_formats
[i
];
1615 if (parse
->drive
== FDRIVE_DRV_NONE
) {
1618 if (drive_in
== parse
->drive
||
1619 drive_in
== FDRIVE_DRV_NONE
) {
1620 size
= (parse
->max_head
+ 1) * parse
->max_track
*
1622 if (nb_sectors
== size
) {
1626 if (first_match
== -1) {
1632 if (first_match
== -1) {
1635 match
= first_match
;
1637 parse
= &fd_formats
[match
];
1639 *nb_heads
= parse
->max_head
+ 1;
1640 *max_track
= parse
->max_track
;
1641 *last_sect
= parse
->last_sect
;
1642 *drive
= parse
->drive
;
1646 int bdrv_get_translation_hint(BlockDriverState
*bs
)
1648 return bs
->translation
;
1651 void bdrv_set_on_error(BlockDriverState
*bs
, BlockErrorAction on_read_error
,
1652 BlockErrorAction on_write_error
)
1654 bs
->on_read_error
= on_read_error
;
1655 bs
->on_write_error
= on_write_error
;
1658 BlockErrorAction
bdrv_get_on_error(BlockDriverState
*bs
, int is_read
)
1660 return is_read
? bs
->on_read_error
: bs
->on_write_error
;
1663 int bdrv_is_read_only(BlockDriverState
*bs
)
1665 return bs
->read_only
;
1668 int bdrv_is_sg(BlockDriverState
*bs
)
1673 int bdrv_enable_write_cache(BlockDriverState
*bs
)
1675 return bs
->enable_write_cache
;
1678 int bdrv_is_encrypted(BlockDriverState
*bs
)
1680 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
1682 return bs
->encrypted
;
1685 int bdrv_key_required(BlockDriverState
*bs
)
1687 BlockDriverState
*backing_hd
= bs
->backing_hd
;
1689 if (backing_hd
&& backing_hd
->encrypted
&& !backing_hd
->valid_key
)
1691 return (bs
->encrypted
&& !bs
->valid_key
);
1694 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
1697 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
) {
1698 ret
= bdrv_set_key(bs
->backing_hd
, key
);
1704 if (!bs
->encrypted
) {
1706 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
1709 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
1712 } else if (!bs
->valid_key
) {
1714 /* call the change callback now, we skipped it on open */
1715 bdrv_dev_change_media_cb(bs
, true);
1720 void bdrv_get_format(BlockDriverState
*bs
, char *buf
, int buf_size
)
1725 pstrcpy(buf
, buf_size
, bs
->drv
->format_name
);
1729 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
1734 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
1735 it(opaque
, drv
->format_name
);
1739 BlockDriverState
*bdrv_find(const char *name
)
1741 BlockDriverState
*bs
;
1743 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1744 if (!strcmp(name
, bs
->device_name
)) {
1751 BlockDriverState
*bdrv_next(BlockDriverState
*bs
)
1754 return QTAILQ_FIRST(&bdrv_states
);
1756 return QTAILQ_NEXT(bs
, list
);
1759 void bdrv_iterate(void (*it
)(void *opaque
, BlockDriverState
*bs
), void *opaque
)
1761 BlockDriverState
*bs
;
1763 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1768 const char *bdrv_get_device_name(BlockDriverState
*bs
)
1770 return bs
->device_name
;
1773 int bdrv_flush(BlockDriverState
*bs
)
1775 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
1779 if (bs
->drv
&& bdrv_has_async_flush(bs
->drv
) && qemu_in_coroutine()) {
1780 return bdrv_co_flush_em(bs
);
1783 if (bs
->drv
&& bs
->drv
->bdrv_flush
) {
1784 return bs
->drv
->bdrv_flush(bs
);
1788 * Some block drivers always operate in either writethrough or unsafe mode
1789 * and don't support bdrv_flush therefore. Usually qemu doesn't know how
1790 * the server works (because the behaviour is hardcoded or depends on
1791 * server-side configuration), so we can't ensure that everything is safe
1792 * on disk. Returning an error doesn't work because that would break guests
1793 * even if the server operates in writethrough mode.
1795 * Let's hope the user knows what he's doing.
1800 void bdrv_flush_all(void)
1802 BlockDriverState
*bs
;
1804 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1805 if (!bdrv_is_read_only(bs
) && bdrv_is_inserted(bs
)) {
1811 int bdrv_has_zero_init(BlockDriverState
*bs
)
1815 if (bs
->drv
->bdrv_has_zero_init
) {
1816 return bs
->drv
->bdrv_has_zero_init(bs
);
1822 int bdrv_discard(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
)
1827 if (!bs
->drv
->bdrv_discard
) {
1830 return bs
->drv
->bdrv_discard(bs
, sector_num
, nb_sectors
);
1834 * Returns true iff the specified sector is present in the disk image. Drivers
1835 * not implementing the functionality are assumed to not support backing files,
1836 * hence all their sectors are reported as allocated.
1838 * 'pnum' is set to the number of sectors (including and immediately following
1839 * the specified sector) that are known to be in the same
1840 * allocated/unallocated state.
1842 * 'nb_sectors' is the max value 'pnum' should be set to.
