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 "qemu/osdep.h"
26 #include "block/block_int.h"
27 #include "block/blockjob.h"
28 #include "block/nbd.h"
29 #include "qemu/error-report.h"
30 #include "module_block.h"
31 #include "qemu/module.h"
32 #include "qapi/qmp/qerror.h"
33 #include "qapi/qmp/qbool.h"
34 #include "qapi/qmp/qjson.h"
35 #include "sysemu/block-backend.h"
36 #include "sysemu/sysemu.h"
37 #include "qemu/notify.h"
38 #include "qemu/coroutine.h"
39 #include "block/qapi.h"
40 #include "qmp-commands.h"
41 #include "qemu/timer.h"
42 #include "qapi-event.h"
43 #include "qemu/cutils.h"
47 #include <sys/ioctl.h>
48 #include <sys/queue.h>
58 #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */
60 static QTAILQ_HEAD(, BlockDriverState
) graph_bdrv_states
=
61 QTAILQ_HEAD_INITIALIZER(graph_bdrv_states
);
63 static QTAILQ_HEAD(, BlockDriverState
) all_bdrv_states
=
64 QTAILQ_HEAD_INITIALIZER(all_bdrv_states
);
66 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
67 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
69 static BlockDriverState
*bdrv_open_inherit(const char *filename
,
70 const char *reference
,
71 QDict
*options
, int flags
,
72 BlockDriverState
*parent
,
73 const BdrvChildRole
*child_role
,
76 /* If non-zero, use only whitelisted block drivers */
77 static int use_bdrv_whitelist
;
80 static int is_windows_drive_prefix(const char *filename
)
82 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
83 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
87 int is_windows_drive(const char *filename
)
89 if (is_windows_drive_prefix(filename
) &&
92 if (strstart(filename
, "\\\\.\\", NULL
) ||
93 strstart(filename
, "//./", NULL
))
99 size_t bdrv_opt_mem_align(BlockDriverState
*bs
)
101 if (!bs
|| !bs
->drv
) {
102 /* page size or 4k (hdd sector size) should be on the safe side */
103 return MAX(4096, getpagesize());
106 return bs
->bl
.opt_mem_alignment
;
109 size_t bdrv_min_mem_align(BlockDriverState
*bs
)
111 if (!bs
|| !bs
->drv
) {
112 /* page size or 4k (hdd sector size) should be on the safe side */
113 return MAX(4096, getpagesize());
116 return bs
->bl
.min_mem_alignment
;
119 /* check if the path starts with "<protocol>:" */
120 int path_has_protocol(const char *path
)
125 if (is_windows_drive(path
) ||
126 is_windows_drive_prefix(path
)) {
129 p
= path
+ strcspn(path
, ":/\\");
131 p
= path
+ strcspn(path
, ":/");
137 int path_is_absolute(const char *path
)
140 /* specific case for names like: "\\.\d:" */
141 if (is_windows_drive(path
) || is_windows_drive_prefix(path
)) {
144 return (*path
== '/' || *path
== '\\');
146 return (*path
== '/');
150 /* if filename is absolute, just copy it to dest. Otherwise, build a
151 path to it by considering it is relative to base_path. URL are
153 void path_combine(char *dest
, int dest_size
,
154 const char *base_path
,
155 const char *filename
)
162 if (path_is_absolute(filename
)) {
163 pstrcpy(dest
, dest_size
, filename
);
165 p
= strchr(base_path
, ':');
170 p1
= strrchr(base_path
, '/');
174 p2
= strrchr(base_path
, '\\');
186 if (len
> dest_size
- 1)
188 memcpy(dest
, base_path
, len
);
190 pstrcat(dest
, dest_size
, filename
);
194 void bdrv_get_full_backing_filename_from_filename(const char *backed
,
196 char *dest
, size_t sz
,
199 if (backing
[0] == '\0' || path_has_protocol(backing
) ||
200 path_is_absolute(backing
))
202 pstrcpy(dest
, sz
, backing
);
203 } else if (backed
[0] == '\0' || strstart(backed
, "json:", NULL
)) {
204 error_setg(errp
, "Cannot use relative backing file names for '%s'",
207 path_combine(dest
, sz
, backed
, backing
);
211 void bdrv_get_full_backing_filename(BlockDriverState
*bs
, char *dest
, size_t sz
,
214 char *backed
= bs
->exact_filename
[0] ? bs
->exact_filename
: bs
->filename
;
216 bdrv_get_full_backing_filename_from_filename(backed
, bs
->backing_file
,
220 void bdrv_register(BlockDriver
*bdrv
)
222 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
225 BlockDriverState
*bdrv_new(void)
227 BlockDriverState
*bs
;
230 bs
= g_new0(BlockDriverState
, 1);
231 QLIST_INIT(&bs
->dirty_bitmaps
);
232 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
233 QLIST_INIT(&bs
->op_blockers
[i
]);
235 notifier_with_return_list_init(&bs
->before_write_notifiers
);
237 bs
->aio_context
= qemu_get_aio_context();
239 qemu_co_queue_init(&bs
->flush_queue
);
241 QTAILQ_INSERT_TAIL(&all_bdrv_states
, bs
, bs_list
);
246 static BlockDriver
*bdrv_do_find_format(const char *format_name
)
250 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
251 if (!strcmp(drv1
->format_name
, format_name
)) {
259 BlockDriver
*bdrv_find_format(const char *format_name
)
264 drv1
= bdrv_do_find_format(format_name
);
269 /* The driver isn't registered, maybe we need to load a module */
270 for (i
= 0; i
< (int)ARRAY_SIZE(block_driver_modules
); ++i
) {
271 if (!strcmp(block_driver_modules
[i
].format_name
, format_name
)) {
272 block_module_load_one(block_driver_modules
[i
].library_name
);
277 return bdrv_do_find_format(format_name
);
280 static int bdrv_is_whitelisted(BlockDriver
*drv
, bool read_only
)
282 static const char *whitelist_rw
[] = {
283 CONFIG_BDRV_RW_WHITELIST
285 static const char *whitelist_ro
[] = {
286 CONFIG_BDRV_RO_WHITELIST
290 if (!whitelist_rw
[0] && !whitelist_ro
[0]) {
291 return 1; /* no whitelist, anything goes */
294 for (p
= whitelist_rw
; *p
; p
++) {
295 if (!strcmp(drv
->format_name
, *p
)) {
300 for (p
= whitelist_ro
; *p
; p
++) {
301 if (!strcmp(drv
->format_name
, *p
)) {
309 bool bdrv_uses_whitelist(void)
311 return use_bdrv_whitelist
;
314 typedef struct CreateCo
{
322 static void coroutine_fn
bdrv_create_co_entry(void *opaque
)
324 Error
*local_err
= NULL
;
327 CreateCo
*cco
= opaque
;
330 ret
= cco
->drv
->bdrv_create(cco
->filename
, cco
->opts
, &local_err
);
331 error_propagate(&cco
->err
, local_err
);
335 int bdrv_create(BlockDriver
*drv
, const char* filename
,
336 QemuOpts
*opts
, Error
**errp
)
343 .filename
= g_strdup(filename
),
349 if (!drv
->bdrv_create
) {
350 error_setg(errp
, "Driver '%s' does not support image creation", drv
->format_name
);
355 if (qemu_in_coroutine()) {
356 /* Fast-path if already in coroutine context */
357 bdrv_create_co_entry(&cco
);
359 co
= qemu_coroutine_create(bdrv_create_co_entry
, &cco
);
360 qemu_coroutine_enter(co
);
361 while (cco
.ret
== NOT_DONE
) {
362 aio_poll(qemu_get_aio_context(), true);
369 error_propagate(errp
, cco
.err
);
371 error_setg_errno(errp
, -ret
, "Could not create image");
376 g_free(cco
.filename
);
380 int bdrv_create_file(const char *filename
, QemuOpts
*opts
, Error
**errp
)
383 Error
*local_err
= NULL
;
386 drv
= bdrv_find_protocol(filename
, true, errp
);
391 ret
= bdrv_create(drv
, filename
, opts
, &local_err
);
392 error_propagate(errp
, local_err
);
397 * Try to get @bs's logical and physical block size.
398 * On success, store them in @bsz struct and return 0.
399 * On failure return -errno.
400 * @bs must not be empty.
402 int bdrv_probe_blocksizes(BlockDriverState
*bs
, BlockSizes
*bsz
)
404 BlockDriver
*drv
= bs
->drv
;
406 if (drv
&& drv
->bdrv_probe_blocksizes
) {
407 return drv
->bdrv_probe_blocksizes(bs
, bsz
);
414 * Try to get @bs's geometry (cyls, heads, sectors).
415 * On success, store them in @geo struct and return 0.
416 * On failure return -errno.
417 * @bs must not be empty.
419 int bdrv_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
421 BlockDriver
*drv
= bs
->drv
;
423 if (drv
&& drv
->bdrv_probe_geometry
) {
424 return drv
->bdrv_probe_geometry(bs
, geo
);
431 * Create a uniquely-named empty temporary file.
432 * Return 0 upon success, otherwise a negative errno value.
434 int get_tmp_filename(char *filename
, int size
)
437 char temp_dir
[MAX_PATH
];
438 /* GetTempFileName requires that its output buffer (4th param)
439 have length MAX_PATH or greater. */
440 assert(size
>= MAX_PATH
);
441 return (GetTempPath(MAX_PATH
, temp_dir
)
442 && GetTempFileName(temp_dir
, "qem", 0, filename
)
443 ? 0 : -GetLastError());
447 tmpdir
= getenv("TMPDIR");
451 if (snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
) >= size
) {
454 fd
= mkstemp(filename
);
458 if (close(fd
) != 0) {
467 * Detect host devices. By convention, /dev/cdrom[N] is always
468 * recognized as a host CDROM.
470 static BlockDriver
*find_hdev_driver(const char *filename
)
472 int score_max
= 0, score
;
473 BlockDriver
*drv
= NULL
, *d
;
475 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
476 if (d
->bdrv_probe_device
) {
477 score
= d
->bdrv_probe_device(filename
);
478 if (score
> score_max
) {
488 static BlockDriver
*bdrv_do_find_protocol(const char *protocol
)
492 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
493 if (drv1
->protocol_name
&& !strcmp(drv1
->protocol_name
, protocol
)) {
501 BlockDriver
*bdrv_find_protocol(const char *filename
,
502 bool allow_protocol_prefix
,
511 /* TODO Drivers without bdrv_file_open must be specified explicitly */
514 * XXX(hch): we really should not let host device detection
515 * override an explicit protocol specification, but moving this
516 * later breaks access to device names with colons in them.
517 * Thanks to the brain-dead persistent naming schemes on udev-
518 * based Linux systems those actually are quite common.
520 drv1
= find_hdev_driver(filename
);
525 if (!path_has_protocol(filename
) || !allow_protocol_prefix
) {
529 p
= strchr(filename
, ':');
532 if (len
> sizeof(protocol
) - 1)
533 len
= sizeof(protocol
) - 1;
534 memcpy(protocol
, filename
, len
);
535 protocol
[len
] = '\0';
537 drv1
= bdrv_do_find_protocol(protocol
);
542 for (i
= 0; i
< (int)ARRAY_SIZE(block_driver_modules
); ++i
) {
543 if (block_driver_modules
[i
].protocol_name
&&
544 !strcmp(block_driver_modules
[i
].protocol_name
, protocol
)) {
545 block_module_load_one(block_driver_modules
[i
].library_name
);
550 drv1
= bdrv_do_find_protocol(protocol
);
552 error_setg(errp
, "Unknown protocol '%s'", protocol
);
558 * Guess image format by probing its contents.
559 * This is not a good idea when your image is raw (CVE-2008-2004), but
560 * we do it anyway for backward compatibility.
562 * @buf contains the image's first @buf_size bytes.
563 * @buf_size is the buffer size in bytes (generally BLOCK_PROBE_BUF_SIZE,
564 * but can be smaller if the image file is smaller)
565 * @filename is its filename.
567 * For all block drivers, call the bdrv_probe() method to get its
569 * Return the first block driver with the highest probing score.
571 BlockDriver
*bdrv_probe_all(const uint8_t *buf
, int buf_size
,
572 const char *filename
)
574 int score_max
= 0, score
;
575 BlockDriver
*drv
= NULL
, *d
;
577 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
579 score
= d
->bdrv_probe(buf
, buf_size
, filename
);
580 if (score
> score_max
) {
590 static int find_image_format(BdrvChild
*file
, const char *filename
,
591 BlockDriver
**pdrv
, Error
**errp
)
593 BlockDriverState
*bs
= file
->bs
;
595 uint8_t buf
[BLOCK_PROBE_BUF_SIZE
];
598 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
599 if (bdrv_is_sg(bs
) || !bdrv_is_inserted(bs
) || bdrv_getlength(bs
) == 0) {
604 ret
= bdrv_pread(file
, 0, buf
, sizeof(buf
));
606 error_setg_errno(errp
, -ret
, "Could not read image for determining its "
612 drv
= bdrv_probe_all(buf
, ret
, filename
);
614 error_setg(errp
, "Could not determine image format: No compatible "
623 * Set the current 'total_sectors' value
624 * Return 0 on success, -errno on error.
626 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
628 BlockDriver
*drv
= bs
->drv
;
630 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
634 /* query actual device if possible, otherwise just trust the hint */
635 if (drv
->bdrv_getlength
) {
636 int64_t length
= drv
->bdrv_getlength(bs
);
640 hint
= DIV_ROUND_UP(length
, BDRV_SECTOR_SIZE
);
643 bs
->total_sectors
= hint
;
648 * Combines a QDict of new block driver @options with any missing options taken
649 * from @old_options, so that leaving out an option defaults to its old value.
