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/dirty-bitmap.h"
39 #include "block/qdict.h"
40 #include "block/throttle-groups.h"
41 #include "monitor/monitor.h"
42 #include "qemu/error-report.h"
43 #include "qemu/option.h"
44 #include "qemu/qemu-print.h"
45 #include "qemu/config-file.h"
46 #include "qapi/qapi-commands-block.h"
47 #include "qapi/qapi-commands-transaction.h"
48 #include "qapi/qapi-visit-block-core.h"
49 #include "qapi/qmp/qdict.h"
50 #include "qapi/qmp/qnum.h"
51 #include "qapi/qmp/qstring.h"
52 #include "qapi/error.h"
53 #include "qapi/qmp/qerror.h"
54 #include "qapi/qmp/qlist.h"
55 #include "qapi/qobject-output-visitor.h"
56 #include "sysemu/sysemu.h"
57 #include "sysemu/iothread.h"
58 #include "block/block_int.h"
59 #include "block/trace.h"
60 #include "sysemu/runstate.h"
61 #include "sysemu/replay.h"
62 #include "qemu/cutils.h"
63 #include "qemu/help_option.h"
64 #include "qemu/main-loop.h"
65 #include "qemu/throttle-options.h"
67 /* Protected by BQL */
68 QTAILQ_HEAD(, BlockDriverState
) monitor_bdrv_states
=
69 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states
);
71 void bdrv_set_monitor_owned(BlockDriverState
*bs
)
74 QTAILQ_INSERT_TAIL(&monitor_bdrv_states
, bs
, monitor_list
);
77 static const char *const if_name
[IF_COUNT
] = {
81 [IF_FLOPPY
] = "floppy",
82 [IF_PFLASH
] = "pflash",
85 [IF_VIRTIO
] = "virtio",
89 static int if_max_devs
[IF_COUNT
] = {
91 * Do not change these numbers! They govern how drive option
92 * index maps to unit and bus. That mapping is ABI.
94 * All controllers used to implement if=T drives need to support
95 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
96 * Otherwise, some index values map to "impossible" bus, unit
99 * For instance, if you change [IF_SCSI] to 255, -drive
100 * if=scsi,index=12 no longer means bus=1,unit=5, but
101 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
102 * the drive can't be set up. Regression.
109 * Boards may call this to offer board-by-board overrides
110 * of the default, global values.
112 void override_max_devs(BlockInterfaceType type
, int max_devs
)
123 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
124 dinfo
= blk_legacy_dinfo(blk
);
125 if (dinfo
->type
== type
) {
126 fprintf(stderr
, "Cannot override units-per-bus property of"
127 " the %s interface, because a drive of that type has"
128 " already been added.\n", if_name
[type
]);
129 g_assert_not_reached();
133 if_max_devs
[type
] = max_devs
;
137 * We automatically delete the drive when a device using it gets
138 * unplugged. Questionable feature, but we can't just drop it.
139 * Device models call blockdev_mark_auto_del() to schedule the
140 * automatic deletion, and generic qdev code calls blockdev_auto_del()
141 * when deletion is actually safe.
143 void blockdev_mark_auto_del(BlockBackend
*blk
)
145 DriveInfo
*dinfo
= blk_legacy_dinfo(blk
);
157 job
= block_job_next_locked(NULL
);
158 while (job
&& (job
->job
.cancelled
||
159 job
->job
.deferred_to_main_loop
||
160 !block_job_has_bdrv(job
, blk_bs(blk
))))
162 job
= block_job_next_locked(job
);
166 * This drops the job lock temporarily and polls, so we need to
167 * restart processing the list from the start after this.
