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/runstate.h"
60 #include "sysemu/replay.h"
61 #include "qemu/cutils.h"
62 #include "qemu/help_option.h"
63 #include "qemu/main-loop.h"
64 #include "qemu/throttle-options.h"
66 /* Protected by BQL */
67 QTAILQ_HEAD(, BlockDriverState
) monitor_bdrv_states
=
68 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states
);
70 void bdrv_set_monitor_owned(BlockDriverState
*bs
)
72 QTAILQ_INSERT_TAIL(&monitor_bdrv_states
, bs
, monitor_list
);
75 static const char *const if_name
[IF_COUNT
] = {
79 [IF_FLOPPY
] = "floppy",
80 [IF_PFLASH
] = "pflash",
83 [IF_VIRTIO
] = "virtio",
87 static int if_max_devs
[IF_COUNT
] = {
89 * Do not change these numbers! They govern how drive option
90 * index maps to unit and bus. That mapping is ABI.
92 * All controllers used to implement if=T drives need to support
93 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
94 * Otherwise, some index values map to "impossible" bus, unit
97 * For instance, if you change [IF_SCSI] to 255, -drive
98 * if=scsi,index=12 no longer means bus=1,unit=5, but
99 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
100 * the drive can't be set up. Regression.
107 * Boards may call this to offer board-by-board overrides
108 * of the default, global values.
110 void override_max_devs(BlockInterfaceType type
, int max_devs
)
119 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
120 dinfo
= blk_legacy_dinfo(blk
);
121 if (dinfo
->type
== type
) {
122 fprintf(stderr
, "Cannot override units-per-bus property of"
123 " the %s interface, because a drive of that type has"
124 " already been added.\n", if_name
[type
]);
125 g_assert_not_reached();
129 if_max_devs
[type
] = max_devs
;
133 * We automatically delete the drive when a device using it gets
134 * unplugged. Questionable feature, but we can't just drop it.
135 * Device models call blockdev_mark_auto_del() to schedule the
136 * automatic deletion, and generic qdev code calls blockdev_auto_del()
137 * when deletion is actually safe.
139 void blockdev_mark_auto_del(BlockBackend
*blk
)
141 DriveInfo
*dinfo
= blk_legacy_dinfo(blk
);
148 for (job
= block_job_next(NULL
); job
; job
= block_job_next(job
)) {
149 if (block_job_has_bdrv(job
, blk_bs(blk
))) {
150 AioContext
*aio_context
= job
->job
.aio_context
;
151 aio_context_acquire(aio_context
);
153 job_cancel(&job
->job
, false);
155 aio_context_release(aio_context
);
162 void blockdev_auto_del(BlockBackend
*blk
)
164 DriveInfo
*dinfo
= blk_legacy_dinfo(blk
);
166 if (dinfo
&& dinfo
->auto_del
) {
167 monitor_remove_blk(blk
);
172 static int drive_index_to_bus_id(BlockInterfaceType type
, int index
)
174 int max_devs
= if_max_devs
[type
];
175 return max_devs
? index
/ max_devs
: 0;
178 static int drive_index_to_unit_id(BlockInterfaceType type
, int index
)
180 int max_devs
= if_max_devs
[type
];
181 return max_devs
? index
% max_devs
: index
;
184 QemuOpts
*drive_add(BlockInterfaceType type
, int index
, const char *file
,
189 opts
= qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr
, false);
193 if (type
!= IF_DEFAULT
) {
194 qemu_opt_set(opts
, "if", if_name
[type
], &error_abort
);
197 qemu_opt_set_number(opts
, "index", index
, &error_abort
);
200 qemu_opt_set(opts
, "file", file
, &error_abort
);
204 DriveInfo
*drive_get(BlockInterfaceType type
, int bus
, int unit
)
209 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
210 dinfo
= blk_legacy_dinfo(blk
);
211 if (dinfo
&& dinfo
->type
== type
212 && dinfo
->bus
== bus
&& dinfo
->unit
== unit
) {
221 * Check board claimed all -drive that are meant to be claimed.
222 * Fatal error if any remain unclaimed.
224 void drive_check_orphaned(void)
229 bool orphans
= false;
231 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
232 dinfo
= blk_legacy_dinfo(blk
);
234 * Ignore default drives, because we create certain default
235 * drives unconditionally, then leave them unclaimed. Not the
237 * Ignore IF_VIRTIO, because it gets desugared into -device,
238 * so we can leave failing to -device.
239 * Ignore IF_NONE, because leaving unclaimed IF_NONE remains
240 * available for device_add is a feature.
242 if (dinfo
->is_default
|| dinfo
->type
== IF_VIRTIO
243 || dinfo
->type
== IF_NONE
) {
246 if (!blk_get_attached_dev(blk
)) {
248 qemu_opts_loc_restore(dinfo
->opts
);
249 error_report("machine type does not support"
250 " if=%s,bus=%d,unit=%d",
251 if_name
[dinfo
->type
], dinfo
->bus
, dinfo
->unit
);
262 DriveInfo
*drive_get_by_index(BlockInterfaceType type
, int index
)
264 return drive_get(type
,
265 drive_index_to_bus_id(type
, index
),
266 drive_index_to_unit_id(type
, index
));
269 int drive_get_max_bus(BlockInterfaceType type
)
276 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
277 dinfo
= blk_legacy_dinfo(blk
);
278 if (dinfo
&& dinfo
->type
== type
&& dinfo
->bus
> max_bus
) {
279 max_bus
= dinfo
->bus
;
285 static void bdrv_format_print(void *opaque
, const char *name
)
287 qemu_printf(" %s", name
);
292 BlockDriverState
*bs
;
295 static int parse_block_error_action(const char *buf
, bool is_read
, Error
**errp
)
297 if (!strcmp(buf
, "ignore")) {
298 return BLOCKDEV_ON_ERROR_IGNORE
;
299 } else if (!is_read
&& !strcmp(buf
, "enospc")) {
300 return BLOCKDEV_ON_ERROR_ENOSPC
;
301 } else if (!strcmp(buf
, "stop")) {
302 return BLOCKDEV_ON_ERROR_STOP
;
303 } else if (!strcmp(buf
, "report")) {
304 return BLOCKDEV_ON_ERROR_REPORT
;
306 error_setg(errp
, "'%s' invalid %s error action",
307 buf
, is_read
? "read" : "write");
312 static bool parse_stats_intervals(BlockAcctStats
*stats
, QList
*intervals
,
315 const QListEntry
*entry
;
316 for (entry
= qlist_first(intervals
); entry
; entry
= qlist_next(entry
)) {
317 switch (qobject_type(entry
->value
)) {
319 case QTYPE_QSTRING
: {
320 unsigned long long length
;
321 const char *str
= qstring_get_str(qobject_to(QString
,
323 if (parse_uint_full(str
, &length
, 10) == 0 &&
324 length
> 0 && length
<= UINT_MAX
) {
325 block_acct_add_interval(stats
, (unsigned) length
);
327 error_setg(errp
, "Invalid interval length: %s", str
);
334 int64_t length
= qnum_get_int(qobject_to(QNum
, entry
->value
));
336 if (length
> 0 && length
<= UINT_MAX
) {
337 block_acct_add_interval(stats
, (unsigned) length
);
339 error_setg(errp
, "Invalid interval length: %" PRId64
, length
);
346 error_setg(errp
, "The specification of stats-intervals is invalid");
353 typedef enum { MEDIA_DISK
, MEDIA_CDROM
} DriveMediaType
;
355 /* All parameters but @opts are optional and may be set to NULL. */
356 static void extract_common_blockdev_options(QemuOpts
*opts
, int *bdrv_flags
,
357 const char **throttling_group
, ThrottleConfig
*throttle_cfg
,
358 BlockdevDetectZeroesOptions
*detect_zeroes
, Error
**errp
)
360 Error
*local_error
= NULL
;
364 if (qemu_opt_get_bool(opts
, "copy-on-read", false)) {
365 *bdrv_flags
|= BDRV_O_COPY_ON_READ
;
368 if ((aio
= qemu_opt_get(opts
, "aio")) != NULL
) {
369 if (bdrv_parse_aio(aio
, bdrv_flags
) < 0) {
370 error_setg(errp
, "invalid aio option");
376 /* disk I/O throttling */
377 if (throttling_group
) {
378 *throttling_group
= qemu_opt_get(opts
, "throttling.group");
382 throttle_config_init(throttle_cfg
);
383 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].avg
=
384 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
385 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].avg
=
386 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
387 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].avg
=
388 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
389 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].avg
=
390 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
391 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].avg
=
392 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
393 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].avg
=
394 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
396 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].max
=
397 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
398 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].max
=
399 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
400 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].max
=
401 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
402 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].max
=
403 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
404 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].max
=
405 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
406 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].max
=
407 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
409 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].burst_length
=
410 qemu_opt_get_number(opts
, "throttling.bps-total-max-length", 1);
411 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].burst_length
=
412 qemu_opt_get_number(opts
, "throttling.bps-read-max-length", 1);
413 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].burst_length
=
414 qemu_opt_get_number(opts
, "throttling.bps-write-max-length", 1);
415 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].burst_length
=
416 qemu_opt_get_number(opts
, "throttling.iops-total-max-length", 1);
417 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].burst_length
=
418 qemu_opt_get_number(opts
, "throttling.iops-read-max-length", 1);
419 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].burst_length
=
420 qemu_opt_get_number(opts
, "throttling.iops-write-max-length", 1);
422 throttle_cfg
->op_size
=
423 qemu_opt_get_number(opts
, "throttling.iops-size", 0);
425 if (!throttle_is_valid(throttle_cfg
, errp
)) {
432 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup
,
433 qemu_opt_get(opts
, "detect-zeroes"),
434 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
437 error_propagate(errp
, local_error
);
443 /* Takes the ownership of bs_opts */
444 static BlockBackend
*blockdev_init(const char *file
, QDict
*bs_opts
,
449 int on_read_error
, on_write_error
;
450 bool account_invalid
, account_failed
;
451 bool writethrough
, read_only
;
453 BlockDriverState
*bs
;
458 QDict
*interval_dict
= NULL
;
459 QList
*interval_list
= NULL
;
461 BlockdevDetectZeroesOptions detect_zeroes
=
462 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
463 const char *throttling_group
= NULL
;
465 /* Check common options by copying from bs_opts to opts, all other options
466 * stay in bs_opts for processing by bdrv_open(). */
467 id
= qdict_get_try_str(bs_opts
, "id");
468 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, errp
);
473 if (!qemu_opts_absorb_qdict(opts
, bs_opts
, errp
)) {
478 qdict_del(bs_opts
, "id");
481 /* extract parameters */
482 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
484 account_invalid
= qemu_opt_get_bool(opts
, "stats-account-invalid", true);
485 account_failed
= qemu_opt_get_bool(opts
, "stats-account-failed", true);
487 writethrough
= !qemu_opt_get_bool(opts
, BDRV_OPT_CACHE_WB
, true);
489 id
= qemu_opts_id(opts
);
491 qdict_extract_subqdict(bs_opts
, &interval_dict
, "stats-intervals.");
492 qdict_array_split(interval_dict
, &interval_list
);
494 if (qdict_size(interval_dict
) != 0) {
495 error_setg(errp
, "Invalid option stats-intervals.%s",
496 qdict_first(interval_dict
)->key
);
500 extract_common_blockdev_options(opts
, &bdrv_flags
, &throttling_group
, &cfg
,
501 &detect_zeroes
, &error
);
503 error_propagate(errp
, error
);
507 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
508 if (is_help_option(buf
)) {
509 qemu_printf("Supported formats:");
510 bdrv_iterate_format(bdrv_format_print
, NULL
, false);
511 qemu_printf("\nSupported formats (read-only):");
512 bdrv_iterate_format(bdrv_format_print
, NULL
, true);
517 if (qdict_haskey(bs_opts
, "driver")) {
518 error_setg(errp
, "Cannot specify both 'driver' and 'format'");
521 qdict_put_str(bs_opts
, "driver", buf
);
524 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
525 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
526 on_write_error
= parse_block_error_action(buf
, 0, &error
);
528 error_propagate(errp
, error
);
533 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
534 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
535 on_read_error
= parse_block_error_action(buf
, 1, &error
);
537 error_propagate(errp
, error
);
543 bdrv_flags
|= BDRV_O_SNAPSHOT
;
546 read_only
= qemu_opt_get_bool(opts
, BDRV_OPT_READ_ONLY
, false);
549 if ((!file
|| !*file
) && !qdict_size(bs_opts
)) {
550 BlockBackendRootState
*blk_rs
;
552 blk
= blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL
);
553 blk_rs
= blk_get_root_state(blk
);
554 blk_rs
->open_flags
= bdrv_flags
| (read_only
? 0 : BDRV_O_RDWR
);
555 blk_rs
->detect_zeroes
= detect_zeroes
;
557 qobject_unref(bs_opts
);
559 if (file
&& !*file
) {
563 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
564 * with other callers) rather than what we want as the real defaults.