1844 int bdrv_is_allocated(BlockDriverState
*bs
, int64_t sector_num
, int nb_sectors
,
1848 if (!bs
->drv
->bdrv_is_allocated
) {
1849 if (sector_num
>= bs
->total_sectors
) {
1853 n
= bs
->total_sectors
- sector_num
;
1854 *pnum
= (n
< nb_sectors
) ? (n
) : (nb_sectors
);
1857 return bs
->drv
->bdrv_is_allocated(bs
, sector_num
, nb_sectors
, pnum
);
1860 void bdrv_mon_event(const BlockDriverState
*bdrv
,
1861 BlockMonEventAction action
, int is_read
)
1864 const char *action_str
;
1867 case BDRV_ACTION_REPORT
:
1868 action_str
= "report";
1870 case BDRV_ACTION_IGNORE
:
1871 action_str
= "ignore";
1873 case BDRV_ACTION_STOP
:
1874 action_str
= "stop";
1880 data
= qobject_from_jsonf("{ 'device': %s, 'action': %s, 'operation': %s }",
1883 is_read
? "read" : "write");
1884 monitor_protocol_event(QEVENT_BLOCK_IO_ERROR
, data
);
1886 qobject_decref(data
);
1889 static void bdrv_print_dict(QObject
*obj
, void *opaque
)
1892 Monitor
*mon
= opaque
;
1894 bs_dict
= qobject_to_qdict(obj
);
1896 monitor_printf(mon
, "%s: removable=%d",
1897 qdict_get_str(bs_dict
, "device"),
1898 qdict_get_bool(bs_dict
, "removable"));
1900 if (qdict_get_bool(bs_dict
, "removable")) {
1901 monitor_printf(mon
, " locked=%d", qdict_get_bool(bs_dict
, "locked"));
1902 monitor_printf(mon
, " tray-open=%d",
1903 qdict_get_bool(bs_dict
, "tray-open"));
1906 if (qdict_haskey(bs_dict
, "io-status")) {
1907 monitor_printf(mon
, " io-status=%s", qdict_get_str(bs_dict
, "io-status"));
1910 if (qdict_haskey(bs_dict
, "inserted")) {
1911 QDict
*qdict
= qobject_to_qdict(qdict_get(bs_dict
, "inserted"));
1913 monitor_printf(mon
, " file=");
1914 monitor_print_filename(mon
, qdict_get_str(qdict
, "file"));
1915 if (qdict_haskey(qdict
, "backing_file")) {
1916 monitor_printf(mon
, " backing_file=");
1917 monitor_print_filename(mon
, qdict_get_str(qdict
, "backing_file"));
1919 monitor_printf(mon
, " ro=%d drv=%s encrypted=%d",
1920 qdict_get_bool(qdict
, "ro"),
1921 qdict_get_str(qdict
, "drv"),
1922 qdict_get_bool(qdict
, "encrypted"));
1924 monitor_printf(mon
, " [not inserted]");
1927 monitor_printf(mon
, "\n");
1930 void bdrv_info_print(Monitor
*mon
, const QObject
*data
)
1932 qlist_iter(qobject_to_qlist(data
), bdrv_print_dict
, mon
);
1935 static const char *const io_status_name
[BDRV_IOS_MAX
] = {
1936 [BDRV_IOS_OK
] = "ok",
1937 [BDRV_IOS_FAILED
] = "failed",
1938 [BDRV_IOS_ENOSPC
] = "nospace",
1941 void bdrv_info(Monitor
*mon
, QObject
**ret_data
)
1944 BlockDriverState
*bs
;
1946 bs_list
= qlist_new();
1948 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
1952 bs_obj
= qobject_from_jsonf("{ 'device': %s, 'type': 'unknown', "
1953 "'removable': %i, 'locked': %i }",
1955 bdrv_dev_has_removable_media(bs
),
1956 bdrv_dev_is_medium_locked(bs
));
1957 bs_dict
= qobject_to_qdict(bs_obj
);
1959 if (bdrv_dev_has_removable_media(bs
)) {
1960 qdict_put(bs_dict
, "tray-open",
1961 qbool_from_int(bdrv_dev_is_tray_open(bs
)));
1964 if (bdrv_iostatus_is_enabled(bs
)) {
1965 qdict_put(bs_dict
, "io-status",
1966 qstring_from_str(io_status_name
[bs
->iostatus
]));
1972 obj
= qobject_from_jsonf("{ 'file': %s, 'ro': %i, 'drv': %s, "
1973 "'encrypted': %i }",
1974 bs
->filename
, bs
->read_only
,
1975 bs
->drv
->format_name
,
1976 bdrv_is_encrypted(bs
));
1977 if (bs
->backing_file
[0] != '\0') {
1978 QDict
*qdict
= qobject_to_qdict(obj
);
1979 qdict_put(qdict
, "backing_file",
1980 qstring_from_str(bs
->backing_file
));
1983 qdict_put_obj(bs_dict
, "inserted", obj
);
1985 qlist_append_obj(bs_list
, bs_obj
);
1988 *ret_data
= QOBJECT(bs_list
);
1991 static void bdrv_stats_iter(QObject
*data
, void *opaque
)
1994 Monitor
*mon
= opaque
;
1996 qdict
= qobject_to_qdict(data
);
1997 monitor_printf(mon
, "%s:", qdict_get_str(qdict
, "device"));
1999 qdict
= qobject_to_qdict(qdict_get(qdict
, "stats"));
2000 monitor_printf(mon
, " rd_bytes=%" PRId64
2001 " wr_bytes=%" PRId64
2002 " rd_operations=%" PRId64
2003 " wr_operations=%" PRId64
2004 " flush_operations=%" PRId64
2005 " wr_total_time_ns=%" PRId64
2006 " rd_total_time_ns=%" PRId64
2007 " flush_total_time_ns=%" PRId64
2009 qdict_get_int(qdict
, "rd_bytes"),
2010 qdict_get_int(qdict
, "wr_bytes"),
2011 qdict_get_int(qdict
, "rd_operations"),
2012 qdict_get_int(qdict
, "wr_operations"),
2013 qdict_get_int(qdict
, "flush_operations"),
2014 qdict_get_int(qdict
, "wr_total_time_ns"),
2015 qdict_get_int(qdict
, "rd_total_time_ns"),
2016 qdict_get_int(qdict
, "flush_total_time_ns"));
2019 void bdrv_stats_print(Monitor
*mon
, const QObject
*data
)
2021 qlist_iter(qobject_to_qlist(data
), bdrv_stats_iter
, mon
);
2024 static QObject
* bdrv_info_stats_bs(BlockDriverState
*bs
)
2029 res
= qobject_from_jsonf("{ 'stats': {"
2030 "'rd_bytes': %" PRId64
","
2031 "'wr_bytes': %" PRId64
","
2032 "'rd_operations': %" PRId64
","
2033 "'wr_operations': %" PRId64
","
2034 "'wr_highest_offset': %" PRId64
","
2035 "'flush_operations': %" PRId64
","
2036 "'wr_total_time_ns': %" PRId64
","
2037 "'rd_total_time_ns': %" PRId64
","
2038 "'flush_total_time_ns': %" PRId64
2040 bs
->nr_bytes
[BDRV_ACCT_READ
],
2041 bs
->nr_bytes
[BDRV_ACCT_WRITE
],
2042 bs
->nr_ops
[BDRV_ACCT_READ
],
2043 bs
->nr_ops
[BDRV_ACCT_WRITE
],
2044 bs
->wr_highest_sector
*
2045 (uint64_t)BDRV_SECTOR_SIZE