169 job_cancel_locked(&job
->job
, false);
176 void blockdev_auto_del(BlockBackend
*blk
)
178 DriveInfo
*dinfo
= blk_legacy_dinfo(blk
);
181 if (dinfo
&& dinfo
->auto_del
) {
182 monitor_remove_blk(blk
);
187 static int drive_index_to_bus_id(BlockInterfaceType type
, int index
)
189 int max_devs
= if_max_devs
[type
];
190 return max_devs
? index
/ max_devs
: 0;
193 static int drive_index_to_unit_id(BlockInterfaceType type
, int index
)
195 int max_devs
= if_max_devs
[type
];
196 return max_devs
? index
% max_devs
: index
;
199 QemuOpts
*drive_add(BlockInterfaceType type
, int index
, const char *file
,
206 opts
= qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr
, false);
210 if (type
!= IF_DEFAULT
) {
211 qemu_opt_set(opts
, "if", if_name
[type
], &error_abort
);
214 qemu_opt_set_number(opts
, "index", index
, &error_abort
);
217 qemu_opt_set(opts
, "file", file
, &error_abort
);
221 DriveInfo
*drive_get(BlockInterfaceType type
, int bus
, int unit
)
228 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
229 dinfo
= blk_legacy_dinfo(blk
);
230 if (dinfo
&& dinfo
->type
== type
231 && dinfo
->bus
== bus
&& dinfo
->unit
== unit
) {
240 * Check board claimed all -drive that are meant to be claimed.
241 * Fatal error if any remain unclaimed.
243 void drive_check_orphaned(void)
248 bool orphans
= false;
252 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
253 dinfo
= blk_legacy_dinfo(blk
);
255 * Ignore default drives, because we create certain default
256 * drives unconditionally, then leave them unclaimed. Not the
258 * Ignore IF_VIRTIO, because it gets desugared into -device,
259 * so we can leave failing to -device.
260 * Ignore IF_NONE, because leaving unclaimed IF_NONE remains
261 * available for device_add is a feature.
263 if (dinfo
->is_default
|| dinfo
->type
== IF_VIRTIO
264 || dinfo
->type
== IF_NONE
) {
267 if (!blk_get_attached_dev(blk
)) {
269 qemu_opts_loc_restore(dinfo
->opts
);
270 error_report("machine type does not support"
271 " if=%s,bus=%d,unit=%d",
272 if_name
[dinfo
->type
], dinfo
->bus
, dinfo
->unit
);
283 DriveInfo
*drive_get_by_index(BlockInterfaceType type
, int index
)
286 return drive_get(type
,
287 drive_index_to_bus_id(type
, index
),
288 drive_index_to_unit_id(type
, index
));
291 int drive_get_max_bus(BlockInterfaceType type
)
300 for (blk
= blk_next(NULL
); blk
; blk
= blk_next(blk
)) {
301 dinfo
= blk_legacy_dinfo(blk
);
302 if (dinfo
&& dinfo
->type
== type
&& dinfo
->bus
> max_bus
) {
303 max_bus
= dinfo
->bus
;
309 static void bdrv_format_print(void *opaque
, const char *name
)
311 qemu_printf(" %s", name
);
316 BlockDriverState
*bs
;
319 static int parse_block_error_action(const char *buf
, bool is_read
, Error
**errp
)
321 if (!strcmp(buf
, "ignore")) {
322 return BLOCKDEV_ON_ERROR_IGNORE
;
323 } else if (!is_read
&& !strcmp(buf
, "enospc")) {
324 return BLOCKDEV_ON_ERROR_ENOSPC
;
325 } else if (!strcmp(buf
, "stop")) {
326 return BLOCKDEV_ON_ERROR_STOP
;
327 } else if (!strcmp(buf
, "report")) {
328 return BLOCKDEV_ON_ERROR_REPORT
;
330 error_setg(errp
, "'%s' invalid %s error action",
331 buf
, is_read
? "read" : "write");
336 static bool parse_stats_intervals(BlockAcctStats
*stats
, QList
*intervals
,
339 const QListEntry
*entry
;
340 for (entry
= qlist_first(intervals
); entry
; entry
= qlist_next(entry
)) {
341 switch (qobject_type(entry
->value
)) {
343 case QTYPE_QSTRING
: {
344 unsigned long long length
;
345 const char *str
= qstring_get_str(qobject_to(QString
,
347 if (parse_uint_full(str
, &length
, 10) == 0 &&
348 length
> 0 && length
<= UINT_MAX
) {
