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 "qemu/error-report.h"
29 #include "qemu/module.h"
30 #include "qapi/qmp/qerror.h"
31 #include "qapi/qmp/qbool.h"
32 #include "qapi/qmp/qjson.h"
33 #include "sysemu/block-backend.h"
34 #include "sysemu/sysemu.h"
35 #include "qemu/notify.h"
36 #include "qemu/coroutine.h"
37 #include "block/qapi.h"
38 #include "qmp-commands.h"
39 #include "qemu/timer.h"
40 #include "qapi-event.h"
41 #include "qemu/cutils.h"
45 #include <sys/ioctl.h>
46 #include <sys/queue.h>
56 #define NOT_DONE 0x7fffffff /* used while emulated sync operation in progress */
58 static QTAILQ_HEAD(, BlockDriverState
) graph_bdrv_states
=
59 QTAILQ_HEAD_INITIALIZER(graph_bdrv_states
);
61 static QTAILQ_HEAD(, BlockDriverState
) all_bdrv_states
=
62 QTAILQ_HEAD_INITIALIZER(all_bdrv_states
);
64 static QLIST_HEAD(, BlockDriver
) bdrv_drivers
=
65 QLIST_HEAD_INITIALIZER(bdrv_drivers
);
67 static BlockDriverState
*bdrv_open_inherit(const char *filename
,
68 const char *reference
,
69 QDict
*options
, int flags
,
70 BlockDriverState
*parent
,
71 const BdrvChildRole
*child_role
,
74 /* If non-zero, use only whitelisted block drivers */
75 static int use_bdrv_whitelist
;
78 static int is_windows_drive_prefix(const char *filename
)
80 return (((filename
[0] >= 'a' && filename
[0] <= 'z') ||
81 (filename
[0] >= 'A' && filename
[0] <= 'Z')) &&
85 int is_windows_drive(const char *filename
)
87 if (is_windows_drive_prefix(filename
) &&
90 if (strstart(filename
, "\\\\.\\", NULL
) ||
91 strstart(filename
, "//./", NULL
))
97 size_t bdrv_opt_mem_align(BlockDriverState
*bs
)
99 if (!bs
|| !bs
->drv
) {
100 /* page size or 4k (hdd sector size) should be on the safe side */
101 return MAX(4096, getpagesize());
104 return bs
->bl
.opt_mem_alignment
;
107 size_t bdrv_min_mem_align(BlockDriverState
*bs
)
109 if (!bs
|| !bs
->drv
) {
110 /* page size or 4k (hdd sector size) should be on the safe side */
111 return MAX(4096, getpagesize());
114 return bs
->bl
.min_mem_alignment
;
117 /* check if the path starts with "<protocol>:" */
118 int path_has_protocol(const char *path
)
123 if (is_windows_drive(path
) ||
124 is_windows_drive_prefix(path
)) {
127 p
= path
+ strcspn(path
, ":/\\");
129 p
= path
+ strcspn(path
, ":/");
135 int path_is_absolute(const char *path
)
138 /* specific case for names like: "\\.\d:" */
139 if (is_windows_drive(path
) || is_windows_drive_prefix(path
)) {
142 return (*path
== '/' || *path
== '\\');
144 return (*path
== '/');
148 /* if filename is absolute, just copy it to dest. Otherwise, build a
149 path to it by considering it is relative to base_path. URL are
151 void path_combine(char *dest
, int dest_size
,
152 const char *base_path
,
153 const char *filename
)
160 if (path_is_absolute(filename
)) {
161 pstrcpy(dest
, dest_size
, filename
);
163 p
= strchr(base_path
, ':');
168 p1
= strrchr(base_path
, '/');
172 p2
= strrchr(base_path
, '\\');
184 if (len
> dest_size
- 1)
186 memcpy(dest
, base_path
, len
);
188 pstrcat(dest
, dest_size
, filename
);
192 void bdrv_get_full_backing_filename_from_filename(const char *backed
,
194 char *dest
, size_t sz
,
197 if (backing
[0] == '\0' || path_has_protocol(backing
) ||
198 path_is_absolute(backing
))
200 pstrcpy(dest
, sz
, backing
);
201 } else if (backed
[0] == '\0' || strstart(backed
, "json:", NULL
)) {
202 error_setg(errp
, "Cannot use relative backing file names for '%s'",
205 path_combine(dest
, sz
, backed
, backing
);
209 void bdrv_get_full_backing_filename(BlockDriverState
*bs
, char *dest
, size_t sz
,
212 char *backed
= bs
->exact_filename
[0] ? bs
->exact_filename
: bs
->filename
;
214 bdrv_get_full_backing_filename_from_filename(backed
, bs
->backing_file
,
218 void bdrv_register(BlockDriver
*bdrv
)
220 QLIST_INSERT_HEAD(&bdrv_drivers
, bdrv
, list
);
223 BlockDriverState
*bdrv_new(void)
225 BlockDriverState
*bs
;
228 bs
= g_new0(BlockDriverState
, 1);
229 QLIST_INIT(&bs
->dirty_bitmaps
);
230 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
231 QLIST_INIT(&bs
->op_blockers
[i
]);
233 notifier_with_return_list_init(&bs
->before_write_notifiers
);
235 bs
->aio_context
= qemu_get_aio_context();
237 QTAILQ_INSERT_TAIL(&all_bdrv_states
, bs
, bs_list
);
242 BlockDriver
*bdrv_find_format(const char *format_name
)
245 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
246 if (!strcmp(drv1
->format_name
, format_name
)) {
253 static int bdrv_is_whitelisted(BlockDriver
*drv
, bool read_only
)
255 static const char *whitelist_rw
[] = {
256 CONFIG_BDRV_RW_WHITELIST
258 static const char *whitelist_ro
[] = {
259 CONFIG_BDRV_RO_WHITELIST
263 if (!whitelist_rw
[0] && !whitelist_ro
[0]) {
264 return 1; /* no whitelist, anything goes */
267 for (p
= whitelist_rw
; *p
; p
++) {
268 if (!strcmp(drv
->format_name
, *p
)) {
273 for (p
= whitelist_ro
; *p
; p
++) {
274 if (!strcmp(drv
->format_name
, *p
)) {
282 bool bdrv_uses_whitelist(void)
284 return use_bdrv_whitelist
;
287 typedef struct CreateCo
{
295 static void coroutine_fn
bdrv_create_co_entry(void *opaque
)
297 Error
*local_err
= NULL
;
300 CreateCo
*cco
= opaque
;
303 ret
= cco
->drv
->bdrv_create(cco
->filename
, cco
->opts
, &local_err
);
304 error_propagate(&cco
->err
, local_err
);
308 int bdrv_create(BlockDriver
*drv
, const char* filename
,
309 QemuOpts
*opts
, Error
**errp
)
316 .filename
= g_strdup(filename
),
322 if (!drv
->bdrv_create
) {
323 error_setg(errp
, "Driver '%s' does not support image creation", drv
->format_name
);
328 if (qemu_in_coroutine()) {
329 /* Fast-path if already in coroutine context */
330 bdrv_create_co_entry(&cco
);
332 co
= qemu_coroutine_create(bdrv_create_co_entry
);
333 qemu_coroutine_enter(co
, &cco
);
334 while (cco
.ret
== NOT_DONE
) {
335 aio_poll(qemu_get_aio_context(), true);
342 error_propagate(errp
, cco
.err
);
344 error_setg_errno(errp
, -ret
, "Could not create image");
349 g_free(cco
.filename
);
353 int bdrv_create_file(const char *filename
, QemuOpts
*opts
, Error
**errp
)
356 Error
*local_err
= NULL
;
359 drv
= bdrv_find_protocol(filename
, true, errp
);
364 ret
= bdrv_create(drv
, filename
, opts
, &local_err
);
365 error_propagate(errp
, local_err
);
370 * Try to get @bs's logical and physical block size.
371 * On success, store them in @bsz struct and return 0.
372 * On failure return -errno.
373 * @bs must not be empty.
375 int bdrv_probe_blocksizes(BlockDriverState
*bs
, BlockSizes
*bsz
)
377 BlockDriver
*drv
= bs
->drv
;
379 if (drv
&& drv
->bdrv_probe_blocksizes
) {
380 return drv
->bdrv_probe_blocksizes(bs
, bsz
);
387 * Try to get @bs's geometry (cyls, heads, sectors).
388 * On success, store them in @geo struct and return 0.
389 * On failure return -errno.
390 * @bs must not be empty.
392 int bdrv_probe_geometry(BlockDriverState
*bs
, HDGeometry
*geo
)
394 BlockDriver
*drv
= bs
->drv
;
396 if (drv
&& drv
->bdrv_probe_geometry
) {
397 return drv
->bdrv_probe_geometry(bs
, geo
);
404 * Create a uniquely-named empty temporary file.
405 * Return 0 upon success, otherwise a negative errno value.
407 int get_tmp_filename(char *filename
, int size
)
410 char temp_dir
[MAX_PATH
];
411 /* GetTempFileName requires that its output buffer (4th param)
412 have length MAX_PATH or greater. */
413 assert(size
>= MAX_PATH
);
414 return (GetTempPath(MAX_PATH
, temp_dir
)
415 && GetTempFileName(temp_dir
, "qem", 0, filename
)
416 ? 0 : -GetLastError());
420 tmpdir
= getenv("TMPDIR");
424 if (snprintf(filename
, size
, "%s/vl.XXXXXX", tmpdir
) >= size
) {
427 fd
= mkstemp(filename
);
431 if (close(fd
) != 0) {
440 * Detect host devices. By convention, /dev/cdrom[N] is always
441 * recognized as a host CDROM.
443 static BlockDriver
*find_hdev_driver(const char *filename
)
445 int score_max
= 0, score
;
446 BlockDriver
*drv
= NULL
, *d
;
448 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
449 if (d
->bdrv_probe_device
) {
450 score
= d
->bdrv_probe_device(filename
);
451 if (score
> score_max
) {
461 BlockDriver
*bdrv_find_protocol(const char *filename
,
462 bool allow_protocol_prefix
,
470 /* TODO Drivers without bdrv_file_open must be specified explicitly */
473 * XXX(hch): we really should not let host device detection
474 * override an explicit protocol specification, but moving this
475 * later breaks access to device names with colons in them.
476 * Thanks to the brain-dead persistent naming schemes on udev-
477 * based Linux systems those actually are quite common.
479 drv1
= find_hdev_driver(filename
);
484 if (!path_has_protocol(filename
) || !allow_protocol_prefix
) {
488 p
= strchr(filename
, ':');
491 if (len
> sizeof(protocol
) - 1)
492 len
= sizeof(protocol
) - 1;
493 memcpy(protocol
, filename
, len
);
494 protocol
[len
] = '\0';
495 QLIST_FOREACH(drv1
, &bdrv_drivers
, list
) {
496 if (drv1
->protocol_name
&&
497 !strcmp(drv1
->protocol_name
, protocol
)) {
502 error_setg(errp
, "Unknown protocol '%s'", protocol
);
507 * Guess image format by probing its contents.
508 * This is not a good idea when your image is raw (CVE-2008-2004), but
509 * we do it anyway for backward compatibility.
511 * @buf contains the image's first @buf_size bytes.
512 * @buf_size is the buffer size in bytes (generally BLOCK_PROBE_BUF_SIZE,
513 * but can be smaller if the image file is smaller)
514 * @filename is its filename.
516 * For all block drivers, call the bdrv_probe() method to get its
518 * Return the first block driver with the highest probing score.
520 BlockDriver
*bdrv_probe_all(const uint8_t *buf
, int buf_size
,
521 const char *filename
)
523 int score_max
= 0, score
;
524 BlockDriver
*drv
= NULL
, *d
;
526 QLIST_FOREACH(d
, &bdrv_drivers
, list
) {
528 score
= d
->bdrv_probe(buf
, buf_size
, filename
);
529 if (score
> score_max
) {
539 static int find_image_format(BlockDriverState
*bs
, const char *filename
,
540 BlockDriver
**pdrv
, Error
**errp
)
543 uint8_t buf
[BLOCK_PROBE_BUF_SIZE
];
546 /* Return the raw BlockDriver * to scsi-generic devices or empty drives */
547 if (bdrv_is_sg(bs
) || !bdrv_is_inserted(bs
) || bdrv_getlength(bs
) == 0) {
552 ret
= bdrv_pread(bs
, 0, buf
, sizeof(buf
));
554 error_setg_errno(errp
, -ret
, "Could not read image for determining its "
560 drv
= bdrv_probe_all(buf
, ret
, filename
);
562 error_setg(errp
, "Could not determine image format: No compatible "
571 * Set the current 'total_sectors' value
572 * Return 0 on success, -errno on error.
574 static int refresh_total_sectors(BlockDriverState
*bs
, int64_t hint
)
576 BlockDriver
*drv
= bs
->drv
;
578 /* Do not attempt drv->bdrv_getlength() on scsi-generic devices */
582 /* query actual device if possible, otherwise just trust the hint */
583 if (drv
->bdrv_getlength
) {
584 int64_t length
= drv
->bdrv_getlength(bs
);
588 hint
= DIV_ROUND_UP(length
, BDRV_SECTOR_SIZE
);
591 bs
->total_sectors
= hint
;
596 * Combines a QDict of new block driver @options with any missing options taken
597 * from @old_options, so that leaving out an option defaults to its old value.
599 static void bdrv_join_options(BlockDriverState
*bs
, QDict
*options
,
602 if (bs
->drv
&& bs
->drv
->bdrv_join_options
) {
603 bs
->drv
->bdrv_join_options(options
, old_options
);
605 qdict_join(options
, old_options
, false);
610 * Set open flags for a given discard mode
612 * Return 0 on success, -1 if the discard mode was invalid.
614 int bdrv_parse_discard_flags(const char *mode
, int *flags
)
616 *flags
&= ~BDRV_O_UNMAP
;
618 if (!strcmp(mode
, "off") || !strcmp(mode
, "ignore")) {
620 } else if (!strcmp(mode
, "on") || !strcmp(mode
, "unmap")) {
621 *flags
|= BDRV_O_UNMAP
;
630 * Set open flags for a given cache mode
632 * Return 0 on success, -1 if the cache mode was invalid.
