2 * QEMU host block devices
4 * Copyright (c) 2003-2008 Fabrice Bellard
6 * This work is licensed under the terms of the GNU GPL, version 2 or
7 * later. See the COPYING file in the top-level directory.
9 * This file incorporates work covered by the following copyright and
12 * Copyright (c) 2003-2008 Fabrice Bellard
14 * Permission is hereby granted, free of charge, to any person obtaining a copy
15 * of this software and associated documentation files (the "Software"), to deal
16 * in the Software without restriction, including without limitation the rights
17 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
18 * copies of the Software, and to permit persons to whom the Software is
19 * furnished to do so, subject to the following conditions:
21 * The above copyright notice and this permission notice shall be included in
22 * all copies or substantial portions of the Software.
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
27 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
33 #include "qemu/osdep.h"
34 #include "sysemu/block-backend.h"
35 #include "sysemu/blockdev.h"
36 #include "hw/block/block.h"
37 #include "block/blockjob.h"
38 #include "block/qdict.h"
39 #include "block/throttle-groups.h"
40 #include "monitor/monitor.h"
41 #include "qemu/error-report.h"
42 #include "qemu/option.h"
43 #include "qemu/qemu-print.h"
44 #include "qemu/config-file.h"
45 #include "qapi/qapi-commands-block.h"
46 #include "qapi/qapi-commands-transaction.h"
47 #include "qapi/qapi-visit-block-core.h"
48 #include "qapi/qmp/qdict.h"
49 #include "qapi/qmp/qnum.h"
50 #include "qapi/qmp/qstring.h"
51 #include "qapi/error.h"
52 #include "qapi/qmp/qerror.h"
53 #include "qapi/qmp/qlist.h"
54 #include "qapi/qobject-output-visitor.h"
55 #include "sysemu/sysemu.h"
56 #include "sysemu/iothread.h"
57 #include "block/block_int.h"
58 #include "block/trace.h"
59 #include "sysemu/arch_init.h"
60 #include "sysemu/qtest.h"
61 #include "sysemu/runstate.h"
62 #include "sysemu/replay.h"
63 #include "qemu/cutils.h"
64 #include "qemu/help_option.h"
65 #include "qemu/main-loop.h"
66 #include "qemu/throttle-options.h"
68 QTAILQ_HEAD(, BlockDriverState
) monitor_bdrv_states
=
69 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states
);
71 void bdrv_set_monitor_owned(BlockDriverState
*bs
)
73 QTAILQ_INSERT_TAIL(&monitor_bdrv_states
, bs
, monitor_list
);
76 static const char *const if_name
[IF_COUNT
] = {
80 [IF_FLOPPY
] = "floppy",
81 [IF_PFLASH
] = "pflash",
84 [IF_VIRTIO
] = "virtio",
88 static int if_max_devs
[IF_COUNT
] = {
90 * Do not change these numbers! They govern how drive option
91 * index maps to unit and bus. That mapping is ABI.
93 * All controllers used to implement if=T drives need to support
94 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
95 * Otherwise, some index values map to "impossible" bus, unit
98 * For instance, if you change [IF_SCSI] to 255, -drive
99 * if=scsi,index=12 no longer means bus=1,unit=5, but
100 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
101 * the drive can't be set up. Regression.
108 * Boards may call this to offer board-by-board overrides
109 * of the default, global values.
111 void override_max_devs(BlockInterfaceType type
, int max_devs
)
120 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
121 dinfo
= blk_legacy_dinfo(blk
);
122 if (dinfo
->type
== type
) {
123 fprintf(stderr
, "Cannot override units-per-bus property of"
124 " the %s interface, because a drive of that type has"
125 " already been added.\n", if_name
[type
]);
126 g_assert_not_reached();
130 if_max_devs
[type
] = max_devs
;
134 * We automatically delete the drive when a device using it gets
135 * unplugged. Questionable feature, but we can't just drop it.
136 * Device models call blockdev_mark_auto_del() to schedule the
137 * automatic deletion, and generic qdev code calls blockdev_auto_del()
138 * when deletion is actually safe.
140 void blockdev_mark_auto_del(BlockBackend
*blk
)
142 DriveInfo
*dinfo
= blk_legacy_dinfo(blk
);
149 for (job
= block_job_next(NULL
); job
; job
= block_job_next(job
)) {
150 if (block_job_has_bdrv(job
, blk_bs(blk
))) {
151 AioContext
*aio_context
= job
->job
.aio_context
;
152 aio_context_acquire(aio_context
);
154 job_cancel(&job
->job
, false);
156 aio_context_release(aio_context
);
163 void blockdev_auto_del(BlockBackend
*blk
)
165 DriveInfo
*dinfo
= blk_legacy_dinfo(blk
);
167 if (dinfo
&& dinfo
->auto_del
) {
168 monitor_remove_blk(blk
);
174 * Returns the current mapping of how many units per bus
175 * a particular interface can support.
177 * A positive integer indicates n units per bus.
178 * 0 implies the mapping has not been established.
179 * -1 indicates an invalid BlockInterfaceType was given.
181 int drive_get_max_devs(BlockInterfaceType type
)
183 if (type
>= IF_IDE
&& type
< IF_COUNT
) {
184 return if_max_devs
[type
];
190 static int drive_index_to_bus_id(BlockInterfaceType type
, int index
)
192 int max_devs
= if_max_devs
[type
];
193 return max_devs
? index
/ max_devs
: 0;
196 static int drive_index_to_unit_id(BlockInterfaceType type
, int index
)
198 int max_devs
= if_max_devs
[type
];
199 return max_devs
? index
% max_devs
: index
;
202 QemuOpts
*drive_def(const char *optstr
)
204 return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr
, false);
207 QemuOpts
*drive_add(BlockInterfaceType type
, int index
, const char *file
,
212 opts
= drive_def(optstr
);
216 if (type
!= IF_DEFAULT
) {
217 qemu_opt_set(opts
, "if", if_name
[type
], &error_abort
);
220 qemu_opt_set_number(opts
, "index", index
, &error_abort
);
223 qemu_opt_set(opts
, "file", file
, &error_abort
);
227 DriveInfo
*drive_get(BlockInterfaceType type
, int bus
, int unit
)
232 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
233 dinfo
= blk_legacy_dinfo(blk
);
234 if (dinfo
&& dinfo
->type
== type
235 && dinfo
->bus
== bus
&& dinfo
->unit
== unit
) {
243 void drive_mark_claimed_by_board(void)
248 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
249 dinfo
= blk_legacy_dinfo(blk
);
250 if (dinfo
&& blk_get_attached_dev(blk
)) {
251 dinfo
->claimed_by_board
= true;
256 void drive_check_orphaned(void)
261 bool orphans
= false;
263 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
264 dinfo
= blk_legacy_dinfo(blk
);
265 if (dinfo
->is_default
|| dinfo
->type
== IF_NONE
) {
268 if (!blk_get_attached_dev(blk
)) {
270 qemu_opts_loc_restore(dinfo
->opts
);
271 error_report("machine type does not support"
272 " if=%s,bus=%d,unit=%d",
273 if_name
[dinfo
->type
], dinfo
->bus
, dinfo
->unit
);
278 if (!dinfo
->claimed_by_board
&& dinfo
->type
!= IF_VIRTIO
) {
280 qemu_opts_loc_restore(dinfo
->opts
);
281 warn_report("bogus if=%s is deprecated, use if=none",
282 if_name
[dinfo
->type
]);
292 DriveInfo
*drive_get_by_index(BlockInterfaceType type
, int index
)
294 return drive_get(type
,
295 drive_index_to_bus_id(type
, index
),
296 drive_index_to_unit_id(type
, index
));
299 int drive_get_max_bus(BlockInterfaceType type
)
306 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
307 dinfo
= blk_legacy_dinfo(blk
);
308 if (dinfo
&& dinfo
->type
== type
&& dinfo
->bus
> max_bus
) {
309 max_bus
= dinfo
->bus
;
315 /* Get a block device. This should only be used for single-drive devices
316 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
318 DriveInfo
*drive_get_next(BlockInterfaceType type
)
320 static int next_block_unit
[IF_COUNT
];
322 return drive_get(type
, 0, next_block_unit
[type
]++);
325 static void bdrv_format_print(void *opaque
, const char *name
)
327 qemu_printf(" %s", name
);
332 BlockDriverState
*bs
;
335 static int parse_block_error_action(const char *buf
, bool is_read
, Error
**errp
)
337 if (!strcmp(buf
, "ignore")) {
338 return BLOCKDEV_ON_ERROR_IGNORE
;
339 } else if (!is_read
&& !strcmp(buf
, "enospc")) {
340 return BLOCKDEV_ON_ERROR_ENOSPC
;
341 } else if (!strcmp(buf
, "stop")) {
342 return BLOCKDEV_ON_ERROR_STOP
;
343 } else if (!strcmp(buf
, "report")) {
344 return BLOCKDEV_ON_ERROR_REPORT
;
346 error_setg(errp
, "'%s' invalid %s error action",
347 buf
, is_read
? "read" : "write");
352 static bool parse_stats_intervals(BlockAcctStats
*stats
, QList
*intervals
,
355 const QListEntry
*entry
;
356 for (entry
= qlist_first(intervals
); entry
; entry
= qlist_next(entry
)) {
357 switch (qobject_type(entry
->value
)) {
359 case QTYPE_QSTRING
: {
360 unsigned long long length
;
361 const char *str
= qstring_get_str(qobject_to(QString
,
363 if (parse_uint_full(str
, &length
, 10) == 0 &&
364 length
> 0 && length
<= UINT_MAX
) {
365 block_acct_add_interval(stats
, (unsigned) length
);
367 error_setg(errp
, "Invalid interval length: %s", str
);
374 int64_t length
= qnum_get_int(qobject_to(QNum
, entry
->value
));
376 if (length
> 0 && length
<= UINT_MAX
) {
377 block_acct_add_interval(stats
, (unsigned) length
);
379 error_setg(errp
, "Invalid interval length: %" PRId64
, length
);
386 error_setg(errp
, "The specification of stats-intervals is invalid");
393 typedef enum { MEDIA_DISK
, MEDIA_CDROM
} DriveMediaType
;
395 /* All parameters but @opts are optional and may be set to NULL. */
396 static void extract_common_blockdev_options(QemuOpts
*opts
, int *bdrv_flags
,
397 const char **throttling_group
, ThrottleConfig
*throttle_cfg
,
398 BlockdevDetectZeroesOptions
*detect_zeroes
, Error
**errp
)
400 Error
*local_error
= NULL
;
404 if (qemu_opt_get_bool(opts
, "copy-on-read", false)) {
405 *bdrv_flags
|= BDRV_O_COPY_ON_READ
;
408 if ((aio
= qemu_opt_get(opts
, "aio")) != NULL
) {
409 if (bdrv_parse_aio(aio
, bdrv_flags
) < 0) {
410 error_setg(errp
, "invalid aio option");
416 /* disk I/O throttling */
417 if (throttling_group
) {
418 *throttling_group
= qemu_opt_get(opts
, "throttling.group");
422 throttle_config_init(throttle_cfg
);
423 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].avg
=
424 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
425 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].avg
=
426 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
427 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].avg
=
428 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
429 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].avg
=
430 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
431 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].avg
=
432 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
433 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].avg
=
434 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
436 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].max
=
437 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
438 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].max
=
439 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
440 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].max
=
441 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
442 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].max
=
443 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
444 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].max
=
445 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
446 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].max
=
447 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
449 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].burst_length
=
450 qemu_opt_get_number(opts
, "throttling.bps-total-max-length", 1);
451 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].burst_length
=
452 qemu_opt_get_number(opts
, "throttling.bps-read-max-length", 1);
453 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].burst_length
=
454 qemu_opt_get_number(opts
, "throttling.bps-write-max-length", 1);
455 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].burst_length
=
456 qemu_opt_get_number(opts
, "throttling.iops-total-max-length", 1);
457 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].burst_length
=
458 qemu_opt_get_number(opts
, "throttling.iops-read-max-length", 1);
459 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].burst_length
=
460 qemu_opt_get_number(opts
, "throttling.iops-write-max-length", 1);
462 throttle_cfg
->op_size
=
463 qemu_opt_get_number(opts
, "throttling.iops-size", 0);
465 if (!throttle_is_valid(throttle_cfg
, errp
)) {
472 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup
,
473 qemu_opt_get(opts
, "detect-zeroes"),
474 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
477 error_propagate(errp
, local_error
);
483 /* Takes the ownership of bs_opts */
484 static BlockBackend
*blockdev_init(const char *file
, QDict
*bs_opts
,
489 int on_read_error
, on_write_error
;
490 bool account_invalid
, account_failed
;
491 bool writethrough
, read_only
;
493 BlockDriverState
*bs
;
498 QDict
*interval_dict
= NULL
;
499 QList
*interval_list
= NULL
;
501 BlockdevDetectZeroesOptions detect_zeroes
=
502 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
503 const char *throttling_group
= NULL
;
505 /* Check common options by copying from bs_opts to opts, all other options
506 * stay in bs_opts for processing by bdrv_open(). */
507 id
= qdict_get_try_str(bs_opts
, "id");
508 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, errp
);
513 if (!qemu_opts_absorb_qdict(opts
, bs_opts
, errp
)) {
518 qdict_del(bs_opts
, "id");
521 /* extract parameters */
522 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
524 account_invalid
= qemu_opt_get_bool(opts
, "stats-account-invalid", true);
525 account_failed
= qemu_opt_get_bool(opts
, "stats-account-failed", true);
527 writethrough
= !qemu_opt_get_bool(opts
, BDRV_OPT_CACHE_WB
, true);
529 id
= qemu_opts_id(opts
);
531 qdict_extract_subqdict(bs_opts
, &interval_dict
, "stats-intervals.");
532 qdict_array_split(interval_dict
, &interval_list
);
534 if (qdict_size(interval_dict
) != 0) {
535 error_setg(errp
, "Invalid option stats-intervals.%s",
536 qdict_first(interval_dict
)->key
);
540 extract_common_blockdev_options(opts
, &bdrv_flags
, &throttling_group
, &cfg
,
541 &detect_zeroes
, &error
);
543 error_propagate(errp
, error
);
547 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
548 if (is_help_option(buf
)) {
549 qemu_printf("Supported formats:");
550 bdrv_iterate_format(bdrv_format_print
, NULL
, false);
551 qemu_printf("\nSupported formats (read-only):");
552 bdrv_iterate_format(bdrv_format_print
, NULL
, true);
557 if (qdict_haskey(bs_opts
, "driver")) {
558 error_setg(errp
, "Cannot specify both 'driver' and 'format'");
561 qdict_put_str(bs_opts
, "driver", buf
);
564 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
565 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
566 on_write_error
= parse_block_error_action(buf
, 0, &error
);
568 error_propagate(errp
, error
);
573 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
574 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
575 on_read_error
= parse_block_error_action(buf
, 1, &error
);
577 error_propagate(errp
, error
);
583 bdrv_flags
|= BDRV_O_SNAPSHOT
;
586 read_only
= qemu_opt_get_bool(opts
, BDRV_OPT_READ_ONLY
, false);
589 if ((!file
|| !*file
) && !qdict_size(bs_opts
)) {
590 BlockBackendRootState
*blk_rs
;
592 blk
= blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL
);
593 blk_rs
= blk_get_root_state(blk
);
594 blk_rs
->open_flags
= bdrv_flags
;
595 blk_rs
->read_only
= read_only
;
596 blk_rs
->detect_zeroes
= detect_zeroes
;
598 qobject_unref(bs_opts
);
600 if (file
&& !*file
) {
604 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
605 * with other callers) rather than what we want as the real defaults.