349 block_acct_add_interval(stats
, (unsigned) length
);
351 error_setg(errp
, "Invalid interval length: %s", str
);
358 int64_t length
= qnum_get_int(qobject_to(QNum
, entry
->value
));
360 if (length
> 0 && length
<= UINT_MAX
) {
361 block_acct_add_interval(stats
, (unsigned) length
);
363 error_setg(errp
, "Invalid interval length: %" PRId64
, length
);
370 error_setg(errp
, "The specification of stats-intervals is invalid");
377 typedef enum { MEDIA_DISK
, MEDIA_CDROM
} DriveMediaType
;
379 /* All parameters but @opts are optional and may be set to NULL. */
380 static void extract_common_blockdev_options(QemuOpts
*opts
, int *bdrv_flags
,
381 const char **throttling_group
, ThrottleConfig
*throttle_cfg
,
382 BlockdevDetectZeroesOptions
*detect_zeroes
, Error
**errp
)
384 Error
*local_error
= NULL
;
388 if (qemu_opt_get_bool(opts
, "copy-on-read", false)) {
389 *bdrv_flags
|= BDRV_O_COPY_ON_READ
;
392 if ((aio
= qemu_opt_get(opts
, "aio")) != NULL
) {
393 if (bdrv_parse_aio(aio
, bdrv_flags
) < 0) {
394 error_setg(errp
, "invalid aio option");
400 /* disk I/O throttling */
401 if (throttling_group
) {
402 *throttling_group
= qemu_opt_get(opts
, "throttling.group");
406 throttle_config_init(throttle_cfg
);
407 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].avg
=
408 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
409 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].avg
=
410 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
411 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].avg
=
412 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
413 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].avg
=
414 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
415 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].avg
=
416 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
417 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].avg
=
418 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
420 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].max
=
421 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
422 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].max
=
423 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
424 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].max
=
425 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
426 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].max
=
427 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
428 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].max
=
429 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
430 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].max
=
431 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
433 throttle_cfg
->buckets
[THROTTLE_BPS_TOTAL
].burst_length
=
434 qemu_opt_get_number(opts
, "throttling.bps-total-max-length", 1);
435 throttle_cfg
->buckets
[THROTTLE_BPS_READ
].burst_length
=
436 qemu_opt_get_number(opts
, "throttling.bps-read-max-length", 1);
437 throttle_cfg
->buckets
[THROTTLE_BPS_WRITE
].burst_length
=
438 qemu_opt_get_number(opts
, "throttling.bps-write-max-length", 1);
439 throttle_cfg
->buckets
[THROTTLE_OPS_TOTAL
].burst_length
=
440 qemu_opt_get_number(opts
, "throttling.iops-total-max-length", 1);
441 throttle_cfg
->buckets
[THROTTLE_OPS_READ
].burst_length
=
442 qemu_opt_get_number(opts
, "throttling.iops-read-max-length", 1);
443 throttle_cfg
->buckets
[THROTTLE_OPS_WRITE
].burst_length
=
444 qemu_opt_get_number(opts
, "throttling.iops-write-max-length", 1);
446 throttle_cfg
->op_size
=
447 qemu_opt_get_number(opts
, "throttling.iops-size", 0);