565 * Apply the defaults here instead. */
566 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_DIRECT
, "off");
567 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_NO_FLUSH
, "off");
568 qdict_set_default_str(bs_opts
, BDRV_OPT_READ_ONLY
,
569 read_only
? "on" : "off");
570 qdict_set_default_str(bs_opts
, BDRV_OPT_AUTO_READ_ONLY
, "on");
571 assert((bdrv_flags
& BDRV_O_CACHE_MASK
) == 0);
573 if (runstate_check(RUN_STATE_INMIGRATE
)) {
574 bdrv_flags
|= BDRV_O_INACTIVE
;
577 blk
= blk_new_open(file
, NULL
, bs_opts
, bdrv_flags
, errp
);
583 bs
->detect_zeroes
= detect_zeroes
;
585 block_acct_setup(blk_get_stats(blk
), account_invalid
, account_failed
);
587 if (!parse_stats_intervals(blk_get_stats(blk
), interval_list
, errp
)) {
594 /* disk I/O throttling */
595 if (throttle_enabled(&cfg
)) {
596 if (!throttling_group
) {
597 throttling_group
= id
;
599 blk_io_limits_enable(blk
, throttling_group
);
600 blk_set_io_limits(blk
, &cfg
);
603 blk_set_enable_write_cache(blk
, !writethrough
);
604 blk_set_on_error(blk
, on_read_error
, on_write_error
);
606 if (!monitor_add_blk(blk
, id
, errp
)) {
614 qobject_unref(interval_dict
);
615 qobject_unref(interval_list
);
620 qobject_unref(interval_dict
);
621 qobject_unref(interval_list
);
623 qobject_unref(bs_opts
);
627 /* Takes the ownership of bs_opts */
628 BlockDriverState
*bds_tree_init(QDict
*bs_opts
, Error
**errp
)
632 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
633 * with other callers) rather than what we want as the real defaults.
634 * Apply the defaults here instead. */
635 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_DIRECT
, "off");
636 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_NO_FLUSH
, "off");
637 qdict_set_default_str(bs_opts
, BDRV_OPT_READ_ONLY
, "off");
639 if (runstate_check(RUN_STATE_INMIGRATE
)) {
640 bdrv_flags
|= BDRV_O_INACTIVE
;
643 return bdrv_open(NULL
, NULL
, bs_opts
, bdrv_flags
, errp
);
646 void blockdev_close_all_bdrv_states(void)
648 BlockDriverState
*bs
, *next_bs
;
650 QTAILQ_FOREACH_SAFE(bs
, &monitor_bdrv_states
, monitor_list
, next_bs
) {
651 AioContext
*ctx
= bdrv_get_aio_context(bs
);
653 aio_context_acquire(ctx
);
655 aio_context_release(ctx
);
659 /* Iterates over the list of monitor-owned BlockDriverStates */
660 BlockDriverState
*bdrv_next_monitor_owned(BlockDriverState
*bs
)
663 return bs
? QTAILQ_NEXT(bs
, monitor_list
)
664 : QTAILQ_FIRST(&monitor_bdrv_states
);
667 static bool qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
,
672 value
= qemu_opt_get(opts
, from
);
674 if (qemu_opt_find(opts
, to
)) {
675 error_setg(errp
, "'%s' and its alias '%s' can't be used at the "
676 "same time", to
, from
);
681 /* rename all items in opts */
682 while ((value
= qemu_opt_get(opts
, from
))) {
683 qemu_opt_set(opts
, to
, value
, &error_abort
);
684 qemu_opt_unset(opts
, from
);
689 QemuOptsList qemu_legacy_drive_opts
= {
691 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
695 .type
= QEMU_OPT_NUMBER
,
696 .help
= "bus number",
699 .type
= QEMU_OPT_NUMBER
,
700 .help
= "unit number (i.e. lun for scsi)",
703 .type
= QEMU_OPT_NUMBER
,
704 .help
= "index number",
707 .type
= QEMU_OPT_STRING
,
708 .help
= "media type (disk, cdrom)",
711 .type
= QEMU_OPT_STRING
,
712 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
715 .type
= QEMU_OPT_STRING
,
719 /* Options that are passed on, but have special semantics with -drive */
721 .name
= BDRV_OPT_READ_ONLY
,
722 .type
= QEMU_OPT_BOOL
,
723 .help
= "open drive file as read-only",
726 .type
= QEMU_OPT_STRING
,
727 .help
= "read error action",
730 .type
= QEMU_OPT_STRING
,
731 .help
= "write error action",
733 .name
= "copy-on-read",
734 .type
= QEMU_OPT_BOOL
,
735 .help
= "copy read data from backing file into image file",
738 { /* end of list */ }
742 DriveInfo
*drive_new(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
,
747 DriveInfo
*dinfo
= NULL
;
749 QemuOpts
*legacy_opts
;
750 DriveMediaType media
= MEDIA_DISK
;
751 BlockInterfaceType type
;
752 int max_devs
, bus_id
, unit_id
, index
;
753 const char *werror
, *rerror
;
754 bool read_only
= false;
756 const char *filename
;
759 /* Change legacy command line options into QMP ones */
760 static const struct {
764 { "iops", "throttling.iops-total" },
765 { "iops_rd", "throttling.iops-read" },
766 { "iops_wr", "throttling.iops-write" },
768 { "bps", "throttling.bps-total" },
769 { "bps_rd", "throttling.bps-read" },
770 { "bps_wr", "throttling.bps-write" },
772 { "iops_max", "throttling.iops-total-max" },
773 { "iops_rd_max", "throttling.iops-read-max" },
774 { "iops_wr_max", "throttling.iops-write-max" },
776 { "bps_max", "throttling.bps-total-max" },
777 { "bps_rd_max", "throttling.bps-read-max" },
778 { "bps_wr_max", "throttling.bps-write-max" },
780 { "iops_size", "throttling.iops-size" },
782 { "group", "throttling.group" },
784 { "readonly", BDRV_OPT_READ_ONLY
},
787 for (i
= 0; i
< ARRAY_SIZE(opt_renames
); i
++) {
788 if (!qemu_opt_rename(all_opts
, opt_renames
[i
].from
,
789 opt_renames
[i
].to
, errp
)) {
794 value
= qemu_opt_get(all_opts
, "cache");
799 if (bdrv_parse_cache_mode(value
, &flags
, &writethrough
) != 0) {
800 error_setg(errp
, "invalid cache option");
804 /* Specific options take precedence */
805 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_WB
)) {
806 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_WB
,
807 !writethrough
, &error_abort
);
809 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_DIRECT
)) {
810 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_DIRECT
,
811 !!(flags
& BDRV_O_NOCACHE
), &error_abort
);
813 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
)) {
814 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
,
815 !!(flags
& BDRV_O_NO_FLUSH
), &error_abort
);
817 qemu_opt_unset(all_opts
, "cache");
820 /* Get a QDict for processing the options */
821 bs_opts
= qdict_new();
822 qemu_opts_to_qdict(all_opts
, bs_opts
);
824 legacy_opts
= qemu_opts_create(&qemu_legacy_drive_opts
, NULL
, 0,
826 if (!qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, errp
)) {
831 value
= qemu_opt_get(legacy_opts
, "media");
833 if (!strcmp(value
, "disk")) {
835 } else if (!strcmp(value
, "cdrom")) {
839 error_setg(errp
, "'%s' invalid media", value
);
844 /* copy-on-read is disabled with a warning for read-only devices */
845 read_only
|= qemu_opt_get_bool(legacy_opts
, BDRV_OPT_READ_ONLY
, false);
846 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
848 if (read_only
&& copy_on_read
) {
849 warn_report("disabling copy-on-read on read-only drive");
850 copy_on_read
= false;
853 qdict_put_str(bs_opts
, BDRV_OPT_READ_ONLY
, read_only
? "on" : "off");
854 qdict_put_str(bs_opts
, "copy-on-read", copy_on_read
? "on" : "off");
856 /* Controller type */
857 value
= qemu_opt_get(legacy_opts
, "if");
860 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
863 if (type
== IF_COUNT
) {
864 error_setg(errp
, "unsupported bus type '%s'", value
);
868 type
= block_default_type
;
871 /* Device address specified by bus/unit or index.
872 * If none was specified, try to find the first free one. */
873 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
874 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
875 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
877 max_devs
= if_max_devs
[type
];
880 if (bus_id
!= 0 || unit_id
!= -1) {
881 error_setg(errp
, "index cannot be used with bus and unit");
884 bus_id
= drive_index_to_bus_id(type
, index
);
885 unit_id
= drive_index_to_unit_id(type
, index
);
890 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
892 if (max_devs
&& unit_id
>= max_devs
) {
899 if (max_devs
&& unit_id
>= max_devs
) {
900 error_setg(errp
, "unit %d too big (max is %d)", unit_id
, max_devs
- 1);
904 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
905 error_setg(errp
, "drive with bus=%d, unit=%d (index=%d) exists",
906 bus_id
, unit_id
, index
);
910 /* no id supplied -> create one */
911 if (qemu_opts_id(all_opts
) == NULL
) {
913 const char *mediastr
= "";
914 if (type
== IF_IDE
|| type
== IF_SCSI
) {
915 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
918 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
921 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
924 qdict_put_str(bs_opts
, "id", new_id
);
928 /* Add virtio block device */
929 if (type
== IF_VIRTIO
) {
931 devopts
= qemu_opts_create(qemu_find_opts("device"), NULL
, 0,
933 qemu_opt_set(devopts
, "driver", "virtio-blk", &error_abort
);
934 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"),
938 filename
= qemu_opt_get(legacy_opts
, "file");
940 /* Check werror/rerror compatibility with if=... */
941 werror
= qemu_opt_get(legacy_opts
, "werror");
942 if (werror
!= NULL
) {
943 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&&
945 error_setg(errp
, "werror is not supported by this bus type");
948 qdict_put_str(bs_opts
, "werror", werror
);
951 rerror
= qemu_opt_get(legacy_opts
, "rerror");
952 if (rerror
!= NULL
) {
953 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&&
955 error_setg(errp
, "rerror is not supported by this bus type");
958 qdict_put_str(bs_opts
, "rerror", rerror
);
961 /* Actual block device init: Functionality shared with blockdev-add */
962 blk
= blockdev_init(filename
, bs_opts
, errp
);
968 /* Create legacy DriveInfo */
969 dinfo
= g_malloc0(sizeof(*dinfo
));
970 dinfo
->opts
= all_opts
;
974 dinfo
->unit
= unit_id
;
976 blk_set_legacy_dinfo(blk
, dinfo
);
983 dinfo
->media_cd
= media
== MEDIA_CDROM
;
990 qemu_opts_del(legacy_opts
);
991 qobject_unref(bs_opts
);
995 static BlockDriverState
*qmp_get_root_bs(const char *name
, Error
**errp
)
997 BlockDriverState
*bs
;
999 bs
= bdrv_lookup_bs(name
, name
, errp
);
1004 if (!bdrv_is_root_node(bs
)) {
1005 error_setg(errp
, "Need a root block node");
1009 if (!bdrv_is_inserted(bs
)) {
1010 error_setg(errp
, "Device has no medium");
1017 static void blockdev_do_action(TransactionAction
*action
, Error
**errp
)
1019 TransactionActionList list
;
1021 list
.value
= action
;
1023 qmp_transaction(&list
, false, NULL
, errp
);
1026 void qmp_blockdev_snapshot_sync(bool has_device
, const char *device
,
1027 bool has_node_name
, const char *node_name
,
1028 const char *snapshot_file
,
1029 bool has_snapshot_node_name
,
1030 const char *snapshot_node_name
,
1031 bool has_format
, const char *format
,
1032 bool has_mode
, NewImageMode mode
, Error
**errp
)
1034 BlockdevSnapshotSync snapshot
= {
1035 .has_device
= has_device
,
1036 .device
= (char *) device
,
1037 .has_node_name
= has_node_name
,
1038 .node_name
= (char *) node_name
,
1039 .snapshot_file
= (char *) snapshot_file
,
1040 .has_snapshot_node_name
= has_snapshot_node_name
,
1041 .snapshot_node_name
= (char *) snapshot_node_name
,
1042 .has_format
= has_format
,
1043 .format
= (char *) format
,
1044 .has_mode
= has_mode
,
1047 TransactionAction action
= {
1048 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
1049 .u
.blockdev_snapshot_sync
.data
= &snapshot
,
1051 blockdev_do_action(&action
, errp
);
1054 void qmp_blockdev_snapshot(const char *node
, const char *overlay
,
1057 BlockdevSnapshot snapshot_data
= {
1058 .node
= (char *) node
,
1059 .overlay
= (char *) overlay
1061 TransactionAction action
= {
1062 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
,
1063 .u
.blockdev_snapshot
.data
= &snapshot_data
,
1065 blockdev_do_action(&action
, errp
);
1068 void qmp_blockdev_snapshot_internal_sync(const char *device
,
1072 BlockdevSnapshotInternal snapshot
= {
1073 .device
= (char *) device
,
1074 .name
= (char *) name
1076 TransactionAction action
= {
1077 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
1078 .u
.blockdev_snapshot_internal_sync
.data
= &snapshot
,
1080 blockdev_do_action(&action
, errp
);
1083 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
1090 BlockDriverState
*bs
;
1091 AioContext
*aio_context
;
1092 QEMUSnapshotInfo sn
;
1093 Error
*local_err
= NULL
;
1094 SnapshotInfo
*info
= NULL
;
1097 bs
= qmp_get_root_bs(device
, errp
);
1101 aio_context
= bdrv_get_aio_context(bs
);
1102 aio_context_acquire(aio_context
);
1113 error_setg(errp
, "Name or id must be provided");
1114 goto out_aio_context
;
1117 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE
, errp
)) {
1118 goto out_aio_context
;
1121 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1123 error_propagate(errp
, local_err
);
1124 goto out_aio_context
;
1128 "Snapshot with id '%s' and name '%s' does not exist on "
1130 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1131 goto out_aio_context
;
1134 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1136 error_propagate(errp
, local_err
);
1137 goto out_aio_context
;
1140 aio_context_release(aio_context
);
1142 info
= g_new0(SnapshotInfo
, 1);
1143 info
->id
= g_strdup(sn
.id_str
);
1144 info
->name
= g_strdup(sn
.name
);
1145 info
->date_nsec
= sn
.date_nsec
;
1146 info
->date_sec
= sn
.date_sec
;
1147 info
->vm_state_size
= sn
.vm_state_size
;
1148 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1149 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1150 if (sn
.icount
!= -1ULL) {
1151 info
->icount
= sn
.icount
;
1152 info
->has_icount
= true;
1158 aio_context_release(aio_context
);
1162 /* New and old BlockDriverState structs for atomic group operations */
1164 typedef struct BlkActionState BlkActionState
;
1168 * Table of operations that define an Action.