606 * Apply the defaults here instead. */
607 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_DIRECT
, "off");
608 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_NO_FLUSH
, "off");
609 qdict_set_default_str(bs_opts
, BDRV_OPT_READ_ONLY
,
610 read_only
? "on" : "off");
611 qdict_set_default_str(bs_opts
, BDRV_OPT_AUTO_READ_ONLY
, "on");
612 assert((bdrv_flags
& BDRV_O_CACHE_MASK
) == 0);
614 if (runstate_check(RUN_STATE_INMIGRATE
)) {
615 bdrv_flags
|= BDRV_O_INACTIVE
;
618 blk
= blk_new_open(file
, NULL
, bs_opts
, bdrv_flags
, errp
);
624 bs
->detect_zeroes
= detect_zeroes
;
626 block_acct_setup(blk_get_stats(blk
), account_invalid
, account_failed
);
628 if (!parse_stats_intervals(blk_get_stats(blk
), interval_list
, errp
)) {
635 /* disk I/O throttling */
636 if (throttle_enabled(&cfg
)) {
637 if (!throttling_group
) {
638 throttling_group
= id
;
640 blk_io_limits_enable(blk
, throttling_group
);
641 blk_set_io_limits(blk
, &cfg
);
644 blk_set_enable_write_cache(blk
, !writethrough
);
645 blk_set_on_error(blk
, on_read_error
, on_write_error
);
647 if (!monitor_add_blk(blk
, id
, errp
)) {
655 qobject_unref(interval_dict
);
656 qobject_unref(interval_list
);
661 qobject_unref(interval_dict
);
662 qobject_unref(interval_list
);
664 qobject_unref(bs_opts
);
668 /* Takes the ownership of bs_opts */
669 BlockDriverState
*bds_tree_init(QDict
*bs_opts
, Error
**errp
)
673 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
674 * with other callers) rather than what we want as the real defaults.
675 * Apply the defaults here instead. */
676 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_DIRECT
, "off");
677 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_NO_FLUSH
, "off");
678 qdict_set_default_str(bs_opts
, BDRV_OPT_READ_ONLY
, "off");
680 if (runstate_check(RUN_STATE_INMIGRATE
)) {
681 bdrv_flags
|= BDRV_O_INACTIVE
;
684 return bdrv_open(NULL
, NULL
, bs_opts
, bdrv_flags
, errp
);
687 void blockdev_close_all_bdrv_states(void)
689 BlockDriverState
*bs
, *next_bs
;
691 QTAILQ_FOREACH_SAFE(bs
, &monitor_bdrv_states
, monitor_list
, next_bs
) {
692 AioContext
*ctx
= bdrv_get_aio_context(bs
);
694 aio_context_acquire(ctx
);
696 aio_context_release(ctx
);
700 /* Iterates over the list of monitor-owned BlockDriverStates */
701 BlockDriverState
*bdrv_next_monitor_owned(BlockDriverState
*bs
)
703 return bs
? QTAILQ_NEXT(bs
, monitor_list
)
704 : QTAILQ_FIRST(&monitor_bdrv_states
);
707 static bool qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
,
712 value
= qemu_opt_get(opts
, from
);
714 if (qemu_opt_find(opts
, to
)) {
715 error_setg(errp
, "'%s' and its alias '%s' can't be used at the "
716 "same time", to
, from
);
721 /* rename all items in opts */
722 while ((value
= qemu_opt_get(opts
, from
))) {
723 qemu_opt_set(opts
, to
, value
, &error_abort
);
724 qemu_opt_unset(opts
, from
);
729 QemuOptsList qemu_legacy_drive_opts
= {
731 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
735 .type
= QEMU_OPT_NUMBER
,
736 .help
= "bus number",
739 .type
= QEMU_OPT_NUMBER
,
740 .help
= "unit number (i.e. lun for scsi)",
743 .type
= QEMU_OPT_NUMBER
,
744 .help
= "index number",
747 .type
= QEMU_OPT_STRING
,
748 .help
= "media type (disk, cdrom)",
751 .type
= QEMU_OPT_STRING
,
752 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
755 .type
= QEMU_OPT_STRING
,
759 /* Options that are passed on, but have special semantics with -drive */
761 .name
= BDRV_OPT_READ_ONLY
,
762 .type
= QEMU_OPT_BOOL
,
763 .help
= "open drive file as read-only",
766 .type
= QEMU_OPT_STRING
,
767 .help
= "read error action",
770 .type
= QEMU_OPT_STRING
,
771 .help
= "write error action",
773 .name
= "copy-on-read",
774 .type
= QEMU_OPT_BOOL
,
775 .help
= "copy read data from backing file into image file",
778 { /* end of list */ }
782 DriveInfo
*drive_new(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
,
787 DriveInfo
*dinfo
= NULL
;
789 QemuOpts
*legacy_opts
;
790 DriveMediaType media
= MEDIA_DISK
;
791 BlockInterfaceType type
;
792 int max_devs
, bus_id
, unit_id
, index
;
793 const char *werror
, *rerror
;
794 bool read_only
= false;
796 const char *filename
;
799 /* Change legacy command line options into QMP ones */
800 static const struct {
804 { "iops", "throttling.iops-total" },
805 { "iops_rd", "throttling.iops-read" },
806 { "iops_wr", "throttling.iops-write" },
808 { "bps", "throttling.bps-total" },
809 { "bps_rd", "throttling.bps-read" },
810 { "bps_wr", "throttling.bps-write" },
812 { "iops_max", "throttling.iops-total-max" },
813 { "iops_rd_max", "throttling.iops-read-max" },
814 { "iops_wr_max", "throttling.iops-write-max" },
816 { "bps_max", "throttling.bps-total-max" },
817 { "bps_rd_max", "throttling.bps-read-max" },
818 { "bps_wr_max", "throttling.bps-write-max" },
820 { "iops_size", "throttling.iops-size" },
822 { "group", "throttling.group" },
824 { "readonly", BDRV_OPT_READ_ONLY
},
827 for (i
= 0; i
< ARRAY_SIZE(opt_renames
); i
++) {
828 if (!qemu_opt_rename(all_opts
, opt_renames
[i
].from
,
829 opt_renames
[i
].to
, errp
)) {
834 value
= qemu_opt_get(all_opts
, "cache");
839 if (bdrv_parse_cache_mode(value
, &flags
, &writethrough
) != 0) {
840 error_setg(errp
, "invalid cache option");
844 /* Specific options take precedence */
845 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_WB
)) {
846 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_WB
,
847 !writethrough
, &error_abort
);
849 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_DIRECT
)) {
850 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_DIRECT
,
851 !!(flags
& BDRV_O_NOCACHE
), &error_abort
);
853 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
)) {
854 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
,
855 !!(flags
& BDRV_O_NO_FLUSH
), &error_abort
);
857 qemu_opt_unset(all_opts
, "cache");
860 /* Get a QDict for processing the options */
861 bs_opts
= qdict_new();
862 qemu_opts_to_qdict(all_opts
, bs_opts
);
864 legacy_opts
= qemu_opts_create(&qemu_legacy_drive_opts
, NULL
, 0,
866 if (!qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, errp
)) {
871 value
= qemu_opt_get(legacy_opts
, "media");
873 if (!strcmp(value
, "disk")) {
875 } else if (!strcmp(value
, "cdrom")) {
879 error_setg(errp
, "'%s' invalid media", value
);
884 /* copy-on-read is disabled with a warning for read-only devices */
885 read_only
|= qemu_opt_get_bool(legacy_opts
, BDRV_OPT_READ_ONLY
, false);
886 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
888 if (read_only
&& copy_on_read
) {
889 warn_report("disabling copy-on-read on read-only drive");
890 copy_on_read
= false;
893 qdict_put_str(bs_opts
, BDRV_OPT_READ_ONLY
, read_only
? "on" : "off");
894 qdict_put_str(bs_opts
, "copy-on-read", copy_on_read
? "on" : "off");
896 /* Controller type */
897 value
= qemu_opt_get(legacy_opts
, "if");
900 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
903 if (type
== IF_COUNT
) {
904 error_setg(errp
, "unsupported bus type '%s'", value
);
908 type
= block_default_type
;
911 /* Device address specified by bus/unit or index.