,
2046 bs
->nr_ops
[BDRV_ACCT_FLUSH
],
2047 bs
->total_time_ns
[BDRV_ACCT_WRITE
],
2048 bs
->total_time_ns
[BDRV_ACCT_READ
],
2049 bs
->total_time_ns
[BDRV_ACCT_FLUSH
]);
2050 dict
= qobject_to_qdict(res
);
2052 if (*bs
->device_name
) {
2053 qdict_put(dict
, "device", qstring_from_str(bs
->device_name
));
2057 QObject
*parent
= bdrv_info_stats_bs(bs
->file
);
2058 qdict_put_obj(dict
, "parent", parent
);
2064 void bdrv_info_stats(Monitor
*mon
, QObject
**ret_data
)
2068 BlockDriverState
*bs
;
2070 devices
= qlist_new();
2072 QTAILQ_FOREACH(bs
, &bdrv_states
, list
) {
2073 obj
= bdrv_info_stats_bs(bs
);
2074 qlist_append_obj(devices
, obj
);
2077 *ret_data
= QOBJECT(devices
);
2080 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
2082 if (bs
->backing_hd
&& bs
->backing_hd
->encrypted
)
2083 return bs
->backing_file
;
2084 else if (bs
->encrypted
)
2085 return bs
->filename
;
2090 void bdrv_get_backing_filename(BlockDriverState
*bs
,
2091 char *filename
, int filename_size
)
2093 if (!bs
->backing_file
) {
2094 pstrcpy(filename
, filename_size
, "");
2096 pstrcpy(filename
, filename_size
, bs
->backing_file
);
2100 int bdrv_write_compressed(BlockDriverState
*bs
, int64_t sector_num
,
2101 const uint8_t *buf
, int nb_sectors
)
2103 BlockDriver
*drv
= bs
->drv
;
2106 if (!drv
->bdrv_write_compressed
)
2108 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2111 if (bs
->dirty_bitmap
) {
2112 set_dirty_bitmap(bs
, sector_num
, nb_sectors
, 1);
2115 return drv
->bdrv_write_compressed(bs
, sector_num
, buf
, nb_sectors
);
2118 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
2120 BlockDriver
*drv
= bs
->drv
;
2123 if (!drv
->bdrv_get_info
)
2125 memset(bdi
, 0, sizeof(*bdi
));
2126 return drv
->bdrv_get_info(bs
, bdi
);
2129 int bdrv_save_vmstate(BlockDriverState
*bs
, const uint8_t *buf
,
2130 int64_t pos
, int size
)
2132 BlockDriver
*drv
= bs
->drv
;
2135 if (drv
->bdrv_save_vmstate
)
2136 return drv
->bdrv_save_vmstate(bs
, buf
, pos
, size
);
2138 return bdrv_save_vmstate(bs
->file
, buf
, pos
, size
);
2142 int bdrv_load_vmstate(BlockDriverState
*bs
, uint8_t *buf
,
2143 int64_t pos
, int size
)
2145 BlockDriver
*drv
= bs
->drv
;
2148 if (drv
->bdrv_load_vmstate
)
2149 return drv
->bdrv_load_vmstate(bs
, buf
, pos
, size
);
2151 return bdrv_load_vmstate(bs
->file
, buf
, pos
, size
);
2155 void bdrv_debug_event(BlockDriverState
*bs
, BlkDebugEvent event
)
2157 BlockDriver
*drv
= bs
->drv
;
2159 if (!drv
|| !drv
->bdrv_debug_event
) {
2163 return drv
->bdrv_debug_event(bs
, event
);
2167 /**************************************************************/
2168 /* handling of snapshots */
2170 int bdrv_can_snapshot(BlockDriverState
*bs
)
2172 BlockDriver
*drv
= bs
->drv
;
2173 if (!drv
|| !bdrv_is_inserted(bs
) || bdrv_is_read_only(bs
)) {
2177 if (!drv
->bdrv_snapshot_create
) {
2178 if (bs
->file
!= NULL
) {
2179 return bdrv_can_snapshot(bs
->file
);
2187 int bdrv_is_snapshot(BlockDriverState
*bs
)
2189 return !!(bs
->open_flags
& BDRV_O_SNAPSHOT
);
2192 BlockDriverState
*bdrv_snapshots(void)
2194 BlockDriverState
*bs
;
2197 return bs_snapshots
;
2201 while ((bs
= bdrv_next(bs
))) {
2202 if (bdrv_can_snapshot(bs
)) {
2210 int bdrv_snapshot_create(BlockDriverState
*bs
,
2211 QEMUSnapshotInfo
*sn_info
)
2213 BlockDriver
*drv
= bs
->drv
;
2216 if (drv
->bdrv_snapshot_create
)
2217 return drv
->bdrv_snapshot_create(bs
, sn_info
);
2219 return bdrv_snapshot_create(bs
->file
, sn_info
);
2223 int bdrv_snapshot_goto(BlockDriverState
*bs
,
2224 const char *snapshot_id
)
2226 BlockDriver
*drv
= bs
->drv
;
2231 if (drv
->bdrv_snapshot_goto
)
2232 return drv
->bdrv_snapshot_goto(bs
, snapshot_id
);
2235 drv
->bdrv_close(bs
);
2236 ret
= bdrv_snapshot_goto(bs
->file
, snapshot_id
);
2237 open_ret
= drv
->bdrv_open(bs
, bs
->open_flags
);
2239 bdrv_delete(bs
->file
);
2249 int bdrv_snapshot_delete(BlockDriverState
*bs
, const char *snapshot_id
)
2251 BlockDriver
*drv
= bs
->drv
;
2254 if (drv
->bdrv_snapshot_delete
)
2255 return drv
->bdrv_snapshot_delete(bs
, snapshot_id
);
2257 return bdrv_snapshot_delete(bs
->file
, snapshot_id
);
2261 int bdrv_snapshot_list(BlockDriverState
*bs
,
2262 QEMUSnapshotInfo
**psn_info
)
2264 BlockDriver
*drv
= bs
->drv
;
2267 if (drv
->bdrv_snapshot_list
)
2268 return drv
->bdrv_snapshot_list(bs
, psn_info
);
2270 return bdrv_snapshot_list(bs
->file
, psn_info
);
2274 int bdrv_snapshot_load_tmp(BlockDriverState
*bs
,
2275 const char *snapshot_name
)
2277 BlockDriver
*drv
= bs
->drv
;
2281 if (!bs
->read_only
) {
2284 if (drv
->bdrv_snapshot_load_tmp
) {
2285 return drv
->bdrv_snapshot_load_tmp(bs
, snapshot_name
);
2290 #define NB_SUFFIXES 4
2292 char *get_human_readable_size(char *buf
, int buf_size
, int64_t size
)
2294 static const char suffixes
[NB_SUFFIXES
] = "KMGT";
2299 snprintf(buf
, buf_size
, "%" PRId64
, size
);
2302 for(i
= 0; i
< NB_SUFFIXES
; i
++) {
2303 if (size
< (10 * base
)) {
2304 snprintf(buf
, buf_size
, "%0.1f%c",
2305 (double)size
/ base
,
2308 } else if (size
< (1000 * base
) || i
== (NB_SUFFIXES
- 1)) {
2309 snprintf(buf
, buf_size
, "%" PRId64
"%c",
2310 ((size
+ (base
>> 1)) / base
),
2320 char *bdrv_snapshot_dump(char *buf
, int buf_size
, QEMUSnapshotInfo
*sn
)
2322 char buf1