651 static void bdrv_join_options(BlockDriverState
*bs
, QDict
*options
,
654 if (bs
->drv
&& bs
->drv
->bdrv_join_options
) {
655 bs
->drv
->bdrv_join_options(options
, old_options
);
657 qdict_join(options
, old_options
, false);
662 * Set open flags for a given discard mode
664 * Return 0 on success, -1 if the discard mode was invalid.
666 int bdrv_parse_discard_flags(const char *mode
, int *flags
)
668 *flags
&= ~BDRV_O_UNMAP
;
670 if (!strcmp(mode
, "off") || !strcmp(mode
, "ignore")) {
672 } else if (!strcmp(mode
, "on") || !strcmp(mode
, "unmap")) {
673 *flags
|= BDRV_O_UNMAP
;
682 * Set open flags for a given cache mode
684 * Return 0 on success, -1 if the cache mode was invalid.
686 int bdrv_parse_cache_mode(const char *mode
, int *flags
, bool *writethrough
)
688 *flags
&= ~BDRV_O_CACHE_MASK
;
690 if (!strcmp(mode
, "off") || !strcmp(mode
, "none")) {
691 *writethrough
= false;
692 *flags
|= BDRV_O_NOCACHE
;
693 } else if (!strcmp(mode
, "directsync")) {
694 *writethrough
= true;
695 *flags
|= BDRV_O_NOCACHE
;
696 } else if (!strcmp(mode
, "writeback")) {
697 *writethrough
= false;
698 } else if (!strcmp(mode
, "unsafe")) {
699 *writethrough
= false;
700 *flags
|= BDRV_O_NO_FLUSH
;
701 } else if (!strcmp(mode
, "writethrough")) {
702 *writethrough
= true;
710 static void bdrv_child_cb_drained_begin(BdrvChild
*child
)
712 BlockDriverState
*bs
= child
->opaque
;
713 bdrv_drained_begin(bs
);
716 static void bdrv_child_cb_drained_end(BdrvChild
*child
)
718 BlockDriverState
*bs
= child
->opaque
;
719 bdrv_drained_end(bs
);
723 * Returns the options and flags that a temporary snapshot should get, based on
724 * the originally requested flags (the originally requested image will have
725 * flags like a backing file)
727 static void bdrv_temp_snapshot_options(int *child_flags
, QDict
*child_options
,
728 int parent_flags
, QDict
*parent_options
)
730 *child_flags
= (parent_flags
& ~BDRV_O_SNAPSHOT
) | BDRV_O_TEMPORARY
;
732 /* For temporary files, unconditional cache=unsafe is fine */
733 qdict_set_default_str(child_options
, BDRV_OPT_CACHE_DIRECT
, "off");
734 qdict_set_default_str(child_options
, BDRV_OPT_CACHE_NO_FLUSH
, "on");
736 /* aio=native doesn't work for cache.direct=off, so disable it for the
737 * temporary snapshot */
738 *child_flags
&= ~BDRV_O_NATIVE_AIO
;
742 * Returns the options and flags that bs->file should get if a protocol driver
743 * is expected, based on the given options and flags for the parent BDS
745 static void bdrv_inherited_options(int *child_flags
, QDict
*child_options
,
746 int parent_flags
, QDict
*parent_options
)
748 int flags
= parent_flags
;
750 /* Enable protocol handling, disable format probing for bs->file */
751 flags
|= BDRV_O_PROTOCOL
;
753 /* If the cache mode isn't explicitly set, inherit direct and no-flush from
755 qdict_copy_default(child_options
, parent_options
, BDRV_OPT_CACHE_DIRECT
);
756 qdict_copy_default(child_options
, parent_options
, BDRV_OPT_CACHE_NO_FLUSH
);
758 /* Our block drivers take care to send flushes and respect unmap policy,
759 * so we can default to enable both on lower layers regardless of the
760 * corresponding parent options. */
761 flags
|= BDRV_O_UNMAP
;
763 /* Clear flags that only apply to the top layer */
764 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
| BDRV_O_COPY_ON_READ
|
767 *child_flags
= flags
;
770 const BdrvChildRole child_file
= {
771 .inherit_options
= bdrv_inherited_options
,
772 .drained_begin
= bdrv_child_cb_drained_begin
,
773 .drained_end
= bdrv_child_cb_drained_end
,
777 * Returns the options and flags that bs->file should get if the use of formats
778 * (and not only protocols) is permitted for it, based on the given options and
779 * flags for the parent BDS
781 static void bdrv_inherited_fmt_options(int *child_flags
, QDict
*child_options
,
782 int parent_flags
, QDict
*parent_options
)
784 child_file
.inherit_options(child_flags
, child_options
,
785 parent_flags
, parent_options
);
787 *child_flags
&= ~(BDRV_O_PROTOCOL
| BDRV_O_NO_IO
);
790 const BdrvChildRole child_format
= {
791 .inherit_options
= bdrv_inherited_fmt_options
,
792 .drained_begin
= bdrv_child_cb_drained_begin
,
793 .drained_end
= bdrv_child_cb_drained_end
,
797 * Returns the options and flags that bs->backing should get, based on the
798 * given options and flags for the parent BDS
800 static void bdrv_backing_options(int *child_flags
, QDict
*child_options
,
801 int parent_flags
, QDict
*parent_options
)
803 int flags
= parent_flags
;
805 /* The cache mode is inherited unmodified for backing files; except WCE,
806 * which is only applied on the top level (BlockBackend) */
807 qdict_copy_default(child_options
, parent_options
, BDRV_OPT_CACHE_DIRECT
);
808 qdict_copy_default(child_options
, parent_options
, BDRV_OPT_CACHE_NO_FLUSH
);
810 /* backing files always opened read-only */
811 flags
&= ~(BDRV_O_RDWR
| BDRV_O_COPY_ON_READ
);
813 /* snapshot=on is handled on the top layer */
814 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_TEMPORARY
);
816 *child_flags
= flags
;
819 static const BdrvChildRole child_backing
= {
820 .inherit_options
= bdrv_backing_options
,
821 .drained_begin
= bdrv_child_cb_drained_begin
,
822 .drained_end
= bdrv_child_cb_drained_end
,
825 static int bdrv_open_flags(BlockDriverState
*bs
, int flags
)
827 int open_flags
= flags
;
830 * Clear flags that are internal to the block layer before opening the
833 open_flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
| BDRV_O_PROTOCOL
);
836 * Snapshots should be writable.
838 if (flags
& BDRV_O_TEMPORARY
) {
839 open_flags
|= BDRV_O_RDWR
;
845 static void update_flags_from_options(int *flags
, QemuOpts
*opts
)
847 *flags
&= ~BDRV_O_CACHE_MASK
;
849 assert(qemu_opt_find(opts
, BDRV_OPT_CACHE_NO_FLUSH
));
850 if (qemu_opt_get_bool(opts
, BDRV_OPT_CACHE_NO_FLUSH
, false)) {
851 *flags
|= BDRV_O_NO_FLUSH
;
854 assert(qemu_opt_find(opts
, BDRV_OPT_CACHE_DIRECT
));
855 if (qemu_opt_get_bool(opts
, BDRV_OPT_CACHE_DIRECT
, false)) {
856 *flags
|= BDRV_O_NOCACHE
;
860 static void update_options_from_flags(QDict
*options
, int flags
)
862 if (!qdict_haskey(options
, BDRV_OPT_CACHE_DIRECT
)) {
863 qdict_put(options
, BDRV_OPT_CACHE_DIRECT
,
864 qbool_from_bool(flags
& BDRV_O_NOCACHE
));
866 if (!qdict_haskey(options
, BDRV_OPT_CACHE_NO_FLUSH
)) {
867 qdict_put(options
, BDRV_OPT_CACHE_NO_FLUSH
,
868 qbool_from_bool(flags
& BDRV_O_NO_FLUSH
));
872 static void bdrv_assign_node_name(BlockDriverState
*bs
,
873 const char *node_name
,
876 char *gen_node_name
= NULL
;
879 node_name
= gen_node_name
= id_generate(ID_BLOCK
);
880 } else if (!id_wellformed(node_name
)) {
882 * Check for empty string or invalid characters, but not if it is
883 * generated (generated names use characters not available to the user)
885 error_setg(errp
, "Invalid node name");
889 /* takes care of avoiding namespaces collisions */
890 if (blk_by_name(node_name
)) {
891 error_setg(errp
, "node-name=%s is conflicting with a device id",
896 /* takes care of avoiding duplicates node names */
897 if (bdrv_find_node(node_name
)) {
898 error_setg(errp
, "Duplicate node name");
902 /* copy node name into the bs and insert it into the graph list */
903 pstrcpy(bs
->node_name
, sizeof(bs
->node_name
), node_name
);
904 QTAILQ_INSERT_TAIL(&graph_bdrv_states
, bs
, node_list
);
906 g_free(gen_node_name
);
909 static QemuOptsList bdrv_runtime_opts
= {
910 .name
= "bdrv_common",
911 .head
= QTAILQ_HEAD_INITIALIZER(bdrv_runtime_opts
.head
),
915 .type
= QEMU_OPT_STRING
,
916 .help
= "Node name of the block device node",
920 .type
= QEMU_OPT_STRING
,
921 .help
= "Block driver to use for the node",
924 .name
= BDRV_OPT_CACHE_DIRECT
,
925 .type
= QEMU_OPT_BOOL
,
926 .help
= "Bypass software writeback cache on the host",
929 .name
= BDRV_OPT_CACHE_NO_FLUSH
,
930 .type
= QEMU_OPT_BOOL
,
931 .help
= "Ignore flush requests",
933 { /* end of list */ }
938 * Common part for opening disk images and files
940 * Removes all processed options from *options.
942 static int bdrv_open_common(BlockDriverState
*bs
, BdrvChild
*file
,
943 QDict
*options
, Error
**errp
)
946 const char *filename
;
947 const char *driver_name
= NULL
;
948 const char *node_name
= NULL
;
951 Error
*local_err
= NULL
;
953 assert(bs
->file
== NULL
);
954 assert(options
!= NULL
&& bs
->options
!= options
);
956 opts
= qemu_opts_create(&bdrv_runtime_opts
, NULL
, 0, &error_abort
);
957 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
959 error_propagate(errp
, local_err
);
964 update_flags_from_options(&bs
->open_flags
, opts
);
966 driver_name
= qemu_opt_get(opts
, "driver");
967 drv
= bdrv_find_format(driver_name
);
971 filename
= file
->bs
->filename
;
973 filename
= qdict_get_try_str(options
, "filename");
976 if (drv
->bdrv_needs_filename
&& !filename
) {
977 error_setg(errp
, "The '%s' block driver requires a file name",
983 trace_bdrv_open_common(bs
, filename
?: "", bs
->open_flags
,
986 node_name
= qemu_opt_get(opts
, "node-name");
987 bdrv_assign_node_name(bs
, node_name
, &local_err
);
989 error_propagate(errp
, local_err
);
994 bs
->read_only
= !(bs
->open_flags
& BDRV_O_RDWR
);
996 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
, bs
->read_only
)) {
998 !bs
->read_only
&& bdrv_is_whitelisted(drv
, true)
999 ? "Driver '%s' can only be used for read-only devices"
1000 : "Driver '%s' is not whitelisted",
1006 assert(bs
->copy_on_read
== 0); /* bdrv_new() and bdrv_close() make it so */
1007 if (bs
->open_flags
& BDRV_O_COPY_ON_READ
) {
1008 if (!bs
->read_only
) {
1009 bdrv_enable_copy_on_read(bs
);
1011 error_setg(errp
, "Can't use copy-on-read on read-only device");
1017 if (filename
!= NULL
) {
1018 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
1020 bs
->filename
[0] = '\0';
1022 pstrcpy(bs
->exact_filename
, sizeof(bs
->exact_filename
), bs
->filename
);
1025 bs
->opaque
= g_malloc0(drv
->instance_size
);
1027 /* Open the image, either directly or using a protocol */
1028 open_flags
= bdrv_open_flags(bs
, bs
->open_flags
);
1029 if (drv
->bdrv_file_open
) {
1030 assert(file
== NULL
);
1031 assert(!drv
->bdrv_needs_filename
|| filename
!= NULL
);
1032 ret
= drv
->bdrv_file_open(bs
, options
, open_flags
, &local_err
);
1035 error_setg(errp
, "Can't use '%s' as a block driver for the "
1036 "protocol level", drv
->format_name
);
1041 ret
= drv
->bdrv_open(bs
, options
, open_flags
, &local_err
);
1046 error_propagate(errp
, local_err
);
1047 } else if (bs
->filename
[0]) {
1048 error_setg_errno(errp
, -ret
, "Could not open '%s'", bs
->filename
);
1050 error_setg_errno(errp
, -ret
, "Could not open image");
1055 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
1057 error_setg_errno(errp
, -ret
, "Could not refresh total sector count");
1061 bdrv_refresh_limits(bs
, &local_err
);
1063 error_propagate(errp
, local_err
);
1068 assert(bdrv_opt_mem_align(bs
) != 0);
1069 assert(bdrv_min_mem_align(bs
) != 0);
1070 assert(is_power_of_2(bs
->bl
.request_alignment
));
1072 qemu_opts_del(opts
);
1081 qemu_opts_del(opts
);
1085 static QDict
*parse_json_filename(const char *filename
, Error
**errp
)
1087 QObject
*options_obj
;
1091 ret
= strstart(filename
, "json:", &filename
);
1094 options_obj
= qobject_from_json(filename
);
1096 error_setg(errp
, "Could not parse the JSON options");
1100 if (qobject_type(options_obj
) != QTYPE_QDICT
) {
1101 qobject_decref(options_obj
);
1102 error_setg(errp
, "Invalid JSON object given");
1106 options
= qobject_to_qdict(options_obj
);
1107 qdict_flatten(options
);
1112 static void parse_json_protocol(QDict
*options
, const char **pfilename
,
1115 QDict
*json_options
;
1116 Error
*local_err
= NULL
;
1118 /* Parse json: pseudo-protocol */
1119 if (!*pfilename
|| !g_str_has_prefix(*pfilename
, "json:")) {
1123 json_options
= parse_json_filename(*pfilename
, &local_err
);
1125 error_propagate(errp
, local_err
);
1129 /* Options given in the filename have lower priority than options
1130 * specified directly */
1131 qdict_join(options
, json_options
, false);
1132 QDECREF(json_options
);
1137 * Fills in default options for opening images and converts the legacy
1138 * filename/flags pair to option QDict entries.