912 * If none was specified, try to find the first free one. */
913 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
914 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
915 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
917 max_devs
= if_max_devs
[type
];
920 if (bus_id
!= 0 || unit_id
!= -1) {
921 error_setg(errp
, "index cannot be used with bus and unit");
924 bus_id
= drive_index_to_bus_id(type
, index
);
925 unit_id
= drive_index_to_unit_id(type
, index
);
930 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
932 if (max_devs
&& unit_id
>= max_devs
) {
939 if (max_devs
&& unit_id
>= max_devs
) {
940 error_setg(errp
, "unit %d too big (max is %d)", unit_id
, max_devs
- 1);
944 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
945 error_setg(errp
, "drive with bus=%d, unit=%d (index=%d) exists",
946 bus_id
, unit_id
, index
);
950 /* no id supplied -> create one */
951 if (qemu_opts_id(all_opts
) == NULL
) {
953 const char *mediastr
= "";
954 if (type
== IF_IDE
|| type
== IF_SCSI
) {
955 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
958 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
961 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
964 qdict_put_str(bs_opts
, "id", new_id
);
968 /* Add virtio block device */
969 if (type
== IF_VIRTIO
) {
971 devopts
= qemu_opts_create(qemu_find_opts("device"), NULL
, 0,
973 if (arch_type
== QEMU_ARCH_S390X
) {
974 qemu_opt_set(devopts
, "driver", "virtio-blk-ccw", &error_abort
);
976 qemu_opt_set(devopts
, "driver", "virtio-blk-pci", &error_abort
);
978 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"),
982 filename
= qemu_opt_get(legacy_opts
, "file");
984 /* Check werror/rerror compatibility with if=... */
985 werror
= qemu_opt_get(legacy_opts
, "werror");
986 if (werror
!= NULL
) {
987 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&&
989 error_setg(errp
, "werror is not supported by this bus type");
992 qdict_put_str(bs_opts
, "werror", werror
);
995 rerror
= qemu_opt_get(legacy_opts
, "rerror");
996 if (rerror
!= NULL
) {
997 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&&
999 error_setg(errp
, "rerror is not supported by this bus type");
1002 qdict_put_str(bs_opts
, "rerror", rerror
);
1005 /* Actual block device init: Functionality shared with blockdev-add */
1006 blk
= blockdev_init(filename
, bs_opts
, errp
);
1012 /* Create legacy DriveInfo */
1013 dinfo
= g_malloc0(sizeof(*dinfo
));
1014 dinfo
->opts
= all_opts
;
1017 dinfo
->bus
= bus_id
;
1018 dinfo
->unit
= unit_id
;
1020 blk_set_legacy_dinfo(blk
, dinfo
);
1027 dinfo
->media_cd
= media
== MEDIA_CDROM
;
1034 qemu_opts_del(legacy_opts
);
1035 qobject_unref(bs_opts
);
1039 static BlockDriverState
*qmp_get_root_bs(const char *name
, Error
**errp
)
1041 BlockDriverState
*bs
;
1043 bs
= bdrv_lookup_bs(name
, name
, errp
);
1048 if (!bdrv_is_root_node(bs
)) {
1049 error_setg(errp
, "Need a root block node");
1053 if (!bdrv_is_inserted(bs
)) {
1054 error_setg(errp
, "Device has no medium");
1061 static void blockdev_do_action(TransactionAction
*action
, Error
**errp
)
1063 TransactionActionList list
;
1065 list
.value
= action
;
1067 qmp_transaction(&list
, false, NULL
, errp
);
1070 void qmp_blockdev_snapshot_sync(bool has_device
, const char *device
,
1071 bool has_node_name
, const char *node_name
,
1072 const char *snapshot_file
,
1073 bool has_snapshot_node_name
,
1074 const char *snapshot_node_name
,
1075 bool has_format
, const char *format
,
1076 bool has_mode
, NewImageMode mode
, Error
**errp
)
1078 BlockdevSnapshotSync snapshot
= {
1079 .has_device
= has_device
,
1080 .device
= (char *) device
,
1081 .has_node_name
= has_node_name
,
1082 .node_name
= (char *) node_name
,
1083 .snapshot_file
= (char *) snapshot_file
,
1084 .has_snapshot_node_name
= has_snapshot_node_name
,
1085 .snapshot_node_name
= (char *) snapshot_node_name
,
1086 .has_format
= has_format
,
1087 .format
= (char *) format
,
1088 .has_mode
= has_mode
,
1091 TransactionAction action
= {
1092 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
1093 .u
.blockdev_snapshot_sync
.data
= &snapshot
,
1095 blockdev_do_action(&action
, errp
);
1098 void qmp_blockdev_snapshot(const char *node
, const char *overlay
,
1101 BlockdevSnapshot snapshot_data
= {
1102 .node
= (char *) node
,
1103 .overlay
= (char *) overlay
1105 TransactionAction action
= {
1106 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
,
1107 .u
.blockdev_snapshot
.data
= &snapshot_data
,
1109 blockdev_do_action(&action
, errp
);
1112 void qmp_blockdev_snapshot_internal_sync(const char *device
,
1116 BlockdevSnapshotInternal snapshot
= {
1117 .device
= (char *) device
,
1118 .name
= (char *) name
1120 TransactionAction action
= {
1121 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
1122 .u
.blockdev_snapshot_internal_sync
.data
= &snapshot
,
1124 blockdev_do_action(&action
, errp
);
1127 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
1134 BlockDriverState
*bs
;
1135 AioContext
*aio_context
;
1136 QEMUSnapshotInfo sn
;
1137 Error
*local_err
= NULL
;
1138 SnapshotInfo
*info
= NULL
;
1141 bs
= qmp_get_root_bs(device
, errp
);
1145 aio_context
= bdrv_get_aio_context(bs
);
1146 aio_context_acquire(aio_context
);
1157 error_setg(errp
, "Name or id must be provided");
1158 goto out_aio_context
;
1161 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE
, errp
)) {
1162 goto out_aio_context
;
1165 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1167 error_propagate(errp
, local_err
);
1168 goto out_aio_context
;
1172 "Snapshot with id '%s' and name '%s' does not exist on "
1174 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1175 goto out_aio_context
;
1178 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1180 error_propagate(errp
, local_err
);
1181 goto out_aio_context
;
1184 aio_context_release(aio_context
);
1186 info
= g_new0(SnapshotInfo
, 1);
1187 info
->id
= g_strdup(sn
.id_str
);
1188 info
->name
= g_strdup(sn
.name
);
1189 info
->date_nsec
= sn
.date_nsec
;
1190 info
->date_sec
= sn
.date_sec
;
1191 info
->vm_state_size
= sn
.vm_state_size
;
1192 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1193 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1194 if (sn
.icount
!= -1ULL) {
1195 info
->icount
= sn
.icount
;
1196 info
->has_icount
= true;
1202 aio_context_release(aio_context
);
1206 /* New and old BlockDriverState structs for atomic group operations */
1208 typedef struct BlkActionState BlkActionState
;
1212 * Table of operations that define an Action.
1214 * @instance_size: Size of state struct, in bytes.
1215 * @prepare: Prepare the work, must NOT be NULL.
1216 * @commit: Commit the changes, can be NULL.
1217 * @abort: Abort the changes on fail, can be NULL.
1218 * @clean: Clean up resources after all transaction actions have called
1219 * commit() or abort(). Can be NULL.
1221 * Only prepare() may fail. In a single transaction, only one of commit() or
1222 * abort() will be called. clean() will always be called if it is present.
1224 typedef struct BlkActionOps
{
1225 size_t instance_size
;
1226 void (*prepare
)(BlkActionState
*common
, Error
**errp
);
1227 void (*commit
)(BlkActionState
*common
);
1228 void (*abort
)(BlkActionState
*common
);
1229 void (*clean
)(BlkActionState
*common
);
1234 * Describes one Action's state within a Transaction.
1236 * @action: QAPI-defined enum identifying which Action to perform.
1237 * @ops: Table of ActionOps this Action can perform.
1238 * @block_job_txn: Transaction which this action belongs to.
1239 * @entry: List membership for all Actions in this Transaction.
1241 * This structure must be arranged as first member in a subclassed type,
1242 * assuming that the compiler will also arrange it to the same offsets as the
1245 struct BlkActionState
{
1246 TransactionAction
*action
;
1247 const BlkActionOps
*ops
;
1248 JobTxn
*block_job_txn
;
1249 TransactionProperties
*txn_props
;
1250 QTAILQ_ENTRY(BlkActionState
) entry
;
1253 /* internal snapshot private data */
1254 typedef struct InternalSnapshotState
{
1255 BlkActionState common
;
1256 BlockDriverState
*bs
;
1257 QEMUSnapshotInfo sn
;
1259 } InternalSnapshotState
;
1262 static int action_check_completion_mode(BlkActionState
*s
, Error
**errp
)
1264 if (s
->txn_props
->completion_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
1266 "Action '%s' does not support Transaction property "
1267 "completion-mode = %s",
1268 TransactionActionKind_str(s
->action
->type
),
1269 ActionCompletionMode_str(s
->txn_props
->completion_mode
));
1275 static void internal_snapshot_prepare(BlkActionState
*common
,
1278 Error
*local_err
= NULL
;
1281 BlockDriverState
*bs
;
1282 QEMUSnapshotInfo old_sn
, *sn
;
1285 BlockdevSnapshotInternal
*internal
;
1286 InternalSnapshotState
*state
;
1287 AioContext
*aio_context
;
1290 g_assert(common
->action
->type
==
1291 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1292 internal
= common
->action
->u
.blockdev_snapshot_internal_sync
.data
;
1293 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1295 /* 1. parse input */
1296 device
= internal
->device
;
1297 name
= internal
->name
;
1299 /* 2. check for validation */
1300 if (action_check_completion_mode(common
, errp
) < 0) {
1304 bs
= qmp_get_root_bs(device
, errp
);
1309 aio_context
= bdrv_get_aio_context(bs
);
1310 aio_context_acquire(aio_context
);
1314 /* Paired with .clean() */
1315 bdrv_drained_begin(bs
);
1317 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT
, errp
)) {
1321 if (bdrv_is_read_only(bs
)) {
1322 error_setg(errp
, "Device '%s' is read only", device
);
1326 if (!bdrv_can_snapshot(bs
)) {
1327 error_setg(errp
, "Block format '%s' used by device '%s' "
1328 "does not support internal snapshots",
1329 bs
->drv
->format_name
, device
);
1333 if (!strlen(name
)) {
1334 error_setg(errp
, "Name is empty");
1338 /* check whether a snapshot with name exist */
1339 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
,
1342 error_propagate(errp
, local_err
);
1346 "Snapshot with name '%s' already exists on device '%s'",
1351 /* 3. take the snapshot */
1353 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1354 qemu_gettimeofday(&tv
);
1355 sn
->date_sec
= tv
.tv_sec
;
1356 sn
->date_nsec
= tv
.tv_usec
* 1000;
1357 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1358 if (replay_mode
!= REPLAY_MODE_NONE
) {
1359 sn
->icount
= replay_get_current_icount();
1364 ret1
= bdrv_snapshot_create(bs
, sn
);
1366 error_setg_errno(errp
, -ret1
,
1367 "Failed to create snapshot '%s' on device '%s'",
1372 /* 4. succeed, mark a snapshot is created */
1373 state
->created
= true;
1376 aio_context_release(aio_context
);
1379 static void internal_snapshot_abort(BlkActionState
*common
)
1381 InternalSnapshotState
*state
=
1382 DO_UPCAST(InternalSnapshotState
, common
, common
);
1383 BlockDriverState
*bs
= state
->bs
;
1384 QEMUSnapshotInfo
*sn
= &state
->sn
;
1385 AioContext
*aio_context
;
1386 Error
*local_error
= NULL
;
1388 if (!state
->created
) {
1392 aio_context
= bdrv_get_aio_context(state
->bs
);
1393 aio_context_acquire(aio_context
);
1395 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1396 error_reportf_err(local_error
,
1397 "Failed to delete snapshot with id '%s' and "
1398 "name '%s' on device '%s' in abort: ",
1399 sn
->id_str
, sn
->name
,
1400 bdrv_get_device_name(bs
));
1403 aio_context_release(aio_context
);
1406 static void internal_snapshot_clean(BlkActionState
*common
)
1408 InternalSnapshotState
*state
= DO_UPCAST(InternalSnapshotState
,
1410 AioContext
*aio_context
;
1416 aio_context
= bdrv_get_aio_context(state
->bs
);
1417 aio_context_acquire(aio_context
);
1419 bdrv_drained_end(state
->bs
);
1421 aio_context_release(aio_context
);
1424 /* external snapshot private data */
1425 typedef struct ExternalSnapshotState
{
1426 BlkActionState common
;
1427 BlockDriverState
*old_bs
;
1428 BlockDriverState
*new_bs
;
1429 bool overlay_appended
;
1430 } ExternalSnapshotState
;
1432 static void external_snapshot_prepare(BlkActionState
*common
,
1436 QDict
*options
= NULL
;
1437 Error
*local_err
= NULL
;
1438 /* Device and node name of the image to generate the snapshot from */
1440 const char *node_name
;
1441 /* Reference to the new image (for 'blockdev-snapshot') */
1442 const char *snapshot_ref
;
1443 /* File name of the new image (for 'blockdev-snapshot-sync') */
1444 const char *new_image_file
;
1445 ExternalSnapshotState
*state
=
1446 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1447 TransactionAction
*action
= common
->action
;
1448 AioContext
*aio_context
;
1449 uint64_t perm
, shared
;
1451 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1452 * purpose but a different set of parameters */
1453 switch (action
->type
) {
1454 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
:
1456 BlockdevSnapshot
*s
= action
->u
.blockdev_snapshot
.data
;
1458 node_name
= s
->node
;
1459 new_image_file
= NULL
;
1460 snapshot_ref
= s
->overlay
;
1463 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
:
1465 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1466 device
= s
->has_device
? s
->device
: NULL
;
1467 node_name
= s
->has_node_name
? s
->node_name
: NULL
;
1468 new_image_file
= s
->snapshot_file
;
1469 snapshot_ref
= NULL
;
1473 g_assert_not_reached();
1476 /* start processing */
1477 if (action_check_completion_mode(common
, errp
) < 0) {
1481 state
->old_bs
= bdrv_lookup_bs(device
, node_name
, errp
);
1482 if (!state
->old_bs
) {
1486 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1487 aio_context_acquire(aio_context
);
1489 /* Paired with .clean() */
1490 bdrv_drained_begin(state
->old_bs
);
1492 if (!bdrv_is_inserted(state
->old_bs
)) {
1493 error_setg(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1497 if (bdrv_op_is_blocked(state
->old_bs
,
1498 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT
, errp
)) {
1502 if (!bdrv_is_read_only(state
->old_bs
)) {
1503 if (bdrv_flush(state
->old_bs
)) {
1504 error_setg(errp
, QERR_IO_ERROR
);
1509 if (action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
) {
1510 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1511 const char *format
= s
->has_format
? s
->format
: "qcow2";
1512 enum NewImageMode mode
;
1513 const char *snapshot_node_name
=
1514 s
->has_snapshot_node_name
? s
->snapshot_node_name
: NULL
;
1516 if (node_name
&& !snapshot_node_name
) {
1517 error_setg(errp
, "New overlay node name missing");
1521 if (snapshot_node_name
&&
1522 bdrv_lookup_bs(snapshot_node_name
, snapshot_node_name
, NULL
)) {
1523 error_setg(errp
, "New overlay node name already in use");
1527 flags
= state
->old_bs
->open_flags
;
1528 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_COPY_ON_READ
);
1529 flags
|= BDRV_O_NO_BACKING
;
1531 /* create new image w/backing file */
1532 mode
= s
->has_mode
? s
->mode
: NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1533 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1534 int64_t size
= bdrv_getlength(state
->old_bs
);
1536 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1539 bdrv_refresh_filename(state
->old_bs
);
1540 bdrv_img_create(new_image_file
, format
,
1541 state
->old_bs
->filename
,
1542 state
->old_bs
->drv
->format_name
,
1543 NULL
, size
, flags
, false, &local_err
);
1545 error_propagate(errp
, local_err
);
1550 options
= qdict_new();
1551 if (snapshot_node_name
) {
1552 qdict_put_str(options
, "node-name", snapshot_node_name
);
1554 qdict_put_str(options
, "driver", format
);
1557 state
->new_bs
= bdrv_open(new_image_file
, snapshot_ref
, options
, flags
,
1559 /* We will manually add the backing_hd field to the bs later */
1560 if (!state
->new_bs
) {
1565 * Allow attaching a backing file to an overlay that's already in use only
1566 * if the parents don't assume that they are already seeing a valid image.