1170 * @instance_size: Size of state struct, in bytes.
1171 * @prepare: Prepare the work, must NOT be NULL.
1172 * @commit: Commit the changes, can be NULL.
1173 * @abort: Abort the changes on fail, can be NULL.
1174 * @clean: Clean up resources after all transaction actions have called
1175 * commit() or abort(). Can be NULL.
1177 * Only prepare() may fail. In a single transaction, only one of commit() or
1178 * abort() will be called. clean() will always be called if it is present.
1180 * Always run under BQL.
1182 typedef struct BlkActionOps
{
1183 size_t instance_size
;
1184 void (*prepare
)(BlkActionState
*common
, Error
**errp
);
1185 void (*commit
)(BlkActionState
*common
);
1186 void (*abort
)(BlkActionState
*common
);
1187 void (*clean
)(BlkActionState
*common
);
1192 * Describes one Action's state within a Transaction.
1194 * @action: QAPI-defined enum identifying which Action to perform.
1195 * @ops: Table of ActionOps this Action can perform.
1196 * @block_job_txn: Transaction which this action belongs to.
1197 * @entry: List membership for all Actions in this Transaction.
1199 * This structure must be arranged as first member in a subclassed type,
1200 * assuming that the compiler will also arrange it to the same offsets as the
1203 struct BlkActionState
{
1204 TransactionAction
*action
;
1205 const BlkActionOps
*ops
;
1206 JobTxn
*block_job_txn
;
1207 TransactionProperties
*txn_props
;
1208 QTAILQ_ENTRY(BlkActionState
) entry
;
1211 /* internal snapshot private data */
1212 typedef struct InternalSnapshotState
{
1213 BlkActionState common
;
1214 BlockDriverState
*bs
;
1215 QEMUSnapshotInfo sn
;
1217 } InternalSnapshotState
;
1220 static int action_check_completion_mode(BlkActionState
*s
, Error
**errp
)
1222 if (s
->txn_props
->completion_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
1224 "Action '%s' does not support Transaction property "
1225 "completion-mode = %s",
1226 TransactionActionKind_str(s
->action
->type
),
1227 ActionCompletionMode_str(s
->txn_props
->completion_mode
));
1233 static void internal_snapshot_prepare(BlkActionState
*common
,
1236 Error
*local_err
= NULL
;
1239 BlockDriverState
*bs
;
1240 QEMUSnapshotInfo old_sn
, *sn
;
1243 BlockdevSnapshotInternal
*internal
;
1244 InternalSnapshotState
*state
;
1245 AioContext
*aio_context
;
1248 g_assert(common
->action
->type
==
1249 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1250 internal
= common
->action
->u
.blockdev_snapshot_internal_sync
.data
;
1251 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1253 /* 1. parse input */
1254 device
= internal
->device
;
1255 name
= internal
->name
;
1257 /* 2. check for validation */
1258 if (action_check_completion_mode(common
, errp
) < 0) {
1262 bs
= qmp_get_root_bs(device
, errp
);
1267 aio_context
= bdrv_get_aio_context(bs
);
1268 aio_context_acquire(aio_context
);
1272 /* Paired with .clean() */
1273 bdrv_drained_begin(bs
);
1275 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT
, errp
)) {
1279 if (bdrv_is_read_only(bs
)) {
1280 error_setg(errp
, "Device '%s' is read only", device
);
1284 if (!bdrv_can_snapshot(bs
)) {
1285 error_setg(errp
, "Block format '%s' used by device '%s' "
1286 "does not support internal snapshots",
1287 bs
->drv
->format_name
, device
);
1291 if (!strlen(name
)) {
1292 error_setg(errp
, "Name is empty");
1296 /* check whether a snapshot with name exist */
1297 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
,
1300 error_propagate(errp
, local_err
);
1304 "Snapshot with name '%s' already exists on device '%s'",
1309 /* 3. take the snapshot */
1311 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1312 qemu_gettimeofday(&tv
);
1313 sn
->date_sec
= tv
.tv_sec
;
1314 sn
->date_nsec
= tv
.tv_usec
* 1000;
1315 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1316 if (replay_mode
!= REPLAY_MODE_NONE
) {
1317 sn
->icount
= replay_get_current_icount();
1322 ret1
= bdrv_snapshot_create(bs
, sn
);
1324 error_setg_errno(errp
, -ret1
,
1325 "Failed to create snapshot '%s' on device '%s'",
1330 /* 4. succeed, mark a snapshot is created */
1331 state
->created
= true;
1334 aio_context_release(aio_context
);
1337 static void internal_snapshot_abort(BlkActionState
*common
)
1339 InternalSnapshotState
*state
=
1340 DO_UPCAST(InternalSnapshotState
, common
, common
);
1341 BlockDriverState
*bs
= state
->bs
;
1342 QEMUSnapshotInfo
*sn
= &state
->sn
;
1343 AioContext
*aio_context
;
1344 Error
*local_error
= NULL
;
1346 if (!state
->created
) {
1350 aio_context
= bdrv_get_aio_context(state
->bs
);
1351 aio_context_acquire(aio_context
);
1353 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1354 error_reportf_err(local_error
,
1355 "Failed to delete snapshot with id '%s' and "
1356 "name '%s' on device '%s' in abort: ",
1357 sn
->id_str
, sn
->name
,
1358 bdrv_get_device_name(bs
));
1361 aio_context_release(aio_context
);
1364 static void internal_snapshot_clean(BlkActionState
*common
)
1366 InternalSnapshotState
*state
= DO_UPCAST(InternalSnapshotState
,
1368 AioContext
*aio_context
;
1374 aio_context
= bdrv_get_aio_context(state
->bs
);
1375 aio_context_acquire(aio_context
);
1377 bdrv_drained_end(state
->bs
);
1379 aio_context_release(aio_context
);
1382 /* external snapshot private data */
1383 typedef struct ExternalSnapshotState
{
1384 BlkActionState common
;
1385 BlockDriverState
*old_bs
;
1386 BlockDriverState
*new_bs
;
1387 bool overlay_appended
;
1388 } ExternalSnapshotState
;
1390 static void external_snapshot_prepare(BlkActionState
*common
,
1395 QDict
*options
= NULL
;
1396 Error
*local_err
= NULL
;
1397 /* Device and node name of the image to generate the snapshot from */
1399 const char *node_name
;
1400 /* Reference to the new image (for 'blockdev-snapshot') */
1401 const char *snapshot_ref
;
1402 /* File name of the new image (for 'blockdev-snapshot-sync') */
1403 const char *new_image_file
;
1404 ExternalSnapshotState
*state
=
1405 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1406 TransactionAction
*action
= common
->action
;
1407 AioContext
*aio_context
;
1408 uint64_t perm
, shared
;
1410 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1411 * purpose but a different set of parameters */
1412 switch (action
->type
) {
1413 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
:
1415 BlockdevSnapshot
*s
= action
->u
.blockdev_snapshot
.data
;
1417 node_name
= s
->node
;
1418 new_image_file
= NULL
;
1419 snapshot_ref
= s
->overlay
;
1422 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
:
1424 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1425 device
= s
->has_device
? s
->device
: NULL
;
1426 node_name
= s
->has_node_name
? s
->node_name
: NULL
;
1427 new_image_file
= s
->snapshot_file
;
1428 snapshot_ref
= NULL
;
1432 g_assert_not_reached();
1435 /* start processing */
1436 if (action_check_completion_mode(common
, errp
) < 0) {
1440 state
->old_bs
= bdrv_lookup_bs(device
, node_name
, errp
);
1441 if (!state
->old_bs
) {
1445 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1446 aio_context_acquire(aio_context
);
1448 /* Paired with .clean() */
1449 bdrv_drained_begin(state
->old_bs
);
1451 if (!bdrv_is_inserted(state
->old_bs
)) {
1452 error_setg(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1456 if (bdrv_op_is_blocked(state
->old_bs
,
1457 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT
, errp
)) {
1461 if (!bdrv_is_read_only(state
->old_bs
)) {
1462 if (bdrv_flush(state
->old_bs
)) {
1463 error_setg(errp
, QERR_IO_ERROR
);
1468 if (action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
) {
1469 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1470 const char *format
= s
->has_format
? s
->format
: "qcow2";
1471 enum NewImageMode mode
;
1472 const char *snapshot_node_name
=
1473 s
->has_snapshot_node_name
? s
->snapshot_node_name
: NULL
;
1475 if (node_name
&& !snapshot_node_name
) {
1476 error_setg(errp
, "New overlay node-name missing");
1480 if (snapshot_node_name
&&
1481 bdrv_lookup_bs(snapshot_node_name
, snapshot_node_name
, NULL
)) {
1482 error_setg(errp
, "New overlay node-name already in use");
1486 flags
= state
->old_bs
->open_flags
;
1487 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_COPY_ON_READ
);
1488 flags
|= BDRV_O_NO_BACKING
;
1490 /* create new image w/backing file */
1491 mode
= s
->has_mode
? s
->mode
: NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1492 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1493 int64_t size
= bdrv_getlength(state
->old_bs
);
1495 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1498 bdrv_refresh_filename(state
->old_bs
);
1499 bdrv_img_create(new_image_file
, format
,
1500 state
->old_bs
->filename
,
1501 state
->old_bs
->drv
->format_name
,
1502 NULL
, size
, flags
, false, &local_err
);
1504 error_propagate(errp
, local_err
);
1509 options
= qdict_new();
1510 if (snapshot_node_name
) {
1511 qdict_put_str(options
, "node-name", snapshot_node_name
);
1513 qdict_put_str(options
, "driver", format
);
1516 state
->new_bs
= bdrv_open(new_image_file
, snapshot_ref
, options
, flags
,
1518 /* We will manually add the backing_hd field to the bs later */
1519 if (!state
->new_bs
) {
1524 * Allow attaching a backing file to an overlay that's already in use only
1525 * if the parents don't assume that they are already seeing a valid image.
1526 * (Specifically, allow it as a mirror target, which is write-only access.)