1139 * The BDRV_O_PROTOCOL flag in *flags will be set or cleared accordingly if a
1140 * block driver has been specified explicitly.
1142 static int bdrv_fill_options(QDict
**options
, const char *filename
,
1143 int *flags
, Error
**errp
)
1145 const char *drvname
;
1146 bool protocol
= *flags
& BDRV_O_PROTOCOL
;
1147 bool parse_filename
= false;
1148 BlockDriver
*drv
= NULL
;
1149 Error
*local_err
= NULL
;
1151 drvname
= qdict_get_try_str(*options
, "driver");
1153 drv
= bdrv_find_format(drvname
);
1155 error_setg(errp
, "Unknown driver '%s'", drvname
);
1158 /* If the user has explicitly specified the driver, this choice should
1159 * override the BDRV_O_PROTOCOL flag */
1160 protocol
= drv
->bdrv_file_open
;
1164 *flags
|= BDRV_O_PROTOCOL
;
1166 *flags
&= ~BDRV_O_PROTOCOL
;
1169 /* Translate cache options from flags into options */
1170 update_options_from_flags(*options
, *flags
);
1172 /* Fetch the file name from the options QDict if necessary */
1173 if (protocol
&& filename
) {
1174 if (!qdict_haskey(*options
, "filename")) {
1175 qdict_put(*options
, "filename", qstring_from_str(filename
));
1176 parse_filename
= true;
1178 error_setg(errp
, "Can't specify 'file' and 'filename' options at "
1184 /* Find the right block driver */
1185 filename
= qdict_get_try_str(*options
, "filename");
1187 if (!drvname
&& protocol
) {
1189 drv
= bdrv_find_protocol(filename
, parse_filename
, errp
);
1194 drvname
= drv
->format_name
;
1195 qdict_put(*options
, "driver", qstring_from_str(drvname
));
1197 error_setg(errp
, "Must specify either driver or file");
1202 assert(drv
|| !protocol
);
1204 /* Driver-specific filename parsing */
1205 if (drv
&& drv
->bdrv_parse_filename
&& parse_filename
) {
1206 drv
->bdrv_parse_filename(filename
, *options
, &local_err
);
1208 error_propagate(errp
, local_err
);
1212 if (!drv
->bdrv_needs_filename
) {
1213 qdict_del(*options
, "filename");
1220 static void bdrv_replace_child(BdrvChild
*child
, BlockDriverState
*new_bs
)
1222 BlockDriverState
*old_bs
= child
->bs
;
1225 if (old_bs
->quiesce_counter
&& child
->role
->drained_end
) {
1226 child
->role
->drained_end(child
);
1228 QLIST_REMOVE(child
, next_parent
);
1234 QLIST_INSERT_HEAD(&new_bs
->parents
, child
, next_parent
);
1235 if (new_bs
->quiesce_counter
&& child
->role
->drained_begin
) {
1236 child
->role
->drained_begin(child
);
1241 BdrvChild
*bdrv_root_attach_child(BlockDriverState
*child_bs
,
1242 const char *child_name
,
1243 const BdrvChildRole
*child_role
,
1246 BdrvChild
*child
= g_new(BdrvChild
, 1);
1247 *child
= (BdrvChild
) {
1249 .name
= g_strdup(child_name
),
1254 bdrv_replace_child(child
, child_bs
);
1259 BdrvChild
*bdrv_attach_child(BlockDriverState
*parent_bs
,
1260 BlockDriverState
*child_bs
,
1261 const char *child_name
,
1262 const BdrvChildRole
*child_role
)
1264 BdrvChild
*child
= bdrv_root_attach_child(child_bs
, child_name
, child_role
,
1266 QLIST_INSERT_HEAD(&parent_bs
->children
, child
, next
);
1270 static void bdrv_detach_child(BdrvChild
*child
)
1272 if (child
->next
.le_prev
) {
1273 QLIST_REMOVE(child
, next
);
1274 child
->next
.le_prev
= NULL
;
1277 bdrv_replace_child(child
, NULL
);
1279 g_free(child
->name
);
1283 void bdrv_root_unref_child(BdrvChild
*child
)
1285 BlockDriverState
*child_bs
;
1287 child_bs
= child
->bs
;
1288 bdrv_detach_child(child
);
1289 bdrv_unref(child_bs
);
1292 void bdrv_unref_child(BlockDriverState
*parent
, BdrvChild
*child
)
1294 if (child
== NULL
) {
1298 if (child
->bs
->inherits_from
== parent
) {
1299 child
->bs
->inherits_from
= NULL
;
1302 bdrv_root_unref_child(child
);
1306 static void bdrv_parent_cb_change_media(BlockDriverState
*bs
, bool load
)
1309 QLIST_FOREACH(c
, &bs
->parents
, next_parent
) {
1310 if (c
->role
->change_media
) {
1311 c
->role
->change_media(c
, load
);
1316 static void bdrv_parent_cb_resize(BlockDriverState
*bs
)
1319 QLIST_FOREACH(c
, &bs
->parents
, next_parent
) {
1320 if (c
->role
->resize
) {
1327 * Sets the backing file link of a BDS. A new reference is created; callers
1328 * which don't need their own reference any more must call bdrv_unref().
1330 void bdrv_set_backing_hd(BlockDriverState
*bs
, BlockDriverState
*backing_hd
)
1333 bdrv_ref(backing_hd
);
1337 assert(bs
->backing_blocker
);
1338 bdrv_op_unblock_all(bs
->backing
->bs
, bs
->backing_blocker
);
1339 bdrv_unref_child(bs
, bs
->backing
);
1340 } else if (backing_hd
) {
1341 error_setg(&bs
->backing_blocker
,
1342 "node is used as backing hd of '%s'",
1343 bdrv_get_device_or_node_name(bs
));
1347 error_free(bs
->backing_blocker
);
1348 bs
->backing_blocker
= NULL
;
1352 bs
->backing
= bdrv_attach_child(bs
, backing_hd
, "backing", &child_backing
);
1353 bs
->open_flags
&= ~BDRV_O_NO_BACKING
;
1354 pstrcpy(bs
->backing_file
, sizeof(bs
->backing_file
), backing_hd
->filename
);
1355 pstrcpy(bs
->backing_format
, sizeof(bs
->backing_format
),
1356 backing_hd
->drv
? backing_hd
->drv
->format_name
: "");
1358 bdrv_op_block_all(backing_hd
, bs
->backing_blocker
);
1359 /* Otherwise we won't be able to commit due to check in bdrv_commit */
1360 bdrv_op_unblock(backing_hd
, BLOCK_OP_TYPE_COMMIT_TARGET
,
1361 bs
->backing_blocker
);
1363 * We do backup in 3 ways:
1365 * The target bs is new opened, and the source is top BDS
1366 * 2. blockdev backup
1367 * Both the source and the target are top BDSes.
1368 * 3. internal backup(used for block replication)
1369 * Both the source and the target are backing file
1371 * In case 1 and 2, neither the source nor the target is the backing file.
1372 * In case 3, we will block the top BDS, so there is only one block job
1373 * for the top BDS and its backing chain.
1375 bdrv_op_unblock(backing_hd
, BLOCK_OP_TYPE_BACKUP_SOURCE
,
1376 bs
->backing_blocker
);
1377 bdrv_op_unblock(backing_hd
, BLOCK_OP_TYPE_BACKUP_TARGET
,
1378 bs
->backing_blocker
);
1380 bdrv_refresh_limits(bs
, NULL
);
1384 * Opens the backing file for a BlockDriverState if not yet open
1386 * bdref_key specifies the key for the image's BlockdevRef in the options QDict.
1387 * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
1388 * itself, all options starting with "${bdref_key}." are considered part of the
1391 * TODO Can this be unified with bdrv_open_image()?
1393 int bdrv_open_backing_file(BlockDriverState
*bs
, QDict
*parent_options
,
1394 const char *bdref_key
, Error
**errp
)
1396 char *backing_filename
= g_malloc0(PATH_MAX
);
1397 char *bdref_key_dot
;
1398 const char *reference
= NULL
;
1400 BlockDriverState
*backing_hd
;
1402 QDict
*tmp_parent_options
= NULL
;
1403 Error
*local_err
= NULL
;
1405 if (bs
->backing
!= NULL
) {
1409 /* NULL means an empty set of options */
1410 if (parent_options
== NULL
) {
1411 tmp_parent_options
= qdict_new();
1412 parent_options
= tmp_parent_options
;
1415 bs
->open_flags
&= ~BDRV_O_NO_BACKING
;
1417 bdref_key_dot
= g_strdup_printf("%s.", bdref_key
);
1418 qdict_extract_subqdict(parent_options
, &options
, bdref_key_dot
);
1419 g_free(bdref_key_dot
);
1421 reference
= qdict_get_try_str(parent_options
, bdref_key
);
1422 if (reference
|| qdict_haskey(options
, "file.filename")) {
1423 backing_filename
[0] = '\0';
1424 } else if (bs
->backing_file
[0] == '\0' && qdict_size(options
) == 0) {
1428 bdrv_get_full_backing_filename(bs
, backing_filename
, PATH_MAX
,
1432 error_propagate(errp
, local_err
);
1438 if (!bs
->drv
|| !bs
->drv
->supports_backing
) {
1440 error_setg(errp
, "Driver doesn't support backing files");
1445 if (bs
->backing_format
[0] != '\0' && !qdict_haskey(options
, "driver")) {
1446 qdict_put(options
, "driver", qstring_from_str(bs
->backing_format
));
1449 backing_hd
= bdrv_open_inherit(*backing_filename
? backing_filename
: NULL
,
1450 reference
, options
, 0, bs
, &child_backing
,
1453 bs
->open_flags
|= BDRV_O_NO_BACKING
;
1454 error_prepend(errp
, "Could not open backing file: ");
1459 /* Hook up the backing file link; drop our reference, bs owns the
1460 * backing_hd reference now */
1461 bdrv_set_backing_hd(bs
, backing_hd
);
1462 bdrv_unref(backing_hd
);
1464 qdict_del(parent_options
, bdref_key
);
1467 g_free(backing_filename
);
1468 QDECREF(tmp_parent_options
);
1473 * Opens a disk image whose options are given as BlockdevRef in another block
1476 * If allow_none is true, no image will be opened if filename is false and no
1477 * BlockdevRef is given. NULL will be returned, but errp remains unset.
1479 * bdrev_key specifies the key for the image's BlockdevRef in the options QDict.
1480 * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
1481 * itself, all options starting with "${bdref_key}." are considered part of the
1484 * The BlockdevRef will be removed from the options QDict.
1486 BdrvChild
*bdrv_open_child(const char *filename
,
1487 QDict
*options
, const char *bdref_key
,
1488 BlockDriverState
* parent
,
1489 const BdrvChildRole
*child_role
,
1490 bool allow_none
, Error
**errp
)
1492 BdrvChild
*c
= NULL
;
1493 BlockDriverState
*bs
;
1494 QDict
*image_options
;
1495 char *bdref_key_dot
;
1496 const char *reference
;
1498 assert(child_role
!= NULL
);
1500 bdref_key_dot
= g_strdup_printf("%s.", bdref_key
);
1501 qdict_extract_subqdict(options
, &image_options
, bdref_key_dot
);
1502 g_free(bdref_key_dot
);
1504 reference
= qdict_get_try_str(options
, bdref_key
);
1505 if (!filename
&& !reference
&& !qdict_size(image_options
)) {
1507 error_setg(errp
, "A block device must be specified for \"%s\"",
1510 QDECREF(image_options
);
1514 bs
= bdrv_open_inherit(filename
, reference
, image_options
, 0,
1515 parent
, child_role
, errp
);
1520 c
= bdrv_attach_child(parent
, bs
, bdref_key
, child_role
);
1523 qdict_del(options
, bdref_key
);
1527 static BlockDriverState
*bdrv_append_temp_snapshot(BlockDriverState
*bs
,
1529 QDict
*snapshot_options
,
1532 /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
1533 char *tmp_filename
= g_malloc0(PATH_MAX
+ 1);
1535 QemuOpts
*opts
= NULL
;
1536 BlockDriverState
*bs_snapshot
;
1539 /* if snapshot, we create a temporary backing file and open it
1540 instead of opening 'filename' directly */
1542 /* Get the required size from the image */
1543 total_size
= bdrv_getlength(bs
);
1544 if (total_size
< 0) {
1545 error_setg_errno(errp
, -total_size
, "Could not get image size");
1549 /* Create the temporary image */
1550 ret
= get_tmp_filename(tmp_filename
, PATH_MAX
+ 1);
1552 error_setg_errno(errp
, -ret
, "Could not get temporary filename");
1556 opts
= qemu_opts_create(bdrv_qcow2
.create_opts
, NULL
, 0,
1558 qemu_opt_set_number(opts
, BLOCK_OPT_SIZE
, total_size
, &error_abort
);
1559 ret
= bdrv_create(&bdrv_qcow2
, tmp_filename
, opts
, errp
);
1560 qemu_opts_del(opts
);
1562 error_prepend(errp
, "Could not create temporary overlay '%s': ",
1567 /* Prepare options QDict for the temporary file */
1568 qdict_put(snapshot_options
, "file.driver",
1569 qstring_from_str("file"));
1570 qdict_put(snapshot_options
, "file.filename",
1571 qstring_from_str(tmp_filename
));
1572 qdict_put(snapshot_options
, "driver",
1573 qstring_from_str("qcow2"));
1575 bs_snapshot
= bdrv_open(NULL
, NULL
, snapshot_options
, flags
, errp
);
1576 snapshot_options
= NULL
;
1582 /* bdrv_append() consumes a strong reference to bs_snapshot (i.e. it will
1583 * call bdrv_unref() on it), so in order to be able to return one, we have
1584 * to increase bs_snapshot's refcount here */
1585 bdrv_ref(bs_snapshot
);
1586 bdrv_append(bs_snapshot
, bs
);
1588 g_free(tmp_filename
);
1592 QDECREF(snapshot_options
);
1593 g_free(tmp_filename
);
1598 * Opens a disk image (raw, qcow2, vmdk, ...)