634 int bdrv_parse_cache_mode(const char *mode
, int *flags
, bool *writethrough
)
636 *flags
&= ~BDRV_O_CACHE_MASK
;
638 if (!strcmp(mode
, "off") || !strcmp(mode
, "none")) {
639 *writethrough
= false;
640 *flags
|= BDRV_O_NOCACHE
;
641 } else if (!strcmp(mode
, "directsync")) {
642 *writethrough
= true;
643 *flags
|= BDRV_O_NOCACHE
;
644 } else if (!strcmp(mode
, "writeback")) {
645 *writethrough
= false;
646 } else if (!strcmp(mode
, "unsafe")) {
647 *writethrough
= false;
648 *flags
|= BDRV_O_NO_FLUSH
;
649 } else if (!strcmp(mode
, "writethrough")) {
650 *writethrough
= true;
658 static void bdrv_child_cb_drained_begin(BdrvChild
*child
)
660 BlockDriverState
*bs
= child
->opaque
;
661 bdrv_drained_begin(bs
);
664 static void bdrv_child_cb_drained_end(BdrvChild
*child
)
666 BlockDriverState
*bs
= child
->opaque
;
667 bdrv_drained_end(bs
);
671 * Returns the options and flags that a temporary snapshot should get, based on
672 * the originally requested flags (the originally requested image will have
673 * flags like a backing file)
675 static void bdrv_temp_snapshot_options(int *child_flags
, QDict
*child_options
,
676 int parent_flags
, QDict
*parent_options
)
678 *child_flags
= (parent_flags
& ~BDRV_O_SNAPSHOT
) | BDRV_O_TEMPORARY
;
680 /* For temporary files, unconditional cache=unsafe is fine */
681 qdict_set_default_str(child_options
, BDRV_OPT_CACHE_DIRECT
, "off");
682 qdict_set_default_str(child_options
, BDRV_OPT_CACHE_NO_FLUSH
, "on");
684 /* aio=native doesn't work for cache.direct=off, so disable it for the
685 * temporary snapshot */
686 *child_flags
&= ~BDRV_O_NATIVE_AIO
;
690 * Returns the options and flags that bs->file should get if a protocol driver
691 * is expected, based on the given options and flags for the parent BDS
693 static void bdrv_inherited_options(int *child_flags
, QDict
*child_options
,
694 int parent_flags
, QDict
*parent_options
)
696 int flags
= parent_flags
;
698 /* Enable protocol handling, disable format probing for bs->file */
699 flags
|= BDRV_O_PROTOCOL
;
701 /* If the cache mode isn't explicitly set, inherit direct and no-flush from
703 qdict_copy_default(child_options
, parent_options
, BDRV_OPT_CACHE_DIRECT
);
704 qdict_copy_default(child_options
, parent_options
, BDRV_OPT_CACHE_NO_FLUSH
);
706 /* Our block drivers take care to send flushes and respect unmap policy,
707 * so we can default to enable both on lower layers regardless of the
708 * corresponding parent options. */
709 flags
|= BDRV_O_UNMAP
;
711 /* Clear flags that only apply to the top layer */
712 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
| BDRV_O_COPY_ON_READ
|
715 *child_flags
= flags
;
718 const BdrvChildRole child_file
= {
719 .inherit_options
= bdrv_inherited_options
,
720 .drained_begin
= bdrv_child_cb_drained_begin
,
721 .drained_end
= bdrv_child_cb_drained_end
,
725 * Returns the options and flags that bs->file should get if the use of formats
726 * (and not only protocols) is permitted for it, based on the given options and
727 * flags for the parent BDS
729 static void bdrv_inherited_fmt_options(int *child_flags
, QDict
*child_options
,
730 int parent_flags
, QDict
*parent_options
)
732 child_file
.inherit_options(child_flags
, child_options
,
733 parent_flags
, parent_options
);
735 *child_flags
&= ~(BDRV_O_PROTOCOL
| BDRV_O_NO_IO
);
738 const BdrvChildRole child_format
= {
739 .inherit_options
= bdrv_inherited_fmt_options
,
740 .drained_begin
= bdrv_child_cb_drained_begin
,
741 .drained_end
= bdrv_child_cb_drained_end
,
745 * Returns the options and flags that bs->backing should get, based on the
746 * given options and flags for the parent BDS
748 static void bdrv_backing_options(int *child_flags
, QDict
*child_options
,
749 int parent_flags
, QDict
*parent_options
)
751 int flags
= parent_flags
;
753 /* The cache mode is inherited unmodified for backing files; except WCE,
754 * which is only applied on the top level (BlockBackend) */
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 /* backing files always opened read-only */
759 flags
&= ~(BDRV_O_RDWR
| BDRV_O_COPY_ON_READ
);
761 /* snapshot=on is handled on the top layer */
762 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_TEMPORARY
);
764 *child_flags
= flags
;
767 static const BdrvChildRole child_backing
= {
768 .inherit_options
= bdrv_backing_options
,
769 .drained_begin
= bdrv_child_cb_drained_begin
,
770 .drained_end
= bdrv_child_cb_drained_end
,
773 static int bdrv_open_flags(BlockDriverState
*bs
, int flags
)
775 int open_flags
= flags
;
778 * Clear flags that are internal to the block layer before opening the
781 open_flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
| BDRV_O_PROTOCOL
);
784 * Snapshots should be writable.
786 if (flags
& BDRV_O_TEMPORARY
) {
787 open_flags
|= BDRV_O_RDWR
;
793 static void update_flags_from_options(int *flags
, QemuOpts
*opts
)
795 *flags
&= ~BDRV_O_CACHE_MASK
;
797 assert(qemu_opt_find(opts
, BDRV_OPT_CACHE_NO_FLUSH
));
798 if (qemu_opt_get_bool(opts
, BDRV_OPT_CACHE_NO_FLUSH
, false)) {
799 *flags
|= BDRV_O_NO_FLUSH
;
802 assert(qemu_opt_find(opts
, BDRV_OPT_CACHE_DIRECT
));
803 if (qemu_opt_get_bool(opts
, BDRV_OPT_CACHE_DIRECT
, false)) {
804 *flags
|= BDRV_O_NOCACHE
;
808 static void update_options_from_flags(QDict
*options
, int flags
)
810 if (!qdict_haskey(options
, BDRV_OPT_CACHE_DIRECT
)) {
811 qdict_put(options
, BDRV_OPT_CACHE_DIRECT
,
812 qbool_from_bool(flags
& BDRV_O_NOCACHE
));
814 if (!qdict_haskey(options
, BDRV_OPT_CACHE_NO_FLUSH
)) {
815 qdict_put(options
, BDRV_OPT_CACHE_NO_FLUSH
,
816 qbool_from_bool(flags
& BDRV_O_NO_FLUSH
));
820 static void bdrv_assign_node_name(BlockDriverState
*bs
,
821 const char *node_name
,
824 char *gen_node_name
= NULL
;
827 node_name
= gen_node_name
= id_generate(ID_BLOCK
);
828 } else if (!id_wellformed(node_name
)) {
830 * Check for empty string or invalid characters, but not if it is
831 * generated (generated names use characters not available to the user)
833 error_setg(errp
, "Invalid node name");
837 /* takes care of avoiding namespaces collisions */
838 if (blk_by_name(node_name
)) {
839 error_setg(errp
, "node-name=%s is conflicting with a device id",
844 /* takes care of avoiding duplicates node names */
845 if (bdrv_find_node(node_name
)) {
846 error_setg(errp
, "Duplicate node name");
850 /* copy node name into the bs and insert it into the graph list */
851 pstrcpy(bs
->node_name
, sizeof(bs
->node_name
), node_name
);
852 QTAILQ_INSERT_TAIL(&graph_bdrv_states
, bs
, node_list
);
854 g_free(gen_node_name
);
857 static QemuOptsList bdrv_runtime_opts
= {
858 .name
= "bdrv_common",
859 .head
= QTAILQ_HEAD_INITIALIZER(bdrv_runtime_opts
.head
),
863 .type
= QEMU_OPT_STRING
,
864 .help
= "Node name of the block device node",
868 .type
= QEMU_OPT_STRING
,
869 .help
= "Block driver to use for the node",
872 .name
= BDRV_OPT_CACHE_DIRECT
,
873 .type
= QEMU_OPT_BOOL
,
874 .help
= "Bypass software writeback cache on the host",
877 .name
= BDRV_OPT_CACHE_NO_FLUSH
,
878 .type
= QEMU_OPT_BOOL
,
879 .help
= "Ignore flush requests",
881 { /* end of list */ }
886 * Common part for opening disk images and files
888 * Removes all processed options from *options.
890 static int bdrv_open_common(BlockDriverState
*bs
, BdrvChild
*file
,
891 QDict
*options
, Error
**errp
)
894 const char *filename
;
895 const char *driver_name
= NULL
;
896 const char *node_name
= NULL
;
899 Error
*local_err
= NULL
;
901 assert(bs
->file
== NULL
);
902 assert(options
!= NULL
&& bs
->options
!= options
);
904 opts
= qemu_opts_create(&bdrv_runtime_opts
, NULL
, 0, &error_abort
);
905 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
907 error_propagate(errp
, local_err
);
912 driver_name
= qemu_opt_get(opts
, "driver");
913 drv
= bdrv_find_format(driver_name
);
917 filename
= file
->bs
->filename
;
919 filename
= qdict_get_try_str(options
, "filename");
922 if (drv
->bdrv_needs_filename
&& !filename
) {
923 error_setg(errp
, "The '%s' block driver requires a file name",
929 trace_bdrv_open_common(bs
, filename
?: "", bs
->open_flags
,
932 node_name
= qemu_opt_get(opts
, "node-name");
933 bdrv_assign_node_name(bs
, node_name
, &local_err
);
935 error_propagate(errp
, local_err
);
940 bs
->read_only
= !(bs
->open_flags
& BDRV_O_RDWR
);
942 if (use_bdrv_whitelist
&& !bdrv_is_whitelisted(drv
, bs
->read_only
)) {
944 !bs
->read_only
&& bdrv_is_whitelisted(drv
, true)
945 ? "Driver '%s' can only be used for read-only devices"
946 : "Driver '%s' is not whitelisted",
952 assert(bs
->copy_on_read
== 0); /* bdrv_new() and bdrv_close() make it so */
953 if (bs
->open_flags
& BDRV_O_COPY_ON_READ
) {
954 if (!bs
->read_only
) {
955 bdrv_enable_copy_on_read(bs
);
957 error_setg(errp
, "Can't use copy-on-read on read-only device");
963 if (filename
!= NULL
) {
964 pstrcpy(bs
->filename
, sizeof(bs
->filename
), filename
);
966 bs
->filename
[0] = '\0';
968 pstrcpy(bs
->exact_filename
, sizeof(bs
->exact_filename
), bs
->filename
);
971 bs
->opaque
= g_malloc0(drv
->instance_size
);
973 /* Apply cache mode options */
974 update_flags_from_options(&bs
->open_flags
, opts
);
976 /* Open the image, either directly or using a protocol */
977 open_flags
= bdrv_open_flags(bs
, bs
->open_flags
);
978 if (drv
->bdrv_file_open
) {
979 assert(file
== NULL
);
980 assert(!drv
->bdrv_needs_filename
|| filename
!= NULL
);
981 ret
= drv
->bdrv_file_open(bs
, options
, open_flags
, &local_err
);
984 error_setg(errp
, "Can't use '%s' as a block driver for the "
985 "protocol level", drv
->format_name
);
990 ret
= drv
->bdrv_open(bs
, options
, open_flags
, &local_err
);
995 error_propagate(errp
, local_err
);
996 } else if (bs
->filename
[0]) {
997 error_setg_errno(errp
, -ret
, "Could not open '%s'", bs
->filename
);
999 error_setg_errno(errp
, -ret
, "Could not open image");
1004 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
1006 error_setg_errno(errp
, -ret
, "Could not refresh total sector count");
1010 bdrv_refresh_limits(bs
, &local_err
);
1012 error_propagate(errp
, local_err
);
1017 assert(bdrv_opt_mem_align(bs
) != 0);
1018 assert(bdrv_min_mem_align(bs
) != 0);
1019 assert(is_power_of_2(bs
->bl
.request_alignment
));
1021 qemu_opts_del(opts
);
1030 qemu_opts_del(opts
);
1034 static QDict
*parse_json_filename(const char *filename
, Error
**errp
)
1036 QObject
*options_obj
;
1040 ret
= strstart(filename
, "json:", &filename
);
1043 options_obj
= qobject_from_json(filename
);
1045 error_setg(errp
, "Could not parse the JSON options");
1049 if (qobject_type(options_obj
) != QTYPE_QDICT
) {
1050 qobject_decref(options_obj
);
1051 error_setg(errp
, "Invalid JSON object given");
1055 options
= qobject_to_qdict(options_obj
);
1056 qdict_flatten(options
);
1061 static void parse_json_protocol(QDict
*options
, const char **pfilename
,
1064 QDict
*json_options
;
1065 Error
*local_err
= NULL
;
1067 /* Parse json: pseudo-protocol */
1068 if (!*pfilename
|| !g_str_has_prefix(*pfilename
, "json:")) {
1072 json_options
= parse_json_filename(*pfilename
, &local_err
);
1074 error_propagate(errp
, local_err
);
1078 /* Options given in the filename have lower priority than options
1079 * specified directly */
1080 qdict_join(options
, json_options
, false);
1081 QDECREF(json_options
);
1086 * Fills in default options for opening images and converts the legacy
1087 * filename/flags pair to option QDict entries.
1088 * The BDRV_O_PROTOCOL flag in *flags will be set or cleared accordingly if a
1089 * block driver has been specified explicitly.