1528 bdrv_get_cumulative_perm(state
->new_bs
, &perm
, &shared
);
1529 if (perm
& BLK_PERM_CONSISTENT_READ
) {
1530 error_setg(errp
, "The overlay is already in use");
1534 if (state
->new_bs
->drv
->is_filter
) {
1535 error_setg(errp
, "Filters cannot be used as overlays");
1539 if (bdrv_cow_child(state
->new_bs
)) {
1540 error_setg(errp
, "The overlay already has a backing image");
1544 if (!state
->new_bs
->drv
->supports_backing
) {
1545 error_setg(errp
, "The overlay does not support backing images");
1549 ret
= bdrv_append(state
->new_bs
, state
->old_bs
, errp
);
1553 state
->overlay_appended
= true;
1556 aio_context_release(aio_context
);
1559 static void external_snapshot_commit(BlkActionState
*common
)
1561 ExternalSnapshotState
*state
=
1562 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1563 AioContext
*aio_context
;
1565 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1566 aio_context_acquire(aio_context
);
1568 /* We don't need (or want) to use the transactional
1569 * bdrv_reopen_multiple() across all the entries at once, because we
1570 * don't want to abort all of them if one of them fails the reopen */
1571 if (!qatomic_read(&state
->old_bs
->copy_on_read
)) {
1572 bdrv_reopen_set_read_only(state
->old_bs
, true, NULL
);
1575 aio_context_release(aio_context
);
1578 static void external_snapshot_abort(BlkActionState
*common
)
1580 ExternalSnapshotState
*state
=
1581 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1582 if (state
->new_bs
) {
1583 if (state
->overlay_appended
) {
1584 AioContext
*aio_context
;
1585 AioContext
*tmp_context
;
1588 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1589 aio_context_acquire(aio_context
);
1591 bdrv_ref(state
->old_bs
); /* we can't let bdrv_set_backind_hd()
1592 close state->old_bs; we need it */
1593 bdrv_set_backing_hd(state
->new_bs
, NULL
, &error_abort
);
1596 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1597 * the main AioContext. As we're still going to be using it, return
1598 * it to the AioContext it was before.
1600 tmp_context
= bdrv_get_aio_context(state
->old_bs
);
1601 if (aio_context
!= tmp_context
) {
1602 aio_context_release(aio_context
);
1603 aio_context_acquire(tmp_context
);
1605 ret
= bdrv_try_set_aio_context(state
->old_bs
,
1609 aio_context_release(tmp_context
);
1610 aio_context_acquire(aio_context
);
1613 bdrv_replace_node(state
->new_bs
, state
->old_bs
, &error_abort
);
1614 bdrv_unref(state
->old_bs
); /* bdrv_replace_node() ref'ed old_bs */
1616 aio_context_release(aio_context
);
1621 static void external_snapshot_clean(BlkActionState
*common
)
1623 ExternalSnapshotState
*state
=
1624 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1625 AioContext
*aio_context
;
1627 if (!state
->old_bs
) {
1631 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1632 aio_context_acquire(aio_context
);
1634 bdrv_drained_end(state
->old_bs
);
1635 bdrv_unref(state
->new_bs
);
1637 aio_context_release(aio_context
);
1640 typedef struct DriveBackupState
{
1641 BlkActionState common
;
1642 BlockDriverState
*bs
;
1646 static BlockJob
*do_backup_common(BackupCommon
*backup
,
1647 BlockDriverState
*bs
,
1648 BlockDriverState
*target_bs
,
1649 AioContext
*aio_context
,
1650 JobTxn
*txn
, Error
**errp
);
1652 static void drive_backup_prepare(BlkActionState
*common
, Error
**errp
)
1654 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1655 DriveBackup
*backup
;
1656 BlockDriverState
*bs
;
1657 BlockDriverState
*target_bs
;
1658 BlockDriverState
*source
= NULL
;
1659 AioContext
*aio_context
;
1660 AioContext
*old_context
;
1662 Error
*local_err
= NULL
;
1665 bool set_backing_hd
= false;
1668 assert(common
->action
->type
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1669 backup
= common
->action
->u
.drive_backup
.data
;
1671 if (!backup
->has_mode
) {
1672 backup
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1675 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1681 error_setg(errp
, "Device has no medium");
1685 aio_context
= bdrv_get_aio_context(bs
);
1686 aio_context_acquire(aio_context
);
1689 /* Paired with .clean() */
1690 bdrv_drained_begin(bs
);
1692 if (!backup
->has_format
) {
1693 backup
->format
= backup
->mode
== NEW_IMAGE_MODE_EXISTING
?
1694 NULL
: (char *) bs
->drv
->format_name
;
1697 /* Early check to avoid creating target */
1698 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_BACKUP_SOURCE
, errp
)) {
1702 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1705 * See if we have a backing HD we can use to create our new image
1708 if (backup
->sync
== MIRROR_SYNC_MODE_TOP
) {
1710 * Backup will not replace the source by the target, so none
1711 * of the filters skipped here will be removed (in contrast to
1712 * mirror). Therefore, we can skip all of them when looking
1713 * for the first COW relationship.
1715 source
= bdrv_cow_bs(bdrv_skip_filters(bs
));
1717 backup
->sync
= MIRROR_SYNC_MODE_FULL
;
1720 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
1722 flags
|= BDRV_O_NO_BACKING
;
1723 set_backing_hd
= true;
1726 size
= bdrv_getlength(bs
);
1728 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1732 if (backup
->mode
!= NEW_IMAGE_MODE_EXISTING
) {
1733 assert(backup
->format
);
1735 /* Implicit filters should not appear in the filename */
1736 BlockDriverState
*explicit_backing
=
1737 bdrv_skip_implicit_filters(source
);
1739 bdrv_refresh_filename(explicit_backing
);
1740 bdrv_img_create(backup
->target
, backup
->format
,
1741 explicit_backing
->filename
,
1742 explicit_backing
->drv
->format_name
, NULL
,
1743 size
, flags
, false, &local_err
);
1745 bdrv_img_create(backup
->target
, backup
->format
, NULL
, NULL
, NULL
,
1746 size
, flags
, false, &local_err
);
1751 error_propagate(errp
, local_err
);
1755 options
= qdict_new();
1756 qdict_put_str(options
, "discard", "unmap");
1757 qdict_put_str(options
, "detect-zeroes", "unmap");
1758 if (backup
->format
) {
1759 qdict_put_str(options
, "driver", backup
->format
);
1762 target_bs
= bdrv_open(backup
->target
, NULL
, options
, flags
, errp
);
1767 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1768 old_context
= bdrv_get_aio_context(target_bs
);
1769 aio_context_release(aio_context
);
1770 aio_context_acquire(old_context
);
1772 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
1774 bdrv_unref(target_bs
);
1775 aio_context_release(old_context
);
1779 aio_context_release(old_context
);
1780 aio_context_acquire(aio_context
);
1782 if (set_backing_hd
) {
1783 if (bdrv_set_backing_hd(target_bs
, source
, errp
) < 0) {
1788 state
->job
= do_backup_common(qapi_DriveBackup_base(backup
),
1789 bs
, target_bs
, aio_context
,
1790 common
->block_job_txn
, errp
);
1793 bdrv_unref(target_bs
);
1795 aio_context_release(aio_context
);
1798 static void drive_backup_commit(BlkActionState
*common
)
1800 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1801 AioContext
*aio_context
;
1803 aio_context
= bdrv_get_aio_context(state
->bs
);
1804 aio_context_acquire(aio_context
);
1807 job_start(&state
->job
->job
);
1809 aio_context_release(aio_context
);
1812 static void drive_backup_abort(BlkActionState
*common
)
1814 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1817 AioContext
*aio_context
;
1819 aio_context
= bdrv_get_aio_context(state
->bs
);
1820 aio_context_acquire(aio_context
);
1822 job_cancel_sync(&state
->job
->job
, true);
1824 aio_context_release(aio_context
);
1828 static void drive_backup_clean(BlkActionState
*common
)
1830 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1831 AioContext
*aio_context
;
1837 aio_context
= bdrv_get_aio_context(state
->bs
);
1838 aio_context_acquire(aio_context
);
1840 bdrv_drained_end(state
->bs
);
1842 aio_context_release(aio_context
);
1845 typedef struct BlockdevBackupState
{
1846 BlkActionState common
;
1847 BlockDriverState
*bs
;
1849 } BlockdevBackupState
;
1851 static void blockdev_backup_prepare(BlkActionState
*common
, Error
**errp
)
1853 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1854 BlockdevBackup
*backup
;
1855 BlockDriverState
*bs
;
1856 BlockDriverState
*target_bs
;
1857 AioContext
*aio_context
;
1858 AioContext
*old_context
;
1861 assert(common
->action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
);
1862 backup
= common
->action
->u
.blockdev_backup
.data
;
1864 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1869 target_bs
= bdrv_lookup_bs(backup
->target
, backup
->target
, errp
);
1874 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1875 aio_context
= bdrv_get_aio_context(bs
);
1876 old_context
= bdrv_get_aio_context(target_bs
);
1877 aio_context_acquire(old_context
);
1879 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
1881 aio_context_release(old_context
);
1885 aio_context_release(old_context
);
1886 aio_context_acquire(aio_context
);
1889 /* Paired with .clean() */
1890 bdrv_drained_begin(state
->bs
);
1892 state
->job
= do_backup_common(qapi_BlockdevBackup_base(backup
),
1893 bs
, target_bs
, aio_context
,
1894 common
->block_job_txn
, errp
);
1896 aio_context_release(aio_context
);
1899 static void blockdev_backup_commit(BlkActionState
*common
)
1901 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1902 AioContext
*aio_context
;
1904 aio_context
= bdrv_get_aio_context(state
->bs
);
1905 aio_context_acquire(aio_context
);
1908 job_start(&state
->job
->job
);
1910 aio_context_release(aio_context
);
1913 static void blockdev_backup_abort(BlkActionState
*common
)
1915 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1918 AioContext
*aio_context
;
1920 aio_context
= bdrv_get_aio_context(state
->bs
);
1921 aio_context_acquire(aio_context
);
1923 job_cancel_sync(&state
->job
->job
, true);
1925 aio_context_release(aio_context
);
1929 static void blockdev_backup_clean(BlkActionState
*common
)
1931 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1932 AioContext
*aio_context
;
1938 aio_context
= bdrv_get_aio_context(state
->bs
);
1939 aio_context_acquire(aio_context
);
1941 bdrv_drained_end(state
->bs
);
1943 aio_context_release(aio_context
);
1946 typedef struct BlockDirtyBitmapState
{
1947 BlkActionState common
;
1948 BdrvDirtyBitmap
*bitmap
;
1949 BlockDriverState
*bs
;
1953 } BlockDirtyBitmapState
;
1955 static void block_dirty_bitmap_add_prepare(BlkActionState
*common
,
1958 Error
*local_err
= NULL
;
1959 BlockDirtyBitmapAdd
*action
;
1960 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
1963 if (action_check_completion_mode(common
, errp
) < 0) {
1967 action
= common
->action
->u
.block_dirty_bitmap_add
.data
;
1968 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1969 qmp_block_dirty_bitmap_add(action
->node
, action
->name
,
1970 action
->has_granularity
, action
->granularity
,
1971 action
->has_persistent
, action
->persistent
,
1972 action
->has_disabled
, action
->disabled
,
1976 state
->prepared
= true;
1978 error_propagate(errp
, local_err
);
1982 static void block_dirty_bitmap_add_abort(BlkActionState
*common
)
1984 BlockDirtyBitmapAdd
*action
;
1985 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
1988 action
= common
->action
->u
.block_dirty_bitmap_add
.data
;
1989 /* Should not be able to fail: IF the bitmap was added via .prepare(),
1990 * then the node reference and bitmap name must have been valid.