1567 * (Specifically, allow it as a mirror target, which is write-only access.)
1569 bdrv_get_cumulative_perm(state
->new_bs
, &perm
, &shared
);
1570 if (perm
& BLK_PERM_CONSISTENT_READ
) {
1571 error_setg(errp
, "The overlay is already in use");
1575 if (state
->new_bs
->drv
->is_filter
) {
1576 error_setg(errp
, "Filters cannot be used as overlays");
1580 if (bdrv_cow_child(state
->new_bs
)) {
1581 error_setg(errp
, "The overlay already has a backing image");
1585 if (!state
->new_bs
->drv
->supports_backing
) {
1586 error_setg(errp
, "The overlay does not support backing images");
1590 /* This removes our old bs and adds the new bs. This is an operation that
1591 * can fail, so we need to do it in .prepare; undoing it for abort is
1592 * always possible. */
1593 bdrv_ref(state
->new_bs
);
1594 bdrv_append(state
->new_bs
, state
->old_bs
, &local_err
);
1596 error_propagate(errp
, local_err
);
1599 state
->overlay_appended
= true;
1602 aio_context_release(aio_context
);
1605 static void external_snapshot_commit(BlkActionState
*common
)
1607 ExternalSnapshotState
*state
=
1608 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1609 AioContext
*aio_context
;
1611 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1612 aio_context_acquire(aio_context
);
1614 /* We don't need (or want) to use the transactional
1615 * bdrv_reopen_multiple() across all the entries at once, because we
1616 * don't want to abort all of them if one of them fails the reopen */
1617 if (!qatomic_read(&state
->old_bs
->copy_on_read
)) {
1618 bdrv_reopen_set_read_only(state
->old_bs
, true, NULL
);
1621 aio_context_release(aio_context
);
1624 static void external_snapshot_abort(BlkActionState
*common
)
1626 ExternalSnapshotState
*state
=
1627 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1628 if (state
->new_bs
) {
1629 if (state
->overlay_appended
) {
1630 AioContext
*aio_context
;
1631 AioContext
*tmp_context
;
1634 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1635 aio_context_acquire(aio_context
);
1637 bdrv_ref(state
->old_bs
); /* we can't let bdrv_set_backind_hd()
1638 close state->old_bs; we need it */
1639 bdrv_set_backing_hd(state
->new_bs
, NULL
, &error_abort
);
1642 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1643 * the main AioContext. As we're still going to be using it, return
1644 * it to the AioContext it was before.
1646 tmp_context
= bdrv_get_aio_context(state
->old_bs
);
1647 if (aio_context
!= tmp_context
) {
1648 aio_context_release(aio_context
);
1649 aio_context_acquire(tmp_context
);
1651 ret
= bdrv_try_set_aio_context(state
->old_bs
,
1655 aio_context_release(tmp_context
);
1656 aio_context_acquire(aio_context
);
1659 bdrv_replace_node(state
->new_bs
, state
->old_bs
, &error_abort
);
1660 bdrv_unref(state
->old_bs
); /* bdrv_replace_node() ref'ed old_bs */
1662 aio_context_release(aio_context
);
1667 static void external_snapshot_clean(BlkActionState
*common
)
1669 ExternalSnapshotState
*state
=
1670 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1671 AioContext
*aio_context
;
1673 if (!state
->old_bs
) {
1677 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1678 aio_context_acquire(aio_context
);
1680 bdrv_drained_end(state
->old_bs
);
1681 bdrv_unref(state
->new_bs
);
1683 aio_context_release(aio_context
);
1686 typedef struct DriveBackupState
{
1687 BlkActionState common
;
1688 BlockDriverState
*bs
;
1692 static BlockJob
*do_backup_common(BackupCommon
*backup
,
1693 BlockDriverState
*bs
,
1694 BlockDriverState
*target_bs
,
1695 AioContext
*aio_context
,
1696 JobTxn
*txn
, Error
**errp
);
1698 static void drive_backup_prepare(BlkActionState
*common
, Error
**errp
)
1700 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1701 DriveBackup
*backup
;
1702 BlockDriverState
*bs
;
1703 BlockDriverState
*target_bs
;
1704 BlockDriverState
*source
= NULL
;
1705 AioContext
*aio_context
;
1706 AioContext
*old_context
;
1708 Error
*local_err
= NULL
;
1711 bool set_backing_hd
= false;
1714 assert(common
->action
->type
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1715 backup
= common
->action
->u
.drive_backup
.data
;
1717 if (!backup
->has_mode
) {
1718 backup
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1721 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1727 error_setg(errp
, "Device has no medium");
1731 aio_context
= bdrv_get_aio_context(bs
);
1732 aio_context_acquire(aio_context
);
1734 /* Paired with .clean() */
1735 bdrv_drained_begin(bs
);
1737 if (!backup
->has_format
) {
1738 backup
->format
= backup
->mode
== NEW_IMAGE_MODE_EXISTING
?
1739 NULL
: (char *) bs
->drv
->format_name
;
1742 /* Early check to avoid creating target */
1743 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_BACKUP_SOURCE
, errp
)) {
1747 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1750 * See if we have a backing HD we can use to create our new image
1753 if (backup
->sync
== MIRROR_SYNC_MODE_TOP
) {
1755 * Backup will not replace the source by the target, so none
1756 * of the filters skipped here will be removed (in contrast to
1757 * mirror). Therefore, we can skip all of them when looking
1758 * for the first COW relationship.
1760 source
= bdrv_cow_bs(bdrv_skip_filters(bs
));
1762 backup
->sync
= MIRROR_SYNC_MODE_FULL
;
1765 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
1767 flags
|= BDRV_O_NO_BACKING
;
1768 set_backing_hd
= true;
1771 size
= bdrv_getlength(bs
);
1773 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1777 if (backup
->mode
!= NEW_IMAGE_MODE_EXISTING
) {
1778 assert(backup
->format
);
1780 /* Implicit filters should not appear in the filename */
1781 BlockDriverState
*explicit_backing
=
1782 bdrv_skip_implicit_filters(source
);
1784 bdrv_refresh_filename(explicit_backing
);
1785 bdrv_img_create(backup
->target
, backup
->format
,
1786 explicit_backing
->filename
,
1787 explicit_backing
->drv
->format_name
, NULL
,
1788 size
, flags
, false, &local_err
);
1790 bdrv_img_create(backup
->target
, backup
->format
, NULL
, NULL
, NULL
,
1791 size
, flags
, false, &local_err
);
1796 error_propagate(errp
, local_err
);
1800 options
= qdict_new();
1801 qdict_put_str(options
, "discard", "unmap");
1802 qdict_put_str(options
, "detect-zeroes", "unmap");
1803 if (backup
->format
) {
1804 qdict_put_str(options
, "driver", backup
->format
);
1807 target_bs
= bdrv_open(backup
->target
, NULL
, options
, flags
, errp
);
1812 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1813 old_context
= bdrv_get_aio_context(target_bs
);
1814 aio_context_release(aio_context
);
1815 aio_context_acquire(old_context
);
1817 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
1819 bdrv_unref(target_bs
);
1820 aio_context_release(old_context
);
1824 aio_context_release(old_context
);
1825 aio_context_acquire(aio_context
);
1827 if (set_backing_hd
) {
1828 bdrv_set_backing_hd(target_bs
, source
, &local_err
);
1830 error_propagate(errp
, local_err
);
1837 state
->job
= do_backup_common(qapi_DriveBackup_base(backup
),
1838 bs
, target_bs
, aio_context
,
1839 common
->block_job_txn
, errp
);
1842 bdrv_unref(target_bs
);
1844 aio_context_release(aio_context
);
1847 static void drive_backup_commit(BlkActionState
*common
)
1849 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1850 AioContext
*aio_context
;
1852 aio_context
= bdrv_get_aio_context(state
->bs
);
1853 aio_context_acquire(aio_context
);
1856 job_start(&state
->job
->job
);
1858 aio_context_release(aio_context
);
1861 static void drive_backup_abort(BlkActionState
*common
)
1863 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1866 AioContext
*aio_context
;
1868 aio_context
= bdrv_get_aio_context(state
->bs
);
1869 aio_context_acquire(aio_context
);
1871 job_cancel_sync(&state
->job
->job
);
1873 aio_context_release(aio_context
);
1877 static void drive_backup_clean(BlkActionState
*common
)
1879 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1880 AioContext
*aio_context
;
1886 aio_context
= bdrv_get_aio_context(state
->bs
);
1887 aio_context_acquire(aio_context
);
1889 bdrv_drained_end(state
->bs
);
1891 aio_context_release(aio_context
);
1894 typedef struct BlockdevBackupState
{
1895 BlkActionState common
;
1896 BlockDriverState
*bs
;
1898 } BlockdevBackupState
;
1900 static void blockdev_backup_prepare(BlkActionState
*common
, Error
**errp
)
1902 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1903 BlockdevBackup
*backup
;
1904 BlockDriverState
*bs
;
1905 BlockDriverState
*target_bs
;
1906 AioContext
*aio_context
;
1907 AioContext
*old_context
;
1910 assert(common
->action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
);
1911 backup
= common
->action
->u
.blockdev_backup
.data
;
1913 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1918 target_bs
= bdrv_lookup_bs(backup
->target
, backup
->target
, errp
);
1923 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1924 aio_context
= bdrv_get_aio_context(bs
);
1925 old_context
= bdrv_get_aio_context(target_bs
);
1926 aio_context_acquire(old_context
);
1928 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
1930 aio_context_release(old_context
);
1934 aio_context_release(old_context
);
1935 aio_context_acquire(aio_context
);
1938 /* Paired with .clean() */
1939 bdrv_drained_begin(state
->bs
);
1941 state
->job
= do_backup_common(qapi_BlockdevBackup_base(backup
),
1942 bs
, target_bs
, aio_context
,
1943 common
->block_job_txn
, errp
);
1945 aio_context_release(aio_context
);
1948 static void blockdev_backup_commit(BlkActionState
*common
)
1950 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1951 AioContext
*aio_context
;
1953 aio_context
= bdrv_get_aio_context(state
->bs
);
1954 aio_context_acquire(aio_context
);
1957 job_start(&state
->job
->job
);
1959 aio_context_release(aio_context
);
1962 static void blockdev_backup_abort(BlkActionState
*common
)
1964 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1967 AioContext
*aio_context
;
1969 aio_context
= bdrv_get_aio_context(state
->bs
);
1970 aio_context_acquire(aio_context
);
1972 job_cancel_sync(&state
->job
->job
);
1974 aio_context_release(aio_context
);
1978 static void blockdev_backup_clean(BlkActionState
*common
)
1980 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1981 AioContext
*aio_context
;
1987 aio_context
= bdrv_get_aio_context(state
->bs
);
1988 aio_context_acquire(aio_context
);
1990 bdrv_drained_end(state
->bs
);
1992 aio_context_release(aio_context
);
1995 typedef struct BlockDirtyBitmapState
{
1996 BlkActionState common
;
1997 BdrvDirtyBitmap
*bitmap
;
1998 BlockDriverState
*bs
;
2002 } BlockDirtyBitmapState
;
2004 static void block_dirty_bitmap_add_prepare(BlkActionState
*common
,
2007 Error
*local_err
= NULL
;
2008 BlockDirtyBitmapAdd
*action
;
2009 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2012 if (action_check_completion_mode(common
, errp
) < 0) {
2016 action
= common
->action
->u
.block_dirty_bitmap_add
.data
;
2017 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2018 qmp_block_dirty_bitmap_add(action
->node
, action
->name
,
2019 action
->has_granularity
, action
->granularity
,
2020 action
->has_persistent
, action
->persistent
,
2021 action
->has_disabled
, action
->disabled
,
2025 state
->prepared
= true;
2027 error_propagate(errp
, local_err
);
2031 static void block_dirty_bitmap_add_abort(BlkActionState
*common
)
2033 BlockDirtyBitmapAdd
*action
;
2034 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2037 action
= common
->action
->u
.block_dirty_bitmap_add
.data
;
2038 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2039 * then the node reference and bitmap name must have been valid.