449 if (!throttle_is_valid(throttle_cfg
, errp
)) {
456 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup
,
457 qemu_opt_get(opts
, "detect-zeroes"),
458 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
,
461 error_propagate(errp
, local_error
);
467 static OnOffAuto
account_get_opt(QemuOpts
*opts
, const char *name
)
469 if (!qemu_opt_find(opts
, name
)) {
470 return ON_OFF_AUTO_AUTO
;
472 if (qemu_opt_get_bool(opts
, name
, true)) {
473 return ON_OFF_AUTO_ON
;
475 return ON_OFF_AUTO_OFF
;
478 /* Takes the ownership of bs_opts */
479 static BlockBackend
*blockdev_init(const char *file
, QDict
*bs_opts
,
484 int on_read_error
, on_write_error
;
485 OnOffAuto account_invalid
, account_failed
;
486 bool writethrough
, read_only
;
488 BlockDriverState
*bs
;
493 QDict
*interval_dict
= NULL
;
494 QList
*interval_list
= NULL
;
496 BlockdevDetectZeroesOptions detect_zeroes
=
497 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF
;
498 const char *throttling_group
= NULL
;
500 /* Check common options by copying from bs_opts to opts, all other options
501 * stay in bs_opts for processing by bdrv_open(). */
502 id
= qdict_get_try_str(bs_opts
, "id");
503 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, errp
);
508 if (!qemu_opts_absorb_qdict(opts
, bs_opts
, errp
)) {
513 qdict_del(bs_opts
, "id");
516 /* extract parameters */
517 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
519 account_invalid
= account_get_opt(opts
, "stats-account-invalid");
520 account_failed
= account_get_opt(opts
, "stats-account-failed");
522 writethrough
= !qemu_opt_get_bool(opts
, BDRV_OPT_CACHE_WB
, true);
524 id
= qemu_opts_id(opts
);
526 qdict_extract_subqdict(bs_opts
, &interval_dict
, "stats-intervals.");
527 qdict_array_split(interval_dict
, &interval_list
);
529 if (qdict_size(interval_dict
) != 0) {
530 error_setg(errp
, "Invalid option stats-intervals.%s",
531 qdict_first(interval_dict
)->key
);
535 extract_common_blockdev_options(opts
, &bdrv_flags
, &throttling_group
, &cfg
,
536 &detect_zeroes
, &error
);
538 error_propagate(errp
, error
);
542 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
543 if (is_help_option(buf
)) {
544 qemu_printf("Supported formats:");
545 bdrv_iterate_format(bdrv_format_print
, NULL
, false);
546 qemu_printf("\nSupported formats (read-only):");
547 bdrv_iterate_format(bdrv_format_print
, NULL
, true);
552 if (qdict_haskey(bs_opts
, "driver")) {
553 error_setg(errp
, "Cannot specify both 'driver' and 'format'");
556 qdict_put_str(bs_opts
, "driver", buf
);
559 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
560 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
561 on_write_error
= parse_block_error_action(buf
, 0, &error
);
563 error_propagate(errp
, error
);
568 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
569 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
570 on_read_error
= parse_block_error_action(buf
, 1, &error
);
572 error_propagate(errp
, error
);
578 bdrv_flags
|= BDRV_O_SNAPSHOT
;
581 read_only
= qemu_opt_get_bool(opts
, BDRV_OPT_READ_ONLY
, false);
584 if ((!file
|| !*file
) && !qdict_size(bs_opts
)) {
585 BlockBackendRootState
*blk_rs
;
587 blk
= blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL
);
588 blk_rs
= blk_get_root_state(blk
);
589 blk_rs
->open_flags
= bdrv_flags
| (read_only
? 0 : BDRV_O_RDWR
);
590 blk_rs
->detect_zeroes
= detect_zeroes
;
592 qobject_unref(bs_opts
);
594 if (file
&& !*file
) {
598 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
599 * with other callers) rather than what we want as the real defaults.