1600 * options is a QDict of options to pass to the block drivers, or NULL for an
1601 * empty set of options. The reference to the QDict belongs to the block layer
1602 * after the call (even on failure), so if the caller intends to reuse the
1603 * dictionary, it needs to use QINCREF() before calling bdrv_open.
1605 * If *pbs is NULL, a new BDS will be created with a pointer to it stored there.
1606 * If it is not NULL, the referenced BDS will be reused.
1608 * The reference parameter may be used to specify an existing block device which
1609 * should be opened. If specified, neither options nor a filename may be given,
1610 * nor can an existing BDS be reused (that is, *pbs has to be NULL).
1612 static BlockDriverState
*bdrv_open_inherit(const char *filename
,
1613 const char *reference
,
1614 QDict
*options
, int flags
,
1615 BlockDriverState
*parent
,
1616 const BdrvChildRole
*child_role
,
1620 BdrvChild
*file
= NULL
;
1621 BlockDriverState
*bs
;
1622 BlockDriver
*drv
= NULL
;
1623 const char *drvname
;
1624 const char *backing
;
1625 Error
*local_err
= NULL
;
1626 QDict
*snapshot_options
= NULL
;
1627 int snapshot_flags
= 0;
1629 assert(!child_role
|| !flags
);
1630 assert(!child_role
== !parent
);
1633 bool options_non_empty
= options
? qdict_size(options
) : false;
1636 if (filename
|| options_non_empty
) {
1637 error_setg(errp
, "Cannot reference an existing block device with "
1638 "additional options or a new filename");
1642 bs
= bdrv_lookup_bs(reference
, reference
, errp
);
1653 /* NULL means an empty set of options */
1654 if (options
== NULL
) {
1655 options
= qdict_new();
1658 /* json: syntax counts as explicit options, as if in the QDict */
1659 parse_json_protocol(options
, &filename
, &local_err
);
1664 bs
->explicit_options
= qdict_clone_shallow(options
);
1667 bs
->inherits_from
= parent
;
1668 child_role
->inherit_options(&flags
, options
,
1669 parent
->open_flags
, parent
->options
);
1672 ret
= bdrv_fill_options(&options
, filename
, &flags
, &local_err
);
1677 if (flags
& BDRV_O_RDWR
) {
1678 flags
|= BDRV_O_ALLOW_RDWR
;
1681 if (flags
& BDRV_O_SNAPSHOT
) {
1682 snapshot_options
= qdict_new();
1683 bdrv_temp_snapshot_options(&snapshot_flags
, snapshot_options
,
1685 bdrv_backing_options(&flags
, options
, flags
, options
);
1688 bs
->open_flags
= flags
;
1689 bs
->options
= options
;
1690 options
= qdict_clone_shallow(options
);
1692 /* Find the right image format driver */
1693 drvname
= qdict_get_try_str(options
, "driver");
1695 drv
= bdrv_find_format(drvname
);
1697 error_setg(errp
, "Unknown driver: '%s'", drvname
);
1702 assert(drvname
|| !(flags
& BDRV_O_PROTOCOL
));
1704 backing
= qdict_get_try_str(options
, "backing");
1705 if (backing
&& *backing
== '\0') {
1706 flags
|= BDRV_O_NO_BACKING
;
1707 qdict_del(options
, "backing");
1710 /* Open image file without format layer */
1711 if ((flags
& BDRV_O_PROTOCOL
) == 0) {
1712 file
= bdrv_open_child(filename
, options
, "file", bs
,
1713 &child_file
, true, &local_err
);
1719 /* Image format probing */
1722 ret
= find_image_format(file
, filename
, &drv
, &local_err
);
1727 * This option update would logically belong in bdrv_fill_options(),
1728 * but we first need to open bs->file for the probing to work, while
1729 * opening bs->file already requires the (mostly) final set of options
1730 * so that cache mode etc. can be inherited.
1732 * Adding the driver later is somewhat ugly, but it's not an option
1733 * that would ever be inherited, so it's correct. We just need to make
1734 * sure to update both bs->options (which has the full effective
1735 * options for bs) and options (which has file.* already removed).
1737 qdict_put(bs
->options
, "driver", qstring_from_str(drv
->format_name
));
1738 qdict_put(options
, "driver", qstring_from_str(drv
->format_name
));
1740 error_setg(errp
, "Must specify either driver or file");
1744 /* BDRV_O_PROTOCOL must be set iff a protocol BDS is about to be created */
1745 assert(!!(flags
& BDRV_O_PROTOCOL
) == !!drv
->bdrv_file_open
);
1746 /* file must be NULL if a protocol BDS is about to be created
1747 * (the inverse results in an error message from bdrv_open_common()) */
1748 assert(!(flags
& BDRV_O_PROTOCOL
) || !file
);
1750 /* Open the image */
1751 ret
= bdrv_open_common(bs
, file
, options
, &local_err
);
1756 if (file
&& (bs
->file
!= file
)) {
1757 bdrv_unref_child(bs
, file
);
1761 /* If there is a backing file, use it */
1762 if ((flags
& BDRV_O_NO_BACKING
) == 0) {
1763 ret
= bdrv_open_backing_file(bs
, options
, "backing", &local_err
);
1765 goto close_and_fail
;
1769 bdrv_refresh_filename(bs
);
1771 /* Check if any unknown options were used */
1772 if (options
&& (qdict_size(options
) != 0)) {
1773 const QDictEntry
*entry
= qdict_first(options
);
1774 if (flags
& BDRV_O_PROTOCOL
) {
1775 error_setg(errp
, "Block protocol '%s' doesn't support the option "
1776 "'%s'", drv
->format_name
, entry
->key
);
1779 "Block format '%s' does not support the option '%s'",
1780 drv
->format_name
, entry
->key
);
1783 goto close_and_fail
;
1786 if (!bdrv_key_required(bs
)) {
1787 bdrv_parent_cb_change_media(bs
, true);
1788 } else if (!runstate_check(RUN_STATE_PRELAUNCH
)
1789 && !runstate_check(RUN_STATE_INMIGRATE
)
1790 && !runstate_check(RUN_STATE_PAUSED
)) { /* HACK */
1792 "Guest must be stopped for opening of encrypted image");
1793 goto close_and_fail
;
1798 /* For snapshot=on, create a temporary qcow2 overlay. bs points to the
1799 * temporary snapshot afterwards. */
1800 if (snapshot_flags
) {
1801 BlockDriverState
*snapshot_bs
;
1802 snapshot_bs
= bdrv_append_temp_snapshot(bs
, snapshot_flags
,
1803 snapshot_options
, &local_err
);
1804 snapshot_options
= NULL
;
1806 goto close_and_fail
;
1808 /* We are not going to return bs but the overlay on top of it
1809 * (snapshot_bs); thus, we have to drop the strong reference to bs
1810 * (which we obtained by calling bdrv_new()). bs will not be deleted,
1811 * though, because the overlay still has a reference to it. */
1820 bdrv_unref_child(bs
, file
);
1822 QDECREF(snapshot_options
);
1823 QDECREF(bs
->explicit_options
);
1824 QDECREF(bs
->options
);
1828 error_propagate(errp
, local_err
);
1833 QDECREF(snapshot_options
);
1835 error_propagate(errp
, local_err
);
1839 BlockDriverState
*bdrv_open(const char *filename
, const char *reference
,
1840 QDict
*options
, int flags
, Error
**errp
)
1842 return bdrv_open_inherit(filename
, reference
, options
, flags
, NULL
,
1846 typedef struct BlockReopenQueueEntry
{
1848 BDRVReopenState state
;
1849 QSIMPLEQ_ENTRY(BlockReopenQueueEntry
) entry
;
1850 } BlockReopenQueueEntry
;
1853 * Adds a BlockDriverState to a simple queue for an atomic, transactional
1854 * reopen of multiple devices.
1856 * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT
1857 * already performed, or alternatively may be NULL a new BlockReopenQueue will
1858 * be created and initialized. This newly created BlockReopenQueue should be
1859 * passed back in for subsequent calls that are intended to be of the same
1862 * bs is the BlockDriverState to add to the reopen queue.
1864 * options contains the changed options for the associated bs
1865 * (the BlockReopenQueue takes ownership)
1867 * flags contains the open flags for the associated bs
1869 * returns a pointer to bs_queue, which is either the newly allocated
1870 * bs_queue, or the existing bs_queue being used.
1873 static BlockReopenQueue
*bdrv_reopen_queue_child(BlockReopenQueue
*bs_queue
,
1874 BlockDriverState
*bs
,
1877 const BdrvChildRole
*role
,
1878 QDict
*parent_options
,
1883 BlockReopenQueueEntry
*bs_entry
;
1885 QDict
*old_options
, *explicit_options
;
1887 if (bs_queue
== NULL
) {
1888 bs_queue
= g_new0(BlockReopenQueue
, 1);
1889 QSIMPLEQ_INIT(bs_queue
);
1893 options
= qdict_new();
1897 * Precedence of options:
1898 * 1. Explicitly passed in options (highest)
1899 * 2. Set in flags (only for top level)
1900 * 3. Retained from explicitly set options of bs
1901 * 4. Inherited from parent node
1902 * 5. Retained from effective options of bs
1905 if (!parent_options
) {
1907 * Any setting represented by flags is always updated. If the
1908 * corresponding QDict option is set, it takes precedence. Otherwise
1909 * the flag is translated into a QDict option. The old setting of bs is
1912 update_options_from_flags(options
, flags
);
1915 /* Old explicitly set values (don't overwrite by inherited value) */
1916 old_options
= qdict_clone_shallow(bs
->explicit_options
);
1917 bdrv_join_options(bs
, options
, old_options
);
1918 QDECREF(old_options
);
1920 explicit_options
= qdict_clone_shallow(options
);
1922 /* Inherit from parent node */
1923 if (parent_options
) {
1925 role
->inherit_options(&flags
, options
, parent_flags
, parent_options
);
1928 /* Old values are used for options that aren't set yet */
1929 old_options
= qdict_clone_shallow(bs
->options
);
1930 bdrv_join_options(bs
, options
, old_options
);
1931 QDECREF(old_options
);
1933 /* bdrv_open() masks this flag out */
1934 flags
&= ~BDRV_O_PROTOCOL
;
1936 QLIST_FOREACH(child
, &bs
->children
, next
) {
1937 QDict
*new_child_options
;
1938 char *child_key_dot
;
1940 /* reopen can only change the options of block devices that were
1941 * implicitly created and inherited options. For other (referenced)
1942 * block devices, a syntax like "backing.foo" results in an error. */
1943 if (child
->bs
->inherits_from
!= bs
) {
1947 child_key_dot
= g_strdup_printf("%s.", child
->name
);
1948 qdict_extract_subqdict(options
, &new_child_options
, child_key_dot
);
1949 g_free(child_key_dot
);
1951 bdrv_reopen_queue_child(bs_queue
, child
->bs
, new_child_options
, 0,
1952 child
->role
, options
, flags
);
1955 bs_entry
= g_new0(BlockReopenQueueEntry
, 1);
1956 QSIMPLEQ_INSERT_TAIL(bs_queue
, bs_entry
, entry
);
1958 bs_entry
->state
.bs
= bs
;
1959 bs_entry
->state
.options
= options
;
1960 bs_entry
->state
.explicit_options
= explicit_options
;
1961 bs_entry
->state
.flags
= flags
;
1966 BlockReopenQueue
*bdrv_reopen_queue(BlockReopenQueue
*bs_queue
,
1967 BlockDriverState
*bs
,
1968 QDict
*options
, int flags
)
1970 return bdrv_reopen_queue_child(bs_queue
, bs
, options
, flags
,
1975 * Reopen multiple BlockDriverStates atomically & transactionally.
1977 * The queue passed in (bs_queue) must have been built up previous
1978 * via bdrv_reopen_queue().