1992 if (state
->prepared
) {
1993 qmp_block_dirty_bitmap_remove(action
->node
, action
->name
, &error_abort
);
1997 static void block_dirty_bitmap_clear_prepare(BlkActionState
*common
,
2000 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2002 BlockDirtyBitmap
*action
;
2004 if (action_check_completion_mode(common
, errp
) < 0) {
2008 action
= common
->action
->u
.block_dirty_bitmap_clear
.data
;
2009 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2013 if (!state
->bitmap
) {
2017 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_DEFAULT
, errp
)) {
2021 bdrv_clear_dirty_bitmap(state
->bitmap
, &state
->backup
);
2024 static void block_dirty_bitmap_restore(BlkActionState
*common
)
2026 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2029 if (state
->backup
) {
2030 bdrv_restore_dirty_bitmap(state
->bitmap
, state
->backup
);
2034 static void block_dirty_bitmap_free_backup(BlkActionState
*common
)
2036 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2039 hbitmap_free(state
->backup
);
2042 static void block_dirty_bitmap_enable_prepare(BlkActionState
*common
,
2045 BlockDirtyBitmap
*action
;
2046 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2049 if (action_check_completion_mode(common
, errp
) < 0) {
2053 action
= common
->action
->u
.block_dirty_bitmap_enable
.data
;
2054 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2058 if (!state
->bitmap
) {
2062 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2066 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2067 bdrv_enable_dirty_bitmap(state
->bitmap
);
2070 static void block_dirty_bitmap_enable_abort(BlkActionState
*common
)
2072 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2075 if (!state
->was_enabled
) {
2076 bdrv_disable_dirty_bitmap(state
->bitmap
);
2080 static void block_dirty_bitmap_disable_prepare(BlkActionState
*common
,
2083 BlockDirtyBitmap
*action
;
2084 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2087 if (action_check_completion_mode(common
, errp
) < 0) {
2091 action
= common
->action
->u
.block_dirty_bitmap_disable
.data
;
2092 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2096 if (!state
->bitmap
) {
2100 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2104 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2105 bdrv_disable_dirty_bitmap(state
->bitmap
);
2108 static void block_dirty_bitmap_disable_abort(BlkActionState
*common
)
2110 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2113 if (state
->was_enabled
) {
2114 bdrv_enable_dirty_bitmap(state
->bitmap
);
2118 static void block_dirty_bitmap_merge_prepare(BlkActionState
*common
,
2121 BlockDirtyBitmapMerge
*action
;
2122 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2125 if (action_check_completion_mode(common
, errp
) < 0) {
2129 action
= common
->action
->u
.block_dirty_bitmap_merge
.data
;
2131 state
->bitmap
= block_dirty_bitmap_merge(action
->node
, action
->target
,
2132 action
->bitmaps
, &state
->backup
,
2136 static void block_dirty_bitmap_remove_prepare(BlkActionState
*common
,
2139 BlockDirtyBitmap
*action
;
2140 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2143 if (action_check_completion_mode(common
, errp
) < 0) {
2147 action
= common
->action
->u
.block_dirty_bitmap_remove
.data
;
2149 state
->bitmap
= block_dirty_bitmap_remove(action
->node
, action
->name
,
2150 false, &state
->bs
, errp
);
2151 if (state
->bitmap
) {
2152 bdrv_dirty_bitmap_skip_store(state
->bitmap
, true);
2153 bdrv_dirty_bitmap_set_busy(state
->bitmap
, true);
2157 static void block_dirty_bitmap_remove_abort(BlkActionState
*common
)
2159 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2162 if (state
->bitmap
) {
2163 bdrv_dirty_bitmap_skip_store(state
->bitmap
, false);
2164 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2168 static void block_dirty_bitmap_remove_commit(BlkActionState
*common
)
2170 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2173 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2174 bdrv_release_dirty_bitmap(state
->bitmap
);
2177 static void abort_prepare(BlkActionState
*common
, Error
**errp
)
2179 error_setg(errp
, "Transaction aborted using Abort action");
2182 static void abort_commit(BlkActionState
*common
)
2184 g_assert_not_reached(); /* this action never succeeds */
2187 static const BlkActionOps actions
[] = {
2188 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
] = {
2189 .instance_size
= sizeof(ExternalSnapshotState
),
2190 .prepare
= external_snapshot_prepare
,
2191 .commit
= external_snapshot_commit
,
2192 .abort
= external_snapshot_abort
,
2193 .clean
= external_snapshot_clean
,
2195 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
2196 .instance_size
= sizeof(ExternalSnapshotState
),
2197 .prepare
= external_snapshot_prepare
,
2198 .commit
= external_snapshot_commit
,
2199 .abort
= external_snapshot_abort
,
2200 .clean
= external_snapshot_clean
,
2202 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
2203 .instance_size
= sizeof(DriveBackupState
),
2204 .prepare
= drive_backup_prepare
,
2205 .commit
= drive_backup_commit
,
2206 .abort
= drive_backup_abort
,
2207 .clean
= drive_backup_clean
,
2209 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
] = {
2210 .instance_size
= sizeof(BlockdevBackupState
),
2211 .prepare
= blockdev_backup_prepare
,
2212 .commit
= blockdev_backup_commit
,
2213 .abort
= blockdev_backup_abort
,
2214 .clean
= blockdev_backup_clean
,
2216 [TRANSACTION_ACTION_KIND_ABORT
] = {
2217 .instance_size
= sizeof(BlkActionState
),
2218 .prepare
= abort_prepare
,
2219 .commit
= abort_commit
,
2221 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
2222 .instance_size
= sizeof(InternalSnapshotState
),
2223 .prepare
= internal_snapshot_prepare
,
2224 .abort
= internal_snapshot_abort
,
2225 .clean
= internal_snapshot_clean
,
2227 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD
] = {
2228 .instance_size
= sizeof(BlockDirtyBitmapState
),
2229 .prepare
= block_dirty_bitmap_add_prepare
,
2230 .abort
= block_dirty_bitmap_add_abort
,
2232 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR
] = {
2233 .instance_size
= sizeof(BlockDirtyBitmapState
),
2234 .prepare
= block_dirty_bitmap_clear_prepare
,
2235 .commit
= block_dirty_bitmap_free_backup
,
2236 .abort
= block_dirty_bitmap_restore
,
2238 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE
] = {
2239 .instance_size
= sizeof(BlockDirtyBitmapState
),
2240 .prepare
= block_dirty_bitmap_enable_prepare
,
2241 .abort
= block_dirty_bitmap_enable_abort
,
2243 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE
] = {
2244 .instance_size
= sizeof(BlockDirtyBitmapState
),
2245 .prepare
= block_dirty_bitmap_disable_prepare
,
2246 .abort
= block_dirty_bitmap_disable_abort
,
2248 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE
] = {
2249 .instance_size
= sizeof(BlockDirtyBitmapState
),
2250 .prepare
= block_dirty_bitmap_merge_prepare
,
2251 .commit
= block_dirty_bitmap_free_backup
,
2252 .abort
= block_dirty_bitmap_restore
,
2254 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE
] = {
2255 .instance_size
= sizeof(BlockDirtyBitmapState
),
2256 .prepare
= block_dirty_bitmap_remove_prepare
,
2257 .commit
= block_dirty_bitmap_remove_commit
,
2258 .abort
= block_dirty_bitmap_remove_abort
,
2260 /* Where are transactions for MIRROR, COMMIT and STREAM?
2261 * Although these blockjobs use transaction callbacks like the backup job,
2262 * these jobs do not necessarily adhere to transaction semantics.
2263 * These jobs may not fully undo all of their actions on abort, nor do they
2264 * necessarily work in transactions with more than one job in them.
2269 * Allocate a TransactionProperties structure if necessary, and fill
2270 * that structure with desired defaults if they are unset.
2272 static TransactionProperties
*get_transaction_properties(
2273 TransactionProperties
*props
)
2276 props
= g_new0(TransactionProperties
, 1);
2279 if (!props
->has_completion_mode
) {
2280 props
->has_completion_mode
= true;
2281 props
->completion_mode
= ACTION_COMPLETION_MODE_INDIVIDUAL
;
2288 * 'Atomic' group operations. The operations are performed as a set, and if
2289 * any fail then we roll back all operations in the group.
2291 * Always run under BQL.
2293 void qmp_transaction(TransactionActionList
*dev_list
,
2295 struct TransactionProperties
*props
,
2298 TransactionActionList
*dev_entry
= dev_list
;
2299 JobTxn
*block_job_txn
= NULL
;
2300 BlkActionState
*state
, *next
;
2301 Error
*local_err
= NULL
;
2303 QTAILQ_HEAD(, BlkActionState
) snap_bdrv_states
;
2304 QTAILQ_INIT(&snap_bdrv_states
);
2306 /* Does this transaction get canceled as a group on failure?
2307 * If not, we don't really need to make a JobTxn.
2309 props
= get_transaction_properties(props
);
2310 if (props
->completion_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
2311 block_job_txn
= job_txn_new();
2314 /* drain all i/o before any operations */
2317 /* We don't do anything in this loop that commits us to the operations */
2318 while (NULL
!= dev_entry
) {
2319 TransactionAction
*dev_info
= NULL
;
2320 const BlkActionOps
*ops
;
2322 dev_info
= dev_entry
->value
;
2323 dev_entry
= dev_entry
->next
;
2325 assert(dev_info
->type
< ARRAY_SIZE(actions
));
2327 ops
= &actions
[dev_info
->type
];
2328 assert(ops
->instance_size
> 0);
2330 state
= g_malloc0(ops
->instance_size
);
2332 state
->action
= dev_info
;
2333 state
->block_job_txn
= block_job_txn
;
2334 state
->txn_props
= props
;
2335 QTAILQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
2337 state
->ops
->prepare(state
, &local_err
);
2339 error_propagate(errp
, local_err
);
2340 goto delete_and_fail
;
2344 QTAILQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
2345 if (state
->ops
->commit
) {
2346 state
->ops
->commit(state
);
2354 /* failure, and it is all-or-none; roll back all operations */
2355 QTAILQ_FOREACH_REVERSE(state
, &snap_bdrv_states
, entry
) {
2356 if (state
->ops
->abort
) {
2357 state
->ops
->abort(state
);
2361 QTAILQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
2362 if (state
->ops
->clean
) {
2363 state
->ops
->clean(state
);
2368 qapi_free_TransactionProperties(props
);
2370 job_txn_unref(block_job_txn
);
2373 BlockDirtyBitmapSha256
*qmp_x_debug_block_dirty_bitmap_sha256(const char *node
,
2377 BdrvDirtyBitmap
*bitmap
;
2378 BlockDriverState
*bs
;
2379 BlockDirtyBitmapSha256
*ret
= NULL
;
2382 bitmap
= block_dirty_bitmap_lookup(node
, name
, &bs
, errp
);
2383 if (!bitmap
|| !bs
) {
2387 sha256
= bdrv_dirty_bitmap_sha256(bitmap
, errp
);
2388 if (sha256
== NULL
) {
2392 ret
= g_new(BlockDirtyBitmapSha256
, 1);
2393 ret
->sha256
= sha256
;
2398 void coroutine_fn
qmp_block_resize(bool has_device
, const char *device
,
2399 bool has_node_name
, const char *node_name
,
2400 int64_t size
, Error
**errp
)
2402 Error
*local_err
= NULL
;
2404 BlockDriverState
*bs
;
2405 AioContext
*old_ctx
;
2407 bs
= bdrv_lookup_bs(has_device
? device
: NULL
,
2408 has_node_name
? node_name
: NULL
,
2411 error_propagate(errp
, local_err
);
2416 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
2420 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_RESIZE
, NULL
)) {
2421 error_setg(errp
, QERR_DEVICE_IN_USE
, device
);
2425 blk
= blk_new_with_bs(bs
, BLK_PERM_RESIZE
, BLK_PERM_ALL
, errp
);
2431 bdrv_drained_begin(bs
);
2434 old_ctx
= bdrv_co_enter(bs
);
2435 blk_truncate(blk
, size
, false, PREALLOC_MODE_OFF
, 0, errp
);
2436 bdrv_co_leave(bs
, old_ctx
);
2439 bdrv_drained_end(bs
);
2444 void qmp_block_stream(bool has_job_id
, const char *job_id
, const char *device
,
2445 bool has_base
, const char *base
,
2446 bool has_base_node
, const char *base_node
,
2447 bool has_backing_file
, const char *backing_file
,
2448 bool has_bottom
, const char *bottom
,
2449 bool has_speed
, int64_t speed
,
2450 bool has_on_error
, BlockdevOnError on_error
,
2451 bool has_filter_node_name
, const char *filter_node_name
,
2452 bool has_auto_finalize
, bool auto_finalize
,
2453 bool has_auto_dismiss
, bool auto_dismiss
,
2456 BlockDriverState
*bs
, *iter
, *iter_end
;
2457 BlockDriverState
*base_bs
= NULL
;
2458 BlockDriverState
*bottom_bs
= NULL
;
2459 AioContext
*aio_context
;
2460 Error
*local_err
= NULL
;
2461 int job_flags
= JOB_DEFAULT
;
2463 if (has_base
&& has_base_node
) {
2464 error_setg(errp
, "'base' and 'base-node' cannot be specified "
2465 "at the same time");
2469 if (has_base