1091 static int bdrv_fill_options(QDict
**options
, const char *filename
,
1092 int *flags
, Error
**errp
)
1094 const char *drvname
;
1095 bool protocol
= *flags
& BDRV_O_PROTOCOL
;
1096 bool parse_filename
= false;
1097 BlockDriver
*drv
= NULL
;
1098 Error
*local_err
= NULL
;
1100 drvname
= qdict_get_try_str(*options
, "driver");
1102 drv
= bdrv_find_format(drvname
);
1104 error_setg(errp
, "Unknown driver '%s'", drvname
);
1107 /* If the user has explicitly specified the driver, this choice should
1108 * override the BDRV_O_PROTOCOL flag */
1109 protocol
= drv
->bdrv_file_open
;
1113 *flags
|= BDRV_O_PROTOCOL
;
1115 *flags
&= ~BDRV_O_PROTOCOL
;
1118 /* Translate cache options from flags into options */
1119 update_options_from_flags(*options
, *flags
);
1121 /* Fetch the file name from the options QDict if necessary */
1122 if (protocol
&& filename
) {
1123 if (!qdict_haskey(*options
, "filename")) {
1124 qdict_put(*options
, "filename", qstring_from_str(filename
));
1125 parse_filename
= true;
1127 error_setg(errp
, "Can't specify 'file' and 'filename' options at "
1133 /* Find the right block driver */
1134 filename
= qdict_get_try_str(*options
, "filename");
1136 if (!drvname
&& protocol
) {
1138 drv
= bdrv_find_protocol(filename
, parse_filename
, errp
);
1143 drvname
= drv
->format_name
;
1144 qdict_put(*options
, "driver", qstring_from_str(drvname
));
1146 error_setg(errp
, "Must specify either driver or file");
1151 assert(drv
|| !protocol
);
1153 /* Driver-specific filename parsing */
1154 if (drv
&& drv
->bdrv_parse_filename
&& parse_filename
) {
1155 drv
->bdrv_parse_filename(filename
, *options
, &local_err
);
1157 error_propagate(errp
, local_err
);
1161 if (!drv
->bdrv_needs_filename
) {
1162 qdict_del(*options
, "filename");
1169 static void bdrv_replace_child(BdrvChild
*child
, BlockDriverState
*new_bs
)
1171 BlockDriverState
*old_bs
= child
->bs
;
1174 if (old_bs
->quiesce_counter
&& child
->role
->drained_end
) {
1175 child
->role
->drained_end(child
);
1177 QLIST_REMOVE(child
, next_parent
);
1183 QLIST_INSERT_HEAD(&new_bs
->parents
, child
, next_parent
);
1184 if (new_bs
->quiesce_counter
&& child
->role
->drained_begin
) {
1185 child
->role
->drained_begin(child
);
1190 BdrvChild
*bdrv_root_attach_child(BlockDriverState
*child_bs
,
1191 const char *child_name
,
1192 const BdrvChildRole
*child_role
,
1195 BdrvChild
*child
= g_new(BdrvChild
, 1);
1196 *child
= (BdrvChild
) {
1198 .name
= g_strdup(child_name
),
1203 bdrv_replace_child(child
, child_bs
);
1208 BdrvChild
*bdrv_attach_child(BlockDriverState
*parent_bs
,
1209 BlockDriverState
*child_bs
,
1210 const char *child_name
,
1211 const BdrvChildRole
*child_role
)
1213 BdrvChild
*child
= bdrv_root_attach_child(child_bs
, child_name
, child_role
,
1215 QLIST_INSERT_HEAD(&parent_bs
->children
, child
, next
);
1219 static void bdrv_detach_child(BdrvChild
*child
)
1221 if (child
->next
.le_prev
) {
1222 QLIST_REMOVE(child
, next
);
1223 child
->next
.le_prev
= NULL
;
1226 bdrv_replace_child(child
, NULL
);
1228 g_free(child
->name
);
1232 void bdrv_root_unref_child(BdrvChild
*child
)
1234 BlockDriverState
*child_bs
;
1236 child_bs
= child
->bs
;
1237 bdrv_detach_child(child
);
1238 bdrv_unref(child_bs
);
1241 void bdrv_unref_child(BlockDriverState
*parent
, BdrvChild
*child
)
1243 if (child
== NULL
) {
1247 if (child
->bs
->inherits_from
== parent
) {
1248 child
->bs
->inherits_from
= NULL
;
1251 bdrv_root_unref_child(child
);
1255 static void bdrv_parent_cb_change_media(BlockDriverState
*bs
, bool load
)
1258 QLIST_FOREACH(c
, &bs
->parents
, next_parent
) {
1259 if (c
->role
->change_media
) {
1260 c
->role
->change_media(c
, load
);
1265 static void bdrv_parent_cb_resize(BlockDriverState
*bs
)
1268 QLIST_FOREACH(c
, &bs
->parents
, next_parent
) {
1269 if (c
->role
->resize
) {
1276 * Sets the backing file link of a BDS. A new reference is created; callers
1277 * which don't need their own reference any more must call bdrv_unref().
1279 void bdrv_set_backing_hd(BlockDriverState
*bs
, BlockDriverState
*backing_hd
)
1282 bdrv_ref(backing_hd
);
1286 assert(bs
->backing_blocker
);
1287 bdrv_op_unblock_all(bs
->backing
->bs
, bs
->backing_blocker
);
1288 bdrv_unref_child(bs
, bs
->backing
);
1289 } else if (backing_hd
) {
1290 error_setg(&bs
->backing_blocker
,
1291 "node is used as backing hd of '%s'",
1292 bdrv_get_device_or_node_name(bs
));
1296 error_free(bs
->backing_blocker
);
1297 bs
->backing_blocker
= NULL
;
1301 bs
->backing
= bdrv_attach_child(bs
, backing_hd
, "backing", &child_backing
);
1302 bs
->open_flags
&= ~BDRV_O_NO_BACKING
;
1303 pstrcpy(bs
->backing_file
, sizeof(bs
->backing_file
), backing_hd
->filename
);
1304 pstrcpy(bs
->backing_format
, sizeof(bs
->backing_format
),
1305 backing_hd
->drv
? backing_hd
->drv
->format_name
: "");
1307 bdrv_op_block_all(backing_hd
, bs
->backing_blocker
);
1308 /* Otherwise we won't be able to commit due to check in bdrv_commit */
1309 bdrv_op_unblock(backing_hd
, BLOCK_OP_TYPE_COMMIT_TARGET
,
1310 bs
->backing_blocker
);
1312 bdrv_refresh_limits(bs
, NULL
);
1316 * Opens the backing file for a BlockDriverState if not yet open
1318 * bdref_key specifies the key for the image's BlockdevRef in the options QDict.
1319 * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
1320 * itself, all options starting with "${bdref_key}." are considered part of the
1323 * TODO Can this be unified with bdrv_open_image()?
1325 int bdrv_open_backing_file(BlockDriverState
*bs
, QDict
*parent_options
,
1326 const char *bdref_key
, Error
**errp
)
1328 char *backing_filename
= g_malloc0(PATH_MAX
);
1329 char *bdref_key_dot
;
1330 const char *reference
= NULL
;
1332 BlockDriverState
*backing_hd
;
1334 QDict
*tmp_parent_options
= NULL
;
1335 Error
*local_err
= NULL
;
1337 if (bs
->backing
!= NULL
) {
1341 /* NULL means an empty set of options */
1342 if (parent_options
== NULL
) {
1343 tmp_parent_options
= qdict_new();
1344 parent_options
= tmp_parent_options
;
1347 bs
->open_flags
&= ~BDRV_O_NO_BACKING
;
1349 bdref_key_dot
= g_strdup_printf("%s.", bdref_key
);
1350 qdict_extract_subqdict(parent_options
, &options
, bdref_key_dot
);
1351 g_free(bdref_key_dot
);
1353 reference
= qdict_get_try_str(parent_options
, bdref_key
);
1354 if (reference
|| qdict_haskey(options
, "file.filename")) {
1355 backing_filename
[0] = '\0';
1356 } else if (bs
->backing_file
[0] == '\0' && qdict_size(options
) == 0) {
1360 bdrv_get_full_backing_filename(bs
, backing_filename
, PATH_MAX
,
1364 error_propagate(errp
, local_err
);
1370 if (!bs
->drv
|| !bs
->drv
->supports_backing
) {
1372 error_setg(errp
, "Driver doesn't support backing files");
1377 if (bs
->backing_format
[0] != '\0' && !qdict_haskey(options
, "driver")) {
1378 qdict_put(options
, "driver", qstring_from_str(bs
->backing_format
));
1381 backing_hd
= bdrv_open_inherit(*backing_filename
? backing_filename
: NULL
,
1382 reference
, options
, 0, bs
, &child_backing
,
1385 bs
->open_flags
|= BDRV_O_NO_BACKING
;
1386 error_prepend(errp
, "Could not open backing file: ");
1391 /* Hook up the backing file link; drop our reference, bs owns the
1392 * backing_hd reference now */
1393 bdrv_set_backing_hd(bs
, backing_hd
);
1394 bdrv_unref(backing_hd
);
1396 qdict_del(parent_options
, bdref_key
);
1399 g_free(backing_filename
);
1400 QDECREF(tmp_parent_options
);
1405 * Opens a disk image whose options are given as BlockdevRef in another block
1408 * If allow_none is true, no image will be opened if filename is false and no
1409 * BlockdevRef is given. NULL will be returned, but errp remains unset.
1411 * bdrev_key specifies the key for the image's BlockdevRef in the options QDict.
1412 * That QDict has to be flattened; therefore, if the BlockdevRef is a QDict
1413 * itself, all options starting with "${bdref_key}." are considered part of the
1416 * The BlockdevRef will be removed from the options QDict.
1418 BdrvChild
*bdrv_open_child(const char *filename
,
1419 QDict
*options
, const char *bdref_key
,
1420 BlockDriverState
* parent
,
1421 const BdrvChildRole
*child_role
,
1422 bool allow_none
, Error
**errp
)
1424 BdrvChild
*c
= NULL
;
1425 BlockDriverState
*bs
;
1426 QDict
*image_options
;
1427 char *bdref_key_dot
;
1428 const char *reference
;
1430 assert(child_role
!= NULL
);
1432 bdref_key_dot
= g_strdup_printf("%s.", bdref_key
);
1433 qdict_extract_subqdict(options
, &image_options
, bdref_key_dot
);
1434 g_free(bdref_key_dot
);
1436 reference
= qdict_get_try_str(options
, bdref_key
);
1437 if (!filename
&& !reference
&& !qdict_size(image_options
)) {
1439 error_setg(errp
, "A block device must be specified for \"%s\"",
1442 QDECREF(image_options
);
1446 bs
= bdrv_open_inherit(filename
, reference
, image_options
, 0,
1447 parent
, child_role
, errp
);
1452 c
= bdrv_attach_child(parent
, bs
, bdref_key
, child_role
);
1455 qdict_del(options
, bdref_key
);
1459 static BlockDriverState
*bdrv_append_temp_snapshot(BlockDriverState
*bs
,
1461 QDict
*snapshot_options
,
1464 /* TODO: extra byte is a hack to ensure MAX_PATH space on Windows. */
1465 char *tmp_filename
= g_malloc0(PATH_MAX
+ 1);
1467 QemuOpts
*opts
= NULL
;
1468 BlockDriverState
*bs_snapshot
;
1471 /* if snapshot, we create a temporary backing file and open it
1472 instead of opening 'filename' directly */
1474 /* Get the required size from the image */
1475 total_size
= bdrv_getlength(bs
);
1476 if (total_size
< 0) {
1477 error_setg_errno(errp
, -total_size
, "Could not get image size");
1481 /* Create the temporary image */
1482 ret
= get_tmp_filename(tmp_filename
, PATH_MAX
+ 1);
1484 error_setg_errno(errp
, -ret
, "Could not get temporary filename");
1488 opts
= qemu_opts_create(bdrv_qcow2
.create_opts
, NULL
, 0,
1490 qemu_opt_set_number(opts
, BLOCK_OPT_SIZE
, total_size
, &error_abort
);
1491 ret
= bdrv_create(&bdrv_qcow2
, tmp_filename
, opts
, errp
);
1492 qemu_opts_del(opts
);
1494 error_prepend(errp
, "Could not create temporary overlay '%s': ",
1499 /* Prepare options QDict for the temporary file */
1500 qdict_put(snapshot_options
, "file.driver",
1501 qstring_from_str("file"));
1502 qdict_put(snapshot_options
, "file.filename",
1503 qstring_from_str(tmp_filename
));
1504 qdict_put(snapshot_options
, "driver",
1505 qstring_from_str("qcow2"));
1507 bs_snapshot
= bdrv_open(NULL
, NULL
, snapshot_options
, flags
, errp
);
1508 snapshot_options
= NULL
;
1514 /* bdrv_append() consumes a strong reference to bs_snapshot (i.e. it will
1515 * call bdrv_unref() on it), so in order to be able to return one, we have
1516 * to increase bs_snapshot's refcount here */
1517 bdrv_ref(bs_snapshot
);
1518 bdrv_append(bs_snapshot
, bs
);
1520 g_free(tmp_filename
);
1524 QDECREF(snapshot_options
);
1525 g_free(tmp_filename
);
1530 * Opens a disk image (raw, qcow2, vmdk, ...)
1532 * options is a QDict of options to pass to the block drivers, or NULL for an
1533 * empty set of options. The reference to the QDict belongs to the block layer
1534 * after the call (even on failure), so if the caller intends to reuse the
1535 * dictionary, it needs to use QINCREF() before calling bdrv_open.
1537 * If *pbs is NULL, a new BDS will be created with a pointer to it stored there.
1538 * If it is not NULL, the referenced BDS will be reused.
1540 * The reference parameter may be used to specify an existing block device which
1541 * should be opened. If specified, neither options nor a filename may be given,
1542 * nor can an existing BDS be reused (that is, *pbs has to be NULL).