600 * Apply the defaults here instead. */
601 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_DIRECT
, "off");
602 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_NO_FLUSH
, "off");
603 qdict_set_default_str(bs_opts
, BDRV_OPT_READ_ONLY
,
604 read_only
? "on" : "off");
605 qdict_set_default_str(bs_opts
, BDRV_OPT_AUTO_READ_ONLY
, "on");
606 assert((bdrv_flags
& BDRV_O_CACHE_MASK
) == 0);
608 if (runstate_check(RUN_STATE_INMIGRATE
)) {
609 bdrv_flags
|= BDRV_O_INACTIVE
;
612 blk
= blk_new_open(file
, NULL
, bs_opts
, bdrv_flags
, errp
);
618 bs
->detect_zeroes
= detect_zeroes
;
620 block_acct_setup(blk_get_stats(blk
), account_invalid
, account_failed
);
622 if (!parse_stats_intervals(blk_get_stats(blk
), interval_list
, errp
)) {
629 /* disk I/O throttling */
630 if (throttle_enabled(&cfg
)) {
631 if (!throttling_group
) {
632 throttling_group
= id
;
634 blk_io_limits_enable(blk
, throttling_group
);
635 blk_set_io_limits(blk
, &cfg
);
638 blk_set_enable_write_cache(blk
, !writethrough
);
639 blk_set_on_error(blk
, on_read_error
, on_write_error
);
641 if (!monitor_add_blk(blk
, id
, errp
)) {
649 qobject_unref(interval_dict
);
650 qobject_unref(interval_list
);
655 qobject_unref(interval_dict
);
656 qobject_unref(interval_list
);
658 qobject_unref(bs_opts
);
662 /* Takes the ownership of bs_opts */
663 BlockDriverState
*bds_tree_init(QDict
*bs_opts
, Error
**errp
)
668 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
669 * with other callers) rather than what we want as the real defaults.
670 * Apply the defaults here instead. */
671 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_DIRECT
, "off");
672 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_NO_FLUSH
, "off");
673 qdict_set_default_str(bs_opts
, BDRV_OPT_READ_ONLY
, "off");
675 if (runstate_check(RUN_STATE_INMIGRATE
)) {
676 bdrv_flags
|= BDRV_O_INACTIVE
;
679 return bdrv_open(NULL
, NULL
, bs_opts
, bdrv_flags
, errp
);
682 void blockdev_close_all_bdrv_states(void)
684 BlockDriverState
*bs
, *next_bs
;
687 QTAILQ_FOREACH_SAFE(bs
, &monitor_bdrv_states
, monitor_list
, next_bs
) {
688 AioContext
*ctx
= bdrv_get_aio_context(bs
);
690 aio_context_acquire(ctx
);
692 aio_context_release(ctx
);
696 /* Iterates over the list of monitor-owned BlockDriverStates */
697 BlockDriverState
*bdrv_next_monitor_owned(BlockDriverState
*bs
)
700 return bs
? QTAILQ_NEXT(bs
, monitor_list
)
701 : QTAILQ_FIRST(&monitor_bdrv_states
);
704 static bool qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
,
709 value
= qemu_opt_get(opts
, from
);
711 if (qemu_opt_find(opts
, to
)) {
712 error_setg(errp
, "'%s' and its alias '%s' can't be used at the "
713 "same time", to
, from
);
718 /* rename all items in opts */
719 while ((value
= qemu_opt_get(opts
, from
))) {
720 qemu_opt_set(opts
, to
, value
, &error_abort
);
721 qemu_opt_unset(opts
, from
);
726 QemuOptsList qemu_legacy_drive_opts
= {
728 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
732 .type
= QEMU_OPT_NUMBER
,
733 .help
= "bus number",
736 .type
= QEMU_OPT_NUMBER
,
737 .help
= "unit number (i.e. lun for scsi)",
740 .type
= QEMU_OPT_NUMBER
,
741 .help
= "index number",
744 .type
= QEMU_OPT_STRING
,
745 .help
= "media type (disk, cdrom)",
748 .type
= QEMU_OPT_STRING
,
749 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
752 .type
= QEMU_OPT_STRING
,
756 /* Options that are passed on, but have special semantics with -drive */
758 .name
= BDRV_OPT_READ_ONLY
,
759 .type
= QEMU_OPT_BOOL
,
760 .help
= "open drive file as read-only",