[128], date_buf
[128], clock_buf
[128];
2332 snprintf(buf
, buf_size
,
2333 "%-10s%-20s%7s%20s%15s",
2334 "ID", "TAG", "VM SIZE", "DATE", "VM CLOCK");
2338 ptm
= localtime(&ti
);
2339 strftime(date_buf
, sizeof(date_buf
),
2340 "%Y-%m-%d %H:%M:%S", ptm
);
2342 localtime_r(&ti
, &tm
);
2343 strftime(date_buf
, sizeof(date_buf
),
2344 "%Y-%m-%d %H:%M:%S", &tm
);
2346 secs
= sn
->vm_clock_nsec
/ 1000000000;
2347 snprintf(clock_buf
, sizeof(clock_buf
),
2348 "%02d:%02d:%02d.%03d",
2350 (int)((secs
/ 60) % 60),
2352 (int)((sn
->vm_clock_nsec
/ 1000000) % 1000));
2353 snprintf(buf
, buf_size
,
2354 "%-10s%-20s%7s%20s%15s",
2355 sn
->id_str
, sn
->name
,
2356 get_human_readable_size(buf1
, sizeof(buf1
), sn
->vm_state_size
),
2363 /**************************************************************/
2366 BlockDriverAIOCB
*bdrv_aio_readv(BlockDriverState
*bs
, int64_t sector_num
,
2367 QEMUIOVector
*qiov
, int nb_sectors
,
2368 BlockDriverCompletionFunc
*cb
, void *opaque
)
2370 BlockDriver
*drv
= bs
->drv
;
2372 trace_bdrv_aio_readv(bs
, sector_num
, nb_sectors
, opaque
);
2376 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2379 return drv
->bdrv_aio_readv(bs
, sector_num
, qiov
, nb_sectors
,
2383 typedef struct BlockCompleteData
{
2384 BlockDriverCompletionFunc
*cb
;
2386 BlockDriverState
*bs
;
2389 } BlockCompleteData
;
2391 static void block_complete_cb(void *opaque
, int ret
)
2393 BlockCompleteData
*b
= opaque
;
2395 if (b
->bs
->dirty_bitmap
) {
2396 set_dirty_bitmap(b
->bs
, b
->sector_num
, b
->nb_sectors
, 1);
2398 b
->cb(b
->opaque
, ret
);
2402 static BlockCompleteData
*blk_dirty_cb_alloc(BlockDriverState
*bs
,
2405 BlockDriverCompletionFunc
*cb
,
2408 BlockCompleteData
*blkdata
= g_malloc0(sizeof(BlockCompleteData
));
2412 blkdata
->opaque
= opaque
;
2413 blkdata
->sector_num
= sector_num
;
2414 blkdata
->nb_sectors
= nb_sectors
;
2419 BlockDriverAIOCB
*bdrv_aio_writev(BlockDriverState
*bs
, int64_t sector_num
,
2420 QEMUIOVector
*qiov
, int nb_sectors
,
2421 BlockDriverCompletionFunc
*cb
, void *opaque
)
2423 BlockDriver
*drv
= bs
->drv
;
2424 BlockDriverAIOCB
*ret
;
2425 BlockCompleteData
*blk_cb_data
;
2427 trace_bdrv_aio_writev(bs
, sector_num
, nb_sectors
, opaque
);
2433 if (bdrv_check_request(bs
, sector_num
, nb_sectors
))
2436 if (bs
->dirty_bitmap
) {
2437 blk_cb_data
= blk_dirty_cb_alloc(bs
, sector_num
, nb_sectors
, cb
,
2439 cb
= &block_complete_cb
;
2440 opaque
= blk_cb_data
;
2443 ret
= drv
->bdrv_aio_writev(bs
, sector_num
, qiov
, nb_sectors
,
2447 if (bs
->wr_highest_sector
< sector_num
+ nb_sectors
- 1) {
2448 bs
->wr_highest_sector
= sector_num
+ nb_sectors
- 1;
2456 typedef struct MultiwriteCB
{
2461 BlockDriverCompletionFunc
*cb
;
2463 QEMUIOVector
*free_qiov
;
2468 static void multiwrite_user_cb(MultiwriteCB
*mcb
)
2472 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2473 mcb
->callbacks
[i
].cb(mcb
->callbacks
[i
].opaque
, mcb
->error
);
2474 if (mcb
->callbacks
[i
].free_qiov
) {
2475 qemu_iovec_destroy(mcb
->callbacks
[i
].free_qiov
);
2477 g_free(mcb
->callbacks
[i
].free_qiov
);
2478 qemu_vfree(mcb
->callbacks
[i
].free_buf
);
2482 static void multiwrite_cb(void *opaque
, int ret
)
2484 MultiwriteCB
*mcb
= opaque
;
2486 trace_multiwrite_cb(mcb
, ret
);
2488 if (ret
< 0 && !mcb
->error
) {
2492 mcb
->num_requests
--;
2493 if (mcb
->num_requests
== 0) {
2494 multiwrite_user_cb(mcb
);
2499 static int multiwrite_req_compare(const void *a
, const void *b
)
2501 const BlockRequest
*req1
= a
, *req2
= b
;
2504 * Note that we can't simply subtract req2->sector from req1->sector
2505 * here as that could overflow the return value.
2507 if (req1
->sector
> req2
->sector
) {
2509 } else if (req1
->sector
< req2
->sector
) {
2517 * Takes a bunch of requests and tries to merge them. Returns the number of
2518 * requests that remain after merging.
2520 static int multiwrite_merge(BlockDriverState
*bs
, BlockRequest
*reqs
,
2521 int num_reqs
, MultiwriteCB
*mcb
)
2525 // Sort requests by start sector
2526 qsort(reqs
, num_reqs
, sizeof(*reqs
), &multiwrite_req_compare
);
2528 // Check if adjacent requests touch the same clusters. If so, combine them,
2529 // filling up gaps with zero sectors.
2531 for (i
= 1; i
< num_reqs
; i
++) {
2533 int64_t oldreq_last
= reqs
[outidx
].sector
+ reqs
[outidx
].nb_sectors
;
2535 // This handles the cases that are valid for all block drivers, namely
2536 // exactly sequential writes and overlapping writes.
2537 if (reqs
[i
].sector
<= oldreq_last
) {
2541 // The block driver may decide that it makes sense to combine requests
2542 // even if there is a gap of some sectors between them. In this case,
2543 // the gap is filled with zeros (therefore only applicable for yet
2544 // unused space in format like qcow2).
2545 if (!merge
&& bs
->drv
->bdrv_merge_requests
) {
2546 merge
= bs
->drv
->bdrv_merge_requests(bs
, &reqs
[outidx
], &reqs
[i
]);
2549 if (reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1 > IOV_MAX
) {
2555 QEMUIOVector
*qiov
= g_malloc0(sizeof(*qiov
));
2556 qemu_iovec_init(qiov
,
2557 reqs
[outidx
].qiov
->niov
+ reqs
[i
].qiov
->niov
+ 1);
2559 // Add the first request to the merged one. If the requests are
2560 // overlapping, drop the last sectors of the first request.