1544 static BlockDriverState
*bdrv_open_inherit(const char *filename
,
1545 const char *reference
,
1546 QDict
*options
, int flags
,
1547 BlockDriverState
*parent
,
1548 const BdrvChildRole
*child_role
,
1552 BdrvChild
*file
= NULL
;
1553 BlockDriverState
*bs
;
1554 BlockDriver
*drv
= NULL
;
1555 const char *drvname
;
1556 const char *backing
;
1557 Error
*local_err
= NULL
;
1558 QDict
*snapshot_options
= NULL
;
1559 int snapshot_flags
= 0;
1561 assert(!child_role
|| !flags
);
1562 assert(!child_role
== !parent
);
1565 bool options_non_empty
= options
? qdict_size(options
) : false;
1568 if (filename
|| options_non_empty
) {
1569 error_setg(errp
, "Cannot reference an existing block device with "
1570 "additional options or a new filename");
1574 bs
= bdrv_lookup_bs(reference
, reference
, errp
);
1585 /* NULL means an empty set of options */
1586 if (options
== NULL
) {
1587 options
= qdict_new();
1590 /* json: syntax counts as explicit options, as if in the QDict */
1591 parse_json_protocol(options
, &filename
, &local_err
);
1596 bs
->explicit_options
= qdict_clone_shallow(options
);
1599 bs
->inherits_from
= parent
;
1600 child_role
->inherit_options(&flags
, options
,
1601 parent
->open_flags
, parent
->options
);
1604 ret
= bdrv_fill_options(&options
, filename
, &flags
, &local_err
);
1609 bs
->open_flags
= flags
;
1610 bs
->options
= options
;
1611 options
= qdict_clone_shallow(options
);
1613 /* Find the right image format driver */
1614 drvname
= qdict_get_try_str(options
, "driver");
1616 drv
= bdrv_find_format(drvname
);
1618 error_setg(errp
, "Unknown driver: '%s'", drvname
);
1623 assert(drvname
|| !(flags
& BDRV_O_PROTOCOL
));
1625 backing
= qdict_get_try_str(options
, "backing");
1626 if (backing
&& *backing
== '\0') {
1627 flags
|= BDRV_O_NO_BACKING
;
1628 qdict_del(options
, "backing");
1631 /* Open image file without format layer */
1632 if ((flags
& BDRV_O_PROTOCOL
) == 0) {
1633 if (flags
& BDRV_O_RDWR
) {
1634 flags
|= BDRV_O_ALLOW_RDWR
;
1636 if (flags
& BDRV_O_SNAPSHOT
) {
1637 snapshot_options
= qdict_new();
1638 bdrv_temp_snapshot_options(&snapshot_flags
, snapshot_options
,
1640 bdrv_backing_options(&flags
, options
, flags
, options
);
1643 bs
->open_flags
= flags
;
1645 file
= bdrv_open_child(filename
, options
, "file", bs
,
1646 &child_file
, true, &local_err
);
1652 /* Image format probing */
1655 ret
= find_image_format(file
->bs
, filename
, &drv
, &local_err
);
1660 * This option update would logically belong in bdrv_fill_options(),
1661 * but we first need to open bs->file for the probing to work, while
1662 * opening bs->file already requires the (mostly) final set of options
1663 * so that cache mode etc. can be inherited.
1665 * Adding the driver later is somewhat ugly, but it's not an option
1666 * that would ever be inherited, so it's correct. We just need to make
1667 * sure to update both bs->options (which has the full effective
1668 * options for bs) and options (which has file.* already removed).
1670 qdict_put(bs
->options
, "driver", qstring_from_str(drv
->format_name
));
1671 qdict_put(options
, "driver", qstring_from_str(drv
->format_name
));
1673 error_setg(errp
, "Must specify either driver or file");
1677 /* BDRV_O_PROTOCOL must be set iff a protocol BDS is about to be created */
1678 assert(!!(flags
& BDRV_O_PROTOCOL
) == !!drv
->bdrv_file_open
);
1679 /* file must be NULL if a protocol BDS is about to be created
1680 * (the inverse results in an error message from bdrv_open_common()) */
1681 assert(!(flags
& BDRV_O_PROTOCOL
) || !file
);
1683 /* Open the image */
1684 ret
= bdrv_open_common(bs
, file
, options
, &local_err
);
1689 if (file
&& (bs
->file
!= file
)) {
1690 bdrv_unref_child(bs
, file
);
1694 /* If there is a backing file, use it */
1695 if ((flags
& BDRV_O_NO_BACKING
) == 0) {
1696 ret
= bdrv_open_backing_file(bs
, options
, "backing", &local_err
);
1698 goto close_and_fail
;
1702 bdrv_refresh_filename(bs
);
1704 /* Check if any unknown options were used */
1705 if (options
&& (qdict_size(options
) != 0)) {
1706 const QDictEntry
*entry
= qdict_first(options
);
1707 if (flags
& BDRV_O_PROTOCOL
) {
1708 error_setg(errp
, "Block protocol '%s' doesn't support the option "
1709 "'%s'", drv
->format_name
, entry
->key
);
1712 "Block format '%s' does not support the option '%s'",
1713 drv
->format_name
, entry
->key
);
1716 goto close_and_fail
;
1719 if (!bdrv_key_required(bs
)) {
1720 bdrv_parent_cb_change_media(bs
, true);
1721 } else if (!runstate_check(RUN_STATE_PRELAUNCH
)
1722 && !runstate_check(RUN_STATE_INMIGRATE
)
1723 && !runstate_check(RUN_STATE_PAUSED
)) { /* HACK */
1725 "Guest must be stopped for opening of encrypted image");
1726 goto close_and_fail
;
1731 /* For snapshot=on, create a temporary qcow2 overlay. bs points to the
1732 * temporary snapshot afterwards. */
1733 if (snapshot_flags
) {
1734 BlockDriverState
*snapshot_bs
;
1735 snapshot_bs
= bdrv_append_temp_snapshot(bs
, snapshot_flags
,
1736 snapshot_options
, &local_err
);
1737 snapshot_options
= NULL
;
1739 goto close_and_fail
;
1741 /* We are not going to return bs but the overlay on top of it
1742 * (snapshot_bs); thus, we have to drop the strong reference to bs
1743 * (which we obtained by calling bdrv_new()). bs will not be deleted,
1744 * though, because the overlay still has a reference to it. */
1753 bdrv_unref_child(bs
, file
);
1755 QDECREF(snapshot_options
);
1756 QDECREF(bs
->explicit_options
);
1757 QDECREF(bs
->options
);
1761 error_propagate(errp
, local_err
);
1766 QDECREF(snapshot_options
);
1768 error_propagate(errp
, local_err
);
1772 BlockDriverState
*bdrv_open(const char *filename
, const char *reference
,
1773 QDict
*options
, int flags
, Error
**errp
)
1775 return bdrv_open_inherit(filename
, reference
, options
, flags
, NULL
,
1779 typedef struct BlockReopenQueueEntry
{
1781 BDRVReopenState state
;
1782 QSIMPLEQ_ENTRY(BlockReopenQueueEntry
) entry
;
1783 } BlockReopenQueueEntry
;
1786 * Adds a BlockDriverState to a simple queue for an atomic, transactional
1787 * reopen of multiple devices.
1789 * bs_queue can either be an existing BlockReopenQueue that has had QSIMPLE_INIT
1790 * already performed, or alternatively may be NULL a new BlockReopenQueue will
1791 * be created and initialized. This newly created BlockReopenQueue should be
1792 * passed back in for subsequent calls that are intended to be of the same
1795 * bs is the BlockDriverState to add to the reopen queue.
1797 * options contains the changed options for the associated bs
1798 * (the BlockReopenQueue takes ownership)
1800 * flags contains the open flags for the associated bs
1802 * returns a pointer to bs_queue, which is either the newly allocated
1803 * bs_queue, or the existing bs_queue being used.
1806 static BlockReopenQueue
*bdrv_reopen_queue_child(BlockReopenQueue
*bs_queue
,
1807 BlockDriverState
*bs
,
1810 const BdrvChildRole
*role
,
1811 QDict
*parent_options
,
1816 BlockReopenQueueEntry
*bs_entry
;
1818 QDict
*old_options
, *explicit_options
;
1820 if (bs_queue
== NULL
) {
1821 bs_queue
= g_new0(BlockReopenQueue
, 1);
1822 QSIMPLEQ_INIT(bs_queue
);
1826 options
= qdict_new();
1830 * Precedence of options:
1831 * 1. Explicitly passed in options (highest)
1832 * 2. Set in flags (only for top level)
1833 * 3. Retained from explicitly set options of bs
1834 * 4. Inherited from parent node
1835 * 5. Retained from effective options of bs
1838 if (!parent_options
) {
1840 * Any setting represented by flags is always updated. If the
1841 * corresponding QDict option is set, it takes precedence. Otherwise
1842 * the flag is translated into a QDict option. The old setting of bs is
1845 update_options_from_flags(options
, flags
);
1848 /* Old explicitly set values (don't overwrite by inherited value) */
1849 old_options
= qdict_clone_shallow(bs
->explicit_options
);
1850 bdrv_join_options(bs
, options
, old_options
);
1851 QDECREF(old_options
);
1853 explicit_options
= qdict_clone_shallow(options
);
1855 /* Inherit from parent node */
1856 if (parent_options
) {
1858 role
->inherit_options(&flags
, options
, parent_flags
, parent_options
);
1861 /* Old values are used for options that aren't set yet */
1862 old_options
= qdict_clone_shallow(bs
->options
);
1863 bdrv_join_options(bs
, options
, old_options
);
1864 QDECREF(old_options
);
1866 /* bdrv_open() masks this flag out */
1867 flags
&= ~BDRV_O_PROTOCOL
;
1869 QLIST_FOREACH(child
, &bs
->children
, next
) {
1870 QDict
*new_child_options
;
1871 char *child_key_dot
;
1873 /* reopen can only change the options of block devices that were
1874 * implicitly created and inherited options. For other (referenced)
1875 * block devices, a syntax like "backing.foo" results in an error. */
1876 if (child
->bs
->inherits_from
!= bs
) {
1880 child_key_dot
= g_strdup_printf("%s.", child
->name
);
1881 qdict_extract_subqdict(options
, &new_child_options
, child_key_dot
);
1882 g_free(child_key_dot
);
1884 bdrv_reopen_queue_child(bs_queue
, child
->bs
, new_child_options
, 0,
1885 child
->role
, options
, flags
);
1888 bs_entry
= g_new0(BlockReopenQueueEntry
, 1);
1889 QSIMPLEQ_INSERT_TAIL(bs_queue
, bs_entry
, entry
);
1891 bs_entry
->state
.bs
= bs
;
1892 bs_entry
->state
.options
= options
;
1893 bs_entry
->state
.explicit_options
= explicit_options
;
1894 bs_entry
->state
.flags
= flags
;
1899 BlockReopenQueue
*bdrv_reopen_queue(BlockReopenQueue
*bs_queue
,
1900 BlockDriverState
*bs
,
1901 QDict
*options
, int flags
)
1903 return bdrv_reopen_queue_child(bs_queue
, bs
, options
, flags
,
1908 * Reopen multiple BlockDriverStates atomically & transactionally.
1910 * The queue passed in (bs_queue) must have been built up previous
1911 * via bdrv_reopen_queue().
1913 * Reopens all BDS specified in the queue, with the appropriate
1914 * flags. All devices are prepared for reopen, and failure of any
1915 * device will cause all device changes to be abandonded, and intermediate
1918 * If all devices prepare successfully, then the changes are committed
1922 int bdrv_reopen_multiple(BlockReopenQueue
*bs_queue
, Error
**errp
)
1925 BlockReopenQueueEntry
*bs_entry
, *next
;
1926 Error
*local_err
= NULL
;
1928 assert(bs_queue
!= NULL
);
1932 QSIMPLEQ_FOREACH(bs_entry
, bs_queue
, entry
) {
1933 if (bdrv_reopen_prepare(&bs_entry
->state
, bs_queue
, &local_err
)) {
1934 error_propagate(errp
, local_err
);
1937 bs_entry
->prepared
= true;
1940 /* If we reach this point, we have success and just need to apply the
1943 QSIMPLEQ_FOREACH(bs_entry
, bs_queue
, entry
) {
1944 bdrv_reopen_commit(&bs_entry
->state
);
1950 QSIMPLEQ_FOREACH_SAFE(bs_entry
, bs_queue
, entry
, next
) {
1951 if (ret
&& bs_entry
->prepared
) {
1952 bdrv_reopen_abort(&bs_entry
->state
);
1954 QDECREF(bs_entry
->state
.explicit_options
);
1956 QDECREF(bs_entry
->state
.options
);
1964 /* Reopen a single BlockDriverState with the specified flags. */
1965 int bdrv_reopen(BlockDriverState
*bs
, int bdrv_flags
, Error
**errp
)
1968 Error
*local_err
= NULL
;
1969 BlockReopenQueue
*queue
= bdrv_reopen_queue(NULL
, bs
, NULL
, bdrv_flags
);
1971 ret
= bdrv_reopen_multiple(queue
, &local_err
);
1972 if (local_err
!= NULL
) {
1973 error_propagate(errp
, local_err
);
1980 * Prepares a BlockDriverState for reopen. All changes are staged in the
1981 * 'opaque' field of the BDRVReopenState, which is used and allocated by
1982 * the block driver layer .bdrv_reopen_prepare()
1984 * bs is the BlockDriverState to reopen
1985 * flags are the new open flags
1986 * queue is the reopen queue
1988 * Returns 0 on success, non-zero on error. On error errp will be set
1991 * On failure, bdrv_reopen_abort() will be called to clean up any data.