2041 if (state
->prepared
) {
2042 qmp_block_dirty_bitmap_remove(action
->node
, action
->name
, &error_abort
);
2046 static void block_dirty_bitmap_clear_prepare(BlkActionState
*common
,
2049 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2051 BlockDirtyBitmap
*action
;
2053 if (action_check_completion_mode(common
, errp
) < 0) {
2057 action
= common
->action
->u
.block_dirty_bitmap_clear
.data
;
2058 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2062 if (!state
->bitmap
) {
2066 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_DEFAULT
, errp
)) {
2070 bdrv_clear_dirty_bitmap(state
->bitmap
, &state
->backup
);
2073 static void block_dirty_bitmap_restore(BlkActionState
*common
)
2075 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2078 if (state
->backup
) {
2079 bdrv_restore_dirty_bitmap(state
->bitmap
, state
->backup
);
2083 static void block_dirty_bitmap_free_backup(BlkActionState
*common
)
2085 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2088 hbitmap_free(state
->backup
);
2091 static void block_dirty_bitmap_enable_prepare(BlkActionState
*common
,
2094 BlockDirtyBitmap
*action
;
2095 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2098 if (action_check_completion_mode(common
, errp
) < 0) {
2102 action
= common
->action
->u
.block_dirty_bitmap_enable
.data
;
2103 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2107 if (!state
->bitmap
) {
2111 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2115 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2116 bdrv_enable_dirty_bitmap(state
->bitmap
);
2119 static void block_dirty_bitmap_enable_abort(BlkActionState
*common
)
2121 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2124 if (!state
->was_enabled
) {
2125 bdrv_disable_dirty_bitmap(state
->bitmap
);
2129 static void block_dirty_bitmap_disable_prepare(BlkActionState
*common
,
2132 BlockDirtyBitmap
*action
;
2133 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2136 if (action_check_completion_mode(common
, errp
) < 0) {
2140 action
= common
->action
->u
.block_dirty_bitmap_disable
.data
;
2141 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2145 if (!state
->bitmap
) {
2149 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2153 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2154 bdrv_disable_dirty_bitmap(state
->bitmap
);
2157 static void block_dirty_bitmap_disable_abort(BlkActionState
*common
)
2159 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2162 if (state
->was_enabled
) {
2163 bdrv_enable_dirty_bitmap(state
->bitmap
);
2167 static void block_dirty_bitmap_merge_prepare(BlkActionState
*common
,
2170 BlockDirtyBitmapMerge
*action
;
2171 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2174 if (action_check_completion_mode(common
, errp
) < 0) {
2178 action
= common
->action
->u
.block_dirty_bitmap_merge
.data
;
2180 state
->bitmap
= block_dirty_bitmap_merge(action
->node
, action
->target
,
2181 action
->bitmaps
, &state
->backup
,
2185 static void block_dirty_bitmap_remove_prepare(BlkActionState
*common
,
2188 BlockDirtyBitmap
*action
;
2189 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2192 if (action_check_completion_mode(common
, errp
) < 0) {
2196 action
= common
->action
->u
.block_dirty_bitmap_remove
.data
;
2198 state
->bitmap
= block_dirty_bitmap_remove(action
->node
, action
->name
,
2199 false, &state
->bs
, errp
);
2200 if (state
->bitmap
) {
2201 bdrv_dirty_bitmap_skip_store(state
->bitmap
, true);
2202 bdrv_dirty_bitmap_set_busy(state
->bitmap
, true);
2206 static void block_dirty_bitmap_remove_abort(BlkActionState
*common
)
2208 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2211 if (state
->bitmap
) {
2212 bdrv_dirty_bitmap_skip_store(state
->bitmap
, false);
2213 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2217 static void block_dirty_bitmap_remove_commit(BlkActionState
*common
)
2219 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2222 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2223 bdrv_release_dirty_bitmap(state
->bitmap
);
2226 static void abort_prepare(BlkActionState
*common
, Error
**errp
)
2228 error_setg(errp
, "Transaction aborted using Abort action");
2231 static void abort_commit(BlkActionState
*common
)
2233 g_assert_not_reached(); /* this action never succeeds */
2236 static const BlkActionOps actions
[] = {
2237 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
] = {
2238 .instance_size
= sizeof(ExternalSnapshotState
),
2239 .prepare
= external_snapshot_prepare
,
2240 .commit
= external_snapshot_commit
,
2241 .abort
= external_snapshot_abort
,
2242 .clean
= external_snapshot_clean
,
2244 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
2245 .instance_size
= sizeof(ExternalSnapshotState
),
2246 .prepare
= external_snapshot_prepare
,
2247 .commit
= external_snapshot_commit
,
2248 .abort
= external_snapshot_abort
,
2249 .clean
= external_snapshot_clean
,
2251 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
2252 .instance_size
= sizeof(DriveBackupState
),
2253 .prepare
= drive_backup_prepare
,
2254 .commit
= drive_backup_commit
,
2255 .abort
= drive_backup_abort
,
2256 .clean
= drive_backup_clean
,
2258 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
] = {
2259 .instance_size
= sizeof(BlockdevBackupState
),
2260 .prepare
= blockdev_backup_prepare
,
2261 .commit
= blockdev_backup_commit
,
2262 .abort
= blockdev_backup_abort
,
2263 .clean
= blockdev_backup_clean
,
2265 [TRANSACTION_ACTION_KIND_ABORT
] = {
2266 .instance_size
= sizeof(BlkActionState
),
2267 .prepare
= abort_prepare
,
2268 .commit
= abort_commit
,
2270 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
2271 .instance_size
= sizeof(InternalSnapshotState
),
2272 .prepare
= internal_snapshot_prepare
,
2273 .abort
= internal_snapshot_abort
,
2274 .clean
= internal_snapshot_clean
,
2276 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD
] = {
2277 .instance_size
= sizeof(BlockDirtyBitmapState
),
2278 .prepare
= block_dirty_bitmap_add_prepare
,
2279 .abort
= block_dirty_bitmap_add_abort
,
2281 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR
] = {
2282 .instance_size
= sizeof(BlockDirtyBitmapState
),
2283 .prepare
= block_dirty_bitmap_clear_prepare
,
2284 .commit
= block_dirty_bitmap_free_backup
,
2285 .abort
= block_dirty_bitmap_restore
,
2287 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE
] = {
2288 .instance_size
= sizeof(BlockDirtyBitmapState
),
2289 .prepare
= block_dirty_bitmap_enable_prepare
,
2290 .abort
= block_dirty_bitmap_enable_abort
,
2292 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE
] = {
2293 .instance_size
= sizeof(BlockDirtyBitmapState
),
2294 .prepare
= block_dirty_bitmap_disable_prepare
,
2295 .abort
= block_dirty_bitmap_disable_abort
,
2297 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE
] = {
2298 .instance_size
= sizeof(BlockDirtyBitmapState
),
2299 .prepare
= block_dirty_bitmap_merge_prepare
,
2300 .commit
= block_dirty_bitmap_free_backup
,
2301 .abort
= block_dirty_bitmap_restore
,
2303 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE
] = {
2304 .instance_size
= sizeof(BlockDirtyBitmapState
),
2305 .prepare
= block_dirty_bitmap_remove_prepare
,
2306 .commit
= block_dirty_bitmap_remove_commit
,
2307 .abort
= block_dirty_bitmap_remove_abort
,
2309 /* Where are transactions for MIRROR, COMMIT and STREAM?
2310 * Although these blockjobs use transaction callbacks like the backup job,
2311 * these jobs do not necessarily adhere to transaction semantics.
2312 * These jobs may not fully undo all of their actions on abort, nor do they
2313 * necessarily work in transactions with more than one job in them.
2318 * Allocate a TransactionProperties structure if necessary, and fill
2319 * that structure with desired defaults if they are unset.
2321 static TransactionProperties
*get_transaction_properties(
2322 TransactionProperties
*props
)
2325 props
= g_new0(TransactionProperties
, 1);
2328 if (!props
->has_completion_mode
) {
2329 props
->has_completion_mode
= true;
2330 props
->completion_mode
= ACTION_COMPLETION_MODE_INDIVIDUAL
;
2337 * 'Atomic' group operations. The operations are performed as a set, and if
2338 * any fail then we roll back all operations in the group.
2340 void qmp_transaction(TransactionActionList
*dev_list
,
2342 struct TransactionProperties
*props
,
2345 TransactionActionList
*dev_entry
= dev_list
;
2346 JobTxn
*block_job_txn
= NULL
;
2347 BlkActionState
*state
, *next
;
2348 Error
*local_err
= NULL
;
2350 QTAILQ_HEAD(, BlkActionState
) snap_bdrv_states
;
2351 QTAILQ_INIT(&snap_bdrv_states
);
2353 /* Does this transaction get canceled as a group on failure?
2354 * If not, we don't really need to make a JobTxn.