&& has_bottom
) {
2470 error_setg(errp
, "'base' and 'bottom' cannot be specified "
2471 "at the same time");
2475 if (has_bottom
&& has_base_node
) {
2476 error_setg(errp
, "'bottom' and 'base-node' cannot be specified "
2477 "at the same time");
2481 if (!has_on_error
) {
2482 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2485 bs
= bdrv_lookup_bs(device
, device
, errp
);
2490 aio_context
= bdrv_get_aio_context(bs
);
2491 aio_context_acquire(aio_context
);
2494 base_bs
= bdrv_find_backing_image(bs
, base
);
2495 if (base_bs
== NULL
) {
2496 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2499 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2502 if (has_base_node
) {
2503 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2507 if (bs
== base_bs
|| !bdrv_chain_contains(bs
, base_bs
)) {
2508 error_setg(errp
, "Node '%s' is not a backing image of '%s'",
2512 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2513 bdrv_refresh_filename(base_bs
);
2517 bottom_bs
= bdrv_lookup_bs(NULL
, bottom
, errp
);
2521 if (!bottom_bs
->drv
) {
2522 error_setg(errp
, "Node '%s' is not open", bottom
);
2525 if (bottom_bs
->drv
->is_filter
) {
2526 error_setg(errp
, "Node '%s' is a filter, use a non-filter node "
2527 "as 'bottom'", bottom
);
2530 if (!bdrv_chain_contains(bs
, bottom_bs
)) {
2531 error_setg(errp
, "Node '%s' is not in a chain starting from '%s'",
2535 assert(bdrv_get_aio_context(bottom_bs
) == aio_context
);
2539 * Check for op blockers in the whole chain between bs and base (or bottom)
2541 iter_end
= has_bottom
? bdrv_filter_or_cow_bs(bottom_bs
) : base_bs
;
2542 for (iter
= bs
; iter
&& iter
!= iter_end
;
2543 iter
= bdrv_filter_or_cow_bs(iter
))
2545 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_STREAM
, errp
)) {
2550 /* if we are streaming the entire chain, the result will have no backing
2551 * file, and specifying one is therefore an error */
2552 if (base_bs
== NULL
&& has_backing_file
) {
2553 error_setg(errp
, "backing file specified, but streaming the "
2558 if (has_auto_finalize
&& !auto_finalize
) {
2559 job_flags
|= JOB_MANUAL_FINALIZE
;
2561 if (has_auto_dismiss
&& !auto_dismiss
) {
2562 job_flags
|= JOB_MANUAL_DISMISS
;
2565 stream_start(has_job_id
? job_id
: NULL
, bs
, base_bs
, backing_file
,
2566 bottom_bs
, job_flags
, has_speed
? speed
: 0, on_error
,
2567 filter_node_name
, &local_err
);
2569 error_propagate(errp
, local_err
);
2573 trace_qmp_block_stream(bs
);
2576 aio_context_release(aio_context
);
2579 void qmp_block_commit(bool has_job_id
, const char *job_id
, const char *device
,
2580 bool has_base_node
, const char *base_node
,
2581 bool has_base
, const char *base
,
2582 bool has_top_node
, const char *top_node
,
2583 bool has_top
, const char *top
,
2584 bool has_backing_file
, const char *backing_file
,
2585 bool has_speed
, int64_t speed
,
2586 bool has_on_error
, BlockdevOnError on_error
,
2587 bool has_filter_node_name
, const char *filter_node_name
,
2588 bool has_auto_finalize
, bool auto_finalize
,
2589 bool has_auto_dismiss
, bool auto_dismiss
,
2592 BlockDriverState
*bs
;
2593 BlockDriverState
*iter
;
2594 BlockDriverState
*base_bs
, *top_bs
;
2595 AioContext
*aio_context
;
2596 Error
*local_err
= NULL
;
2597 int job_flags
= JOB_DEFAULT
;
2598 uint64_t top_perm
, top_shared
;
2603 if (!has_on_error
) {
2604 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2606 if (!has_filter_node_name
) {
2607 filter_node_name
= NULL
;
2609 if (has_auto_finalize
&& !auto_finalize
) {
2610 job_flags
|= JOB_MANUAL_FINALIZE
;
2612 if (has_auto_dismiss
&& !auto_dismiss
) {
2613 job_flags
|= JOB_MANUAL_DISMISS
;
2617 * libvirt relies on the DeviceNotFound error class in order to probe for
2618 * live commit feature versions; for this to work, we must make sure to
2619 * perform the device lookup before any generic errors that may occur in a
2620 * scenario in which all optional arguments are omitted. */
2621 bs
= qmp_get_root_bs(device
, &local_err
);
2623 bs
= bdrv_lookup_bs(device
, device
, NULL
);
2625 error_free(local_err
);
2626 error_set(errp
, ERROR_CLASS_DEVICE_NOT_FOUND
,
2627 "Device '%s' not found", device
);
2629 error_propagate(errp
, local_err
);
2634 aio_context
= bdrv_get_aio_context(bs
);
2635 aio_context_acquire(aio_context
);
2637 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_COMMIT_SOURCE
, errp
)) {
2641 /* default top_bs is the active layer */
2644 if (has_top_node
&& has_top
) {
2645 error_setg(errp
, "'top-node' and 'top' are mutually exclusive");
2647 } else if (has_top_node
) {
2648 top_bs
= bdrv_lookup_bs(NULL
, top_node
, errp
);
2649 if (top_bs
== NULL
) {
2652 if (!bdrv_chain_contains(bs
, top_bs
)) {
2653 error_setg(errp
, "'%s' is not in this backing file chain",
2657 } else if (has_top
&& top
) {
2658 /* This strcmp() is just a shortcut, there is no need to
2659 * refresh @bs's filename. If it mismatches,
2660 * bdrv_find_backing_image() will do the refresh and may still
2662 if (strcmp(bs
->filename
, top
) != 0) {
2663 top_bs
= bdrv_find_backing_image(bs
, top
);
2667 if (top_bs
== NULL
) {
2668 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
2672 assert(bdrv_get_aio_context(top_bs
) == aio_context
);
2674 if (has_base_node
&& has_base
) {
2675 error_setg(errp
, "'base-node' and 'base' are mutually exclusive");
2677 } else if (has_base_node
) {
2678 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2679 if (base_bs
== NULL
) {
2682 if (!bdrv_chain_contains(top_bs
, base_bs
)) {
2683 error_setg(errp
, "'%s' is not in this backing file chain",
2687 } else if (has_base
&& base
) {
2688 base_bs
= bdrv_find_backing_image(top_bs
, base
);
2689 if (base_bs
== NULL
) {
2690 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2694 base_bs
= bdrv_find_base(top_bs
);
2695 if (base_bs
== NULL
) {
2696 error_setg(errp
, "There is no backimg image");
2701 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2703 for (iter
= top_bs
; iter
!= bdrv_filter_or_cow_bs(base_bs
);
2704 iter
= bdrv_filter_or_cow_bs(iter
))
2706 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2711 /* Do not allow attempts to commit an image into itself */
2712 if (top_bs
== base_bs
) {
2713 error_setg(errp
, "cannot commit an image into itself");
2718 * Active commit is required if and only if someone has taken a
2719 * WRITE permission on the top node. Historically, we have always
2720 * used active commit for top nodes, so continue that practice
2721 * lest we possibly break clients that rely on this behavior, e.g.
2722 * to later attach this node to a writing parent.
2723 * (Active commit is never really wrong.)
2725 bdrv_get_cumulative_perm(top_bs
, &top_perm
, &top_shared
);
2726 if (top_perm
& BLK_PERM_WRITE
||
2727 bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
))
2729 if (has_backing_file
) {
2730 if (bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
)) {
2731 error_setg(errp
, "'backing-file' specified,"
2732 " but 'top' is the active layer");
2734 error_setg(errp
, "'backing-file' specified, but 'top' has a "
2741 * Emulate here what block_job_create() does, because it
2742 * is possible that @bs != @top_bs (the block job should
2743 * be named after @bs, even if @top_bs is the actual
2746 job_id
= bdrv_get_device_name(bs
);
2748 commit_active_start(job_id
, top_bs
, base_bs
, job_flags
, speed
, on_error
,
2749 filter_node_name
, NULL
, NULL
, false, &local_err
);
2751 BlockDriverState
*overlay_bs
= bdrv_find_overlay(bs
, top_bs
);
2752 if (bdrv_op_is_blocked(overlay_bs
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2755 commit_start(has_job_id
? job_id
: NULL
, bs
, base_bs
, top_bs
, job_flags
,
2756 speed
, on_error
, has_backing_file
? backing_file
: NULL
,
2757 filter_node_name
, &local_err
);
2759 if (local_err
!= NULL
) {
2760 error_propagate(errp
, local_err
);
2765 aio_context_release(aio_context
);
2768 /* Common QMP interface for drive-backup and blockdev-backup */
2769 static BlockJob
*do_backup_common(BackupCommon
*backup
,
2770 BlockDriverState
*bs
,
2771 BlockDriverState
*target_bs
,
2772 AioContext
*aio_context
,
2773 JobTxn
*txn
, Error
**errp
)
2775 BlockJob
*job
= NULL
;
2776 BdrvDirtyBitmap
*bmap
= NULL
;
2777 BackupPerf perf
= { .max_workers
= 64 };
2778 int job_flags
= JOB_DEFAULT
;
2780 if (!backup
->has_speed
) {
2783 if (!backup
->has_on_source_error
) {
2784 backup
->on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2786 if (!backup
->has_on_target_error
) {
2787 backup
->on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2789 if (!backup
->has_job_id
) {
2790 backup
->job_id
= NULL
;
2792 if (!backup
->has_auto_finalize
) {
2793 backup
->auto_finalize
= true;
2795 if (!backup
->has_auto_dismiss
) {
2796 backup
->auto_dismiss
= true;
2798 if (!backup
->has_compress
) {
2799 backup
->compress
= false;
2802 if (backup
->x_perf
) {
2803 if (backup
->x_perf
->has_use_copy_range
) {
2804 perf
.use_copy_range
= backup
->x_perf
->use_copy_range
;
2806 if (backup
->x_perf
->has_max_workers
) {
2807 perf
.max_workers
= backup
->x_perf
->max_workers
;
2809 if (backup
->x_perf
->has_max_chunk
) {
2810 perf
.max_chunk
= backup
->x_perf
->max_chunk
;
2814 if ((backup
->sync
== MIRROR_SYNC_MODE_BITMAP
) ||
2815 (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
)) {
2816 /* done before desugaring 'incremental' to print the right message */
2817 if (!backup
->has_bitmap
) {
2818 error_setg(errp
, "must provide a valid bitmap name for "
2819 "'%s' sync mode", MirrorSyncMode_str(backup
->sync
));
2824 if (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
) {
2825 if (backup
->has_bitmap_mode
&&
2826 backup
->bitmap_mode
!= BITMAP_SYNC_MODE_ON_SUCCESS
) {
2827 error_setg(errp
, "Bitmap sync mode must be '%s' "
2828 "when using sync mode '%s'",
2829 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS
),
2830 MirrorSyncMode_str(backup
->sync
));
2833 backup
->has_bitmap_mode
= true;
2834 backup
->sync
= MIRROR_SYNC_MODE_BITMAP
;
2835 backup
->bitmap_mode
= BITMAP_SYNC_MODE_ON_SUCCESS
;
2838 if (backup
->has_bitmap
) {
2839 bmap
= bdrv_find_dirty_bitmap(bs
, backup
->bitmap
);
2841 error_setg(errp
, "Bitmap '%s' could not be found", backup
->bitmap
);
2844 if (!backup
->has_bitmap_mode
) {
2845 error_setg(errp
, "Bitmap sync mode must be given "
2846 "when providing a bitmap");
2849 if (bdrv_dirty_bitmap_check(bmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2853 /* This does not produce a useful bitmap artifact: */
2854 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
2855 error_setg(errp
, "sync mode '%s' does not produce meaningful bitmap"
2856 " outputs", MirrorSyncMode_str(backup
->sync
));
2860 /* If the bitmap isn't used for input or output, this is useless: */
2861 if (backup
->bitmap_mode
== BITMAP_SYNC_MODE_NEVER
&&
2862 backup
->sync
!= MIRROR_SYNC_MODE_BITMAP
) {
2863 error_setg(errp
, "Bitmap sync mode '%s' has no meaningful effect"
2864 " when combined with sync mode '%s'",
2865 BitmapSyncMode_str(backup
->bitmap_mode
),
2866 MirrorSyncMode_str(backup
->sync
));
2871 if (!backup
->has_bitmap
&& backup
->has_bitmap_mode
) {
2872 error_setg(errp
, "Cannot specify bitmap sync mode without a bitmap");
2876 if (!backup
->auto_finalize
) {
2877 job_flags
|= JOB_MANUAL_FINALIZE
;
2879 if (!backup
->auto_dismiss
) {
2880 job_flags
|= JOB_MANUAL_DISMISS
;
2883 job
= backup_job_create(backup
->job_id
, bs
, target_bs
, backup
->speed
,
2884 backup
->sync
, bmap
, backup
->bitmap_mode
,
2886 backup
->filter_node_name
,
2888 backup
->on_source_error
,
2889 backup
->on_target_error
,
2890 job_flags
, NULL
, NULL
, txn
, errp
);
2894 void qmp_drive_backup(DriveBackup
*backup
, Error
**errp
)
2896 TransactionAction action
= {
2897 .type
= TRANSACTION_ACTION_KIND_DRIVE_BACKUP
,
2898 .u
.drive_backup
.data
= backup
,
2900 blockdev_do_action(&action
, errp
);
2903 BlockDeviceInfoList
*qmp_query_named_block_nodes(bool has_flat
,
2907 bool return_flat
= has_flat
&& flat
;
2909 return bdrv_named_nodes_list(return_flat
, errp
);
2912 XDbgBlockGraph
*qmp_x_debug_query_block_graph(Error
**errp
)
2914 return bdrv_get_xdbg_block_graph(errp
);
2917 void qmp_blockdev_backup(BlockdevBackup
*backup
, Error
**errp
)
2919 TransactionAction action
= {
2920 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
,
2921 .u
.blockdev_backup
.data
= backup
,
2923 blockdev_do_action(&action
, errp
);
2926 /* Parameter check and block job starting for drive mirroring.