1992 * It is the responsibility of the caller to then call the abort() or
1993 * commit() for any other BDS that have been left in a prepare() state
1996 int bdrv_reopen_prepare(BDRVReopenState
*reopen_state
, BlockReopenQueue
*queue
,
2000 Error
*local_err
= NULL
;
2005 assert(reopen_state
!= NULL
);
2006 assert(reopen_state
->bs
->drv
!= NULL
);
2007 drv
= reopen_state
->bs
->drv
;
2009 /* Process generic block layer options */
2010 opts
= qemu_opts_create(&bdrv_runtime_opts
, NULL
, 0, &error_abort
);
2011 qemu_opts_absorb_qdict(opts
, reopen_state
->options
, &local_err
);
2013 error_propagate(errp
, local_err
);
2018 update_flags_from_options(&reopen_state
->flags
, opts
);
2020 /* node-name and driver must be unchanged. Put them back into the QDict, so
2021 * that they are checked at the end of this function. */
2022 value
= qemu_opt_get(opts
, "node-name");
2024 qdict_put(reopen_state
->options
, "node-name", qstring_from_str(value
));
2027 value
= qemu_opt_get(opts
, "driver");
2029 qdict_put(reopen_state
->options
, "driver", qstring_from_str(value
));
2032 /* if we are to stay read-only, do not allow permission change
2034 if (!(reopen_state
->bs
->open_flags
& BDRV_O_ALLOW_RDWR
) &&
2035 reopen_state
->flags
& BDRV_O_RDWR
) {
2036 error_setg(errp
, "Node '%s' is read only",
2037 bdrv_get_device_or_node_name(reopen_state
->bs
));
2042 ret
= bdrv_flush(reopen_state
->bs
);
2044 error_setg_errno(errp
, -ret
, "Error flushing drive");
2048 if (drv
->bdrv_reopen_prepare
) {
2049 ret
= drv
->bdrv_reopen_prepare(reopen_state
, queue
, &local_err
);
2051 if (local_err
!= NULL
) {
2052 error_propagate(errp
, local_err
);
2054 error_setg(errp
, "failed while preparing to reopen image '%s'",
2055 reopen_state
->bs
->filename
);
2060 /* It is currently mandatory to have a bdrv_reopen_prepare()
2061 * handler for each supported drv. */
2062 error_setg(errp
, "Block format '%s' used by node '%s' "
2063 "does not support reopening files", drv
->format_name
,
2064 bdrv_get_device_or_node_name(reopen_state
->bs
));
2069 /* Options that are not handled are only okay if they are unchanged
2070 * compared to the old state. It is expected that some options are only
2071 * used for the initial open, but not reopen (e.g. filename) */
2072 if (qdict_size(reopen_state
->options
)) {
2073 const QDictEntry
*entry
= qdict_first(reopen_state
->options
);
2076 QString
*new_obj
= qobject_to_qstring(entry
->value
);
2077 const char *new = qstring_get_str(new_obj
);
2078 const char *old
= qdict_get_try_str(reopen_state
->bs
->options
,
2081 if (!old
|| strcmp(new, old
)) {
2082 error_setg(errp
, "Cannot change the option '%s'", entry
->key
);
2086 } while ((entry
= qdict_next(reopen_state
->options
, entry
)));
2092 qemu_opts_del(opts
);
2097 * Takes the staged changes for the reopen from bdrv_reopen_prepare(), and
2098 * makes them final by swapping the staging BlockDriverState contents into
2099 * the active BlockDriverState contents.
2101 void bdrv_reopen_commit(BDRVReopenState
*reopen_state
)
2105 assert(reopen_state
!= NULL
);
2106 drv
= reopen_state
->bs
->drv
;
2107 assert(drv
!= NULL
);
2109 /* If there are any driver level actions to take */
2110 if (drv
->bdrv_reopen_commit
) {
2111 drv
->bdrv_reopen_commit(reopen_state
);
2114 /* set BDS specific flags now */
2115 QDECREF(reopen_state
->bs
->explicit_options
);
2117 reopen_state
->bs
->explicit_options
= reopen_state
->explicit_options
;
2118 reopen_state
->bs
->open_flags
= reopen_state
->flags
;
2119 reopen_state
->bs
->read_only
= !(reopen_state
->flags
& BDRV_O_RDWR
);
2121 bdrv_refresh_limits(reopen_state
->bs
, NULL
);
2125 * Abort the reopen, and delete and free the staged changes in
2128 void bdrv_reopen_abort(BDRVReopenState
*reopen_state
)
2132 assert(reopen_state
!= NULL
);
2133 drv
= reopen_state
->bs
->drv
;
2134 assert(drv
!= NULL
);
2136 if (drv
->bdrv_reopen_abort
) {
2137 drv
->bdrv_reopen_abort(reopen_state
);
2140 QDECREF(reopen_state
->explicit_options
);
2144 static void bdrv_close(BlockDriverState
*bs
)
2146 BdrvAioNotifier
*ban
, *ban_next
;
2149 assert(!bs
->refcnt
);
2151 bdrv_drained_begin(bs
); /* complete I/O */
2153 bdrv_drain(bs
); /* in case flush left pending I/O */
2155 bdrv_release_named_dirty_bitmaps(bs
);
2156 assert(QLIST_EMPTY(&bs
->dirty_bitmaps
));
2159 BdrvChild
*child
, *next
;
2161 bs
->drv
->bdrv_close(bs
);
2164 bdrv_set_backing_hd(bs
, NULL
);
2166 if (bs
->file
!= NULL
) {
2167 bdrv_unref_child(bs
, bs
->file
);
2171 QLIST_FOREACH_SAFE(child
, &bs
->children
, next
, next
) {
2172 /* TODO Remove bdrv_unref() from drivers' close function and use
2173 * bdrv_unref_child() here */
2174 if (child
->bs
->inherits_from
== bs
) {
2175 child
->bs
->inherits_from
= NULL
;
2177 bdrv_detach_child(child
);
2182 bs
->copy_on_read
= 0;
2183 bs
->backing_file
[0] = '\0';
2184 bs
->backing_format
[0] = '\0';
2185 bs
->total_sectors
= 0;
2186 bs
->encrypted
= false;
2187 bs
->valid_key
= false;
2189 QDECREF(bs
->options
);
2190 QDECREF(bs
->explicit_options
);
2192 QDECREF(bs
->full_open_options
);
2193 bs
->full_open_options
= NULL
;
2196 QLIST_FOREACH_SAFE(ban
, &bs
->aio_notifiers
, list
, ban_next
) {
2199 QLIST_INIT(&bs
->aio_notifiers
);
2200 bdrv_drained_end(bs
);
2203 void bdrv_close_all(void)
2205 block_job_cancel_sync_all();
2207 /* Drop references from requests still in flight, such as canceled block
2208 * jobs whose AIO context has not been polled yet */
2211 blk_remove_all_bs();
2212 blockdev_close_all_bdrv_states();
2214 assert(QTAILQ_EMPTY(&all_bdrv_states
));
2217 static void change_parent_backing_link(BlockDriverState
*from
,
2218 BlockDriverState
*to
)
2220 BdrvChild
*c
, *next
, *to_c
;
2222 QLIST_FOREACH_SAFE(c
, &from
->parents
, next_parent
, next
) {
2223 if (c
->role
== &child_backing
) {
2224 /* @from is generally not allowed to be a backing file, except for
2225 * when @to is the overlay. In that case, @from may not be replaced
2226 * by @to as @to's backing node. */
2227 QLIST_FOREACH(to_c
, &to
->children
, next
) {
2237 assert(c
->role
!= &child_backing
);
2239 bdrv_replace_child(c
, to
);
2245 * Add new bs contents at the top of an image chain while the chain is
2246 * live, while keeping required fields on the top layer.
2248 * This will modify the BlockDriverState fields, and swap contents
2249 * between bs_new and bs_top. Both bs_new and bs_top are modified.
2251 * bs_new must not be attached to a BlockBackend.
2253 * This function does not create any image files.
2255 * bdrv_append() takes ownership of a bs_new reference and unrefs it because
2256 * that's what the callers commonly need. bs_new will be referenced by the old
2257 * parents of bs_top after bdrv_append() returns. If the caller needs to keep a
2258 * reference of its own, it must call bdrv_ref().
2260 void bdrv_append(BlockDriverState
*bs_new
, BlockDriverState
*bs_top
)
2262 assert(!bdrv_requests_pending(bs_top
));
2263 assert(!bdrv_requests_pending(bs_new
));
2267 change_parent_backing_link(bs_top
, bs_new
);
2268 bdrv_set_backing_hd(bs_new
, bs_top
);
2271 /* bs_new is now referenced by its new parents, we don't need the
2272 * additional reference any more. */
2276 void bdrv_replace_in_backing_chain(BlockDriverState
*old
, BlockDriverState
*new)
2278 assert(!bdrv_requests_pending(old
));
2279 assert(!bdrv_requests_pending(new));
2283 change_parent_backing_link(old
, new);
2288 static void bdrv_delete(BlockDriverState
*bs
)
2291 assert(bdrv_op_blocker_is_empty(bs
));
2292 assert(!bs
->refcnt
);
2296 /* remove from list, if necessary */
2297 if (bs
->node_name
[0] != '\0') {
2298 QTAILQ_REMOVE(&graph_bdrv_states
, bs
, node_list
);
2300 QTAILQ_REMOVE(&all_bdrv_states
, bs
, bs_list
);
2306 * Run consistency checks on an image
2308 * Returns 0 if the check could be completed (it doesn't mean that the image is
2309 * free of errors) or -errno when an internal error occurred. The results of the
2310 * check are stored in res.
2312 int bdrv_check(BlockDriverState
*bs
, BdrvCheckResult
*res
, BdrvCheckMode fix
)
2314 if (bs
->drv
== NULL
) {
2317 if (bs
->drv
->bdrv_check
== NULL
) {
2321 memset(res
, 0, sizeof(*res
));
2322 return bs
->drv
->bdrv_check(bs
, res
, fix
);
2325 #define COMMIT_BUF_SECTORS 2048
2327 /* commit COW file into the raw image */
2328 int bdrv_commit(BlockDriverState
*bs
)
2330 BlockDriver
*drv
= bs
->drv
;
2331 int64_t sector
, total_sectors
, length
, backing_length
;
2332 int n
, ro
, open_flags
;
2334 uint8_t *buf
= NULL
;
2343 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_COMMIT_SOURCE
, NULL
) ||
2344 bdrv_op_is_blocked(bs
->backing
->bs
, BLOCK_OP_TYPE_COMMIT_TARGET
, NULL
)) {
2348 ro
= bs
->backing
->bs
->read_only
;
2349 open_flags
= bs
->backing
->bs
->open_flags
;
2352 if (bdrv_reopen(bs
->backing
->bs
, open_flags
| BDRV_O_RDWR
, NULL
)) {
2357 length
= bdrv_getlength(bs
);
2363 backing_length
= bdrv_getlength(bs
->backing
->bs
);
2364 if (backing_length
< 0) {
2365 ret
= backing_length
;
2369 /* If our top snapshot is larger than the backing file image,
2370 * grow the backing file image if possible. If not possible,
2371 * we must return an error */
2372 if (length
> backing_length
) {
2373 ret
= bdrv_truncate(bs
->backing
->bs
, length
);
2379 total_sectors
= length
>> BDRV_SECTOR_BITS
;
2381 /* qemu_try_blockalign() for bs will choose an alignment that works for
2382 * bs->backing->bs as well, so no need to compare the alignment manually. */
2383 buf
= qemu_try_blockalign(bs
, COMMIT_BUF_SECTORS
* BDRV_SECTOR_SIZE
);
2389 for (sector
= 0; sector
< total_sectors
; sector
+= n
) {
2390 ret
= bdrv_is_allocated(bs
, sector
, COMMIT_BUF_SECTORS
, &n
);
2395 ret
= bdrv_read(bs
, sector
, buf
, n
);
2400 ret
= bdrv_write(bs
->backing
->bs
, sector
, buf
, n
);
2407 if (drv
->bdrv_make_empty
) {
2408 ret
= drv
->bdrv_make_empty(bs
);
2416 * Make sure all data we wrote to the backing device is actually
2420 bdrv_flush(bs
->backing
->bs
);
2428 /* ignoring error return here */
2429 bdrv_reopen(bs
->backing
->bs
, open_flags
& ~BDRV_O_RDWR
, NULL
);
2438 * -EINVAL - backing format specified, but no file
2439 * -ENOSPC - can't update the backing file because no space is left in the
2441 * -ENOTSUP - format driver doesn't support changing the backing file
2443 int bdrv_change_backing_file(BlockDriverState
*bs
,
2444 const char *backing_file
, const char *backing_fmt
)
2446 BlockDriver
*drv
= bs
->drv
;
2449 /* Backing file format doesn't make sense without a backing file */
2450 if (backing_fmt
&& !backing_file
) {
2454 if (drv
->bdrv_change_backing_file
!= NULL
) {
2455 ret
= drv
->bdrv_change_backing_file(bs
, backing_file
, backing_fmt
);
2461 pstrcpy(bs
->backing_file
, sizeof(bs
->backing_file
), backing_file
?: "");
2462 pstrcpy(bs
->backing_format
, sizeof(bs
->backing_format
), backing_fmt
?: "");
2468 * Finds the image layer in the chain that has 'bs' as its backing file.
2470 * active is the current topmost image.
2472 * Returns NULL if bs is not found in active's image chain,
2473 * or if active == bs.
2475 * Returns the bottommost base image if bs == NULL.
2477 BlockDriverState
*bdrv_find_overlay(BlockDriverState
*active
,
2478 BlockDriverState
*bs
)
2480 while (active
&& bs
!= backing_bs(active
)) {
2481 active
= backing_bs(active
);
2487 /* Given a BDS, searches for the base layer. */
2488 BlockDriverState
*bdrv_find_base(BlockDriverState
*bs
)
2490 return bdrv_find_overlay(bs
, NULL
);
2494 * Drops images above 'base' up to and including 'top', and sets the image
2495 * above 'top' to have base as its backing file.
2497 * Requires that the overlay to 'top' is opened r/w, so that the backing file
2498 * information in 'bs' can be properly updated.
2500 * E.g., this will convert the following chain:
2501 * bottom <- base <- intermediate <- top <- active
2505 * bottom <- base <- active
2507 * It is allowed for bottom==base, in which case it converts:
2509 * base <- intermediate <- top <- active
2515 * If backing_file_str is non-NULL, it will be used when modifying top's
2516 * overlay image metadata.