2561 size
= (reqs
[i
].sector
- reqs
[outidx
].sector
) << 9;
2562 qemu_iovec_concat(qiov
, reqs
[outidx
].qiov
, size
);
2564 // We might need to add some zeros between the two requests
2565 if (reqs
[i
].sector
> oldreq_last
) {
2566 size_t zero_bytes
= (reqs
[i
].sector
- oldreq_last
) << 9;
2567 uint8_t *buf
= qemu_blockalign(bs
, zero_bytes
);
2568 memset(buf
, 0, zero_bytes
);
2569 qemu_iovec_add(qiov
, buf
, zero_bytes
);
2570 mcb
->callbacks
[i
].free_buf
= buf
;
2573 // Add the second request
2574 qemu_iovec_concat(qiov
, reqs
[i
].qiov
, reqs
[i
].qiov
->size
);
2576 reqs
[outidx
].nb_sectors
= qiov
->size
>> 9;
2577 reqs
[outidx
].qiov
= qiov
;
2579 mcb
->callbacks
[i
].free_qiov
= reqs
[outidx
].qiov
;
2582 reqs
[outidx
].sector
= reqs
[i
].sector
;
2583 reqs
[outidx
].nb_sectors
= reqs
[i
].nb_sectors
;
2584 reqs
[outidx
].qiov
= reqs
[i
].qiov
;
2592 * Submit multiple AIO write requests at once.
2594 * On success, the function returns 0 and all requests in the reqs array have
2595 * been submitted. In error case this function returns -1, and any of the
2596 * requests may or may not be submitted yet. In particular, this means that the
2597 * callback will be called for some of the requests, for others it won't. The
2598 * caller must check the error field of the BlockRequest to wait for the right
2599 * callbacks (if error != 0, no callback will be called).
2601 * The implementation may modify the contents of the reqs array, e.g. to merge
2602 * requests. However, the fields opaque and error are left unmodified as they
2603 * are used to signal failure for a single request to the caller.
2605 int bdrv_aio_multiwrite(BlockDriverState
*bs
, BlockRequest
*reqs
, int num_reqs
)
2607 BlockDriverAIOCB
*acb
;
2611 /* don't submit writes if we don't have a medium */
2612 if (bs
->drv
== NULL
) {
2613 for (i
= 0; i
< num_reqs
; i
++) {
2614 reqs
[i
].error
= -ENOMEDIUM
;
2619 if (num_reqs
== 0) {
2623 // Create MultiwriteCB structure
2624 mcb
= g_malloc0(sizeof(*mcb
) + num_reqs
* sizeof(*mcb
->callbacks
));
2625 mcb
->num_requests
= 0;
2626 mcb
->num_callbacks
= num_reqs
;
2628 for (i
= 0; i
< num_reqs
; i
++) {
2629 mcb
->callbacks
[i
].cb
= reqs
[i
].cb
;
2630 mcb
->callbacks
[i
].opaque
= reqs
[i
].opaque
;
2633 // Check for mergable requests
2634 num_reqs
= multiwrite_merge(bs
, reqs
, num_reqs
, mcb
);
2636 trace_bdrv_aio_multiwrite(mcb
, mcb
->num_callbacks
, num_reqs
);
2639 * Run the aio requests. As soon as one request can't be submitted
2640 * successfully, fail all requests that are not yet submitted (we must
2641 * return failure for all requests anyway)
2643 * num_requests cannot be set to the right value immediately: If
2644 * bdrv_aio_writev fails for some request, num_requests would be too high
2645 * and therefore multiwrite_cb() would never recognize the multiwrite
2646 * request as completed. We also cannot use the loop variable i to set it
2647 * when the first request fails because the callback may already have been
2648 * called for previously submitted requests. Thus, num_requests must be
2649 * incremented for each request that is submitted.
2651 * The problem that callbacks may be called early also means that we need
2652 * to take care that num_requests doesn't become 0 before all requests are
2653 * submitted - multiwrite_cb() would consider the multiwrite request
2654 * completed. A dummy request that is "completed" by a manual call to
2655 * multiwrite_cb() takes care of this.
2657 mcb
->num_requests
= 1;
2659 // Run the aio requests
2660 for (i
= 0; i
< num_reqs
; i
++) {
2661 mcb
->num_requests
++;
2662 acb
= bdrv_aio_writev(bs
, reqs
[i
].sector
, reqs
[i
].qiov
,
2663 reqs
[i
].nb_sectors
, multiwrite_cb
, mcb
);
2666 // We can only fail the whole thing if no request has been
2667 // submitted yet. Otherwise we'll wait for the submitted AIOs to
2668 // complete and report the error in the callback.
2670 trace_bdrv_aio_multiwrite_earlyfail(mcb
);
2673 trace_bdrv_aio_multiwrite_latefail(mcb
, i
);
2674 multiwrite_cb(mcb
, -EIO
);
2680 /* Complete the dummy request */
2681 multiwrite_cb(mcb
, 0);
2686 for (i
= 0; i
< mcb
->num_callbacks
; i
++) {
2687 reqs
[i
].error
= -EIO
;
2693 BlockDriverAIOCB
*bdrv_aio_flush(BlockDriverState
*bs
,
2694 BlockDriverCompletionFunc
*cb
, void *opaque
)
2696 BlockDriver
*drv
= bs
->drv
;
2698 trace_bdrv_aio_flush(bs
, opaque
);
2700 if (bs
->open_flags
& BDRV_O_NO_FLUSH
) {
2701 return bdrv_aio_noop_em(bs
, cb
, opaque
);
2706 return drv
->bdrv_aio_flush(bs
, cb
, opaque
);
2709 void bdrv_aio_cancel(BlockDriverAIOCB
*acb
)
2711 acb
->pool
->cancel(acb
);
2715 /**************************************************************/
2716 /* async block device emulation */
2718 typedef struct BlockDriverAIOCBSync
{
2719 BlockDriverAIOCB common
;
2722 /* vector translation state */
2726 } BlockDriverAIOCBSync
;
2728 static void bdrv_aio_cancel_em(BlockDriverAIOCB
*blockacb
)
2730 BlockDriverAIOCBSync
*acb
=
2731 container_of(blockacb
, BlockDriverAIOCBSync
, common
);
2732 qemu_bh_delete(acb
->bh
);
2734 qemu_aio_release(acb
);
2737 static AIOPool bdrv_em_aio_pool
= {
2738 .aiocb_size
= sizeof(BlockDriverAIOCBSync
),
2739 .cancel
= bdrv_aio_cancel_em
,
2742 static void bdrv_aio_bh_cb(void *opaque
)