763 .type
= QEMU_OPT_STRING
,
764 .help
= "read error action",
767 .type
= QEMU_OPT_STRING
,
768 .help
= "write error action",
770 .name
= "copy-on-read",
771 .type
= QEMU_OPT_BOOL
,
772 .help
= "copy read data from backing file into image file",
775 { /* end of list */ }
779 DriveInfo
*drive_new(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
,
784 DriveInfo
*dinfo
= NULL
;
786 QemuOpts
*legacy_opts
;
787 DriveMediaType media
= MEDIA_DISK
;
788 BlockInterfaceType type
;
789 int max_devs
, bus_id
, unit_id
, index
;
790 const char *werror
, *rerror
;
791 bool read_only
= false;
793 const char *filename
;
798 /* Change legacy command line options into QMP ones */
799 static const struct {
803 { "iops", "throttling.iops-total" },
804 { "iops_rd", "throttling.iops-read" },
805 { "iops_wr", "throttling.iops-write" },
807 { "bps", "throttling.bps-total" },
808 { "bps_rd", "throttling.bps-read" },
809 { "bps_wr", "throttling.bps-write" },
811 { "iops_max", "throttling.iops-total-max" },
812 { "iops_rd_max", "throttling.iops-read-max" },
813 { "iops_wr_max", "throttling.iops-write-max" },
815 { "bps_max", "throttling.bps-total-max" },
816 { "bps_rd_max", "throttling.bps-read-max" },
817 { "bps_wr_max", "throttling.bps-write-max" },
819 { "iops_size", "throttling.iops-size" },
821 { "group", "throttling.group" },
823 { "readonly", BDRV_OPT_READ_ONLY
},
826 for (i
= 0; i
< ARRAY_SIZE(opt_renames
); i
++) {
827 if (!qemu_opt_rename(all_opts
, opt_renames
[i
].from
,
828 opt_renames
[i
].to
, errp
)) {
833 value
= qemu_opt_get(all_opts
, "cache");
838 if (bdrv_parse_cache_mode(value
, &flags
, &writethrough
) != 0) {
839 error_setg(errp
, "invalid cache option");
843 /* Specific options take precedence */
844 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_WB
)) {
845 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_WB
,
846 !writethrough
, &error_abort
);
848 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_DIRECT
)) {
849 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_DIRECT
,
850 !!(flags
& BDRV_O_NOCACHE
), &error_abort
);
852 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
)) {
853 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
,
854 !!(flags
& BDRV_O_NO_FLUSH
), &error_abort
);
856 qemu_opt_unset(all_opts
, "cache");
859 /* Get a QDict for processing the options */
860 bs_opts
= qdict_new();
861 qemu_opts_to_qdict(all_opts
, bs_opts
);
863 legacy_opts
= qemu_opts_create(&qemu_legacy_drive_opts
, NULL
, 0,
865 if (!qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, errp
)) {
870 value
= qemu_opt_get(legacy_opts
, "media");
872 if (!strcmp(value
, "disk")) {
874 } else if (!strcmp(value
, "cdrom")) {
878 error_setg(errp
, "'%s' invalid media", value
);
883 /* copy-on-read is disabled with a warning for read-only devices */
884 read_only
|= qemu_opt_get_bool(legacy_opts
, BDRV_OPT_READ_ONLY
, false);
885 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
887 if (read_only
&& copy_on_read
) {
888 warn_report("disabling copy-on-read on read-only drive");
889 copy_on_read
= false;
892 qdict_put_str(bs_opts
, BDRV_OPT_READ_ONLY
, read_only
? "on" : "off");
893 qdict_put_str(bs_opts
, "copy-on-read", copy_on_read
? "on" : "off");
895 /* Controller type */
896 value
= qemu_opt_get(legacy_opts
, "if");
899 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
902 if (type
== IF_COUNT
) {
903 error_setg(errp
, "unsupported bus type '%s'", value
);
907 type
= block_default_type
;
910 /* Device address specified by bus/unit or index.