2927 * Caller should hold @device and @target's aio context (must be the same).
2929 static void blockdev_mirror_common(const char *job_id
, BlockDriverState
*bs
,
2930 BlockDriverState
*target
,
2931 bool has_replaces
, const char *replaces
,
2932 enum MirrorSyncMode sync
,
2933 BlockMirrorBackingMode backing_mode
,
2935 bool has_speed
, int64_t speed
,
2936 bool has_granularity
, uint32_t granularity
,
2937 bool has_buf_size
, int64_t buf_size
,
2938 bool has_on_source_error
,
2939 BlockdevOnError on_source_error
,
2940 bool has_on_target_error
,
2941 BlockdevOnError on_target_error
,
2942 bool has_unmap
, bool unmap
,
2943 bool has_filter_node_name
,
2944 const char *filter_node_name
,
2945 bool has_copy_mode
, MirrorCopyMode copy_mode
,
2946 bool has_auto_finalize
, bool auto_finalize
,
2947 bool has_auto_dismiss
, bool auto_dismiss
,
2950 BlockDriverState
*unfiltered_bs
;
2951 int job_flags
= JOB_DEFAULT
;
2956 if (!has_on_source_error
) {
2957 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2959 if (!has_on_target_error
) {
2960 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2962 if (!has_granularity
) {
2965 if (!has_buf_size
) {
2971 if (!has_filter_node_name
) {
2972 filter_node_name
= NULL
;
2974 if (!has_copy_mode
) {
2975 copy_mode
= MIRROR_COPY_MODE_BACKGROUND
;
2977 if (has_auto_finalize
&& !auto_finalize
) {
2978 job_flags
|= JOB_MANUAL_FINALIZE
;
2980 if (has_auto_dismiss
&& !auto_dismiss
) {
2981 job_flags
|= JOB_MANUAL_DISMISS
;
2984 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
2985 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2986 "a value in range [512B, 64MB]");
2989 if (granularity
& (granularity
- 1)) {
2990 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2995 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
2998 if (bdrv_op_is_blocked(target
, BLOCK_OP_TYPE_MIRROR_TARGET
, errp
)) {
3002 if (!bdrv_backing_chain_next(bs
) && sync
== MIRROR_SYNC_MODE_TOP
) {
3003 sync
= MIRROR_SYNC_MODE_FULL
;
3006 if (!has_replaces
) {
3007 /* We want to mirror from @bs, but keep implicit filters on top */
3008 unfiltered_bs
= bdrv_skip_implicit_filters(bs
);
3009 if (unfiltered_bs
!= bs
) {
3010 replaces
= unfiltered_bs
->node_name
;
3011 has_replaces
= true;
3016 BlockDriverState
*to_replace_bs
;
3017 AioContext
*replace_aio_context
;
3018 int64_t bs_size
, replace_size
;
3020 bs_size
= bdrv_getlength(bs
);
3022 error_setg_errno(errp
, -bs_size
, "Failed to query device's size");
3026 to_replace_bs
= check_to_replace_node(bs
, replaces
, errp
);
3027 if (!to_replace_bs
) {
3031 replace_aio_context
= bdrv_get_aio_context(to_replace_bs
);
3032 aio_context_acquire(replace_aio_context
);
3033 replace_size
= bdrv_getlength(to_replace_bs
);
3034 aio_context_release(replace_aio_context
);
3036 if (replace_size
< 0) {
3037 error_setg_errno(errp
, -replace_size
,
3038 "Failed to query the replacement node's size");
3041 if (bs_size
!= replace_size
) {
3042 error_setg(errp
, "cannot replace image with a mirror image of "
3048 /* pass the node name to replace to mirror start since it's loose coupling
3049 * and will allow to check whether the node still exist at mirror completion
3051 mirror_start(job_id
, bs
, target
,
3052 has_replaces
? replaces
: NULL
, job_flags
,
3053 speed
, granularity
, buf_size
, sync
, backing_mode
, zero_target
,
3054 on_source_error
, on_target_error
, unmap
, filter_node_name
,
3058 void qmp_drive_mirror(DriveMirror
*arg
, Error
**errp
)
3060 BlockDriverState
*bs
;
3061 BlockDriverState
*target_backing_bs
, *target_bs
;
3062 AioContext
*aio_context
;
3063 AioContext
*old_context
;
3064 BlockMirrorBackingMode backing_mode
;
3065 Error
*local_err
= NULL
;
3066 QDict
*options
= NULL
;
3069 const char *format
= arg
->format
;
3073 bs
= qmp_get_root_bs(arg
->device
, errp
);
3078 /* Early check to avoid creating target */
3079 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3083 aio_context
= bdrv_get_aio_context(bs
);
3084 aio_context_acquire(aio_context
);
3086 if (!arg
->has_mode
) {
3087 arg
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
3090 if (!arg
->has_format
) {
3091 format
= (arg
->mode
== NEW_IMAGE_MODE_EXISTING
3092 ? NULL
: bs
->drv
->format_name
);
3095 flags
= bs
->open_flags
| BDRV_O_RDWR
;
3096 target_backing_bs
= bdrv_cow_bs(bdrv_skip_filters(bs
));
3097 if (!target_backing_bs
&& arg
->sync
== MIRROR_SYNC_MODE_TOP
) {
3098 arg
->sync
= MIRROR_SYNC_MODE_FULL
;
3100 if (arg
->sync
== MIRROR_SYNC_MODE_NONE
) {
3101 target_backing_bs
= bs
;
3104 size
= bdrv_getlength(bs
);
3106 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
3110 if (arg
->has_replaces
) {
3111 if (!arg
->has_node_name
) {
3112 error_setg(errp
, "a node-name must be provided when replacing a"
3113 " named node of the graph");
3118 if (arg
->mode
== NEW_IMAGE_MODE_ABSOLUTE_PATHS
) {
3119 backing_mode
= MIRROR_SOURCE_BACKING_CHAIN
;
3121 backing_mode
= MIRROR_OPEN_BACKING_CHAIN
;
3124 /* Don't open backing image in create() */
3125 flags
|= BDRV_O_NO_BACKING
;
3127 if ((arg
->sync
== MIRROR_SYNC_MODE_FULL
|| !target_backing_bs
)
3128 && arg
->mode
!= NEW_IMAGE_MODE_EXISTING
)
3130 /* create new image w/o backing file */
3132 bdrv_img_create(arg
->target
, format
,
3133 NULL
, NULL
, NULL
, size
, flags
, false, &local_err
);
3135 /* Implicit filters should not appear in the filename */
3136 BlockDriverState
*explicit_backing
=
3137 bdrv_skip_implicit_filters(target_backing_bs
);
3139 switch (arg
->mode
) {
3140 case NEW_IMAGE_MODE_EXISTING
:
3142 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
3143 /* create new image with backing file */
3144 bdrv_refresh_filename(explicit_backing
);
3145 bdrv_img_create(arg
->target
, format
,
3146 explicit_backing
->filename
,
3147 explicit_backing
->drv
->format_name
,
3148 NULL
, size
, flags
, false, &local_err
);
3156 error_propagate(errp
, local_err
);
3160 options
= qdict_new();
3161 if (arg
->has_node_name
) {
3162 qdict_put_str(options
, "node-name", arg
->node_name
);
3165 qdict_put_str(options
, "driver", format
);
3168 /* Mirroring takes care of copy-on-write using the source's backing
3171 target_bs
= bdrv_open(arg
->target
, NULL
, options
, flags
, errp
);
3176 zero_target
= (arg
->sync
== MIRROR_SYNC_MODE_FULL
&&
3177 (arg
->mode
== NEW_IMAGE_MODE_EXISTING
||
3178 !bdrv_has_zero_init(target_bs
)));
3181 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3182 old_context
= bdrv_get_aio_context(target_bs
);
3183 aio_context_release(aio_context
);
3184 aio_context_acquire(old_context
);
3186 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
3188 bdrv_unref(target_bs
);
3189 aio_context_release(old_context
);
3193 aio_context_release(old_context
);
3194 aio_context_acquire(aio_context
);
3196 blockdev_mirror_common(arg
->has_job_id
? arg
->job_id
: NULL
, bs
, target_bs
,
3197 arg
->has_replaces
, arg
->replaces
, arg
->sync
,
3198 backing_mode
, zero_target
,
3199 arg
->has_speed
, arg
->speed
,
3200 arg
->has_granularity
, arg
->granularity
,
3201 arg
->has_buf_size
, arg
->buf_size
,
3202 arg
->has_on_source_error
, arg
->on_source_error
,
3203 arg
->has_on_target_error
, arg
->on_target_error
,
3204 arg
->has_unmap
, arg
->unmap
,
3206 arg
->has_copy_mode
, arg
->copy_mode
,
3207 arg
->has_auto_finalize
, arg
->auto_finalize
,
3208 arg
->has_auto_dismiss
, arg
->auto_dismiss
,
3210 bdrv_unref(target_bs
);
3212 aio_context_release(aio_context
);
3215 void qmp_blockdev_mirror(bool has_job_id
, const char *job_id
,
3216 const char *device
, const char *target
,
3217 bool has_replaces
, const char *replaces
,
3218 MirrorSyncMode sync
,
3219 bool has_speed
, int64_t speed
,
3220 bool has_granularity
, uint32_t granularity
,
3221 bool has_buf_size
, int64_t buf_size
,
3222 bool has_on_source_error
,
3223 BlockdevOnError on_source_error
,
3224 bool has_on_target_error
,
3225 BlockdevOnError on_target_error
,
3226 bool has_filter_node_name
,
3227 const char *filter_node_name
,
3228 bool has_copy_mode
, MirrorCopyMode copy_mode
,
3229 bool has_auto_finalize
, bool auto_finalize
,
3230 bool has_auto_dismiss
, bool auto_dismiss
,
3233 BlockDriverState
*bs
;
3234 BlockDriverState
*target_bs
;
3235 AioContext
*aio_context
;
3236 AioContext
*old_context
;
3237 BlockMirrorBackingMode backing_mode
= MIRROR_LEAVE_BACKING_CHAIN
;
3241 bs
= qmp_get_root_bs(device
, errp
);
3246 target_bs
= bdrv_lookup_bs(target
, target
, errp
);
3251 zero_target
= (sync
== MIRROR_SYNC_MODE_FULL
);
3253 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3254 old_context
= bdrv_get_aio_context(target_bs
);
3255 aio_context
= bdrv_get_aio_context(bs
);
3256 aio_context_acquire(old_context
);
3258 ret
= bdrv_try_set_aio_context(target_bs
, aio_context
, errp
);
3260 aio_context_release(old_context
);
3261 aio_context_acquire(aio_context
);
3267 blockdev_mirror_common(has_job_id
? job_id
: NULL
, bs
, target_bs
,
3268 has_replaces
, replaces
, sync
, backing_mode
,
3269 zero_target
, has_speed
, speed
,
3270 has_granularity
, granularity
,
3271 has_buf_size
, buf_size
,
3272 has_on_source_error
, on_source_error
,
3273 has_on_target_error
, on_target_error
,
3275 has_filter_node_name
, filter_node_name
,
3276 has_copy_mode
, copy_mode
,
3277 has_auto_finalize
, auto_finalize
,
3278 has_auto_dismiss
, auto_dismiss
,
3281 aio_context_release(aio_context
);
3284 /* Get a block job using its ID and acquire its AioContext */
3285 static BlockJob
*find_block_job(const char *id
, AioContext
**aio_context
,
3292 *aio_context
= NULL
;
3294 job
= block_job_get(id
);
3297 error_set(errp
, ERROR_CLASS_DEVICE_NOT_ACTIVE
,
3298 "Block job '%s' not found", id
);
3302 *aio_context
= block_job_get_aio_context(job
);
3303 aio_context_acquire(*aio_context
);
3308 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
3310 AioContext
*aio_context
;
3311 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3317 block_job_set_speed(job
, speed
, errp
);
3318 aio_context_release(aio_context
);
3321 void qmp_block_job_cancel(const char *device
,
3322 bool has_force
, bool force
, Error
**errp
)
3324 AioContext
*aio_context
;
3325 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3335 if (job_user_paused(&job
->job
) && !force
) {
3336 error_setg(errp
, "The block job for device '%s' is currently paused",