1980 * Reopens all BDS specified in the queue, with the appropriate
1981 * flags. All devices are prepared for reopen, and failure of any
1982 * device will cause all device changes to be abandonded, and intermediate
1985 * If all devices prepare successfully, then the changes are committed
1989 int bdrv_reopen_multiple(BlockReopenQueue
*bs_queue
, Error
**errp
)
1992 BlockReopenQueueEntry
*bs_entry
, *next
;
1993 Error
*local_err
= NULL
;
1995 assert(bs_queue
!= NULL
);
1999 QSIMPLEQ_FOREACH(bs_entry
, bs_queue
, entry
) {
2000 if (bdrv_reopen_prepare(&bs_entry
->state
, bs_queue
, &local_err
)) {
2001 error_propagate(errp
, local_err
);
2004 bs_entry
->prepared
= true;
2007 /* If we reach this point, we have success and just need to apply the
2010 QSIMPLEQ_FOREACH(bs_entry
, bs_queue
, entry
) {
2011 bdrv_reopen_commit(&bs_entry
->state
);
2017 QSIMPLEQ_FOREACH_SAFE(bs_entry
, bs_queue
, entry
, next
) {
2018 if (ret
&& bs_entry
->prepared
) {
2019 bdrv_reopen_abort(&bs_entry
->state
);
2021 QDECREF(bs_entry
->state
.explicit_options
);
2023 QDECREF(bs_entry
->state
.options
);
2031 /* Reopen a single BlockDriverState with the specified flags. */
2032 int bdrv_reopen(BlockDriverState
*bs
, int bdrv_flags
, Error
**errp
)
2035 Error
*local_err
= NULL
;
2036 BlockReopenQueue
*queue
= bdrv_reopen_queue(NULL
, bs
, NULL
, bdrv_flags
);
2038 ret
= bdrv_reopen_multiple(queue
, &local_err
);
2039 if (local_err
!= NULL
) {
2040 error_propagate(errp
, local_err
);
2047 * Prepares a BlockDriverState for reopen. All changes are staged in the
2048 * 'opaque' field of the BDRVReopenState, which is used and allocated by
2049 * the block driver layer .bdrv_reopen_prepare()
2051 * bs is the BlockDriverState to reopen
2052 * flags are the new open flags
2053 * queue is the reopen queue
2055 * Returns 0 on success, non-zero on error. On error errp will be set
2058 * On failure, bdrv_reopen_abort() will be called to clean up any data.
2059 * It is the responsibility of the caller to then call the abort() or
2060 * commit() for any other BDS that have been left in a prepare() state
2063 int bdrv_reopen_prepare(BDRVReopenState
*reopen_state
, BlockReopenQueue
*queue
,
2067 Error
*local_err
= NULL
;
2072 assert(reopen_state
!= NULL
);
2073 assert(reopen_state
->bs
->drv
!= NULL
);
2074 drv
= reopen_state
->bs
->drv
;
2076 /* Process generic block layer options */
2077 opts
= qemu_opts_create(&bdrv_runtime_opts
, NULL
, 0, &error_abort
);
2078 qemu_opts_absorb_qdict(opts
, reopen_state
->options
, &local_err
);
2080 error_propagate(errp
, local_err
);
2085 update_flags_from_options(&reopen_state
->flags
, opts
);
2087 /* node-name and driver must be unchanged. Put them back into the QDict, so
2088 * that they are checked at the end of this function. */
2089 value
= qemu_opt_get(opts
, "node-name");
2091 qdict_put(reopen_state
->options
, "node-name", qstring_from_str(value
));
2094 value
= qemu_opt_get(opts
, "driver");
2096 qdict_put(reopen_state
->options
, "driver", qstring_from_str(value
));
2099 /* if we are to stay read-only, do not allow permission change
2101 if (!(reopen_state
->bs
->open_flags
& BDRV_O_ALLOW_RDWR
) &&
2102 reopen_state
->flags
& BDRV_O_RDWR
) {
2103 error_setg(errp
, "Node '%s' is read only",
2104 bdrv_get_device_or_node_name(reopen_state
->bs
));
2109 ret
= bdrv_flush(reopen_state
->bs
);
2111 error_setg_errno(errp
, -ret
, "Error flushing drive");
2115 if (drv
->bdrv_reopen_prepare
) {
2116 ret
= drv
->bdrv_reopen_prepare(reopen_state
, queue
, &local_err
);
2118 if (local_err
!= NULL
) {
2119 error_propagate(errp
, local_err
);
2121 error_setg(errp
, "failed while preparing to reopen image '%s'",
2122 reopen_state
->bs
->filename
);
2127 /* It is currently mandatory to have a bdrv_reopen_prepare()
2128 * handler for each supported drv. */
2129 error_setg(errp
, "Block format '%s' used by node '%s' "
2130 "does not support reopening files", drv
->format_name
,
2131 bdrv_get_device_or_node_name(reopen_state
->bs
));
2136 /* Options that are not handled are only okay if they are unchanged
2137 * compared to the old state. It is expected that some options are only
2138 * used for the initial open, but not reopen (e.g. filename) */
2139 if (qdict_size(reopen_state
->options
)) {
2140 const QDictEntry
*entry
= qdict_first(reopen_state
->options
);
2143 QString
*new_obj
= qobject_to_qstring(entry
->value
);
2144 const char *new = qstring_get_str(new_obj
);
2145 const char *old
= qdict_get_try_str(reopen_state
->bs
->options
,
2148 if (!old
|| strcmp(new, old
)) {
2149 error_setg(errp
, "Cannot change the option '%s'", entry
->key
);
2153 } while ((entry
= qdict_next(reopen_state
->options
, entry
)));
2159 qemu_opts_del(opts
);
2164 * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and
2165 * makes them final by swapping the staging BlockDriverState contents into
2166 * the active BlockDriverState contents.
2168 void bdrv_reopen_commit(BDRVReopenState
*reopen_state
)
2172 assert(reopen_state
!= NULL
);
2173 drv
= reopen_state
->bs
->drv
;
2174 assert(drv
!= NULL
);
2176 /* If there are any driver level actions to take */
2177 if (drv
->bdrv_reopen_commit
) {
2178 drv
->bdrv_reopen_commit(reopen_state
);
2181 /* set BDS specific flags now */
2182 QDECREF(reopen_state
->bs
->explicit_options
);
2184 reopen_state
->bs
->explicit_options
= reopen_state
->explicit_options
;
2185 reopen_state
->bs
->open_flags
= reopen_state
->flags
;
2186 reopen_state
->bs
->read_only
= !(reopen_state
->flags
& BDRV_O_RDWR
);
2188 bdrv_refresh_limits(reopen_state
->bs
, NULL
);
2192 * Abort the reopen, and delete and free the staged changes in
2195 void bdrv_reopen_abort(BDRVReopenState
*reopen_state
)
2199 assert(reopen_state
!= NULL
);
2200 drv
= reopen_state
->bs
->drv
;
2201 assert(drv
!= NULL
);
2203 if (drv
->bdrv_reopen_abort
) {
2204 drv
->bdrv_reopen_abort(reopen_state
);
2207 QDECREF(reopen_state
->explicit_options
);
2211 static void bdrv_close(BlockDriverState
*bs
)
2213 BdrvAioNotifier
*ban
, *ban_next
;
2216 assert(!bs
->refcnt
);
2218 bdrv_drained_begin(bs
); /* complete I/O */
2220 bdrv_drain(bs
); /* in case flush left pending I/O */
2222 bdrv_release_named_dirty_bitmaps(bs
);
2223 assert(QLIST_EMPTY(&bs
->dirty_bitmaps
));
2226 BdrvChild
*child
, *next
;
2228 bs
->drv
->bdrv_close(bs
);
2231 bdrv_set_backing_hd(bs
, NULL
);
2233 if (bs
->file
!= NULL
) {
2234 bdrv_unref_child(bs
, bs
->file
);
2238 QLIST_FOREACH_SAFE(child
, &bs
->children
, next
, next
) {
2239 /* TODO Remove bdrv_unref() from drivers' close function and use
2240 * bdrv_unref_child() here */
2241 if (child
->bs
->inherits_from
== bs
) {
2242 child
->bs
->inherits_from
= NULL
;
2244 bdrv_detach_child(child
);
2249 bs
->copy_on_read
= 0;
2250 bs
->backing_file
[0] = '\0';
2251 bs
->backing_format
[0] = '\0';
2252 bs
->total_sectors
= 0;
2253 bs
->encrypted
= false;
2254 bs
->valid_key
= false;
2256 QDECREF(bs
->options
);
2257 QDECREF(bs
->explicit_options
);
2259 QDECREF(bs
->full_open_options
);
2260 bs
->full_open_options
= NULL
;
2263 QLIST_FOREACH_SAFE(ban
, &bs
->aio_notifiers
, list
, ban_next
) {
2266 QLIST_INIT(&bs
->aio_notifiers
);
2267 bdrv_drained_end(bs
);
2270 void bdrv_close_all(void)
2272 block_job_cancel_sync_all();
2273 nbd_export_close_all();
2275 /* Drop references from requests still in flight, such as canceled block
2276 * jobs whose AIO context has not been polled yet */
2279 blk_remove_all_bs();
2280 blockdev_close_all_bdrv_states();
2282 assert(QTAILQ_EMPTY(&all_bdrv_states
));
2285 static void change_parent_backing_link(BlockDriverState
*from
,
2286 BlockDriverState
*to
)
2288 BdrvChild
*c
, *next
, *to_c
;
2290 QLIST_FOREACH_SAFE(c
, &from
->parents
, next_parent
, next
) {
2291 if (c
->role
== &child_backing
) {
2292 /* @from is generally not allowed to be a backing file, except for
2293 * when @to is the overlay. In that case, @from may not be replaced
2294 * by @to as @to's backing node. */
2295 QLIST_FOREACH(to_c
, &to
->children
, next
) {
2305 assert(c
->role
!= &child_backing
);
2307 bdrv_replace_child(c
, to
);
2313 * Add new bs contents at the top of an image chain while the chain is
2314 * live, while keeping required fields on the top layer.
2316 * This will modify the BlockDriverState fields, and swap contents
2317 * between bs_new and bs_top. Both bs_new and bs_top are modified.
2319 * bs_new must not be attached to a BlockBackend.
2321 * This function does not create any image files.
2323 * bdrv_append() takes ownership of a bs_new reference and unrefs it because
2324 * that's what the callers commonly need. bs_new will be referenced by the old
2325 * parents of bs_top after bdrv_append() returns. If the caller needs to keep a
2326 * reference of its own, it must call bdrv_ref().
2328 void bdrv_append(BlockDriverState
*bs_new
, BlockDriverState
*bs_top
)
2330 assert(!bdrv_requests_pending(bs_top
));
2331 assert(!bdrv_requests_pending(bs_new
));
2335 change_parent_backing_link(bs_top
, bs_new
);
2336 bdrv_set_backing_hd(bs_new
, bs_top
);
2339 /* bs_new is now referenced by its new parents, we don't need the
2340 * additional reference any more. */
2344 void bdrv_replace_in_backing_chain(BlockDriverState
*old
, BlockDriverState
*new)
2346 assert(!bdrv_requests_pending(old
));
2347 assert(!bdrv_requests_pending(new));
2351 change_parent_backing_link(old
, new);
2356 static void bdrv_delete(BlockDriverState
*bs
)
2359 assert(bdrv_op_blocker_is_empty(bs
));
2360 assert(!bs
->refcnt
);
2364 /* remove from list, if necessary */
2365 if (bs
->node_name
[0] != '\0') {
2366 QTAILQ_REMOVE(&graph_bdrv_states
, bs
, node_list
);
2368 QTAILQ_REMOVE(&all_bdrv_states
, bs
, bs_list
);
2374 * Run consistency checks on an image
2376 * Returns 0 if the check could be completed (it doesn't mean that the image is
2377 * free of errors) or -errno when an internal error occurred. The results of the
2378 * check are stored in res.
2380 int bdrv_check(BlockDriverState
*bs
, BdrvCheckResult
*res
, BdrvCheckMode fix
)
2382 if (bs
->drv
== NULL
) {
2385 if (bs
->drv
->bdrv_check
== NULL
) {
2389 memset(res
, 0, sizeof(*res
));
2390 return bs
->drv
->bdrv_check(bs
, res
, fix
);
2396 * -EINVAL - backing format specified, but no file
2397 * -ENOSPC - can't update the backing file because no space is left in the
2399 * -ENOTSUP - format driver doesn't support changing the backing file
2401 int bdrv_change_backing_file(BlockDriverState
*bs
,
2402 const char *backing_file
, const char *backing_fmt
)
2404 BlockDriver
*drv
= bs
->drv
;
2407 /* Backing file format doesn't make sense without a backing file */
2408 if (backing_fmt
&& !backing_file
) {
2412 if (drv
->bdrv_change_backing_file
!= NULL
) {
2413 ret
= drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
2419 pstrcpy(bs
->backing_file
, sizeof(bs
->backing_file
), backing_file
?: "");
2420 pstrcpy(bs
->backing_format
, sizeof(bs
->backing_format
), backing_fmt
?: "");
2426 * Finds the image layer in the chain that has 'bs' as its backing file.
2428 * active is the current topmost image.
2430 * Returns NULL if bs is not found in active's image chain,
2431 * or if active == bs.
2433 * Returns the bottommost base image if bs == NULL.
2435 BlockDriverState
*bdrv_find_overlay(BlockDriverState
*active
,
2436 BlockDriverState
*bs
)
2438 while (active
&& bs
!= backing_bs(active
)) {
2439 active
= backing_bs(active
);
2445 /* Given a BDS, searches for the base layer. */
2446 BlockDriverState
*bdrv_find_base(BlockDriverState
*bs
)
2448 return bdrv_find_overlay(bs
, NULL
);
2452 * Drops images above 'base' up to and including 'top', and sets the image
2453 * above 'top' to have base as its backing file.
2455 * Requires that the overlay to 'top' is opened r/w, so that the backing file
2456 * information in 'bs' can be properly updated.
2458 * E.g., this will convert the following chain:
2459 * bottom <- base <- intermediate <- top <- active
2463 * bottom <- base <- active
2465 * It is allowed for bottom==base, in which case it converts:
2467 * base <- intermediate <- top <- active
2473 * If backing_file_str is non-NULL, it will be used when modifying top's
2474 * overlay image metadata.