911 * If none was specified, try to find the first free one. */
912 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
913 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
914 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
916 max_devs
= if_max_devs
[type
];
919 if (bus_id
!= 0 || unit_id
!= -1) {
920 error_setg(errp
, "index cannot be used with bus and unit");
923 bus_id
= drive_index_to_bus_id(type
, index
);
924 unit_id
= drive_index_to_unit_id(type
, index
);
929 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
931 if (max_devs
&& unit_id
>= max_devs
) {
938 if (max_devs
&& unit_id
>= max_devs
) {
939 error_setg(errp
, "unit %d too big (max is %d)", unit_id
, max_devs
- 1);
943 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
944 error_setg(errp
, "drive with bus=%d, unit=%d (index=%d) exists",
945 bus_id
, unit_id
, index
);
949 /* no id supplied -> create one */
950 if (qemu_opts_id(all_opts
) == NULL
) {
952 const char *mediastr
= "";
953 if (type
== IF_IDE
|| type
== IF_SCSI
) {
954 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
957 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
960 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
963 qdict_put_str(bs_opts
, "id", new_id
);
967 /* Add virtio block device */
968 if (type
== IF_VIRTIO
) {
970 devopts
= qemu_opts_create(qemu_find_opts("device"), NULL
, 0,
972 qemu_opt_set(devopts
, "driver", "virtio-blk", &error_abort
);
973 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"),
977 filename
= qemu_opt_get(legacy_opts
, "file");
979 /* Check werror/rerror compatibility with if=... */
980 werror
= qemu_opt_get(legacy_opts
, "werror");
981 if (werror
!= NULL
) {
982 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&&
984 error_setg(errp
, "werror is not supported by this bus type");
987 qdict_put_str(bs_opts
, "werror", werror
);
990 rerror
= qemu_opt_get(legacy_opts
, "rerror");
991 if (rerror
!= NULL
) {
992 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&&
994 error_setg(errp
, "rerror is not supported by this bus type");
997 qdict_put_str(bs_opts
, "rerror", rerror
);
1000 /* Actual block device init: Functionality shared with blockdev-add */
1001 blk
= blockdev_init(filename
, bs_opts
, errp
);
1007 /* Create legacy DriveInfo */
1008 dinfo
= g_malloc0(sizeof(*dinfo
));
1009 dinfo
->opts
= all_opts
;
1012 dinfo
->bus
= bus_id
;
1013 dinfo
->unit
= unit_id
;
1015 blk_set_legacy_dinfo(blk
, dinfo
);
1022 dinfo
->media_cd
= media
== MEDIA_CDROM
;
1029 qemu_opts_del(legacy_opts
);
1030 qobject_unref(bs_opts
);
1034 static BlockDriverState
*qmp_get_root_bs(const char *name
, Error
**errp
)
1036 BlockDriverState
*bs
;
1037 AioContext
*aio_context
;
1039 bs
= bdrv_lookup_bs(name
, name
, errp
);
1044 if (!bdrv_is_root_node(bs
)) {
1045 error_setg(errp
, "Need a root block node");
1049 aio_context
= bdrv_get_aio_context(bs
);
1050 aio_context_acquire(aio_context
);
1052 if (!bdrv_is_inserted(bs
)) {
1053 error_setg(errp
, "Device has no medium");
1057 aio_context_release(aio_context
);
1062 static void blockdev_do_action(TransactionAction
*action
, Error
**errp
)
1064 TransactionActionList list
;
1066 list
.value
= action
;
1068 qmp_transaction(&list
, NULL
, errp
);
1071 void qmp_blockdev_snapshot_sync(const char *device
, const char *node_name
,
1072 const char *snapshot_file
,
1073 const char *snapshot_node_name
,
1075 bool has_mode
, NewImageMode mode
, Error
**errp
)
1077 BlockdevSnapshotSync snapshot
= {
1078 .device