2519 * if active == top, that is considered an error
2522 int bdrv_drop_intermediate(BlockDriverState
*active
, BlockDriverState
*top
,
2523 BlockDriverState
*base
, const char *backing_file_str
)
2525 BlockDriverState
*new_top_bs
= NULL
;
2528 if (!top
->drv
|| !base
->drv
) {
2532 new_top_bs
= bdrv_find_overlay(active
, top
);
2534 if (new_top_bs
== NULL
) {
2535 /* we could not find the image above 'top', this is an error */
2539 /* special case of new_top_bs->backing->bs already pointing to base - nothing
2540 * to do, no intermediate images */
2541 if (backing_bs(new_top_bs
) == base
) {
2546 /* Make sure that base is in the backing chain of top */
2547 if (!bdrv_chain_contains(top
, base
)) {
2551 /* success - we can delete the intermediate states, and link top->base */
2552 backing_file_str
= backing_file_str
? backing_file_str
: base
->filename
;
2553 ret
= bdrv_change_backing_file(new_top_bs
, backing_file_str
,
2554 base
->drv
? base
->drv
->format_name
: "");
2558 bdrv_set_backing_hd(new_top_bs
, base
);
2566 * Truncate file to 'offset' bytes (needed only for file protocols)
2568 int bdrv_truncate(BlockDriverState
*bs
, int64_t offset
)
2570 BlockDriver
*drv
= bs
->drv
;
2574 if (!drv
->bdrv_truncate
)
2579 ret
= drv
->bdrv_truncate(bs
, offset
);
2581 ret
= refresh_total_sectors(bs
, offset
>> BDRV_SECTOR_BITS
);
2582 bdrv_dirty_bitmap_truncate(bs
);
2583 bdrv_parent_cb_resize(bs
);
2589 * Length of a allocated file in bytes. Sparse files are counted by actual
2590 * allocated space. Return < 0 if error or unknown.
2592 int64_t bdrv_get_allocated_file_size(BlockDriverState
*bs
)
2594 BlockDriver
*drv
= bs
->drv
;
2598 if (drv
->bdrv_get_allocated_file_size
) {
2599 return drv
->bdrv_get_allocated_file_size(bs
);
2602 return bdrv_get_allocated_file_size(bs
->file
->bs
);
2608 * Return number of sectors on success, -errno on error.
2610 int64_t bdrv_nb_sectors(BlockDriverState
*bs
)
2612 BlockDriver
*drv
= bs
->drv
;
2617 if (drv
->has_variable_length
) {
2618 int ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
2623 return bs
->total_sectors
;
2627 * Return length in bytes on success, -errno on error.
2628 * The length is always a multiple of BDRV_SECTOR_SIZE.
2630 int64_t bdrv_getlength(BlockDriverState
*bs
)
2632 int64_t ret
= bdrv_nb_sectors(bs
);
2634 ret
= ret
> INT64_MAX
/ BDRV_SECTOR_SIZE
? -EFBIG
: ret
;
2635 return ret
< 0 ? ret
: ret
* BDRV_SECTOR_SIZE
;
2638 /* return 0 as number of sectors if no device present or error */
2639 void bdrv_get_geometry(BlockDriverState
*bs
, uint64_t *nb_sectors_ptr
)
2641 int64_t nb_sectors
= bdrv_nb_sectors(bs
);
2643 *nb_sectors_ptr
= nb_sectors
< 0 ? 0 : nb_sectors
;
2646 bool bdrv_is_read_only(BlockDriverState
*bs
)
2648 return bs
->read_only
;
2651 bool bdrv_is_sg(BlockDriverState
*bs
)
2656 bool bdrv_is_encrypted(BlockDriverState
*bs
)
2658 if (bs
->backing
&& bs
->backing
->bs
->encrypted
) {
2661 return bs
->encrypted
;
2664 bool bdrv_key_required(BlockDriverState
*bs
)
2666 BdrvChild
*backing
= bs
->backing
;
2668 if (backing
&& backing
->bs
->encrypted
&& !backing
->bs
->valid_key
) {
2671 return (bs
->encrypted
&& !bs
->valid_key
);
2674 int bdrv_set_key(BlockDriverState
*bs
, const char *key
)
2677 if (bs
->backing
&& bs
->backing
->bs
->encrypted
) {
2678 ret
= bdrv_set_key(bs
->backing
->bs
, key
);
2684 if (!bs
->encrypted
) {
2686 } else if (!bs
->drv
|| !bs
->drv
->bdrv_set_key
) {
2689 ret
= bs
->drv
->bdrv_set_key(bs
, key
);
2691 bs
->valid_key
= false;
2692 } else if (!bs
->valid_key
) {
2693 /* call the change callback now, we skipped it on open */
2694 bs
->valid_key
= true;
2695 bdrv_parent_cb_change_media(bs
, true);
2701 * Provide an encryption key for @bs.
2702 * If @key is non-null:
2703 * If @bs is not encrypted, fail.
2704 * Else if the key is invalid, fail.
2705 * Else set @bs's key to @key, replacing the existing key, if any.
2707 * If @bs is encrypted and still lacks a key, fail.
2709 * On failure, store an error object through @errp if non-null.
2711 void bdrv_add_key(BlockDriverState
*bs
, const char *key
, Error
**errp
)
2714 if (!bdrv_is_encrypted(bs
)) {
2715 error_setg(errp
, "Node '%s' is not encrypted",
2716 bdrv_get_device_or_node_name(bs
));
2717 } else if (bdrv_set_key(bs
, key
) < 0) {
2718 error_setg(errp
, QERR_INVALID_PASSWORD
);
2721 if (bdrv_key_required(bs
)) {
2722 error_set(errp
, ERROR_CLASS_DEVICE_ENCRYPTED
,
2723 "'%s' (%s) is encrypted",
2724 bdrv_get_device_or_node_name(bs
),
2725 bdrv_get_encrypted_filename(bs
));
2730 const char *bdrv_get_format_name(BlockDriverState
*bs
)
2732 return bs
->drv
? bs
->drv
->format_name
: NULL
;
2735 static int qsort_strcmp(const void *a
, const void *b
)
2737 return strcmp(a
, b
);
2740 void bdrv_iterate_format(void (*it
)(void *opaque
, const char *name
),
2746 const char **formats
= NULL
;
2748 QLIST_FOREACH(drv
, &bdrv_drivers
, list
) {
2749 if (drv
->format_name
) {
2752 while (formats
&& i
&& !found
) {
2753 found
= !strcmp(formats
[--i
], drv
->format_name
);
2757 formats
= g_renew(const char *, formats
, count
+ 1);
2758 formats
[count
++] = drv
->format_name
;
2763 qsort(formats
, count
, sizeof(formats
[0]), qsort_strcmp
);
2765 for (i
= 0; i
< count
; i
++) {
2766 it(opaque
, formats
[i
]);
2772 /* This function is to find a node in the bs graph */
2773 BlockDriverState
*bdrv_find_node(const char *node_name
)
2775 BlockDriverState
*bs
;
2779 QTAILQ_FOREACH(bs
, &graph_bdrv_states
, node_list
) {
2780 if (!strcmp(node_name
, bs
->node_name
)) {
2787 /* Put this QMP function here so it can access the static graph_bdrv_states. */
2788 BlockDeviceInfoList
*bdrv_named_nodes_list(Error
**errp
)
2790 BlockDeviceInfoList
*list
, *entry
;
2791 BlockDriverState
*bs
;
2794 QTAILQ_FOREACH(bs
, &graph_bdrv_states
, node_list
) {
2795 BlockDeviceInfo
*info
= bdrv_block_device_info(NULL
, bs
, errp
);
2797 qapi_free_BlockDeviceInfoList(list
);
2800 entry
= g_malloc0(sizeof(*entry
));
2801 entry
->value
= info
;
2809 BlockDriverState
*bdrv_lookup_bs(const char *device
,
2810 const char *node_name
,
2814 BlockDriverState
*bs
;
2817 blk
= blk_by_name(device
);
2822 error_setg(errp
, "Device '%s' has no medium", device
);
2830 bs
= bdrv_find_node(node_name
);
2837 error_setg(errp
, "Cannot find device=%s nor node_name=%s",
2838 device
? device
: "",
2839 node_name
? node_name
: "");
2843 /* If 'base' is in the same chain as 'top', return true. Otherwise,
2844 * return false. If either argument is NULL, return false. */
2845 bool bdrv_chain_contains(BlockDriverState
*top
, BlockDriverState
*base
)
2847 while (top
&& top
!= base
) {
2848 top
= backing_bs(top
);
2854 BlockDriverState
*bdrv_next_node(BlockDriverState
*bs
)
2857 return QTAILQ_FIRST(&graph_bdrv_states
);
2859 return QTAILQ_NEXT(bs
, node_list
);
2862 const char *bdrv_get_node_name(const BlockDriverState
*bs
)
2864 return bs
->node_name
;
2867 const char *bdrv_get_parent_name(const BlockDriverState
*bs
)
2872 /* If multiple parents have a name, just pick the first one. */
2873 QLIST_FOREACH(c
, &bs
->parents
, next_parent
) {
2874 if (c
->role
->get_name
) {
2875 name
= c
->role
->get_name(c
);
2876 if (name
&& *name
) {
2885 /* TODO check what callers really want: bs->node_name or blk_name() */
2886 const char *bdrv_get_device_name(const BlockDriverState
*bs
)
2888 return bdrv_get_parent_name(bs
) ?: "";
2891 /* This can be used to identify nodes that might not have a device
2892 * name associated. Since node and device names live in the same
2893 * namespace, the result is unambiguous. The exception is if both are
2894 * absent, then this returns an empty (non-null) string. */
2895 const char *bdrv_get_device_or_node_name(const BlockDriverState
*bs
)
2897 return bdrv_get_parent_name(bs
) ?: bs
->node_name
;
2900 int bdrv_get_flags(BlockDriverState
*bs
)
2902 return bs
->open_flags
;
2905 int bdrv_has_zero_init_1(BlockDriverState
*bs
)
2910 int bdrv_has_zero_init(BlockDriverState
*bs
)
2914 /* If BS is a copy on write image, it is initialized to
2915 the contents of the base image, which may not be zeroes. */
2919 if (bs
->drv
->bdrv_has_zero_init
) {
2920 return bs
->drv
->bdrv_has_zero_init(bs
);
2927 bool bdrv_unallocated_blocks_are_zero(BlockDriverState
*bs
)
2929 BlockDriverInfo bdi
;
2935 if (bdrv_get_info(bs
, &bdi
) == 0) {
2936 return bdi
.unallocated_blocks_are_zero
;
2942 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState
*bs
)
2944 BlockDriverInfo bdi
;
2946 if (bs
->backing
|| !(bs
->open_flags
& BDRV_O_UNMAP
)) {
2950 if (bdrv_get_info(bs
, &bdi
) == 0) {
2951 return bdi
.can_write_zeroes_with_unmap
;
2957 const char *bdrv_get_encrypted_filename(BlockDriverState
*bs
)
2959 if (bs
->backing
&& bs
->backing
->bs
->encrypted
)
2960 return bs
->backing_file
;
2961 else if (bs
->encrypted
)
2962 return bs
->filename
;
2967 void bdrv_get_backing_filename(BlockDriverState
*bs
,
2968 char *filename
, int filename_size
)
2970 pstrcpy(filename
, filename_size
, bs
->backing_file
);
2973 int bdrv_get_info(BlockDriverState
*bs
, BlockDriverInfo
*bdi
)
2975 BlockDriver
*drv
= bs
->drv
;
2978 if (!drv
->bdrv_get_info
)
2980 memset(bdi
, 0, sizeof(*bdi
));
2981 return drv
->bdrv_get_info(bs
, bdi
);
2984 ImageInfoSpecific
*bdrv_get_specific_info(BlockDriverState
*bs
)
2986 BlockDriver
*drv
= bs
->drv
;
2987 if (drv
&& drv
->bdrv_get_specific_info
) {
2988 return drv
->bdrv_get_specific_info(bs
);
2993 void bdrv_debug_event(BlockDriverState
*bs
, BlkdebugEvent event
)
2995 if (!bs
|| !bs
->drv
|| !bs
->drv
->bdrv_debug_event
) {
2999 bs
->drv
->bdrv_debug_event(bs
, event
);
3002 int bdrv_debug_breakpoint(BlockDriverState
*bs
, const char *event
,
3005 while (bs
&& bs
->drv
&& !bs
->drv
->bdrv_debug_breakpoint
) {
3006 bs
= bs
->file
? bs
->file
->bs
: NULL
;
3009 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_breakpoint
) {
3010 return bs
->drv
->bdrv_debug_breakpoint(bs
, event
, tag
);
3016 int bdrv_debug_remove_breakpoint(BlockDriverState
*bs
, const char *tag
)
3018 while (bs
&& bs
->drv
&& !bs
->drv
->bdrv_debug_remove_breakpoint
) {
3019 bs
= bs
->file
? bs
->file
->bs
: NULL
;
3022 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_remove_breakpoint
) {
3023 return bs
->drv
->bdrv_debug_remove_breakpoint(bs
, tag
);
3029 int bdrv_debug_resume(BlockDriverState
*bs
, const char *tag
)
3031 while (bs
&& (!bs
->drv
|| !bs
->drv
->bdrv_debug_resume
)) {
3032 bs
= bs
->file
? bs
->file
->bs
: NULL
;
3035 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_resume
) {
3036 return bs
->drv
->bdrv_debug_resume(bs
, tag
);
3042 bool bdrv_debug_is_suspended(BlockDriverState
*bs
, const char *tag
)
3044 while (bs
&& bs
->drv
&& !bs
->drv
->bdrv_debug_is_suspended
) {
3045 bs
= bs
->file
? bs
->file
->bs
: NULL
;
3048 if (bs
&& bs
->drv
&& bs
->drv
->bdrv_debug_is_suspended
) {
3049 return bs
->drv
->bdrv_debug_is_suspended(bs
, tag
);
3055 int bdrv_is_snapshot(BlockDriverState
*bs
)
3057 return !!(bs
->open_flags
& BDRV_O_SNAPSHOT
);
3060 /* backing_file can either be relative, or absolute, or a protocol. If it is
3061 * relative, it must be relative to the chain. So, passing in bs->filename
3062 * from a BDS as backing_file should not be done, as that may be relative to
3063 * the CWD rather than the chain. */
3064 BlockDriverState
*bdrv_find_backing_image(BlockDriverState
*bs
,
3065 const char *backing_file
)
3067 char *filename_full
= NULL
;
3068 char *backing_file_full
= NULL
;
3069 char *filename_tmp
= NULL
;
3070 int is_protocol
= 0;
3071 BlockDriverState
*curr_bs
= NULL
;
3072 BlockDriverState
*retval
= NULL
;
3074 if (!bs
|| !bs
->drv
|| !backing_file
) {
3078 filename_full
= g_malloc(PATH_MAX
);
3079 backing_file_full
= g_malloc(PATH_MAX