2477 * if active == top, that is considered an error
2480 int bdrv_drop_intermediate(BlockDriverState
*active
, BlockDriverState
*top
,
2481 BlockDriverState
*base
, const char *backing_file_str
)
2483 BlockDriverState
*new_top_bs
= NULL
;
2486 if (!top
->drv
|| !base
->drv
) {
2490 new_top_bs
= bdrv_find_overlay(active
, top
);
2492 if (new_top_bs
== NULL
) {
2493 /* we could not find the image above 'top', this is an error */
2497 /* special case of new_top_bs->backing->bs already pointing to base - nothing
2498 * to do, no intermediate images */
2499 if (backing_bs(new_top_bs
) == base
) {
2504 /* Make sure that base is in the backing chain of top */
2505 if (!bdrv_chain_contains(top
, base
)) {
2509 /* success - we can delete the intermediate states, and link top->base */
2510 backing_file_str
= backing_file_str
? backing_file_str
: base
->filename
;
2511 ret
= bdrv_change_backing_file(new_top_bs
, backing_file_str
,
2512 base
->drv
? base
->drv
->format_name
: "");
2516 bdrv_set_backing_hd(new_top_bs
, base
);
2524 * Truncate file to 'offset' bytes (needed only for file protocols)
2526 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
2528 BlockDriver
*drv
= bs
->drv
;
2532 if (!drv
->bdrv_truncate
)
2537 ret
= drv
->bdrv_truncate(bs
, offset
);
2539 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
2540 bdrv_dirty_bitmap_truncate(bs
);
2541 bdrv_parent_cb_resize(bs
);
2548 * Length of a allocated file in bytes. Sparse files are counted by actual
2549 * allocated space. Return < 0 if error or unknown.
2551 int64_t bdrv_get_allocated_file_size(BlockDriverState
*bs
)
2553 BlockDriver
*drv
= bs
->drv
;
2557 if (drv
->bdrv_get_allocated_file_size
) {
2558 return drv
->bdrv_get_allocated_file_size(bs
);
2561 return bdrv_get_allocated_file_size(bs
->file
->bs
);
2567 * Return number of sectors on success, -errno on error.
2569 int64_t bdrv_nb_sectors(BlockDriverState
*bs
)
2571 BlockDriver
*drv
= bs
->drv
;
2576 if (drv
->has_variable_length
) {
2577 int ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
2582 return bs
->total_sectors
;
2586 * Return length in bytes on success, -errno on error.
2587 * The length is always a multiple of BDRV_SECTOR_SIZE.
2589 int64_t bdrv_getlength(BlockDriverState
*bs
)
2591 int64_t ret
= bdrv_nb_sectors(bs
);
2593 ret
= ret
> INT64_MAX
/ BDRV_SECTOR_SIZE
? -EFBIG
: ret
;
2594 return ret
< 0 ? ret
: ret
* BDRV_SECTOR_SIZE
;
2597 /* return 0 as number of sectors if no device present or error */
2598 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
2600 int64_t nb_sectors
= bdrv_nb_sectors(bs
);
2602 *nb_sectors_ptr
= nb_sectors
< 0 ? 0 : nb_sectors
;
2605 bool bdrv_is_read_only(BlockDriverState
*bs
)
2607 return bs
->read_only
;
2610 bool bdrv_is_sg(BlockDriverState
*bs
)
2615 bool bdrv_is_encrypted(BlockDriverState
*bs
)
2617 if (bs
->backing
&& bs
->backing
->bs
->encrypted
) {
2620 return bs
->encrypted
;
2623 bool bdrv_key_required(BlockDriverState
*bs
)
2625 BdrvChild
*backing
= bs
->backing
;
2627 if (backing
&& backing
->bs
->encrypted
&& !backing
->bs
->valid_key
) {
2630 return (bs
->encrypted
&& !bs
->valid_key
);
2633 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
2636 if (bs
->backing
&& bs
->backing
->bs
->encrypted
) {
2637 ret
= bdrv_set_key(bs
->backing
->bs
, key
);
2643 if (!bs
->encrypted
) {
2645 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
2648 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
2650 bs
->valid_key
= false;
2651 } else if (!bs
->valid_key
) {
2652 /* call the change callback now, we skipped it on open */
2653 bs
->valid_key
= true;
2654 bdrv_parent_cb_change_media(bs
, true);
2660 * Provide an encryption key for @bs.
2661 * If @key is non-null:
2662 * If @bs is not encrypted, fail.
2663 * Else if the key is invalid, fail.
2664 * Else set @bs's key to @key, replacing the existing key, if any.
2666 * If @bs is encrypted and still lacks a key, fail.
2668 * On failure, store an error object through @errp if non-null.
2670 void bdrv_add_key(BlockDriverState
*bs
, const char *key
, Error
**errp
)
2673 if (!bdrv_is_encrypted(bs
)) {
2674 error_setg(errp
, "Node '%s' is not encrypted",
2675 bdrv_get_device_or_node_name(bs
));
2676 } else if (bdrv_set_key(bs
, key
) < 0) {
2677 error_setg(errp
, QERR_INVALID_PASSWORD
);
2680 if (bdrv_key_required(bs
)) {
2681 error_set(errp
, ERROR_CLASS_DEVICE_ENCRYPTED
,
2682 "'%s' (%s) is encrypted",
2683 bdrv_get_device_or_node_name(bs
),
2684 bdrv_get_encrypted_filename(bs
));
2689 const char *bdrv_get_format_name(BlockDriverState
*bs
)
2691 return bs
->drv
? bs
->drv
->format_name
: NULL
;
2694 static int qsort_strcmp(const void *a
, const void *b
)
2696 return strcmp(a
, b
);
2699 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
2705 const char **formats
= NULL
;
2707 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
2708 if (drv
->format_name
) {
2711 while (formats
&& i
&& !found
) {
2712 found
= !strcmp(formats
[--i
], drv
->format_name
);
2716 formats
= g_renew(const char *, formats
, count
+ 1);
2717 formats
[count
++] = drv
->format_name
;
2722 qsort(formats
, count
, sizeof(formats
[0]), qsort_strcmp
);
2724 for (i
= 0; i
< count
; i
++) {
2725 it(opaque
, formats
[i
]);
2731 /* This function is to find a node in the bs graph */
2732 BlockDriverState
*bdrv_find_node(const char *node_name
)
2734 BlockDriverState
*bs
;
2738 QTAILQ_FOREACH(bs
, &graph_bdrv_states
, node_list
) {
2739 if (!strcmp(node_name
, bs
->node_name
)) {
2746 /* Put this QMP function here so it can access the static graph_bdrv_states. */
2747 BlockDeviceInfoList
*bdrv_named_nodes_list(Error
**errp
)
2749 BlockDeviceInfoList
*list
, *entry
;
2750 BlockDriverState
*bs
;
2753 QTAILQ_FOREACH(bs
, &graph_bdrv_states
, node_list
) {
2754 BlockDeviceInfo
*info
= bdrv_block_device_info(NULL
, bs
, errp
);
2756 qapi_free_BlockDeviceInfoList(list
);
2759 entry
= g_malloc0(sizeof(*entry
));
2760 entry
->value
= info
;
2768 BlockDriverState
*bdrv_lookup_bs(const char *device
,
2769 const char *node_name
,
2773 BlockDriverState
*bs
;
2776 blk
= blk_by_name(device
);
2781 error_setg(errp
, "Device '%s' has no medium", device
);
2789 bs
= bdrv_find_node(node_name
);
2796 error_setg(errp
, "Cannot find device=%s nor node_name=%s",
2797 device
? device
: "",
2798 node_name
? node_name
: "");
2802 /* If 'base' is in the same chain as 'top', return true. Otherwise,
2803 * return false. If either argument is NULL, return false. */
2804 bool bdrv_chain_contains(BlockDriverState
*top
, BlockDriverState
*base
)
2806 while (top
&& top
!= base
) {
2807 top
= backing_bs(top
);
2813 BlockDriverState
*bdrv_next_node(BlockDriverState
*bs
)
2816 return QTAILQ_FIRST(&graph_bdrv_states
);
2818 return QTAILQ_NEXT(bs
, node_list
);
2821 const char *bdrv_get_node_name(const BlockDriverState
*bs
)
2823 return bs
->node_name
;
2826 const char *bdrv_get_parent_name(const BlockDriverState
*bs
)
2831 /* If multiple parents have a name, just pick the first one. */
2832 QLIST_FOREACH(c
, &bs
->parents
, next_parent
) {
2833 if (c
->role
->get_name
) {
2834 name
= c
->role
->get_name(c
);
2835 if (name
&& *name
) {
2844 /* TODO check what callers really want: bs->node_name or blk_name() */
2845 const char *bdrv_get_device_name(const BlockDriverState
*bs
)
2847 return bdrv_get_parent_name(bs
) ?: "";
2850 /* This can be used to identify nodes that might not have a device
2851 * name associated. Since node and device names live in the same
2852 * namespace, the result is unambiguous. The exception is if both are
2853 * absent, then this returns an empty (non-null) string. */
2854 const char *bdrv_get_device_or_node_name(const BlockDriverState
*bs
)
2856 return bdrv_get_parent_name(bs
) ?: bs
->node_name
;
2859 int bdrv_get_flags(BlockDriverState
*bs
)
2861 return bs
->open_flags
;
2864 int bdrv_has_zero_init_1(BlockDriverState
*bs
)
2869 int bdrv_has_zero_init(BlockDriverState
*bs
)
2873 /* If BS is a copy on write image, it is initialized to
2874 the contents of the base image, which may not be zeroes. */
2878 if (bs
->drv
->bdrv_has_zero_init
) {
2879 return bs
->drv
->bdrv_has_zero_init(bs
);
2886 bool bdrv_unallocated_blocks_are_zero(BlockDriverState
*bs
)
2888 BlockDriverInfo bdi
;
2894 if (bdrv_get_info(bs
, &bdi
) == 0) {
2895 return bdi
.unallocated_blocks_are_zero
;
2901 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState
*bs
)
2903 BlockDriverInfo bdi
;
2905 if (!(bs
->open_flags
& BDRV_O_UNMAP
)) {
2909 if (bdrv_get_info(bs
, &bdi
) == 0) {
2910 return bdi
.can_write_zeroes_with_unmap
;
2916 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
2918 if (bs
->backing
&& bs
->backing
->bs
->encrypted
)
2919 return bs
->backing_file
;
2920 else if (bs
->encrypted
)
2921 return bs
->filename
;
2926 void bdrv_get_backing_filename(BlockDriverState
*bs
,
2927 char *filename
, int filename_size
)
2929 pstrcpy(filename
, filename_size
, bs
->backing_file
);
2932 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
2934 BlockDriver
*drv
= bs
->drv
;
2937 if (!drv
->bdrv_get_info
)
2939 memset(bdi
, 0, sizeof(*bdi
));
2940 return drv
->bdrv_get_info(bs
, bdi
);
2943 ImageInfoSpecific
*bdrv_get_specific_info(BlockDriverState
*bs
)
2945 BlockDriver
*drv
= bs
->drv
;
2946 if (drv
&& drv
->bdrv_get_specific_info
) {
2947 return drv
->bdrv_get_specific_info(bs
);
2952 void bdrv_debug_event(BlockDriverState
*bs
, BlkdebugEvent event
)
2954 if (!bs
|| !bs
->drv
|| !bs
->drv
->bdrv_debug_event
) {
2958 bs
->drv
->bdrv_debug_event(bs
, event
);
2961 int bdrv_debug_breakpoint(BlockDriverState
*bs
, const char *event
,
2964 while (bs
&& bs
->drv
&& !bs
->drv
->bdrv_debug_breakpoint
) {
2965 bs
= bs
->file
? bs
->file
->bs
: NULL
;
2968 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_breakpoint
) {
2969 return bs
->drv
->bdrv_debug_breakpoint(bs
, event
, tag
);
2975 int bdrv_debug_remove_breakpoint(BlockDriverState
*bs
, const char *tag
)
2977 while (bs
&& bs
->drv
&& !bs
->drv
->bdrv_debug_remove_breakpoint
) {
2978 bs
= bs
->file
? bs
->file
->bs
: NULL
;
2981 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_remove_breakpoint
) {
2982 return bs
->drv
->bdrv_debug_remove_breakpoint(bs
, tag
);
2988 int bdrv_debug_resume(BlockDriverState
*bs
, const char *tag
)
2990 while (bs
&& (!bs
->drv
|| !bs
->drv
->bdrv_debug_resume
)) {
2991 bs
= bs
->file
? bs
->file
->bs
: NULL
;
2994 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_resume
) {
2995 return bs
->drv
->bdrv_debug_resume(bs
, tag
);
3001 bool bdrv_debug_is_suspended(BlockDriverState
*bs
, const char *tag
)
3003 while (bs
&& bs
->drv
&& !bs
->drv
->bdrv_debug_is_suspended
) {
3004 bs
= bs
->file
? bs
->file
->bs
: NULL
;
3007 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_is_suspended
) {
3008 return bs
->drv
->bdrv_debug_is_suspended(bs
, tag
);
3014 /* backing_file can either be relative, or absolute, or a protocol. If it is
3015 * relative, it must be relative to the chain. So, passing in bs->filename
3016 * from a BDS as backing_file should not be done, as that may be relative to
3017 * the CWD rather than the chain. */
3018 BlockDriverState
*bdrv_find_backing_image(BlockDriverState
*bs
,
3019 const char *backing_file
)
3021 char *filename_full
= NULL
;
3022 char *backing_file_full
= NULL
;
3023 char *filename_tmp
= NULL
;
3024 int is_protocol
= 0;
3025 BlockDriverState
*curr_bs
= NULL
;
3026 BlockDriverState
*retval
= NULL
;
3028 if (!bs
|| !bs
->drv
|| !backing_file
) {
3032 filename_full
= g_malloc(PATH_MAX
);
3033 backing_file_full
= g_malloc(PATH_MAX
);
3034 filename_tmp
= g_malloc(PATH_MAX
);
3036 is_protocol