2356 props
= get_transaction_properties(props
);
2357 if (props
->completion_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
2358 block_job_txn
= job_txn_new();
2361 /* drain all i/o before any operations */
2364 /* We don't do anything in this loop that commits us to the operations */
2365 while (NULL
!= dev_entry
) {
2366 TransactionAction
*dev_info
= NULL
;
2367 const BlkActionOps
*ops
;
2369 dev_info
= dev_entry
->value
;
2370 dev_entry
= dev_entry
->next
;
2372 assert(dev_info
->type
< ARRAY_SIZE(actions
));
2374 ops
= &actions
[dev_info
->type
];
2375 assert(ops
->instance_size
> 0);
2377 state
= g_malloc0(ops
->instance_size
);
2379 state
->action
= dev_info
;
2380 state
->block_job_txn
= block_job_txn
;
2381 state
->txn_props
= props
;
2382 QTAILQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
2384 state
->ops
->prepare(state
, &local_err
);
2386 error_propagate(errp
, local_err
);
2387 goto delete_and_fail
;
2391 QTAILQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
2392 if (state
->ops
->commit
) {
2393 state
->ops
->commit(state
);
2401 /* failure, and it is all-or-none; roll back all operations */
2402 QTAILQ_FOREACH_REVERSE(state
, &snap_bdrv_states
, entry
) {
2403 if (state
->ops
->abort
) {
2404 state
->ops
->abort(state
);
2408 QTAILQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
2409 if (state
->ops
->clean
) {
2410 state
->ops
->clean(state
);
2415 qapi_free_TransactionProperties(props
);
2417 job_txn_unref(block_job_txn
);
2420 void qmp_block_passwd(bool has_device
, const char *device
,
2421 bool has_node_name
, const char *node_name
,
2422 const char *password
, Error
**errp
)
2425 "Setting block passwords directly is no longer supported");
2428 BlockDirtyBitmapSha256
*qmp_x_debug_block_dirty_bitmap_sha256(const char *node
,
2432 BdrvDirtyBitmap
*bitmap
;
2433 BlockDriverState
*bs
;
2434 BlockDirtyBitmapSha256
*ret
= NULL
;
2437 bitmap
= block_dirty_bitmap_lookup(node
, name
, &bs
, errp
);
2438 if (!bitmap
|| !bs
) {
2442 sha256
= bdrv_dirty_bitmap_sha256(bitmap
, errp
);
2443 if (sha256
== NULL
) {
2447 ret
= g_new(BlockDirtyBitmapSha256
, 1);
2448 ret
->sha256
= sha256
;
2453 void coroutine_fn
qmp_block_resize(bool has_device
, const char *device
,
2454 bool has_node_name
, const char *node_name
,
2455 int64_t size
, Error
**errp
)
2457 Error
*local_err
= NULL
;
2459 BlockDriverState
*bs
;
2460 AioContext
*old_ctx
;
2462 bs
= bdrv_lookup_bs(has_device
? device
: NULL
,
2463 has_node_name
? node_name
: NULL
,
2466 error_propagate(errp
, local_err
);
2471 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
2475 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_RESIZE
, NULL
)) {
2476 error_setg(errp
, QERR_DEVICE_IN_USE
, device
);
2480 blk
= blk_new_with_bs(bs
, BLK_PERM_RESIZE
, BLK_PERM_ALL
, errp
);
2486 bdrv_drained_begin(bs
);
2489 old_ctx
= bdrv_co_enter(bs
);
2490 blk_truncate(blk
, size
, false, PREALLOC_MODE_OFF
, 0, errp
);
2491 bdrv_co_leave(bs
, old_ctx
);
2494 bdrv_drained_end(bs
);
2499 void qmp_block_stream(bool has_job_id
, const char *job_id
, const char *device
,
2500 bool has_base
, const char *base
,
2501 bool has_base_node
, const char *base_node
,
2502 bool has_backing_file
, const char *backing_file
,
2503 bool has_speed
, int64_t speed
,
2504 bool has_on_error
, BlockdevOnError on_error
,
2505 bool has_filter_node_name
, const char *filter_node_name
,
2506 bool has_auto_finalize
, bool auto_finalize
,
2507 bool has_auto_dismiss
, bool auto_dismiss
,
2510 BlockDriverState
*bs
, *iter
;
2511 BlockDriverState
*base_bs
= NULL
;
2512 AioContext
*aio_context
;
2513 Error
*local_err
= NULL
;
2514 int job_flags
= JOB_DEFAULT
;
2516 if (!has_on_error
) {
2517 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2520 bs
= bdrv_lookup_bs(device
, device
, errp
);
2525 aio_context
= bdrv_get_aio_context(bs
);
2526 aio_context_acquire(aio_context
);
2528 if (has_base
&& has_base_node
) {
2529 error_setg(errp
, "'base' and 'base-node' cannot be specified "
2530 "at the same time");
2535 base_bs
= bdrv_find_backing_image(bs
, base
);
2536 if (base_bs
== NULL
) {
2537 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2540 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2543 if (has_base_node
) {
2544 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2548 if (bs
== base_bs
|| !bdrv_chain_contains(bs
, base_bs
)) {
2549 error_setg(errp
, "Node '%s' is not a backing image of '%s'",
2553 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2554 bdrv_refresh_filename(base_bs
);
2557 /* Check for op blockers in the whole chain between bs and base */
2558 for (iter
= bs
; iter
&& iter
!= base_bs
;
2559 iter
= bdrv_filter_or_cow_bs(iter
))
2561 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_STREAM
, errp
)) {
2566 /* if we are streaming the entire chain, the result will have no backing
2567 * file, and specifying one is therefore an error */
2568 if (base_bs
== NULL
&& has_backing_file
) {
2569 error_setg(errp
, "backing file specified, but streaming the "
2574 if (has_auto_finalize
&& !auto_finalize
) {
2575 job_flags
|= JOB_MANUAL_FINALIZE
;
2577 if (has_auto_dismiss
&& !auto_dismiss
) {
2578 job_flags
|= JOB_MANUAL_DISMISS
;
2581 stream_start(has_job_id
? job_id
: NULL
, bs
, base_bs
, backing_file
,
2582 job_flags
, has_speed
? speed
: 0, on_error
,
2583 filter_node_name
, &local_err
);
2585 error_propagate(errp
, local_err
);
2589 trace_qmp_block_stream(bs
);
2592 aio_context_release(aio_context
);
2595 void qmp_block_commit(bool has_job_id
, const char *job_id
, const char *device
,
2596 bool has_base_node
, const char *base_node
,
2597 bool has_base
, const char *base
,
2598 bool has_top_node
, const char *top_node
,
2599 bool has_top
, const char *top
,
2600 bool has_backing_file
, const char *backing_file
,
2601 bool has_speed
, int64_t speed
,
2602 bool has_on_error
, BlockdevOnError on_error
,
2603 bool has_filter_node_name
, const char *filter_node_name
,
2604 bool has_auto_finalize
, bool auto_finalize
,
2605 bool has_auto_dismiss
, bool auto_dismiss
,
2608 BlockDriverState
*bs
;
2609 BlockDriverState
*iter
;
2610 BlockDriverState
*base_bs
, *top_bs
;
2611 AioContext
*aio_context
;
2612 Error
*local_err
= NULL
;
2613 int job_flags
= JOB_DEFAULT
;
2614 uint64_t top_perm
, top_shared
;
2619 if (!has_on_error
) {
2620 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2622 if (!has_filter_node_name
) {
2623 filter_node_name
= NULL
;
2625 if (has_auto_finalize
&& !auto_finalize
) {
2626 job_flags
|= JOB_MANUAL_FINALIZE
;
2628 if (has_auto_dismiss
&& !auto_dismiss
) {
2629 job_flags
|= JOB_MANUAL_DISMISS
;
2633 * libvirt relies on the DeviceNotFound error class in order to probe for
2634 * live commit feature versions; for this to work, we must make sure to
2635 * perform the device lookup before any generic errors that may occur in a
2636 * scenario in which all optional arguments are omitted. */
2637 bs
= qmp_get_root_bs(device
, &local_err
);
2639 bs
= bdrv_lookup_bs(device
, device
, NULL
);
2641 error_free(local_err
);
2642 error_set(errp
, ERROR_CLASS_DEVICE_NOT_FOUND
,
2643 "Device '%s' not found", device
);
2645 error_propagate(errp
, local_err
);
2650 aio_context
= bdrv_get_aio_context(bs
);
2651 aio_context_acquire(aio_context
);
2653 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_COMMIT_SOURCE
, errp
)) {
2657 /* default top_bs is the active layer */
2660 if (has_top_node
&& has_top
) {
2661 error_setg(errp
, "'top-node' and 'top' are mutually exclusive");
2663 } else if (has_top_node
) {
2664 top_bs
= bdrv_lookup_bs(NULL
, top_node
, errp
);
2665 if (top_bs
== NULL
) {
2668 if (!bdrv_chain_contains(bs
, top_bs
)) {
2669 error_setg(errp
, "'%s' is not in this backing file chain",
2673 } else if (has_top
&& top
) {
2674 /* This strcmp() is just a shortcut, there is no need to
2675 * refresh @bs's filename. If it mismatches,
2676 * bdrv_find_backing_image() will do the refresh and may still
2678 if (strcmp(bs
->filename
, top
) != 0) {
2679 top_bs
= bdrv_find_backing_image(bs
, top
);
2683 if (top_bs
== NULL
) {
2684 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
2688 assert(bdrv_get_aio_context(top_bs
) == aio_context
);
2690 if (has_base_node
&& has_base
) {
2691 error_setg(errp
, "'base-node' and 'base' are mutually exclusive");
2693 } else if (has_base_node
) {
2694 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2695 if (base_bs
== NULL
) {
2698 if (!bdrv_chain_contains(top_bs
, base_bs
)) {
2699 error_setg(errp
, "'%s' is not in this backing file chain",
2703 } else if (has_base
&& base
) {
2704 base_bs
= bdrv_find_backing_image(top_bs
, base
);
2705 if (base_bs
== NULL
) {
2706 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2710 base_bs
= bdrv_find_base(top_bs
);
2711 if (base_bs
== NULL
) {
2712 error_setg(errp
, "There is no backimg image");
2717 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2719 for (iter
= top_bs
; iter
!= bdrv_filter_or_cow_bs(base_bs
);
2720 iter
= bdrv_filter_or_cow_bs(iter
))
2722 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2727 /* Do not allow attempts to commit an image into itself */
2728 if (top_bs
== base_bs
) {
2729 error_setg(errp
, "cannot commit an image into itself");
2734 * Active commit is required if and only if someone has taken a
2735 * WRITE permission on the top node. Historically, we have always
2736 * used active commit for top nodes, so continue that practice
2737 * lest we possibly break clients that rely on this behavior, e.g.
2738 * to later attach this node to a writing parent.
2739 * (Active commit is never really wrong.)
2741 bdrv_get_cumulative_perm(top_bs
, &top_perm
, &top_shared
);
2742 if (top_perm
& BLK_PERM_WRITE
||
2743 bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
))
2745 if (has_backing_file
) {
2746 if (bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
)) {
2747 error_setg(errp
, "'backing-file' specified,"
2748 " but 'top' is the active layer");
2750 error_setg(errp
, "'backing-file' specified, but 'top' has a "
2757 * Emulate here what block_job_create() does, because it
2758 * is possible that @bs != @top_bs (the block job should
2759 * be named after @bs, even if @top_bs is the actual
2762 job_id
= bdrv_get_device_name(bs
);
2764 commit_active_start(job_id
, top_bs
, base_bs
, job_flags
, speed
, on_error
,
2765 filter_node_name
, NULL
, NULL
, false, &local_err
);
2767 BlockDriverState
*overlay_bs
= bdrv_find_overlay(bs
, top_bs
);
2768 if (bdrv_op_is_blocked(overlay_bs
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2771 commit_start(has_job_id
? job_id
: NULL
, bs
, base_bs
, top_bs
, job_flags
,
2772 speed
, on_error
, has_backing_file
? backing_file
: NULL
,
2773 filter_node_name
, &local_err
);
2775 if (local_err
!= NULL
) {
2776 error_propagate(errp
, local_err
);
2781 aio_context_release(aio_context
);
2784 /* Common QMP interface for drive-backup and blockdev-backup */
2785 static BlockJob
*do_backup_common(BackupCommon
*backup
,
2786 BlockDriverState
*bs
,
2787 BlockDriverState
*target_bs
,
2788 AioContext
*aio_context
,
2789 JobTxn
*txn
, Error
**errp
)
2791 BlockJob
*job
= NULL
;
2792 BdrvDirtyBitmap
*bmap
= NULL
;
2793 int job_flags
= JOB_DEFAULT
;
2795 if (!backup
->has_speed
) {
2798 if (!backup
->has_on_source_error
) {
2799 backup
->on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2801 if (!backup
->has_on_target_error
) {
2802 backup
->on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2804 if (!backup
->has_job_id
) {
2805 backup
->job_id
= NULL
;
2807 if (!backup
->has_auto_finalize
) {
2808 backup
->auto_finalize
= true;
2810 if (!backup
->has_auto_dismiss
) {
2811 backup
->auto_dismiss
= true;
2813 if (!backup
->has_compress
) {
2814 backup
->compress
= false;
2817 if ((backup
->sync
== MIRROR_SYNC_MODE_BITMAP
) ||
2818 (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
)) {
2819 /* done before desugaring 'incremental' to print the right message */
2820 if (!backup
->has_bitmap
) {
2821 error_setg(errp
, "must provide a valid bitmap name for "
2822 "'%s' sync mode", MirrorSyncMode_str(backup
->sync
));
2827 if (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
) {
2828 if (backup
->has_bitmap_mode
&&
2829 backup
->bitmap_mode
!= BITMAP_SYNC_MODE_ON_SUCCESS
) {
2830 error_setg(errp
, "Bitmap sync mode must be '%s' "
2831 "when using sync mode '%s'",
2832 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS
),
2833 MirrorSyncMode_str(backup
->sync
));
2836 backup
->has_bitmap_mode
= true;
2837 backup
->sync
= MIRROR_SYNC_MODE_BITMAP
;
2838 backup
->bitmap_mode
= BITMAP_SYNC_MODE_ON_SUCCESS
;
2841 if (backup
->has_bitmap
) {
2842 bmap
= bdrv_find_dirty_bitmap(bs
, backup
->bitmap
);
2844 error_setg(errp
, "Bitmap '%s' could not be found", backup
->bitmap
);
2847 if (!backup
->has_bitmap_mode
) {
2848 error_setg(errp
, "Bitmap sync mode must be given "
2849 "when providing a bitmap");
2852 if (bdrv_dirty_bitmap_check(bmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2856 /* This does not produce a useful bitmap artifact: */
2857 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
2858 error_setg(errp
, "sync mode '%s' does not produce meaningful bitmap"
2859 " outputs", MirrorSyncMode_str(backup
->sync
));
2863 /* If the bitmap isn't used for input or output, this is useless: */
2864 if (backup
->bitmap_mode
== BITMAP_SYNC_MODE_NEVER
&&
2865 backup
->sync
!= MIRROR_SYNC_MODE_BITMAP
) {
2866 error_setg(errp
, "Bitmap sync mode '%s' has no meaningful effect"
2867 " when combined with sync mode '%s'",
2868 BitmapSyncMode_str(backup
->bitmap_mode
),
2869 MirrorSyncMode_str(backup
->sync
));
2874 if (!backup
->has_bitmap
&& backup
->has_bitmap_mode
) {
2875 error_setg(errp
, "Cannot specify bitmap sync mode without a bitmap");
2879 if (!backup
->auto_finalize
) {
2880 job_flags
|= JOB_MANUAL_FINALIZE
;
2882 if (!backup
->auto_dismiss
) {
2883 job_flags
|= JOB_MANUAL_DISMISS
;
2886 job
= backup_job_create(backup
->job_id
, bs
, target_bs
, backup
->speed
,
2887 backup
->sync
, bmap
, backup
->bitmap_mode
,
2889 backup
->filter_node_name
,
2890 backup
->on_source_error
,
2891 backup
->on_target_error
,
2892 job_flags
, NULL
, NULL
, txn
, errp
);
2896 void qmp_drive_backup(DriveBackup
*backup
, Error
**errp
)
2898 TransactionAction action
= {
2899 .type
= TRANSACTION_ACTION_KIND_DRIVE_BACKUP
,
2900 .u
.drive_backup
.data
= backup
,
2902 blockdev_do_action(&action
, errp
);
2905 BlockDeviceInfoList
*qmp_query_named_block_nodes(bool has_flat
,
2909 bool return_flat
= has_flat
&& flat
;
2911 return bdrv_named_nodes_list(return_flat
, errp
);
2914 XDbgBlockGraph
*qmp_x_debug_query_block_graph(Error
**errp
)
2916 return bdrv_get_xdbg_block_graph(errp
);
2919 void qmp_blockdev_backup(BlockdevBackup
*backup
, Error
**errp
)
2921 TransactionAction action
= {
2922 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
,
2923 .u
.blockdev_backup
.data
= backup
,
2925 blockdev_do_action(&action
, errp
);
2928 /* Parameter check and block job starting for drive mirroring.