);
3080 filename_tmp
= g_malloc(PATH_MAX
);
3082 is_protocol
= path_has_protocol(backing_file
);
3084 for (curr_bs
= bs
; curr_bs
->backing
; curr_bs
= curr_bs
->backing
->bs
) {
3086 /* If either of the filename paths is actually a protocol, then
3087 * compare unmodified paths; otherwise make paths relative */
3088 if (is_protocol
|| path_has_protocol(curr_bs
->backing_file
)) {
3089 if (strcmp(backing_file
, curr_bs
->backing_file
) == 0) {
3090 retval
= curr_bs
->backing
->bs
;
3094 /* If not an absolute filename path, make it relative to the current
3095 * image's filename path */
3096 path_combine(filename_tmp
, PATH_MAX
, curr_bs
->filename
,
3099 /* We are going to compare absolute pathnames */
3100 if (!realpath(filename_tmp
, filename_full
)) {
3104 /* We need to make sure the backing filename we are comparing against
3105 * is relative to the current image filename (or absolute) */
3106 path_combine(filename_tmp
, PATH_MAX
, curr_bs
->filename
,
3107 curr_bs
->backing_file
);
3109 if (!realpath(filename_tmp
, backing_file_full
)) {
3113 if (strcmp(backing_file_full
, filename_full
) == 0) {
3114 retval
= curr_bs
->backing
->bs
;
3120 g_free(filename_full
);
3121 g_free(backing_file_full
);
3122 g_free(filename_tmp
);
3126 int bdrv_get_backing_file_depth(BlockDriverState
*bs
)
3136 return 1 + bdrv_get_backing_file_depth(bs
->backing
->bs
);
3139 void bdrv_init(void)
3141 module_call_init(MODULE_INIT_BLOCK
);
3144 void bdrv_init_with_whitelist(void)
3146 use_bdrv_whitelist
= 1;
3150 void bdrv_invalidate_cache(BlockDriverState
*bs
, Error
**errp
)
3153 Error
*local_err
= NULL
;
3160 if (!(bs
->open_flags
& BDRV_O_INACTIVE
)) {
3163 bs
->open_flags
&= ~BDRV_O_INACTIVE
;
3165 if (bs
->drv
->bdrv_invalidate_cache
) {
3166 bs
->drv
->bdrv_invalidate_cache(bs
, &local_err
);
3168 bs
->open_flags
|= BDRV_O_INACTIVE
;
3169 error_propagate(errp
, local_err
);
3174 QLIST_FOREACH(child
, &bs
->children
, next
) {
3175 bdrv_invalidate_cache(child
->bs
, &local_err
);
3177 bs
->open_flags
|= BDRV_O_INACTIVE
;
3178 error_propagate(errp
, local_err
);
3183 ret
= refresh_total_sectors(bs
, bs
->total_sectors
);
3185 bs
->open_flags
|= BDRV_O_INACTIVE
;
3186 error_setg_errno(errp
, -ret
, "Could not refresh total sector count");
3191 void bdrv_invalidate_cache_all(Error
**errp
)
3193 BlockDriverState
*bs
;
3194 Error
*local_err
= NULL
;
3195 BdrvNextIterator it
;
3197 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3198 AioContext
*aio_context
= bdrv_get_aio_context(bs
);
3200 aio_context_acquire(aio_context
);
3201 bdrv_invalidate_cache(bs
, &local_err
);
3202 aio_context_release(aio_context
);
3204 error_propagate(errp
, local_err
);
3210 static int bdrv_inactivate_recurse(BlockDriverState
*bs
,
3216 if (!setting_flag
&& bs
->drv
->bdrv_inactivate
) {
3217 ret
= bs
->drv
->bdrv_inactivate(bs
);
3223 QLIST_FOREACH(child
, &bs
->children
, next
) {
3224 ret
= bdrv_inactivate_recurse(child
->bs
, setting_flag
);
3231 bs
->open_flags
|= BDRV_O_INACTIVE
;
3236 int bdrv_inactivate_all(void)
3238 BlockDriverState
*bs
= NULL
;
3239 BdrvNextIterator it
;
3243 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3244 aio_context_acquire(bdrv_get_aio_context(bs
));
3247 /* We do two passes of inactivation. The first pass calls to drivers'
3248 * .bdrv_inactivate callbacks recursively so all cache is flushed to disk;
3249 * the second pass sets the BDRV_O_INACTIVE flag so that no further write
3251 for (pass
= 0; pass
< 2; pass
++) {
3252 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3253 ret
= bdrv_inactivate_recurse(bs
, pass
);
3261 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3262 aio_context_release(bdrv_get_aio_context(bs
));
3268 /**************************************************************/
3269 /* removable device support */
3272 * Return TRUE if the media is present
3274 bool bdrv_is_inserted(BlockDriverState
*bs
)
3276 BlockDriver
*drv
= bs
->drv
;
3282 if (drv
->bdrv_is_inserted
) {
3283 return drv
->bdrv_is_inserted(bs
);
3285 QLIST_FOREACH(child
, &bs
->children
, next
) {
3286 if (!bdrv_is_inserted(child
->bs
)) {
3294 * Return whether the media changed since the last call to this
3295 * function, or -ENOTSUP if we don't know. Most drivers don't know.
3297 int bdrv_media_changed(BlockDriverState
*bs
)
3299 BlockDriver
*drv
= bs
->drv
;
3301 if (drv
&& drv
->bdrv_media_changed
) {
3302 return drv
->bdrv_media_changed(bs
);
3308 * If eject_flag is TRUE, eject the media. Otherwise, close the tray
3310 void bdrv_eject(BlockDriverState
*bs
, bool eject_flag
)
3312 BlockDriver
*drv
= bs
->drv
;
3313 const char *device_name
;
3315 if (drv
&& drv
->bdrv_eject
) {
3316 drv
->bdrv_eject(bs
, eject_flag
);
3319 device_name
= bdrv_get_device_name(bs
);
3320 if (device_name
[0] != '\0') {
3321 qapi_event_send_device_tray_moved(device_name
,
3322 eject_flag
, &error_abort
);
3327 * Lock or unlock the media (if it is locked, the user won't be able
3328 * to eject it manually).
3330 void bdrv_lock_medium(BlockDriverState
*bs
, bool locked
)
3332 BlockDriver
*drv
= bs
->drv
;
3334 trace_bdrv_lock_medium(bs
, locked
);
3336 if (drv
&& drv
->bdrv_lock_medium
) {
3337 drv
->bdrv_lock_medium(bs
, locked
);
3341 /* Get a reference to bs */
3342 void bdrv_ref(BlockDriverState
*bs
)
3347 /* Release a previously grabbed reference to bs.
3348 * If after releasing, reference count is zero, the BlockDriverState is
3350 void bdrv_unref(BlockDriverState
*bs
)
3355 assert(bs
->refcnt
> 0);
3356 if (--bs
->refcnt
== 0) {
3361 struct BdrvOpBlocker
{
3363 QLIST_ENTRY(BdrvOpBlocker
) list
;
3366 bool bdrv_op_is_blocked(BlockDriverState
*bs
, BlockOpType op
, Error
**errp
)
3368 BdrvOpBlocker
*blocker
;
3369 assert((int) op
>= 0 && op
< BLOCK_OP_TYPE_MAX
);
3370 if (!QLIST_EMPTY(&bs
->op_blockers
[op
])) {
3371 blocker
= QLIST_FIRST(&bs
->op_blockers
[op
]);
3373 *errp
= error_copy(blocker
->reason
);
3374 error_prepend(errp
, "Node '%s' is busy: ",
3375 bdrv_get_device_or_node_name(bs
));
3382 void bdrv_op_block(BlockDriverState
*bs
, BlockOpType op
, Error
*reason
)
3384 BdrvOpBlocker
*blocker
;
3385 assert((int) op
>= 0 && op
< BLOCK_OP_TYPE_MAX
);
3387 blocker
= g_new0(BdrvOpBlocker
, 1);
3388 blocker
->reason
= reason
;
3389 QLIST_INSERT_HEAD(&bs
->op_blockers
[op
], blocker
, list
);
3392 void bdrv_op_unblock(BlockDriverState
*bs
, BlockOpType op
, Error
*reason
)
3394 BdrvOpBlocker
*blocker
, *next
;
3395 assert((int) op
>= 0 && op
< BLOCK_OP_TYPE_MAX
);
3396 QLIST_FOREACH_SAFE(blocker
, &bs
->op_blockers
[op
], list
, next
) {
3397 if (blocker
->reason
== reason
) {
3398 QLIST_REMOVE(blocker
, list
);
3404 void bdrv_op_block_all(BlockDriverState
*bs
, Error
*reason
)
3407 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
3408 bdrv_op_block(bs
, i
, reason
);
3412 void bdrv_op_unblock_all(BlockDriverState
*bs
, Error
*reason
)
3415 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
3416 bdrv_op_unblock(bs
, i
, reason
);
3420 bool bdrv_op_blocker_is_empty(BlockDriverState
*bs
)
3424 for (i
= 0; i
< BLOCK_OP_TYPE_MAX
; i
++) {
3425 if (!QLIST_EMPTY(&bs
->op_blockers
[i
])) {
3432 void bdrv_img_create(const char *filename
, const char *fmt
,
3433 const char *base_filename
, const char *base_fmt
,
3434 char *options
, uint64_t img_size
, int flags
,
3435 Error
**errp
, bool quiet
)
3437 QemuOptsList
*create_opts
= NULL
;
3438 QemuOpts
*opts
= NULL
;
3439 const char *backing_fmt
, *backing_file
;
3441 BlockDriver
*drv
, *proto_drv
;
3442 Error
*local_err
= NULL
;
3445 /* Find driver and parse its options */
3446 drv
= bdrv_find_format(fmt
);
3448 error_setg(errp
, "Unknown file format '%s'", fmt
);
3452 proto_drv
= bdrv_find_protocol(filename
, true, errp
);
3457 if (!drv
->create_opts
) {
3458 error_setg(errp
, "Format driver '%s' does not support image creation",
3463 if (!proto_drv
->create_opts
) {
3464 error_setg(errp
, "Protocol driver '%s' does not support image creation",
3465 proto_drv
->format_name
);
3469 create_opts
= qemu_opts_append(create_opts
, drv
->create_opts
);
3470 create_opts
= qemu_opts_append(create_opts
, proto_drv
->create_opts
);
3472 /* Create parameter list with default values */
3473 opts
= qemu_opts_create(create_opts
, NULL
, 0, &error_abort
);
3474 qemu_opt_set_number(opts
, BLOCK_OPT_SIZE
, img_size
, &error_abort
);
3476 /* Parse -o options */
3478 qemu_opts_do_parse(opts
, options
, NULL
, &local_err
);
3480 error_report_err(local_err
);
3482 error_setg(errp
, "Invalid options for file format '%s'", fmt
);
3487 if (base_filename
) {
3488 qemu_opt_set(opts
, BLOCK_OPT_BACKING_FILE
, base_filename
, &local_err
);
3490 error_setg(errp
, "Backing file not supported for file format '%s'",
3497 qemu_opt_set(opts
, BLOCK_OPT_BACKING_FMT
, base_fmt
, &local_err
);
3499 error_setg(errp
, "Backing file format not supported for file "
3500 "format '%s'", fmt
);
3505 backing_file
= qemu_opt_get(opts
, BLOCK_OPT_BACKING_FILE
);
3507 if (!strcmp(filename
, backing_file
)) {
3508 error_setg(errp
, "Error: Trying to create an image with the "
3509 "same filename as the backing file");
3514 backing_fmt
= qemu_opt_get(opts
, BLOCK_OPT_BACKING_FMT
);
3516 // The size for the image must always be specified, with one exception:
3517 // If we are using a backing file, we can obtain the size from there
3518 size
= qemu_opt_get_size(opts
, BLOCK_OPT_SIZE
, 0);
3521 BlockDriverState
*bs
;
3522 char *full_backing
= g_new0(char, PATH_MAX
);
3525 QDict
*backing_options
= NULL
;
3527 bdrv_get_full_backing_filename_from_filename(filename
, backing_file
,
3528 full_backing
, PATH_MAX
,
3531 g_free(full_backing
);
3535 /* backing files always opened read-only */
3537 back_flags
&= ~(BDRV_O_RDWR
| BDRV_O_SNAPSHOT
| BDRV_O_NO_BACKING
);
3540 backing_options
= qdict_new();
3541 qdict_put(backing_options
, "driver",
3542 qstring_from_str(backing_fmt
));
3545 bs
= bdrv_open(full_backing
, NULL
, backing_options
, back_flags
,
3547 g_free(full_backing
);
3551 size
= bdrv_getlength(bs
);
3553 error_setg_errno(errp
, -size
, "Could not get size of '%s'",
3559 qemu_opt_set_number(opts
, BLOCK_OPT_SIZE
, size
, &error_abort
);
3563 error_setg(errp
, "Image creation needs a size parameter");
3569 printf("Formatting '%s', fmt=%s ", filename
, fmt
);
3570 qemu_opts_print(opts
, " ");
3574 ret
= bdrv_create(drv
, filename
, opts
, &local_err
);
3576 if (ret
== -EFBIG
) {
3577 /* This is generally a better message than whatever the driver would
3578 * deliver (especially because of the cluster_size_hint), since that
3579 * is most probably not much different from "image too large". */
3580 const char *cluster_size_hint
= "";
3581 if (qemu_opt_get_size(opts
, BLOCK_OPT_CLUSTER_SIZE
, 0)) {
3582 cluster_size_hint
= " (try using a larger cluster size)";
3584 error_setg(errp
, "The image size is too large for file format '%s'"
3585 "%s", fmt
, cluster_size_hint
);
3586 error_free(local_err
);
3591 qemu_opts_del(opts
);
3592 qemu_opts_free(create_opts
);
3593 error_propagate(errp
, local_err
);
3596 AioContext
*bdrv_get_aio_context(BlockDriverState
*bs
)
3598 return bs
->aio_context
;
3601 static void bdrv_do_remove_aio_context_notifier(BdrvAioNotifier
*ban
)
3603 QLIST_REMOVE(ban
, list
);
3607 void bdrv_detach_aio_context(BlockDriverState
*bs
)
3609 BdrvAioNotifier
*baf
, *baf_tmp
;
3616 assert(!bs
->walking_aio_notifiers
);
3617 bs
->walking_aio_notifiers
= true;
3618 QLIST_FOREACH_SAFE(baf
, &bs
->aio_notifiers
, list
, baf_tmp
) {
3620 bdrv_do_remove_aio_context_notifier(baf
);
3622 baf
->detach_aio_context(baf
->opaque
);
3625 /* Never mind iterating again to check for ->deleted. bdrv_close() will
3626 * remove remaining aio notifiers if we aren't called again.