= path_has_protocol(backing_file
);
3038 for (curr_bs
= bs
; curr_bs
->backing
; curr_bs
= curr_bs
->backing
->bs
) {
3040 /* If either of the filename paths is actually a protocol, then
3041 * compare unmodified paths; otherwise make paths relative */
3042 if (is_protocol
|| path_has_protocol(curr_bs
->backing_file
)) {
3043 if (strcmp(backing_file
, curr_bs
->backing_file
) == 0) {
3044 retval
= curr_bs
->backing
->bs
;
3048 /* If not an absolute filename path, make it relative to the current
3049 * image's filename path */
3050 path_combine(filename_tmp
, PATH_MAX
, curr_bs
->filename
,
3053 /* We are going to compare absolute pathnames */
3054 if (!realpath(filename_tmp
, filename_full
)) {
3058 /* We need to make sure the backing filename we are comparing against
3059 * is relative to the current image filename (or absolute) */
3060 path_combine(filename_tmp
, PATH_MAX
, curr_bs
->filename
,
3061 curr_bs
->backing_file
);
3063 if (!realpath(filename_tmp
, backing_file_full
)) {
3067 if (strcmp(backing_file_full
, filename_full
) == 0) {
3068 retval
= curr_bs
->backing
->bs
;
3074 g_free(filename_full
);
3075 g_free(backing_file_full
);
3076 g_free(filename_tmp
);
3080 int bdrv_get_backing_file_depth(BlockDriverState
*bs
)
3090 return 1 + bdrv_get_backing_file_depth(bs
->backing
->bs
);
3093 void bdrv_init(void)
3095 module_call_init(MODULE_INIT_BLOCK
);
3098 void bdrv_init_with_whitelist(void)
3100 use_bdrv_whitelist
= 1;
3104 void bdrv_invalidate_cache(BlockDriverState
*bs
, Error
**errp
)
3107 Error
*local_err
= NULL
;
3114 if (!(bs
->open_flags
& BDRV_O_INACTIVE
)) {
3117 bs
->open_flags
&= ~BDRV_O_INACTIVE
;
3119 if (bs
->drv
->bdrv_invalidate_cache
) {
3120 bs
->drv
->bdrv_invalidate_cache(bs
, &local_err
);
3122 bs
->open_flags
|= BDRV_O_INACTIVE
;
3123 error_propagate(errp
, local_err
);
3128 QLIST_FOREACH(child
, &bs
->children
, next
) {
3129 bdrv_invalidate_cache(child
->bs
, &local_err
);
3131 bs
->open_flags
|= BDRV_O_INACTIVE
;
3132 error_propagate(errp
, local_err
);
3137 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
3139 bs
->open_flags
|= BDRV_O_INACTIVE
;
3140 error_setg_errno(errp
, -ret
, "Could not refresh total sector count");
3145 void bdrv_invalidate_cache_all(Error
**errp
)
3147 BlockDriverState
*bs
;
3148 Error
*local_err
= NULL
;
3149 BdrvNextIterator it
;
3151 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3152 AioContext
*aio_context
= bdrv_get_aio_context(bs
);
3154 aio_context_acquire(aio_context
);
3155 bdrv_invalidate_cache(bs
, &local_err
);
3156 aio_context_release(aio_context
);
3158 error_propagate(errp
, local_err
);
3164 static int bdrv_inactivate_recurse(BlockDriverState
*bs
,
3170 if (!setting_flag
&& bs
->drv
->bdrv_inactivate
) {
3171 ret
= bs
->drv
->bdrv_inactivate(bs
);
3177 QLIST_FOREACH(child
, &bs
->children
, next
) {
3178 ret
= bdrv_inactivate_recurse(child
->bs
, setting_flag
);
3185 bs
->open_flags
|= BDRV_O_INACTIVE
;
3190 int bdrv_inactivate_all(void)
3192 BlockDriverState
*bs
= NULL
;
3193 BdrvNextIterator it
;
3197 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3198 aio_context_acquire(bdrv_get_aio_context(bs
));
3201 /* We do two passes of inactivation. The first pass calls to drivers'
3202 * .bdrv_inactivate callbacks recursively so all cache is flushed to disk;
3203 * the second pass sets the BDRV_O_INACTIVE flag so that no further write
3205 for (pass
= 0; pass
< 2; pass
++) {
3206 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3207 ret
= bdrv_inactivate_recurse(bs
, pass
);
3215 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3216 aio_context_release(bdrv_get_aio_context(bs
));
3222 /**************************************************************/
3223 /* removable device support */
3226 * Return TRUE if the media is present
3228 bool bdrv_is_inserted(BlockDriverState
*bs
)
3230 BlockDriver
*drv
= bs
->drv
;
3236 if (drv
->bdrv_is_inserted
) {
3237 return drv
->bdrv_is_inserted(bs
);
3239 QLIST_FOREACH(child
, &bs
->children
, next
) {
3240 if (!bdrv_is_inserted(child
->bs
)) {
3248 * Return whether the media changed since the last call to this
3249 * function, or -ENOTSUP if we don't know. Most drivers don't know.
3251 int bdrv_media_changed(BlockDriverState
*bs
)
3253 BlockDriver
*drv
= bs
->drv
;
3255 if (drv
&& drv
->bdrv_media_changed
) {
3256 return drv
->bdrv_media_changed(bs
);
3262 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
3264 void bdrv_eject(BlockDriverState
*bs
, bool eject_flag
)
3266 BlockDriver
*drv
= bs
->drv
;
3267 const char *device_name
;
3269 if (drv
&& drv
->bdrv_eject
) {
3270 drv
->bdrv_eject(bs
, eject_flag
);
3273 device_name
= bdrv_get_device_name(bs
);
3274 if (device_name
[0] != '\0') {
3275 qapi_event_send_device_tray_moved(device_name
,
3276 eject_flag
, &error_abort
);
3281 * Lock or unlock the media (if it is locked, the user won't be able
3282 * to eject it manually).
3284 void bdrv_lock_medium(BlockDriverState
*bs
, bool locked
)
3286 BlockDriver
*drv
= bs
->drv
;
3288 trace_bdrv_lock_medium(bs
, locked
);
3290 if (drv
&& drv
->bdrv_lock_medium
) {
3291 drv
->bdrv_lock_medium(bs
, locked
);
3295 /* Get a reference to bs */
3296 void bdrv_ref(BlockDriverState
*bs
)
3301 /* Release a previously grabbed reference to bs.
3302 * If after releasing, reference count is zero, the BlockDriverState is
3304 void bdrv_unref(BlockDriverState
*bs
)
3309 assert(bs
->refcnt
> 0);
3310 if (--bs
->refcnt
== 0) {
3315 struct BdrvOpBlocker
{
3317 QLIST_ENTRY(BdrvOpBlocker
) list
;
3320 bool bdrv_op_is_blocked(BlockDriverState
*bs
, BlockOpType op
, Error
**errp
)
3322 BdrvOpBlocker
*blocker
;
3323 assert((int) op
>= 0 && op
< BLOCK_OP_TYPE_MAX
);
3324 if (!QLIST_EMPTY(&bs
->op_blockers
[op
])) {
3325 blocker
= QLIST_FIRST(&bs
->op_blockers
[op
]);
3327 *errp
= error_copy(blocker
->reason
);
3328 error_prepend(errp
, "Node '%s' is busy: ",
3329 bdrv_get_device_or_node_name(bs
));
3336 void bdrv_op_block(BlockDriverState
*bs
, BlockOpType op
, Error
*reason
)
3338 BdrvOpBlocker
*blocker
;
3339 assert((int) op
>= 0 && op
< BLOCK_OP_TYPE_MAX
);
3341 blocker
= g_new0(BdrvOpBlocker
, 1);
3342 blocker
->reason
= reason
;
3343 QLIST_INSERT_HEAD(&bs
->op_blockers
[op
], blocker
, list
);
3346 void bdrv_op_unblock(BlockDriverState
*bs
, BlockOpType op
, Error
*reason
)
3348 BdrvOpBlocker
*blocker
, *next
;
3349 assert((int) op
>= 0 && op
< BLOCK_OP_TYPE_MAX
);
3350 QLIST_FOREACH_SAFE(blocker
, &bs
->op_blockers
[op
], list
, next
) {
3351 if (blocker
->reason
== reason
) {
3352 QLIST_REMOVE(blocker
, list
);
3358 void bdrv_op_block_all(BlockDriverState
*bs
, Error
*reason
)
3361 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
3362 bdrv_op_block(bs
, i
, reason
);
3366 void bdrv_op_unblock_all(BlockDriverState
*bs
, Error
*reason
)
3369 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
3370 bdrv_op_unblock(bs
, i
, reason
);
3374 bool bdrv_op_blocker_is_empty(BlockDriverState
*bs
)
3378 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
3379 if (!QLIST_EMPTY(&bs
->op_blockers
[i
])) {
3386 void bdrv_img_create(const char *filename
, const char *fmt
,
3387 const char *base_filename
, const char *base_fmt
,
3388 char *options
, uint64_t img_size
, int flags
,
3389 Error
**errp
, bool quiet
)
3391 QemuOptsList
*create_opts
= NULL
;
3392 QemuOpts
*opts
= NULL
;
3393 const char *backing_fmt
, *backing_file
;
3395 BlockDriver
*drv
, *proto_drv
;
3396 Error
*local_err
= NULL
;
3399 /* Find driver and parse its options */
3400 drv
= bdrv_find_format(fmt
);
3402 error_setg(errp
, "Unknown file format '%s'", fmt
);
3406 proto_drv
= bdrv_find_protocol(filename
, true, errp
);
3411 if (!drv
->create_opts
) {
3412 error_setg(errp
, "Format driver '%s' does not support image creation",
3417 if (!proto_drv
->create_opts
) {
3418 error_setg(errp
, "Protocol driver '%s' does not support image creation",
3419 proto_drv
->format_name
);
3423 create_opts
= qemu_opts_append(create_opts
, drv
->create_opts
);
3424 create_opts
= qemu_opts_append(create_opts
, proto_drv
->create_opts
);
3426 /* Create parameter list with default values */
3427 opts
= qemu_opts_create(create_opts
, NULL
, 0, &error_abort
);
3428 qemu_opt_set_number(opts
, BLOCK_OPT_SIZE
, img_size
, &error_abort
);
3430 /* Parse -o options */
3432 qemu_opts_do_parse(opts
, options
, NULL
, &local_err
);
3434 error_report_err(local_err
);
3436 error_setg(errp
, "Invalid options for file format '%s'", fmt
);
3441 if (base_filename
) {
3442 qemu_opt_set(opts
, BLOCK_OPT_BACKING_FILE
, base_filename
, &local_err
);
3444 error_setg(errp
, "Backing file not supported for file format '%s'",
3451 qemu_opt_set(opts
, BLOCK_OPT_BACKING_FMT
, base_fmt
, &local_err
);
3453 error_setg(errp
, "Backing file format not supported for file "
3454 "format '%s'", fmt
);
3459 backing_file
= qemu_opt_get(opts
, BLOCK_OPT_BACKING_FILE
);
3461 if (!strcmp(filename
, backing_file
)) {
3462 error_setg(errp
, "Error: Trying to create an image with the "
3463 "same filename as the backing file");
3468 backing_fmt
= qemu_opt_get(opts
, BLOCK_OPT_BACKING_FMT
);
3470 // The size for the image must always be specified, with one exception:
3471 // If we are using a backing file, we can obtain the size from there
3472 size
= qemu_opt_get_size(opts
, BLOCK_OPT_SIZE
, 0);
3475 BlockDriverState
*bs
;
3476 char *full_backing
= g_new0(char, PATH_MAX
);
3479 QDict
*backing_options
= NULL
;
3481 bdrv_get_full_backing_filename_from_filename(filename
, backing_file
,
3482 full_backing
, PATH_MAX
,
3485 g_free(full_backing
);
3489 /* backing files always opened read-only */
3491 back_flags
&= ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
3494 backing_options
= qdict_new();
3495 qdict_put(backing_options
, "driver",
3496 qstring_from_str(backing_fmt
));
3499 bs
= bdrv_open(full_backing
, NULL
, backing_options
, back_flags
,
3501 g_free(full_backing
);
3505 size
= bdrv_getlength(bs
);
3507 error_setg_errno(errp
, -size
, "Could not get size of '%s'",
3513 qemu_opt_set_number(opts
, BLOCK_OPT_SIZE
, size
, &error_abort
);
3517 error_setg(errp
, "Image creation needs a size parameter");
3523 printf("Formatting '%s', fmt=%s ", filename
, fmt
);
3524 qemu_opts_print(opts
, " ");
3528 ret
= bdrv_create(drv
, filename
, opts
, &local_err
);
3530 if (ret
== -EFBIG
) {
3531 /* This is generally a better message than whatever the driver would
3532 * deliver (especially because of the cluster_size_hint), since that
3533 * is most probably not much different from "image too large". */
3534 const char *cluster_size_hint
= "";
3535 if (qemu_opt_get_size(opts
, BLOCK_OPT_CLUSTER_SIZE
, 0)) {
3536 cluster_size_hint
= " (try using a larger cluster size)";
3538 error_setg(errp
, "The image size is too large for file format '%s'"
3539 "%s", fmt
, cluster_size_hint
);
3540 error_free(local_err
);
3545 qemu_opts_del(opts
);
3546 qemu_opts_free(create_opts
);
3547 error_propagate(errp
, local_err
);
3550 AioContext
*bdrv_get_aio_context(BlockDriverState
*bs
)
3552 return bs
->aio_context
;
3555 static void bdrv_do_remove_aio_context_notifier(BdrvAioNotifier
*ban
)
3557 QLIST_REMOVE(ban
, list
);
3561 void bdrv_detach_aio_context(BlockDriverState
*bs
)
3563 BdrvAioNotifier
*baf
, *baf_tmp
;
3570 assert(!bs
->walking_aio_notifiers
);
3571 bs
->walking_aio_notifiers
= true;
3572 QLIST_FOREACH_SAFE(baf
, &bs
->aio_notifiers
, list
, baf_tmp
) {
3574 bdrv_do_remove_aio_context_notifier(baf
);
3576 baf
->detach_aio_context(baf
->opaque
);
3579 /* Never mind iterating again to check for ->deleted. bdrv_close() will
3580 * remove remaining aio notifiers if we aren't called again.