2929 * Caller should hold @device and @target's aio context (must be the same).
2931 static void blockdev_mirror_common(const char *job_id
, BlockDriverState
*bs
,
2932 BlockDriverState
*target
,
2933 bool has_replaces
, const char *replaces
,
2934 enum MirrorSyncMode sync
,
2935 BlockMirrorBackingMode backing_mode
,
2937 bool has_speed
, int64_t speed
,
2938 bool has_granularity
, uint32_t granularity
,
2939 bool has_buf_size
, int64_t buf_size
,
2940 bool has_on_source_error
,
2941 BlockdevOnError on_source_error
,
2942 bool has_on_target_error
,
2943 BlockdevOnError on_target_error
,
2944 bool has_unmap
, bool unmap
,
2945 bool has_filter_node_name
,
2946 const char *filter_node_name
,
2947 bool has_copy_mode
, MirrorCopyMode copy_mode
,
2948 bool has_auto_finalize
, bool auto_finalize
,
2949 bool has_auto_dismiss
, bool auto_dismiss
,
2952 BlockDriverState
*unfiltered_bs
;
2953 int job_flags
= JOB_DEFAULT
;
2958 if (!has_on_source_error
) {
2959 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2961 if (!has_on_target_error
) {
2962 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2964 if (!has_granularity
) {
2967 if (!has_buf_size
) {
2973 if (!has_filter_node_name
) {
2974 filter_node_name
= NULL
;
2976 if (!has_copy_mode
) {
2977 copy_mode
= MIRROR_COPY_MODE_BACKGROUND
;
2979 if (has_auto_finalize
&& !auto_finalize
) {
2980 job_flags
|= JOB_MANUAL_FINALIZE
;
2982 if (has_auto_dismiss
&& !auto_dismiss
) {
2983 job_flags
|= JOB_MANUAL_DISMISS
;
2986 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
2987 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2988 "a value in range [512B, 64MB]");
2991 if (granularity
& (granularity
- 1)) {
2992 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2997 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3000 if (bdrv_op_is_blocked(target
, BLOCK_OP_TYPE_MIRROR_TARGET
, errp
)) {
3004 if (!bdrv_backing_chain_next(bs
) && sync
== MIRROR_SYNC_MODE_TOP
) {
3005 sync
= MIRROR_SYNC_MODE_FULL
;
3008 if (!has_replaces
) {
3009 /* We want to mirror from @bs, but keep implicit filters on top */
3010 unfiltered_bs
= bdrv_skip_implicit_filters(bs
);
3011 if (unfiltered_bs
!= bs
) {
3012 replaces
= unfiltered_bs
->node_name
;
3013 has_replaces
= true;
3018 BlockDriverState
*to_replace_bs
;
3019 AioContext
*replace_aio_context
;
3020 int64_t bs_size
, replace_size
;
3022 bs_size
= bdrv_getlength(bs
);
3024 error_setg_errno(errp
, -bs_size
, "Failed to query device's size");
3028 to_replace_bs
= check_to_replace_node(bs
, replaces
, errp
);
3029 if (!to_replace_bs
) {
3033 replace_aio_context
= bdrv_get_aio_context(to_replace_bs
);
3034 aio_context_acquire(replace_aio_context
);
3035 replace_size
= bdrv_getlength(to_replace_bs
);
3036 aio_context_release(replace_aio_context
);
3038 if (replace_size
< 0) {
3039 error_setg_errno(errp
, -replace_size
,
3040 "Failed to query the replacement node's size");
3043 if (bs_size
!= replace_size
) {
3044 error_setg(errp
, "cannot replace image with a mirror image of "
3050 /* pass the node name to replace to mirror start since it's loose coupling
3051 * and will allow to check whether the node still exist at mirror completion
3053 mirror_start(job_id
, bs
, target
,
3054 has_replaces
? replaces
: NULL
, job_flags
,
3055 speed
, granularity
, buf_size
, sync
, backing_mode
, zero_target
,
3056 on_source_error
, on_target_error
, unmap
, filter_node_name
,
3060 void qmp_drive_mirror(DriveMirror
*arg
, Error
**errp
)
3062 BlockDriverState
*bs
;
3063 BlockDriverState
*target_backing_bs
, *target_bs
;
3064 AioContext
*aio_context
;
3065 AioContext
*old_context
;
3066 BlockMirrorBackingMode backing_mode
;
3067 Error
*local_err
= NULL
;
3068 QDict
*options
= NULL
;
3071 const char *format
= arg
->format
;
3075 bs
= qmp_get_root_bs(arg
->device
, errp
);
3080 /* Early check to avoid creating target */
3081 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3085 aio_context
= bdrv_get_aio_context(bs
);
3086 aio_context_acquire(aio_context
);
3088 if (!arg
->has_mode
) {
3089 arg
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
3092 if (!arg
->has_format
) {
3093 format
= (arg
->mode
== NEW_IMAGE_MODE_EXISTING
3094 ? NULL
: bs
->drv
->format_name
);
3097 flags
= bs
->open_flags
| BDRV_O_RDWR
;
3098 target_backing_bs
= bdrv_cow_bs(bdrv_skip_filters(bs
));
3099 if (!target_backing_bs
&& arg
->sync
== MIRROR_SYNC_MODE_TOP
) {
3100 arg
->sync
= MIRROR_SYNC_MODE_FULL
;
3102 if (arg
->sync
== MIRROR_SYNC_MODE_NONE
) {
3103 target_backing_bs
= bs
;
3106 size
= bdrv_getlength(bs
);
3108 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
3112 if (arg
->has_replaces
) {
3113 if (!arg
->has_node_name
) {
3114 error_setg(errp
, "a node-name must be provided when replacing a"
3115 " named node of the graph");
3120 if (arg
->mode
== NEW_IMAGE_MODE_ABSOLUTE_PATHS
) {
3121 backing_mode
= MIRROR_SOURCE_BACKING_CHAIN
;
3123 backing_mode
= MIRROR_OPEN_BACKING_CHAIN
;
3126 /* Don't open backing image in create() */
3127 flags
|= BDRV_O_NO_BACKING
;
3129 if ((arg
->sync
== MIRROR_SYNC_MODE_FULL
|| !target_backing_bs
)
3130 && arg
->mode
!= NEW_IMAGE_MODE_EXISTING
)
3132 /* create new image w/o backing file */
3134 bdrv_img_create(arg
->target
, format
,
3135 NULL
, NULL
, NULL
, size
, flags
, false, &local_err
);
3137 /* Implicit filters should not appear in the filename */
3138 BlockDriverState
*explicit_backing
=
3139 bdrv_skip_implicit_filters(target_backing_bs
);
3141 switch (arg
->mode
) {
3142 case NEW_IMAGE_MODE_EXISTING
:
3144 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
3145 /* create new image with backing file */
3146 bdrv_refresh_filename(explicit_backing
);
3147 bdrv_img_create(arg
->target
, format
,
3148 explicit_backing
->filename
,
3149 explicit_backing
->drv
->format_name
,
3150 NULL
, size
, flags
, false, &local_err
);
3158 error_propagate(errp
, local_err
);
3162 options
= qdict_new();
3163 if (arg
->has_node_name
) {
3164 qdict_put_str(options
, "node-name", arg
->node_name
);
3167 qdict_put_str(options
, "driver", format
);
3170 /* Mirroring takes care of copy-on-write using the source's backing
3173 target_bs
= bdrv_open(arg
->target
, NULL
, options
, flags
, errp
);
3178 zero_target
= (arg
->sync
== MIRROR_SYNC_MODE_FULL
&&
3179 (arg
->mode
== NEW_IMAGE_MODE_EXISTING
||
3180 !bdrv_has_zero_init(target_bs
)));
3183 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3184 old_context
= bdrv_get_aio_context(target_bs
);
3185 aio_context_release(aio_context
);
3186 aio_context_acquire(old_context
);
3188 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
3190 bdrv_unref(target_bs
);
3191 aio_context_release(old_context
);
3195 aio_context_release(old_context
);
3196 aio_context_acquire(aio_context
);
3198 blockdev_mirror_common(arg
->has_job_id
? arg
->job_id
: NULL
, bs
, target_bs
,
3199 arg
->has_replaces
, arg
->replaces
, arg
->sync
,
3200 backing_mode
, zero_target
,
3201 arg
->has_speed
, arg
->speed
,
3202 arg
->has_granularity
, arg
->granularity
,
3203 arg
->has_buf_size
, arg
->buf_size
,
3204 arg
->has_on_source_error
, arg
->on_source_error
,
3205 arg
->has_on_target_error
, arg
->on_target_error
,
3206 arg
->has_unmap
, arg
->unmap
,
3208 arg
->has_copy_mode
, arg
->copy_mode
,
3209 arg
->has_auto_finalize
, arg
->auto_finalize
,
3210 arg
->has_auto_dismiss
, arg
->auto_dismiss
,
3212 bdrv_unref(target_bs
);
3214 aio_context_release(aio_context
);
3217 void qmp_blockdev_mirror(bool has_job_id
, const char *job_id
,
3218 const char *device
, const char *target
,
3219 bool has_replaces
, const char *replaces
,
3220 MirrorSyncMode sync
,
3221 bool has_speed
, int64_t speed
,
3222 bool has_granularity
, uint32_t granularity
,
3223 bool has_buf_size
, int64_t buf_size
,
3224 bool has_on_source_error
,
3225 BlockdevOnError on_source_error
,
3226 bool has_on_target_error
,
3227 BlockdevOnError on_target_error
,
3228 bool has_filter_node_name
,
3229 const char *filter_node_name
,
3230 bool has_copy_mode
, MirrorCopyMode copy_mode
,
3231 bool has_auto_finalize
, bool auto_finalize
,
3232 bool has_auto_dismiss
, bool auto_dismiss
,
3235 BlockDriverState
*bs
;
3236 BlockDriverState
*target_bs
;
3237 AioContext
*aio_context
;
3238 AioContext
*old_context
;
3239 BlockMirrorBackingMode backing_mode
= MIRROR_LEAVE_BACKING_CHAIN
;
3243 bs
= qmp_get_root_bs(device
, errp
);
3248 target_bs
= bdrv_lookup_bs(target
, target
, errp
);
3253 zero_target
= (sync
== MIRROR_SYNC_MODE_FULL
);
3255 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3256 old_context
= bdrv_get_aio_context(target_bs
);
3257 aio_context
= bdrv_get_aio_context(bs
);
3258 aio_context_acquire(old_context
);
3260 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
3262 aio_context_release(old_context
);
3263 aio_context_acquire(aio_context
);
3269 blockdev_mirror_common(has_job_id
? job_id
: NULL
, bs
, target_bs
,
3270 has_replaces
, replaces
, sync
, backing_mode
,
3271 zero_target
, has_speed
, speed
,
3272 has_granularity
, granularity
,
3273 has_buf_size
, buf_size
,
3274 has_on_source_error
, on_source_error
,
3275 has_on_target_error
, on_target_error
,
3277 has_filter_node_name
, filter_node_name
,
3278 has_copy_mode
, copy_mode
,
3279 has_auto_finalize
, auto_finalize
,
3280 has_auto_dismiss
, auto_dismiss
,
3283 aio_context_release(aio_context
);
3286 /* Get a block job using its ID and acquire its AioContext */
3287 static BlockJob
*find_block_job(const char *id
, AioContext
**aio_context
,
3294 *aio_context
= NULL
;
3296 job
= block_job_get(id
);
3299 error_set(errp
, ERROR_CLASS_DEVICE_NOT_ACTIVE
,
3300 "Block job '%s' not found", id
);
3304 *aio_context
= blk_get_aio_context(job
->blk
);
3305 aio_context_acquire(*aio_context
);
3310 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
3312 AioContext
*aio_context
;
3313 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3319 block_job_set_speed(job
, speed
, errp
);
3320 aio_context_release(aio_context
);
3323 void qmp_block_job_cancel(const char *device
,