2744 BlockDriverAIOCBSync
*acb
= opaque
;
2747 qemu_iovec_from_buffer(acb
->qiov
, acb
->bounce
, acb
->qiov
->size
);
2748 qemu_vfree(acb
->bounce
);
2749 acb
->common
.cb(acb
->common
.opaque
, acb
->ret
);
2750 qemu_bh_delete(acb
->bh
);
2752 qemu_aio_release(acb
);
2755 static BlockDriverAIOCB
*bdrv_aio_rw_vector(BlockDriverState
*bs
,
2759 BlockDriverCompletionFunc
*cb
,
2764 BlockDriverAIOCBSync
*acb
;
2766 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2767 acb
->is_write
= is_write
;
2769 acb
->bounce
= qemu_blockalign(bs
, qiov
->size
);
2772 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2775 qemu_iovec_to_buffer(acb
->qiov
, acb
->bounce
);
2776 acb
->ret
= bs
->drv
->bdrv_write(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2778 acb
->ret
= bs
->drv
->bdrv_read(bs
, sector_num
, acb
->bounce
, nb_sectors
);
2781 qemu_bh_schedule(acb
->bh
);
2783 return &acb
->common
;
2786 static BlockDriverAIOCB
*bdrv_aio_readv_em(BlockDriverState
*bs
,
2787 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2788 BlockDriverCompletionFunc
*cb
, void *opaque
)
2790 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 0);
2793 static BlockDriverAIOCB
*bdrv_aio_writev_em(BlockDriverState
*bs
,
2794 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2795 BlockDriverCompletionFunc
*cb
, void *opaque
)
2797 return bdrv_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
, 1);
2801 typedef struct BlockDriverAIOCBCoroutine
{
2802 BlockDriverAIOCB common
;
2806 } BlockDriverAIOCBCoroutine
;
2808 static void bdrv_aio_co_cancel_em(BlockDriverAIOCB
*blockacb
)
2813 static AIOPool bdrv_em_co_aio_pool
= {
2814 .aiocb_size
= sizeof(BlockDriverAIOCBCoroutine
),
2815 .cancel
= bdrv_aio_co_cancel_em
,
2818 static void bdrv_co_rw_bh(void *opaque
)
2820 BlockDriverAIOCBCoroutine
*acb
= opaque
;
2822 acb
->common
.cb(acb
->common
.opaque
, acb
->req
.error
);
2823 qemu_bh_delete(acb
->bh
);
2824 qemu_aio_release(acb
);
2827 static void coroutine_fn
bdrv_co_rw(void *opaque
)
2829 BlockDriverAIOCBCoroutine
*acb
= opaque
;
2830 BlockDriverState
*bs
= acb
->common
.bs
;
2832 if (!acb
->is_write
) {
2833 acb
->req
.error
= bs
->drv
->bdrv_co_readv(bs
, acb
->req
.sector
,
2834 acb
->req
.nb_sectors
, acb
->req
.qiov
);
2836 acb
->req
.error
= bs
->drv
->bdrv_co_writev(bs
, acb
->req
.sector
,
2837 acb
->req
.nb_sectors
, acb
->req
.qiov
);
2840 acb
->bh
= qemu_bh_new(bdrv_co_rw_bh
, acb
);
2841 qemu_bh_schedule(acb
->bh
);
2844 static BlockDriverAIOCB
*bdrv_co_aio_rw_vector(BlockDriverState
*bs
,
2848 BlockDriverCompletionFunc
*cb
,
2853 BlockDriverAIOCBCoroutine
*acb
;
2855 acb
= qemu_aio_get(&bdrv_em_co_aio_pool
, bs
, cb
, opaque
);
2856 acb
->req
.sector
= sector_num
;
2857 acb
->req
.nb_sectors
= nb_sectors
;
2858 acb
->req
.qiov
= qiov
;
2859 acb
->is_write
= is_write
;
2861 co
= qemu_coroutine_create(bdrv_co_rw
);
2862 qemu_coroutine_enter(co
, acb
);
2864 return &acb
->common
;
2867 static BlockDriverAIOCB
*bdrv_co_aio_readv_em(BlockDriverState
*bs
,
2868 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2869 BlockDriverCompletionFunc
*cb
, void *opaque
)
2871 return bdrv_co_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
,
2875 static BlockDriverAIOCB
*bdrv_co_aio_writev_em(BlockDriverState
*bs
,
2876 int64_t sector_num
, QEMUIOVector
*qiov
, int nb_sectors
,
2877 BlockDriverCompletionFunc
*cb
, void *opaque
)
2879 return bdrv_co_aio_rw_vector(bs
, sector_num
, qiov
, nb_sectors
, cb
, opaque
,
2883 static BlockDriverAIOCB
*bdrv_aio_flush_em(BlockDriverState
*bs
,
2884 BlockDriverCompletionFunc
*cb
, void *opaque
)
2886 BlockDriverAIOCBSync
*acb
;
2888 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2889 acb
->is_write
= 1; /* don't bounce in the completion hadler */
2895 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2898 qemu_bh_schedule(acb
->bh
);
2899 return &acb
->common
;
2902 static BlockDriverAIOCB
*bdrv_aio_noop_em(BlockDriverState
*bs
,
2903 BlockDriverCompletionFunc
*cb
, void *opaque
)
2905 BlockDriverAIOCBSync
*acb
;
2907 acb
= qemu_aio_get(&bdrv_em_aio_pool
, bs
, cb
, opaque
);
2908 acb
->is_write
= 1; /* don't bounce in the completion handler */
2914 acb
->bh
= qemu_bh_new(bdrv_aio_bh_cb
, acb
);
2917 qemu_bh_schedule(acb
->bh
);
2918 return &acb
->common
;
2921 /**************************************************************/
2922 /* sync block device emulation */
2924 static void bdrv_rw_em_cb(void *opaque
, int ret
)
2926 *(int *)opaque
= ret
;
2929 static int bdrv_read_em(BlockDriverState
*bs
, int64_t sector_num
,
2930 uint8_t *buf
, int nb_sectors
)
2933 BlockDriverAIOCB
*acb
;
2937 async_ret
= NOT_DONE
;
2938 iov
.iov_base
= (void *)buf
;
2939 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2940 qemu_iovec_init_external(&qiov
, &iov
, 1);
2942 acb
= bs
->drv
->bdrv_aio_readv(bs
, sector_num
, &qiov
, nb_sectors
,
2943 bdrv_rw_em_cb
, &async_ret
);
2949 while (async_ret
== NOT_DONE
) {
2958 static int bdrv_write_em(BlockDriverState
*bs
, int64_t sector_num
,
2959 const uint8_t *buf
, int nb_sectors
)
2962 BlockDriverAIOCB
*acb
;
2966 async_ret
= NOT_DONE
;
2967 iov
.iov_base
= (void *)buf
;
2968 iov
.iov_len
= nb_sectors
* BDRV_SECTOR_SIZE
;
2969 qemu_iovec_init_external(&qiov
, &iov
, 1);
2971 acb
= bs
->drv
->bdrv_aio_writev(bs