3582 bs
->walking_aio_notifiers
= false;
3584 if (bs
->drv
->bdrv_detach_aio_context
) {
3585 bs
->drv
->bdrv_detach_aio_context(bs
);
3587 QLIST_FOREACH(child
, &bs
->children
, next
) {
3588 bdrv_detach_aio_context(child
->bs
);
3591 bs
->aio_context
= NULL
;
3594 void bdrv_attach_aio_context(BlockDriverState
*bs
,
3595 AioContext
*new_context
)
3597 BdrvAioNotifier
*ban
, *ban_tmp
;
3604 bs
->aio_context
= new_context
;
3606 QLIST_FOREACH(child
, &bs
->children
, next
) {
3607 bdrv_attach_aio_context(child
->bs
, new_context
);
3609 if (bs
->drv
->bdrv_attach_aio_context
) {
3610 bs
->drv
->bdrv_attach_aio_context(bs
, new_context
);
3613 assert(!bs
->walking_aio_notifiers
);
3614 bs
->walking_aio_notifiers
= true;
3615 QLIST_FOREACH_SAFE(ban
, &bs
->aio_notifiers
, list
, ban_tmp
) {
3617 bdrv_do_remove_aio_context_notifier(ban
);
3619 ban
->attached_aio_context(new_context
, ban
->opaque
);
3622 bs
->walking_aio_notifiers
= false;
3625 void bdrv_set_aio_context(BlockDriverState
*bs
, AioContext
*new_context
)
3627 bdrv_drain(bs
); /* ensure there are no in-flight requests */
3629 bdrv_detach_aio_context(bs
);
3631 /* This function executes in the old AioContext so acquire the new one in
3632 * case it runs in a different thread.
3634 aio_context_acquire(new_context
);
3635 bdrv_attach_aio_context(bs
, new_context
);
3636 aio_context_release(new_context
);
3639 void bdrv_add_aio_context_notifier(BlockDriverState
*bs
,
3640 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
),
3641 void (*detach_aio_context
)(void *opaque
), void *opaque
)
3643 BdrvAioNotifier
*ban
= g_new(BdrvAioNotifier
, 1);
3644 *ban
= (BdrvAioNotifier
){
3645 .attached_aio_context
= attached_aio_context
,
3646 .detach_aio_context
= detach_aio_context
,
3650 QLIST_INSERT_HEAD(&bs
->aio_notifiers
, ban
, list
);
3653 void bdrv_remove_aio_context_notifier(BlockDriverState
*bs
,
3654 void (*attached_aio_context
)(AioContext
*,
3656 void (*detach_aio_context
)(void *),
3659 BdrvAioNotifier
*ban
, *ban_next
;
3661 QLIST_FOREACH_SAFE(ban
, &bs
->aio_notifiers
, list
, ban_next
) {
3662 if (ban
->attached_aio_context
== attached_aio_context
&&
3663 ban
->detach_aio_context
== detach_aio_context
&&
3664 ban
->opaque
== opaque
&&
3665 ban
->deleted
== false)
3667 if (bs
->walking_aio_notifiers
) {
3668 ban
->deleted
= true;
3670 bdrv_do_remove_aio_context_notifier(ban
);
3679 int bdrv_amend_options(BlockDriverState
*bs
, QemuOpts
*opts
,
3680 BlockDriverAmendStatusCB
*status_cb
, void *cb_opaque
)
3682 if (!bs
->drv
->bdrv_amend_options
) {
3685 return bs
->drv
->bdrv_amend_options(bs
, opts
, status_cb
, cb_opaque
);
3688 /* This function will be called by the bdrv_recurse_is_first_non_filter method
3689 * of block filter and by bdrv_is_first_non_filter.
3690 * It is used to test if the given bs is the candidate or recurse more in the
3693 bool bdrv_recurse_is_first_non_filter(BlockDriverState
*bs
,
3694 BlockDriverState
*candidate
)
3696 /* return false if basic checks fails */
3697 if (!bs
|| !bs
->drv
) {
3701 /* the code reached a non block filter driver -> check if the bs is
3702 * the same as the candidate. It's the recursion termination condition.
3704 if (!bs
->drv
->is_filter
) {
3705 return bs
== candidate
;
3707 /* Down this path the driver is a block filter driver */
3709 /* If the block filter recursion method is defined use it to recurse down
3712 if (bs
->drv
->bdrv_recurse_is_first_non_filter
) {
3713 return bs
->drv
->bdrv_recurse_is_first_non_filter(bs
, candidate
);
3716 /* the driver is a block filter but don't allow to recurse -> return false
3721 /* This function checks if the candidate is the first non filter bs down it's
3722 * bs chain. Since we don't have pointers to parents it explore all bs chains
3723 * from the top. Some filters can choose not to pass down the recursion.
3725 bool bdrv_is_first_non_filter(BlockDriverState
*candidate
)
3727 BlockDriverState
*bs
;
3728 BdrvNextIterator it
;
3730 /* walk down the bs forest recursively */
3731 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3734 /* try to recurse in this top level bs */
3735 perm
= bdrv_recurse_is_first_non_filter(bs
, candidate
);
3737 /* candidate is the first non filter */
3746 BlockDriverState
*check_to_replace_node(BlockDriverState
*parent_bs
,
3747 const char *node_name
, Error
**errp
)
3749 BlockDriverState
*to_replace_bs
= bdrv_find_node(node_name
);
3750 AioContext
*aio_context
;
3752 if (!to_replace_bs
) {
3753 error_setg(errp
, "Node name '%s' not found", node_name
);
3757 aio_context
= bdrv_get_aio_context(to_replace_bs
);
3758 aio_context_acquire(aio_context
);
3760 if (bdrv_op_is_blocked(to_replace_bs
, BLOCK_OP_TYPE_REPLACE
, errp
)) {
3761 to_replace_bs
= NULL
;
3765 /* We don't want arbitrary node of the BDS chain to be replaced only the top
3766 * most non filter in order to prevent data corruption.
3767 * Another benefit is that this tests exclude backing files which are
3768 * blocked by the backing blockers.
3770 if (!bdrv_recurse_is_first_non_filter(parent_bs
, to_replace_bs
)) {
3771 error_setg(errp
, "Only top most non filter can be replaced");
3772 to_replace_bs
= NULL
;
3777 aio_context_release(aio_context
);
3778 return to_replace_bs
;
3781 static bool append_open_options(QDict
*d
, BlockDriverState
*bs
)
3783 const QDictEntry
*entry
;
3786 bool found_any
= false;
3789 for (entry
= qdict_first(bs
->options
); entry
;
3790 entry
= qdict_next(bs
->options
, entry
))
3792 /* Exclude options for children */
3793 QLIST_FOREACH(child
, &bs
->children
, next
) {
3794 if (strstart(qdict_entry_key(entry
), child
->name
, &p
)
3795 && (!*p
|| *p
== '.'))
3804 /* And exclude all non-driver-specific options */
3805 for (desc
= bdrv_runtime_opts
.desc
; desc
->name
; desc
++) {
3806 if (!strcmp(qdict_entry_key(entry
), desc
->name
)) {
3814 qobject_incref(qdict_entry_value(entry
));
3815 qdict_put_obj(d
, qdict_entry_key(entry
), qdict_entry_value(entry
));
3822 /* Updates the following BDS fields:
3823 * - exact_filename: A filename which may be used for opening a block device
3824 * which (mostly) equals the given BDS (even without any
3825 * other options; so reading and writing must return the same
3826 * results, but caching etc. may be different)
3827 * - full_open_options: Options which, when given when opening a block device
3828 * (without a filename), result in a BDS (mostly)
3829 * equalling the given one
3830 * - filename: If exact_filename is set, it is copied here. Otherwise,
3831 * full_open_options is converted to a JSON object, prefixed with
3832 * "json:" (for use through the JSON pseudo protocol) and put here.
3834 void bdrv_refresh_filename(BlockDriverState
*bs
)
3836 BlockDriver
*drv
= bs
->drv
;
3843 /* This BDS's file name will most probably depend on its file's name, so
3844 * refresh that first */
3846 bdrv_refresh_filename(bs
->file
->bs
);
3849 if (drv
->bdrv_refresh_filename
) {
3850 /* Obsolete information is of no use here, so drop the old file name
3851 * information before refreshing it */
3852 bs
->exact_filename
[0] = '\0';
3853 if (bs
->full_open_options
) {
3854 QDECREF(bs
->full_open_options
);
3855 bs
->full_open_options
= NULL
;
3859 append_open_options(opts
, bs
);
3860 drv
->bdrv_refresh_filename(bs
, opts
);
3862 } else if (bs
->file
) {
3863 /* Try to reconstruct valid information from the underlying file */
3864 bool has_open_options
;
3866 bs
->exact_filename
[0] = '\0';
3867 if (bs
->full_open_options
) {
3868 QDECREF(bs
->full_open_options
);
3869 bs
->full_open_options
= NULL
;
3873 has_open_options
= append_open_options(opts
, bs
);
3875 /* If no specific options have been given for this BDS, the filename of
3876 * the underlying file should suffice for this one as well */
3877 if (bs
->file
->bs
->exact_filename
[0] && !has_open_options
) {
3878 strcpy(bs
->exact_filename
, bs
->file
->bs
->exact_filename
);
3880 /* Reconstructing the full options QDict is simple for most format block
3881 * drivers, as long as the full options are known for the underlying
3882 * file BDS. The full options QDict of that file BDS should somehow
3883 * contain a representation of the filename, therefore the following
3884 * suffices without querying the (exact_)filename of this BDS. */
3885 if (bs
->file
->bs
->full_open_options
) {
3886 qdict_put_obj(opts
, "driver",
3887 QOBJECT(qstring_from_str(drv
->format_name
)));
3888 QINCREF(bs
->file
->bs
->full_open_options
);
3889 qdict_put_obj(opts
, "file",
3890 QOBJECT(bs
->file
->bs
->full_open_options
));
3892 bs
->full_open_options
= opts
;
3896 } else if (!bs
->full_open_options
&& qdict_size(bs
->options
)) {
3897 /* There is no underlying file BDS (at least referenced by BDS.file),
3898 * so the full options QDict should be equal to the options given
3899 * specifically for this block device when it was opened (plus the
3900 * driver specification).
3901 * Because those options don't change, there is no need to update
3902 * full_open_options when it's already set. */
3905 append_open_options(opts
, bs
);
3906 qdict_put_obj(opts
, "driver",
3907 QOBJECT(qstring_from_str(drv
->format_name
)));
3909 if (bs
->exact_filename
[0]) {
3910 /* This may not work for all block protocol drivers (some may
3911 * require this filename to be parsed), but we have to find some
3912 * default solution here, so just include it. If some block driver
3913 * does not support pure options without any filename at all or
3914 * needs some special format of the options QDict, it needs to
3915 * implement the driver-specific bdrv_refresh_filename() function.
3917 qdict_put_obj(opts
, "filename",
3918 QOBJECT(qstring_from_str(bs
->exact_filename
)));
3921 bs
->full_open_options
= opts
;
3924 if (bs
->exact_filename
[0]) {
3925 pstrcpy(bs
->filename
, sizeof(bs
->filename
), bs
->exact_filename
);
3926 } else if (bs
->full_open_options
) {
3927 QString
*json
= qobject_to_json(QOBJECT(bs
->full_open_options
));
3928 snprintf(bs
->filename
, sizeof(bs
->filename
), "json:%s",
3929 qstring_get_str(json
));
3935 * Hot add/remove a BDS's child. So the user can take a child offline when
3936 * it is broken and take a new child online
3938 void bdrv_add_child(BlockDriverState
*parent_bs
, BlockDriverState
*child_bs
,
3942 if (!parent_bs
->drv
|| !parent_bs
->drv
->bdrv_add_child
) {
3943 error_setg(errp
, "The node %s does not support adding a child",
3944 bdrv_get_device_or_node_name(parent_bs
));
3948 if (!QLIST_EMPTY(&child_bs
->parents
)) {
3949 error_setg(errp
, "The node %s already has a parent",
3950 child_bs
->node_name
);
3954 parent_bs
->drv
->bdrv_add_child(parent_bs
, child_bs
, errp
);
3957 void bdrv_del_child(BlockDriverState
*parent_bs
, BdrvChild
*child
, Error
**errp
)
3961 if (!parent_bs
->drv
|| !parent_bs
->drv
->bdrv_del_child
) {
3962 error_setg(errp
, "The node %s does not support removing a child",
3963 bdrv_get_device_or_node_name(parent_bs
));
3967 QLIST_FOREACH(tmp
, &parent_bs
->children
, next
) {
3974 error_setg(errp
, "The node %s does not have a child named %s",
3975 bdrv_get_device_or_node_name(parent_bs
),
3976 bdrv_get_device_or_node_name(child
->bs
));
3980 parent_bs
->drv
->bdrv_del_child(parent_bs
, child
, errp
);