3341 trace_qmp_block_job_cancel(job
);
3342 job_user_cancel(&job
->job
, force
, errp
);
3344 aio_context_release(aio_context
);
3347 void qmp_block_job_pause(const char *device
, Error
**errp
)
3349 AioContext
*aio_context
;
3350 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3356 trace_qmp_block_job_pause(job
);
3357 job_user_pause(&job
->job
, errp
);
3358 aio_context_release(aio_context
);
3361 void qmp_block_job_resume(const char *device
, Error
**errp
)
3363 AioContext
*aio_context
;
3364 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3370 trace_qmp_block_job_resume(job
);
3371 job_user_resume(&job
->job
, errp
);
3372 aio_context_release(aio_context
);
3375 void qmp_block_job_complete(const char *device
, Error
**errp
)
3377 AioContext
*aio_context
;
3378 BlockJob
*job
= find_block_job(device
, &aio_context
, errp
);
3384 trace_qmp_block_job_complete(job
);
3385 job_complete(&job
->job
, errp
);
3386 aio_context_release(aio_context
);
3389 void qmp_block_job_finalize(const char *id
, Error
**errp
)
3391 AioContext
*aio_context
;
3392 BlockJob
*job
= find_block_job(id
, &aio_context
, errp
);
3398 trace_qmp_block_job_finalize(job
);
3400 job_finalize(&job
->job
, errp
);
3403 * Job's context might have changed via job_finalize (and job_txn_apply
3404 * automatically acquires the new one), so make sure we release the correct
3407 aio_context
= block_job_get_aio_context(job
);
3408 job_unref(&job
->job
);
3409 aio_context_release(aio_context
);
3412 void qmp_block_job_dismiss(const char *id
, Error
**errp
)
3414 AioContext
*aio_context
;
3415 BlockJob
*bjob
= find_block_job(id
, &aio_context
, errp
);
3422 trace_qmp_block_job_dismiss(bjob
);
3424 job_dismiss(&job
, errp
);
3425 aio_context_release(aio_context
);
3428 void qmp_change_backing_file(const char *device
,
3429 const char *image_node_name
,
3430 const char *backing_file
,
3433 BlockDriverState
*bs
= NULL
;
3434 AioContext
*aio_context
;
3435 BlockDriverState
*image_bs
= NULL
;
3436 Error
*local_err
= NULL
;
3440 bs
= qmp_get_root_bs(device
, errp
);
3445 aio_context
= bdrv_get_aio_context(bs
);
3446 aio_context_acquire(aio_context
);
3448 image_bs
= bdrv_lookup_bs(NULL
, image_node_name
, &local_err
);
3450 error_propagate(errp
, local_err
);
3455 error_setg(errp
, "image file not found");
3459 if (bdrv_find_base(image_bs
) == image_bs
) {
3460 error_setg(errp
, "not allowing backing file change on an image "
3461 "without a backing file");
3465 /* even though we are not necessarily operating on bs, we need it to
3466 * determine if block ops are currently prohibited on the chain */
3467 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_CHANGE
, errp
)) {
3471 /* final sanity check */
3472 if (!bdrv_chain_contains(bs
, image_bs
)) {
3473 error_setg(errp
, "'%s' and image file are not in the same chain",
3478 /* if not r/w, reopen to make r/w */
3479 ro
= bdrv_is_read_only(image_bs
);
3482 if (bdrv_reopen_set_read_only(image_bs
, false, errp
) != 0) {
3487 ret
= bdrv_change_backing_file(image_bs
, backing_file
,
3488 image_bs
->drv
? image_bs
->drv
->format_name
: "",
3492 error_setg_errno(errp
, -ret
, "Could not change backing file to '%s'",
3494 /* don't exit here, so we can try to restore open flags if
3499 bdrv_reopen_set_read_only(image_bs
, true, errp
);
3503 aio_context_release(aio_context
);
3506 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
3508 BlockDriverState
*bs
;
3510 Visitor
*v
= qobject_output_visitor_new(&obj
);
3513 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3514 visit_complete(v
, &obj
);
3515 qdict
= qobject_to(QDict
, obj
);
3517 qdict_flatten(qdict
);
3519 if (!qdict_get_try_str(qdict
, "node-name")) {
3520 error_setg(errp
, "'node-name' must be specified for the root node");
3524 bs
= bds_tree_init(qdict
, errp
);
3529 bdrv_set_monitor_owned(bs
);
3535 void qmp_blockdev_reopen(BlockdevOptionsList
*reopen_list
, Error
**errp
)
3537 BlockReopenQueue
*queue
= NULL
;
3538 GSList
*drained
= NULL
;
3541 /* Add each one of the BDS that we want to reopen to the queue */
3542 for (; reopen_list
!= NULL
; reopen_list
= reopen_list
->next
) {
3543 BlockdevOptions
*options
= reopen_list
->value
;
3544 BlockDriverState
*bs
;
3550 /* Check for the selected node name */
3551 if (!options
->has_node_name
) {
3552 error_setg(errp
, "node-name not specified");
3556 bs
= bdrv_find_node(options
->node_name
);
3558 error_setg(errp
, "Failed to find node with node-name='%s'",
3559 options
->node_name
);
3563 /* Put all options in a QDict and flatten it */
3564 v
= qobject_output_visitor_new(&obj
);
3565 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3566 visit_complete(v
, &obj
);
3569 qdict
= qobject_to(QDict
, obj
);
3571 qdict_flatten(qdict
);
3573 ctx
= bdrv_get_aio_context(bs
);
3574 aio_context_acquire(ctx
);
3576 bdrv_subtree_drained_begin(bs
);
3577 queue
= bdrv_reopen_queue(queue
, bs
, qdict
, false);
3578 drained
= g_slist_prepend(drained
, bs
);
3580 aio_context_release(ctx
);
3583 /* Perform the reopen operation */
3584 bdrv_reopen_multiple(queue
, errp
);
3588 bdrv_reopen_queue_free(queue
);
3589 for (p
= drained
; p
; p
= p
->next
) {
3590 BlockDriverState
*bs
= p
->data
;
3591 AioContext
*ctx
= bdrv_get_aio_context(bs
);
3593 aio_context_acquire(ctx
);
3594 bdrv_subtree_drained_end(bs
);
3595 aio_context_release(ctx
);
3597 g_slist_free(drained
);
3600 void qmp_blockdev_del(const char *node_name
, Error
**errp
)
3602 AioContext
*aio_context
;
3603 BlockDriverState
*bs
;
3605 bs
= bdrv_find_node(node_name
);
3607 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3610 if (bdrv_has_blk(bs
)) {
3611 error_setg(errp
, "Node %s is in use", node_name
);
3614 aio_context
= bdrv_get_aio_context(bs
);
3615 aio_context_acquire(aio_context
);
3617 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_DRIVE_DEL
, errp
)) {
3621 if (!QTAILQ_IN_USE(bs
, monitor_list
)) {
3622 error_setg(errp
, "Node %s is not owned by the monitor",
3627 if (bs
->refcnt
> 1) {
3628 error_setg(errp
, "Block device %s is in use",
3629 bdrv_get_device_or_node_name(bs
));
3633 QTAILQ_REMOVE(&monitor_bdrv_states
, bs
, monitor_list
);
3637 aio_context_release(aio_context
);
3640 static BdrvChild
*bdrv_find_child(BlockDriverState
*parent_bs
,
3641 const char *child_name
)
3645 QLIST_FOREACH(child
, &parent_bs
->children
, next
) {
3646 if (strcmp(child
->name
, child_name
) == 0) {
3654 void qmp_x_blockdev_change(const char *parent
, bool has_child
,
3655 const char *child
, bool has_node
,
3656 const char *node
, Error
**errp
)
3658 BlockDriverState
*parent_bs
, *new_bs
= NULL
;
3661 parent_bs
= bdrv_lookup_bs(parent
, parent
, errp
);
3666 if (has_child
== has_node
) {
3668 error_setg(errp
, "The parameters child and node are in conflict");
3670 error_setg(errp
, "Either child or node must be specified");
3676 p_child
= bdrv_find_child(parent_bs
, child
);
3678 error_setg(errp
, "Node '%s' does not have child '%s'",
3682 bdrv_del_child(parent_bs
, p_child
, errp
);
3686 new_bs
= bdrv_find_node(node
);
3688 error_setg(errp
, "Node '%s' not found", node
);
3691 bdrv_add_child(parent_bs
, new_bs
, errp
);
3695 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
3697 BlockJobInfoList
*head
= NULL
, **tail
= &head
;
3700 for (job
= block_job_next(NULL
); job
; job
= block_job_next(job
)) {
3701 BlockJobInfo
*value
;
3702 AioContext
*aio_context
;
3704 if (block_job_is_internal(job
)) {
3707 aio_context
= block_job_get_aio_context(job
);
3708 aio_context_acquire(aio_context
);
3709 value
= block_job_query(job
, errp
);
3710 aio_context_release(aio_context
);
3712 qapi_free_BlockJobInfoList(head
);
3715 QAPI_LIST_APPEND(tail
, value
);
3721 void qmp_x_blockdev_set_iothread(const char *node_name
, StrOrNull
*iothread
,
3722 bool has_force
, bool force
, Error
**errp
)
3724 AioContext
*old_context
;
3725 AioContext
*new_context
;
3726 BlockDriverState
*bs
;
3728 bs
= bdrv_find_node(node_name
);
3730 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3734 /* Protects against accidents. */
3735 if (!(has_force
&& force
) && bdrv_has_blk(bs
)) {
3736 error_setg(errp
, "Node %s is associated with a BlockBackend and could "
3737 "be in use (use force=true to override this check)",
3742 if (iothread
->type
== QTYPE_QSTRING
) {
3743 IOThread
*obj
= iothread_by_id(iothread
->u
.s
);
3745 error_setg(errp
, "Cannot find iothread %s", iothread
->u
.s
);
3749 new_context
= iothread_get_aio_context(obj
);
3751 new_context
= qemu_get_aio_context();
3754 old_context
= bdrv_get_aio_context(bs
);
3755 aio_context_acquire(old_context
);
3757 bdrv_try_set_aio_context(bs
, new_context
, errp
);
3759 aio_context_release(old_context
);
3762 QemuOptsList qemu_common_drive_opts
= {
3764 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
3768 .type
= QEMU_OPT_BOOL
,
3769 .help
= "enable/disable snapshot mode",
3772 .type
= QEMU_OPT_STRING
,
3773 .help
= "host AIO implementation (threads, native, io_uring)",
3775 .name
= BDRV_OPT_CACHE_WB
,
3776 .type
= QEMU_OPT_BOOL
,
3777 .help
= "Enable writeback mode",
3780 .type
= QEMU_OPT_STRING
,
3781 .help
= "disk format (raw, qcow2, ...)",
3784 .type
= QEMU_OPT_STRING
,
3785 .help
= "read error action",
3788 .type
= QEMU_OPT_STRING
,
3789 .help
= "write error action",
3791 .name
= BDRV_OPT_READ_ONLY
,
3792 .type
= QEMU_OPT_BOOL
,
3793 .help
= "open drive file as read-only",
3799 .name
= "throttling.group",
3800 .type
= QEMU_OPT_STRING
,
3801 .help
= "name of the block throttling group",
3803 .name
= "copy-on-read",
3804 .type
= QEMU_OPT_BOOL
,
3805 .help
= "copy read data from backing file into image file",
3807 .name
= "detect-zeroes",
3808 .type
= QEMU_OPT_STRING
,
3809 .help
= "try to optimize zero writes (off, on, unmap)",
3811 .name
= "stats-account-invalid",
3812 .type
= QEMU_OPT_BOOL
,
3813 .help
= "whether to account for invalid I/O operations "
3814 "in the statistics",
3816 .name
= "stats-account-failed",
3817 .type
= QEMU_OPT_BOOL
,
3818 .help
= "whether to account for failed I/O operations "
3819 "in the statistics",
3821 { /* end of list */ }
3825 QemuOptsList qemu_drive_opts
= {
3827 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
3830 * no elements => accept any params
3831 * validation will happen later
3833 { /* end of list */ }