3628 bs
->walking_aio_notifiers
= false;
3630 if (bs
->drv
->bdrv_detach_aio_context
) {
3631 bs
->drv
->bdrv_detach_aio_context(bs
);
3633 QLIST_FOREACH(child
, &bs
->children
, next
) {
3634 bdrv_detach_aio_context(child
->bs
);
3637 bs
->aio_context
= NULL
;
3640 void bdrv_attach_aio_context(BlockDriverState
*bs
,
3641 AioContext
*new_context
)
3643 BdrvAioNotifier
*ban
, *ban_tmp
;
3650 bs
->aio_context
= new_context
;
3652 QLIST_FOREACH(child
, &bs
->children
, next
) {
3653 bdrv_attach_aio_context(child
->bs
, new_context
);
3655 if (bs
->drv
->bdrv_attach_aio_context
) {
3656 bs
->drv
->bdrv_attach_aio_context(bs
, new_context
);
3659 assert(!bs
->walking_aio_notifiers
);
3660 bs
->walking_aio_notifiers
= true;
3661 QLIST_FOREACH_SAFE(ban
, &bs
->aio_notifiers
, list
, ban_tmp
) {
3663 bdrv_do_remove_aio_context_notifier(ban
);
3665 ban
->attached_aio_context(new_context
, ban
->opaque
);
3668 bs
->walking_aio_notifiers
= false;
3671 void bdrv_set_aio_context(BlockDriverState
*bs
, AioContext
*new_context
)
3673 bdrv_drain(bs
); /* ensure there are no in-flight requests */
3675 bdrv_detach_aio_context(bs
);
3677 /* This function executes in the old AioContext so acquire the new one in
3678 * case it runs in a different thread.
3680 aio_context_acquire(new_context
);
3681 bdrv_attach_aio_context(bs
, new_context
);
3682 aio_context_release(new_context
);
3685 void bdrv_add_aio_context_notifier(BlockDriverState
*bs
,
3686 void (*attached_aio_context
)(AioContext
*new_context
, void *opaque
),
3687 void (*detach_aio_context
)(void *opaque
), void *opaque
)
3689 BdrvAioNotifier
*ban
= g_new(BdrvAioNotifier
, 1);
3690 *ban
= (BdrvAioNotifier
){
3691 .attached_aio_context
= attached_aio_context
,
3692 .detach_aio_context
= detach_aio_context
,
3696 QLIST_INSERT_HEAD(&bs
->aio_notifiers
, ban
, list
);
3699 void bdrv_remove_aio_context_notifier(BlockDriverState
*bs
,
3700 void (*attached_aio_context
)(AioContext
*,
3702 void (*detach_aio_context
)(void *),
3705 BdrvAioNotifier
*ban
, *ban_next
;
3707 QLIST_FOREACH_SAFE(ban
, &bs
->aio_notifiers
, list
, ban_next
) {
3708 if (ban
->attached_aio_context
== attached_aio_context
&&
3709 ban
->detach_aio_context
== detach_aio_context
&&
3710 ban
->opaque
== opaque
&&
3711 ban
->deleted
== false)
3713 if (bs
->walking_aio_notifiers
) {
3714 ban
->deleted
= true;
3716 bdrv_do_remove_aio_context_notifier(ban
);
3725 int bdrv_amend_options(BlockDriverState
*bs
, QemuOpts
*opts
,
3726 BlockDriverAmendStatusCB
*status_cb
, void *cb_opaque
)
3728 if (!bs
->drv
->bdrv_amend_options
) {
3731 return bs
->drv
->bdrv_amend_options(bs
, opts
, status_cb
, cb_opaque
);
3734 /* This function will be called by the bdrv_recurse_is_first_non_filter method
3735 * of block filter and by bdrv_is_first_non_filter.
3736 * It is used to test if the given bs is the candidate or recurse more in the
3739 bool bdrv_recurse_is_first_non_filter(BlockDriverState
*bs
,
3740 BlockDriverState
*candidate
)
3742 /* return false if basic checks fails */
3743 if (!bs
|| !bs
->drv
) {
3747 /* the code reached a non block filter driver -> check if the bs is
3748 * the same as the candidate. It's the recursion termination condition.
3750 if (!bs
->drv
->is_filter
) {
3751 return bs
== candidate
;
3753 /* Down this path the driver is a block filter driver */
3755 /* If the block filter recursion method is defined use it to recurse down
3758 if (bs
->drv
->bdrv_recurse_is_first_non_filter
) {
3759 return bs
->drv
->bdrv_recurse_is_first_non_filter(bs
, candidate
);
3762 /* the driver is a block filter but don't allow to recurse -> return false
3767 /* This function checks if the candidate is the first non filter bs down it's
3768 * bs chain. Since we don't have pointers to parents it explore all bs chains
3769 * from the top. Some filters can choose not to pass down the recursion.
3771 bool bdrv_is_first_non_filter(BlockDriverState
*candidate
)
3773 BlockDriverState
*bs
;
3774 BdrvNextIterator it
;
3776 /* walk down the bs forest recursively */
3777 for (bs
= bdrv_first(&it
); bs
; bs
= bdrv_next(&it
)) {
3780 /* try to recurse in this top level bs */
3781 perm
= bdrv_recurse_is_first_non_filter(bs
, candidate
);
3783 /* candidate is the first non filter */
3792 BlockDriverState
*check_to_replace_node(BlockDriverState
*parent_bs
,
3793 const char *node_name
, Error
**errp
)
3795 BlockDriverState
*to_replace_bs
= bdrv_find_node(node_name
);
3796 AioContext
*aio_context
;
3798 if (!to_replace_bs
) {
3799 error_setg(errp
, "Node name '%s' not found", node_name
);
3803 aio_context
= bdrv_get_aio_context(to_replace_bs
);
3804 aio_context_acquire(aio_context
);
3806 if (bdrv_op_is_blocked(to_replace_bs
, BLOCK_OP_TYPE_REPLACE
, errp
)) {
3807 to_replace_bs
= NULL
;
3811 /* We don't want arbitrary node of the BDS chain to be replaced only the top
3812 * most non filter in order to prevent data corruption.
3813 * Another benefit is that this tests exclude backing files which are
3814 * blocked by the backing blockers.
3816 if (!bdrv_recurse_is_first_non_filter(parent_bs
, to_replace_bs
)) {
3817 error_setg(errp
, "Only top most non filter can be replaced");
3818 to_replace_bs
= NULL
;
3823 aio_context_release(aio_context
);
3824 return to_replace_bs
;
3827 static bool append_open_options(QDict
*d
, BlockDriverState
*bs
)
3829 const QDictEntry
*entry
;
3832 bool found_any
= false;
3835 for (entry
= qdict_first(bs
->options
); entry
;
3836 entry
= qdict_next(bs
->options
, entry
))
3838 /* Exclude options for children */
3839 QLIST_FOREACH(child
, &bs
->children
, next
) {
3840 if (strstart(qdict_entry_key(entry
), child
->name
, &p
)
3841 && (!*p
|| *p
== '.'))
3850 /* And exclude all non-driver-specific options */
3851 for (desc
= bdrv_runtime_opts
.desc
; desc
->name
; desc
++) {
3852 if (!strcmp(qdict_entry_key(entry
), desc
->name
)) {
3860 qobject_incref(qdict_entry_value(entry
));
3861 qdict_put_obj(d
, qdict_entry_key(entry
), qdict_entry_value(entry
));
3868 /* Updates the following BDS fields:
3869 * - exact_filename: A filename which may be used for opening a block device
3870 * which (mostly) equals the given BDS (even without any
3871 * other options; so reading and writing must return the same
3872 * results, but caching etc. may be different)
3873 * - full_open_options: Options which, when given when opening a block device
3874 * (without a filename), result in a BDS (mostly)
3875 * equalling the given one
3876 * - filename: If exact_filename is set, it is copied here. Otherwise,
3877 * full_open_options is converted to a JSON object, prefixed with
3878 * "json:" (for use through the JSON pseudo protocol) and put here.
3880 void bdrv_refresh_filename(BlockDriverState
*bs
)
3882 BlockDriver
*drv
= bs
->drv
;
3889 /* This BDS's file name will most probably depend on its file's name, so
3890 * refresh that first */
3892 bdrv_refresh_filename(bs
->file
->bs
);
3895 if (drv
->bdrv_refresh_filename
) {
3896 /* Obsolete information is of no use here, so drop the old file name
3897 * information before refreshing it */
3898 bs
->exact_filename
[0] = '\0';
3899 if (bs
->full_open_options
) {
3900 QDECREF(bs
->full_open_options
);
3901 bs
->full_open_options
= NULL
;
3905 append_open_options(opts
, bs
);
3906 drv
->bdrv_refresh_filename(bs
, opts
);
3908 } else if (bs
->file
) {
3909 /* Try to reconstruct valid information from the underlying file */
3910 bool has_open_options
;
3912 bs
->exact_filename
[0] = '\0';
3913 if (bs
->full_open_options
) {
3914 QDECREF(bs
->full_open_options
);
3915 bs
->full_open_options
= NULL
;
3919 has_open_options
= append_open_options(opts
, bs
);
3921 /* If no specific options have been given for this BDS, the filename of
3922 * the underlying file should suffice for this one as well */
3923 if (bs
->file
->bs
->exact_filename
[0] && !has_open_options
) {
3924 strcpy(bs
->exact_filename
, bs
->file
->bs
->exact_filename
);
3926 /* Reconstructing the full options QDict is simple for most format block
3927 * drivers, as long as the full options are known for the underlying
3928 * file BDS. The full options QDict of that file BDS should somehow
3929 * contain a representation of the filename, therefore the following
3930 * suffices without querying the (exact_)filename of this BDS. */
3931 if (bs
->file
->bs
->full_open_options
) {
3932 qdict_put_obj(opts
, "driver",
3933 QOBJECT(qstring_from_str(drv
->format_name
)));
3934 QINCREF(bs
->file
->bs
->full_open_options
);
3935 qdict_put_obj(opts
, "file",
3936 QOBJECT(bs
->file
->bs
->full_open_options
));
3938 bs
->full_open_options
= opts
;
3942 } else if (!bs
->full_open_options
&& qdict_size(bs
->options
)) {
3943 /* There is no underlying file BDS (at least referenced by BDS.file),
3944 * so the full options QDict should be equal to the options given
3945 * specifically for this block device when it was opened (plus the
3946 * driver specification).
3947 * Because those options don't change, there is no need to update
3948 * full_open_options when it's already set. */
3951 append_open_options(opts
, bs
);
3952 qdict_put_obj(opts
, "driver",
3953 QOBJECT(qstring_from_str(drv
->format_name
)));
3955 if (bs
->exact_filename
[0]) {
3956 /* This may not work for all block protocol drivers (some may
3957 * require this filename to be parsed), but we have to find some
3958 * default solution here, so just include it. If some block driver
3959 * does not support pure options without any filename at all or
3960 * needs some special format of the options QDict, it needs to
3961 * implement the driver-specific bdrv_refresh_filename() function.
3963 qdict_put_obj(opts
, "filename",
3964 QOBJECT(qstring_from_str(bs
->exact_filename
)));
3967 bs
->full_open_options
= opts
;
3970 if (bs
->exact_filename
[0]) {
3971 pstrcpy(bs
->filename
, sizeof(bs
->filename
), bs
->exact_filename
);
3972 } else if (bs
->full_open_options
) {
3973 QString
*json
= qobject_to_json(QOBJECT(bs
->full_open_options
));
3974 snprintf(bs
->filename
, sizeof(bs
->filename
), "json:%s",
3975 qstring_get_str(json
));
3981 * Hot add/remove a BDS's child. So the user can take a child offline when
3982 * it is broken and take a new child online
3984 void bdrv_add_child(BlockDriverState
*parent_bs
, BlockDriverState
*child_bs
,
3988 if (!parent_bs
->drv
|| !parent_bs
->drv
->bdrv_add_child
) {
3989 error_setg(errp
, "The node %s does not support adding a child",
3990 bdrv_get_device_or_node_name(parent_bs
));
3994 if (!QLIST_EMPTY(&child_bs
->parents
)) {
3995 error_setg(errp
, "The node %s already has a parent",
3996 child_bs
->node_name
);
4000 parent_bs
->drv
->bdrv_add_child(parent_bs
, child_bs
, errp
);
4003 void bdrv_del_child(BlockDriverState
*parent_bs
, BdrvChild
*child
, Error
**errp
)
4007 if (!parent_bs
->drv
|| !parent_bs
->drv
->bdrv_del_child
) {
4008 error_setg(errp
, "The node %s does not support removing a child",
4009 bdrv_get_device_or_node_name(parent_bs
));
4013 QLIST_FOREACH(tmp
, &parent_bs
->children
, next
) {
4020 error_setg(errp
, "The node %s does not have a child named %s",
4021 bdrv_get_device_or_node_name(parent_bs
),
4022 bdrv_get_device_or_node_name(child
->bs
));
4026 parent_bs
->drv
->bdrv_del_child(parent_bs
, child
, errp
);