3324 bool has_force
, bool force
, Error
**errp
)
3326 AioContext
*aio_context
;
3327 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3337 if (job_user_paused(&job
->job
) && !force
) {
3338 error_setg(errp
, "The block job for device '%s' is currently paused",
3343 trace_qmp_block_job_cancel(job
);
3344 job_user_cancel(&job
->job
, force
, errp
);
3346 aio_context_release(aio_context
);
3349 void qmp_block_job_pause(const char *device
, Error
**errp
)
3351 AioContext
*aio_context
;
3352 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3358 trace_qmp_block_job_pause(job
);
3359 job_user_pause(&job
->job
, errp
);
3360 aio_context_release(aio_context
);
3363 void qmp_block_job_resume(const char *device
, Error
**errp
)
3365 AioContext
*aio_context
;
3366 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3372 trace_qmp_block_job_resume(job
);
3373 job_user_resume(&job
->job
, errp
);
3374 aio_context_release(aio_context
);
3377 void qmp_block_job_complete(const char *device
, Error
**errp
)
3379 AioContext
*aio_context
;
3380 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3386 trace_qmp_block_job_complete(job
);
3387 job_complete(&job
->job
, errp
);
3388 aio_context_release(aio_context
);
3391 void qmp_block_job_finalize(const char *id
, Error
**errp
)
3393 AioContext
*aio_context
;
3394 BlockJob
*job
= find_block_job(id
, &aio_context
, errp
);
3400 trace_qmp_block_job_finalize(job
);
3402 job_finalize(&job
->job
, errp
);
3405 * Job's context might have changed via job_finalize (and job_txn_apply
3406 * automatically acquires the new one), so make sure we release the correct
3409 aio_context
= blk_get_aio_context(job
->blk
);
3410 job_unref(&job
->job
);
3411 aio_context_release(aio_context
);
3414 void qmp_block_job_dismiss(const char *id
, Error
**errp
)
3416 AioContext
*aio_context
;
3417 BlockJob
*bjob
= find_block_job(id
, &aio_context
, errp
);
3424 trace_qmp_block_job_dismiss(bjob
);
3426 job_dismiss(&job
, errp
);
3427 aio_context_release(aio_context
);
3430 void qmp_change_backing_file(const char *device
,
3431 const char *image_node_name
,
3432 const char *backing_file
,
3435 BlockDriverState
*bs
= NULL
;
3436 AioContext
*aio_context
;
3437 BlockDriverState
*image_bs
= NULL
;
3438 Error
*local_err
= NULL
;
3442 bs
= qmp_get_root_bs(device
, errp
);
3447 aio_context
= bdrv_get_aio_context(bs
);
3448 aio_context_acquire(aio_context
);
3450 image_bs
= bdrv_lookup_bs(NULL
, image_node_name
, &local_err
);
3452 error_propagate(errp
, local_err
);
3457 error_setg(errp
, "image file not found");
3461 if (bdrv_find_base(image_bs
) == image_bs
) {
3462 error_setg(errp
, "not allowing backing file change on an image "
3463 "without a backing file");
3467 /* even though we are not necessarily operating on bs, we need it to
3468 * determine if block ops are currently prohibited on the chain */
3469 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_CHANGE
, errp
)) {
3473 /* final sanity check */
3474 if (!bdrv_chain_contains(bs
, image_bs
)) {
3475 error_setg(errp
, "'%s' and image file are not in the same chain",
3480 /* if not r/w, reopen to make r/w */
3481 ro
= bdrv_is_read_only(image_bs
);
3484 if (bdrv_reopen_set_read_only(image_bs
, false, errp
) != 0) {
3489 ret
= bdrv_change_backing_file(image_bs
, backing_file
,
3490 image_bs
->drv
? image_bs
->drv
->format_name
: "",
3494 error_setg_errno(errp
, -ret
, "Could not change backing file to '%s'",
3496 /* don't exit here, so we can try to restore open flags if
3501 bdrv_reopen_set_read_only(image_bs
, true, errp
);
3505 aio_context_release(aio_context
);
3508 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
3510 BlockDriverState
*bs
;
3512 Visitor
*v
= qobject_output_visitor_new(&obj
);
3515 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3516 visit_complete(v
, &obj
);
3517 qdict
= qobject_to(QDict
, obj
);
3519 qdict_flatten(qdict
);
3521 if (!qdict_get_try_str(qdict
, "node-name")) {
3522 error_setg(errp
, "'node-name' must be specified for the root node");
3526 bs
= bds_tree_init(qdict
, errp
);
3531 bdrv_set_monitor_owned(bs
);
3537 void qmp_x_blockdev_reopen(BlockdevOptions
*options
, Error
**errp
)
3539 BlockDriverState
*bs
;
3542 Visitor
*v
= qobject_output_visitor_new(&obj
);
3543 BlockReopenQueue
*queue
;
3546 /* Check for the selected node name */
3547 if (!options
->has_node_name
) {
3548 error_setg(errp
, "Node name not specified");
3552 bs
= bdrv_find_node(options
->node_name
);
3554 error_setg(errp
, "Cannot find node named '%s'", options
->node_name
);
3558 /* Put all options in a QDict and flatten it */
3559 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3560 visit_complete(v
, &obj
);
3561 qdict
= qobject_to(QDict
, obj
);
3563 qdict_flatten(qdict
);
3565 /* Perform the reopen operation */
3566 ctx
= bdrv_get_aio_context(bs
);
3567 aio_context_acquire(ctx
);
3568 bdrv_subtree_drained_begin(bs
);
3569 queue
= bdrv_reopen_queue(NULL
, bs
, qdict
, false);
3570 bdrv_reopen_multiple(queue
, errp
);
3571 bdrv_subtree_drained_end(bs
);
3572 aio_context_release(ctx
);
3578 void qmp_blockdev_del(const char *node_name
, Error
**errp
)
3580 AioContext
*aio_context
;
3581 BlockDriverState
*bs
;
3583 bs
= bdrv_find_node(node_name
);
3585 error_setg(errp
, "Cannot find node %s", node_name
);
3588 if (bdrv_has_blk(bs
)) {
3589 error_setg(errp
, "Node %s is in use", node_name
);
3592 aio_context
= bdrv_get_aio_context(bs
);
3593 aio_context_acquire(aio_context
);
3595 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_DRIVE_DEL
, errp
)) {
3599 if (!QTAILQ_IN_USE(bs
, monitor_list
)) {
3600 error_setg(errp
, "Node %s is not owned by the monitor",
3605 if (bs
->refcnt
> 1) {
3606 error_setg(errp
, "Block device %s is in use",
3607 bdrv_get_device_or_node_name(bs
));
3611 QTAILQ_REMOVE(&monitor_bdrv_states
, bs
, monitor_list
);
3615 aio_context_release(aio_context
);
3618 static BdrvChild
*bdrv_find_child(BlockDriverState
*parent_bs
,
3619 const char *child_name
)
3623 QLIST_FOREACH(child
, &parent_bs
->children
, next
) {
3624 if (strcmp(child
->name
, child_name
) == 0) {
3632 void qmp_x_blockdev_change(const char *parent
, bool has_child
,
3633 const char *child
, bool has_node
,
3634 const char *node
, Error
**errp
)
3636 BlockDriverState
*parent_bs
, *new_bs
= NULL
;
3639 parent_bs
= bdrv_lookup_bs(parent
, parent
, errp
);
3644 if (has_child
== has_node
) {
3646 error_setg(errp
, "The parameters child and node are in conflict");
3648 error_setg(errp
, "Either child or node must be specified");
3654 p_child
= bdrv_find_child(parent_bs
, child
);
3656 error_setg(errp
, "Node '%s' does not have child '%s'",
3660 bdrv_del_child(parent_bs
, p_child
, errp
);
3664 new_bs
= bdrv_find_node(node
);
3666 error_setg(errp
, "Node '%s' not found", node
);
3669 bdrv_add_child(parent_bs
, new_bs
, errp
);
3673 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
3675 BlockJobInfoList
*head
= NULL
, **p_next
= &head
;
3678 for (job
= block_job_next(NULL
); job
; job
= block_job_next(job
)) {
3679 BlockJobInfoList
*elem
;
3680 AioContext
*aio_context
;
3682 if (block_job_is_internal(job
)) {
3685 elem
= g_new0(BlockJobInfoList
, 1);
3686 aio_context
= blk_get_aio_context(job
->blk
);
3687 aio_context_acquire(aio_context
);
3688 elem
->value
= block_job_query(job
, errp
);
3689 aio_context_release(aio_context
);
3692 qapi_free_BlockJobInfoList(head
);
3696 p_next
= &elem
->next
;
3702 void qmp_x_blockdev_set_iothread(const char *node_name
, StrOrNull
*iothread
,
3703 bool has_force
, bool force
, Error
**errp
)
3705 AioContext
*old_context
;
3706 AioContext
*new_context
;
3707 BlockDriverState
*bs
;
3709 bs
= bdrv_find_node(node_name
);
3711 error_setg(errp
, "Cannot find node %s", node_name
);
3715 /* Protects against accidents. */
3716 if (!(has_force
&& force
) && bdrv_has_blk(bs
)) {
3717 error_setg(errp
, "Node %s is associated with a BlockBackend and could "
3718 "be in use (use force=true to override this check)",
3723 if (iothread
->type
== QTYPE_QSTRING
) {
3724 IOThread
*obj
= iothread_by_id(iothread
->u
.s
);
3726 error_setg(errp
, "Cannot find iothread %s", iothread
->u
.s
);
3730 new_context
= iothread_get_aio_context(obj
);
3732 new_context
= qemu_get_aio_context();
3735 old_context
= bdrv_get_aio_context(bs
);
3736 aio_context_acquire(old_context
);
3738 bdrv_try_set_aio_context(bs
, new_context
, errp
);
3740 aio_context_release(old_context
);
3743 QemuOptsList qemu_common_drive_opts
= {
3745 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
3749 .type
= QEMU_OPT_BOOL
,
3750 .help
= "enable/disable snapshot mode",
3753 .type
= QEMU_OPT_STRING
,
3754 .help
= "host AIO implementation (threads, native, io_uring)",
3756 .name
= BDRV_OPT_CACHE_WB
,
3757 .type
= QEMU_OPT_BOOL
,
3758 .help
= "Enable writeback mode",
3761 .type
= QEMU_OPT_STRING
,
3762 .help
= "disk format (raw, qcow2, ...)",
3765 .type
= QEMU_OPT_STRING
,
3766 .help
= "read error action",
3769 .type
= QEMU_OPT_STRING
,
3770 .help
= "write error action",
3772 .name
= BDRV_OPT_READ_ONLY
,
3773 .type
= QEMU_OPT_BOOL
,
3774 .help
= "open drive file as read-only",
3780 .name
= "throttling.group",
3781 .type
= QEMU_OPT_STRING
,
3782 .help
= "name of the block throttling group",
3784 .name
= "copy-on-read",
3785 .type
= QEMU_OPT_BOOL
,
3786 .help
= "copy read data from backing file into image file",
3788 .name
= "detect-zeroes",
3789 .type
= QEMU_OPT_STRING
,
3790 .help
= "try to optimize zero writes (off, on, unmap)",
3792 .name
= "stats-account-invalid",
3793 .type
= QEMU_OPT_BOOL
,
3794 .help
= "whether to account for invalid I/O operations "
3795 "in the statistics",
3797 .name
= "stats-account-failed",
3798 .type
= QEMU_OPT_BOOL
,
3799 .help
= "whether to account for failed I/O operations "
3800 "in the statistics",
3802 { /* end of list */ }
3806 QemuOptsList qemu_drive_opts
= {
3808 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
3811 * no elements => accept any params
3812 * validation will happen later
3814 { /* end of list */ }