, sector_num
, &qiov
, nb_sectors
,
2972 bdrv_rw_em_cb
, &async_ret
);
2977 while (async_ret
== NOT_DONE
) {
2985 void bdrv_init(void)
2987 module_call_init(MODULE_INIT_BLOCK
);
2990 void bdrv_init_with_whitelist(void)
2992 use_bdrv_whitelist
= 1;
2996 void *qemu_aio_get(AIOPool
*pool
, BlockDriverState
*bs
,
2997 BlockDriverCompletionFunc
*cb
, void *opaque
)
2999 BlockDriverAIOCB
*acb
;
3001 if (pool
->free_aiocb
) {
3002 acb
= pool
->free_aiocb
;
3003 pool
->free_aiocb
= acb
->next
;
3005 acb
= g_malloc0(pool
->aiocb_size
);
3010 acb
->opaque
= opaque
;
3014 void qemu_aio_release(void *p
)
3016 BlockDriverAIOCB
*acb
= (BlockDriverAIOCB
*)p
;
3017 AIOPool
*pool
= acb
->pool
;
3018 acb
->next
= pool
->free_aiocb
;
3019 pool
->free_aiocb
= acb
;
3022 /**************************************************************/
3023 /* Coroutine block device emulation */
3025 typedef struct CoroutineIOCompletion
{
3026 Coroutine
*coroutine
;
3028 } CoroutineIOCompletion
;
3030 static void bdrv_co_io_em_complete(void *opaque
, int ret
)
3032 CoroutineIOCompletion
*co
= opaque
;
3035 qemu_coroutine_enter(co
->coroutine
, NULL
);
3038 static int coroutine_fn
bdrv_co_io_em(BlockDriverState
*bs
, int64_t sector_num
,
3039 int nb_sectors
, QEMUIOVector
*iov
,
3042 CoroutineIOCompletion co
= {
3043 .coroutine
= qemu_coroutine_self(),
3045 BlockDriverAIOCB
*acb
;
3048 acb
= bs
->drv
->bdrv_aio_writev(bs
, sector_num
, iov
, nb_sectors
,
3049 bdrv_co_io_em_complete
, &co
);
3051 acb
= bs
->drv
->bdrv_aio_readv(bs
, sector_num
, iov
, nb_sectors
,
3052 bdrv_co_io_em_complete
, &co
);
3055 trace_bdrv_co_io_em(bs
, sector_num
, nb_sectors
, is_write
, acb
);
3059 qemu_coroutine_yield();
3064 static int coroutine_fn
bdrv_co_readv_em(BlockDriverState
*bs
,
3065 int64_t sector_num
, int nb_sectors
,
3068 return bdrv_co_io_em(bs
, sector_num
, nb_sectors
, iov
, false);
3071 static int coroutine_fn
bdrv_co_writev_em(BlockDriverState
*bs
,
3072 int64_t sector_num
, int nb_sectors
,
3075 return bdrv_co_io_em(bs
, sector_num
, nb_sectors
, iov
, true);
3078 static int coroutine_fn
bdrv_co_flush_em(BlockDriverState
*bs
)
3080 CoroutineIOCompletion co
= {
3081 .coroutine
= qemu_coroutine_self(),
3083 BlockDriverAIOCB
*acb
;
3085 acb
= bdrv_aio_flush(bs
, bdrv_co_io_em_complete
, &co
);
3089 qemu_coroutine_yield();
3093 /**************************************************************/
3094 /* removable device support */
3097 * Return TRUE if the media is present
3099 int bdrv_is_inserted(BlockDriverState
*bs
)
3101 BlockDriver
*drv
= bs
->drv
;
3105 if (!drv
->bdrv_is_inserted
)
3107 return drv
->bdrv_is_inserted(bs
);
3111 * Return whether the media changed since the last call to this
3112 * function, or -ENOTSUP if we don't know. Most drivers don't know.
3114 int bdrv_media_changed(BlockDriverState
*bs
)
3116 BlockDriver
*drv
= bs
->drv
;
3118 if (drv
&& drv
->bdrv_media_changed
) {
3119 return drv
->bdrv_media_changed(bs
);
3125 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
3127 void bdrv_eject(BlockDriverState
*bs
, int eject_flag
)
3129 BlockDriver
*drv
= bs
->drv
;
3131 if (drv
&& drv
->bdrv_eject
) {
3132 drv
->bdrv_eject(bs
, eject_flag
);
3137 * Lock or unlock the media (if it is locked, the user won't be able
3138 * to eject it manually).
3140 void bdrv_lock_medium(BlockDriverState
*bs
, bool locked
)
3142 BlockDriver
*drv
= bs
->drv
;
3144 trace_bdrv_lock_medium(bs
, locked
);
3146 if (drv
&& drv
->bdrv_lock_medium
) {
3147 drv
->bdrv_lock_medium(bs
, locked
);
3151 /* needed for generic scsi interface */
3153 int bdrv_ioctl(BlockDriverState
*bs
, unsigned long int req
, void *buf
)
3155 BlockDriver
*drv
= bs
->drv
;
3157 if (drv
&& drv
->bdrv_ioctl
)
3158 return drv
->bdrv_ioctl(bs
, req
, buf
);
3162 BlockDriverAIOCB
*bdrv_aio_ioctl(BlockDriverState
*bs
,
3163 unsigned long int req
, void *buf
,
3164 BlockDriverCompletionFunc
*cb
, void *opaque
)
3166 BlockDriver
*drv
= bs
->drv
;
3168 if (drv
&& drv
->bdrv_aio_ioctl
)
3169 return drv
->bdrv_aio_ioctl(bs
, req
, buf
, cb
, opaque
);
3173 void bdrv_set_buffer_alignment(BlockDriverState
*bs
, int align
)
3175 bs
->buffer_alignment
= align
;
3178 void *qemu_blockalign(BlockDriverState
*bs
, size_t size
)
3180 return qemu_memalign((bs
&& bs
->buffer_alignment
) ? bs
->buffer_alignment
: 512, size
);
3183 void bdrv_set_dirty_tracking(BlockDriverState
*bs
, int enable
)
3185 int64_t bitmap_size
;
3187 bs
->dirty_count
= 0;
3189 if (!bs
->dirty_bitmap
) {
3190 bitmap_size
= (bdrv_getlength(bs
) >> BDRV_SECTOR_BITS
) +
3191 BDRV_SECTORS_PER_DIRTY_CHUNK
* 8 - 1;
3192 bitmap_size
/= BDRV_SECTORS_PER_DIRTY_CHUNK
* 8;
3194 bs
->dirty_bitmap
= g_malloc0(bitmap_size
);
3197 if (bs
->dirty_bitmap
) {
3198 g_free(bs
->dirty_bitmap
);
3199 bs
->dirty_bitmap
= NULL
;
3204 int bdrv_get_dirty(BlockDriverState
*bs
, int64_t sector
)
3206 int64_t chunk
= sector
/ (int64_t)BDRV_SECTORS_PER_DIRTY_CHUNK
;
3208 if (bs
->dirty_bitmap
&&
3209 (sector
<< BDRV_SECTOR_BITS
) < bdrv_getlength(bs
)) {
3210 return !!(bs
->dirty_bitmap
[chunk
/ (sizeof(unsigned long) * 8)] &
3211 (1UL << (chunk
% (sizeof(unsigned long) * 8))));
3217 void bdrv_reset_dirty(BlockDriverState
*bs
, int64_t cur_sector
,
3220 set_dirty_bitmap(bs
, cur_sector
, nr_sectors
, 0);
3223 int64_t bdrv_get_dirty_count(BlockDriverState
*bs
)
3225 return bs
->dirty_count
;
3228 void bdrv_set_in_use(BlockDriverState
*bs
, int in_use
)
3230 assert(bs
->in_use
!= in_use
);
3231 bs
->in_use
= in_use
;
3234 int bdrv_in_use(BlockDriverState
*bs
)
3239 void bdrv_iostatus_enable(BlockDriverState
*bs
)
3241 bs
->iostatus
= BDRV_IOS_OK
;
3244 /* The I/O status is only enabled if the drive explicitly
3245 * enables it _and_ the VM is configured to stop on errors */
3246 bool bdrv_iostatus_is_enabled(const BlockDriverState
*bs
)
3248 return (bs
->iostatus
!= BDRV_IOS_INVAL
&&
3249 (bs
->on_write_error
== BLOCK_ERR_STOP_ENOSPC
||
3250 bs
->on_write_error
== BLOCK_ERR_STOP_ANY
||
3251 bs
->on_read_error
== BLOCK_ERR_STOP_ANY
));
3254 void bdrv_iostatus_disable(BlockDriverState
*bs
)
3256 bs
->iostatus
= BDRV_IOS_INVAL
;
3259 void bdrv_iostatus_reset(BlockDriverState
*bs
)
3261 if (bdrv_iostatus_is_enabled(bs
)) {
3262 bs
->iostatus
= BDRV_IOS_OK
;
3266 /* XXX: Today this is set by device models because it makes the implementation
3267 quite simple. However, the block layer knows about the error, so it's
3268 possible to implement this without device models being involved */
3269 void bdrv_iostatus_set_err(BlockDriverState
*bs
, int error
)
3271 if (bdrv_iostatus_is_enabled(bs
) && bs
->iostatus
== BDRV_IOS_OK
) {
3273 bs
->iostatus
= error
== ENOSPC
? BDRV_IOS_ENOSPC
: BDRV_IOS_FAILED
;
3278 bdrv_acct_start(BlockDriverState
*bs
, BlockAcctCookie
*cookie
, int64_t bytes
,
3279 enum BlockAcctType type
)
3281 assert(type
< BDRV_MAX_IOTYPE
);
3283 cookie
->bytes
= bytes
;
3284 cookie
->start_time_ns
= get_clock();
3285 cookie
->type
= type
;
3289 bdrv_acct_done(BlockDriverState
*bs
, BlockAcctCookie
*cookie
)
3291 assert(cookie
->type
< BDRV_MAX_IOTYPE
);
3293 bs
->nr_bytes
[cookie
->type
] += cookie
->bytes
;
3294 bs
->nr_ops
[cookie
->type
]++;
3295 bs
->total_time_ns
[cookie
->type
] += get_clock() - cookie
->start_time_ns
;
3298 int bdrv_img_create(const char *filename
, const char *fmt
,
3299 const char *base_filename
, const char *base_fmt
,
3300 char *options
, uint64_t img_size
, int flags
)
3302 QEMUOptionParameter
*param
= NULL
, *create_options
= NULL
;
3303 QEMUOptionParameter
*backing_fmt
, *backing_file
, *size
;
3304 BlockDriverState
*bs
= NULL
;
3305 BlockDriver
*drv
, *proto_drv
;
3306 BlockDriver
*backing_drv
= NULL
;
3309 /* Find driver and parse its options */
3310 drv
= bdrv_find_format(fmt
);
3312 error_report("Unknown file format '%s'", fmt
);
3317 proto_drv
= bdrv_find_protocol(filename
);
3319 error_report("Unknown protocol '%s'", filename
);
3324 create_options
= append_option_parameters(create_options
,
3325 drv
->create_options
);
3326 create_options
= append_option_parameters(create_options
,
3327 proto_drv
->create_options
);
3329 /* Create parameter list with default values */
3330 param
= parse_option_parameters("", create_options
, param
);
3332 set_option_parameter_int(param
, BLOCK_OPT_SIZE
, img_size
);
3334 /* Parse -o options */
3336 param
= parse_option_parameters(options
, create_options
, param
);
3337 if (param
== NULL
) {
3338 error_report("Invalid options for file format '%s'.", fmt
);
3344 if (base_filename
) {
3345 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FILE
,
3347 error_report("Backing file not supported for file format '%s'",
3355 if (set_option_parameter(param
, BLOCK_OPT_BACKING_FMT
, base_fmt
)) {
3356 error_report("Backing file format not supported for file "
3357 "format '%s'", fmt
);
3363 backing_file
= get_option_parameter(param
, BLOCK_OPT_BACKING_FILE
);
3364 if (backing_file
&& backing_file
->value
.s
) {
3365 if (!strcmp(filename
, backing_file
->value
.s
)) {
3366 error_report("Error: Trying to create an image with the "
3367 "same filename as the backing file");
3373 backing_fmt
= get_option_parameter(param
, BLOCK_OPT_BACKING_FMT
);
3374 if (backing_fmt
&& backing_fmt
->value
.s
) {
3375 backing_drv
= bdrv_find_format(backing_fmt
->value
.s
);
3377 error_report("Unknown backing file format '%s'",
3378 backing_fmt
->value
.s
);
3384 // The size for the image must always be specified, with one exception:
3385 // If we are using a backing file, we can obtain the size from there
3386 size
= get_option_parameter(param
, BLOCK_OPT_SIZE
);
3387 if (size
&& size
->value
.n
== -1) {
3388 if (backing_file
&& backing_file
->value
.s
) {
3394 ret
= bdrv_open(bs
, backing_file
->value
.s
, flags
, backing_drv
);
3396 error_report("Could not open '%s'", backing_file
->value
.s
);
3399 bdrv_get_geometry(bs
, &size
);
3402 snprintf(buf
, sizeof(buf
), "%" PRId64
, size
);
3403 set_option_parameter(param
, BLOCK_OPT_SIZE
, buf
);
3405 error_report("Image creation needs a size parameter");
3411 printf("Formatting '%s', fmt=%s ", filename
, fmt
);
3412 print_option_parameters(param
);
3415 ret
= bdrv_create(drv
, filename
, param
);
3418 if (ret
== -ENOTSUP
) {
3419 error_report("Formatting or formatting option not supported for "
3420 "file format '%s'", fmt
);
3421 } else if (ret
== -EFBIG
) {
3422 error_report("The image size is too large for file format '%s'",
3425 error_report("%s: error while creating %s: %s", filename
, fmt
,
3431 free_option_parameters(create_options
);
3432 free_option_parameters(param
);