= (char *) device
,
1079 .node_name
= (char *) node_name
,
1080 .snapshot_file
= (char *) snapshot_file
,
1081 .snapshot_node_name
= (char *) snapshot_node_name
,
1082 .format
= (char *) format
,
1083 .has_mode
= has_mode
,
1086 TransactionAction action
= {
1087 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
1088 .u
.blockdev_snapshot_sync
.data
= &snapshot
,
1090 blockdev_do_action(&action
, errp
);
1093 void qmp_blockdev_snapshot(const char *node
, const char *overlay
,
1096 BlockdevSnapshot snapshot_data
= {
1097 .node
= (char *) node
,
1098 .overlay
= (char *) overlay
1100 TransactionAction action
= {
1101 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
,
1102 .u
.blockdev_snapshot
.data
= &snapshot_data
,
1104 blockdev_do_action(&action
, errp
);
1107 void qmp_blockdev_snapshot_internal_sync(const char *device
,
1111 BlockdevSnapshotInternal snapshot
= {
1112 .device
= (char *) device
,
1113 .name
= (char *) name
1115 TransactionAction action
= {
1116 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
1117 .u
.blockdev_snapshot_internal_sync
.data
= &snapshot
,
1119 blockdev_do_action(&action
, errp
);
1122 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
1127 BlockDriverState
*bs
;
1128 AioContext
*aio_context
;
1129 QEMUSnapshotInfo sn
;
1130 Error
*local_err
= NULL
;
1131 SnapshotInfo
*info
= NULL
;
1134 bs
= qmp_get_root_bs(device
, errp
);
1138 aio_context
= bdrv_get_aio_context(bs
);
1139 aio_context_acquire(aio_context
);
1142 error_setg(errp
, "Name or id must be provided");
1143 goto out_aio_context
;
1146 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE
, errp
)) {
1147 goto out_aio_context
;
1150 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1152 error_propagate(errp
, local_err
);
1153 goto out_aio_context
;
1157 "Snapshot with id '%s' and name '%s' does not exist on "
1159 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1160 goto out_aio_context
;
1163 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1165 error_propagate(errp
, local_err
);
1166 goto out_aio_context
;
1169 aio_context_release(aio_context
);
1171 info
= g_new0(SnapshotInfo
, 1);
1172 info
->id
= g_strdup(sn
.id_str
);
1173 info
->name
= g_strdup(sn
.name
);
1174 info
->date_nsec
= sn
.date_nsec
;
1175 info
->date_sec
= sn
.date_sec
;
1176 info
->vm_state_size
= sn
.vm_state_size
;
1177 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1178 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1179 if (sn
.icount
!= -1ULL) {
1180 info
->icount
= sn
.icount
;
1181 info
->has_icount
= true;
1187 aio_context_release(aio_context
);
1191 /* New and old BlockDriverState structs for atomic group operations */
1193 typedef struct BlkActionState BlkActionState
;
1197 * Table of operations that define an Action.
1199 * @instance_size: Size of state struct, in bytes.
1200 * @prepare: Prepare the work, must NOT be NULL.
1201 * @commit: Commit the changes, can be NULL.
1202 * @abort: Abort the changes on fail, can be NULL.
1203 * @clean: Clean up resources after all transaction actions have called
1204 * commit() or abort(). Can be NULL.
1206 * Only prepare() may fail. In a single transaction, only one of commit() or
1207 * abort() will be called. clean() will always be called if it is present.
1209 * Always run under BQL.
1211 typedef struct BlkActionOps
{
1212 size_t instance_size
;
1213 void (*prepare
)(BlkActionState
*common
, Error
**errp
);
1214 void (*commit
)(BlkActionState
*common
);
1215 void (*abort
)(BlkActionState
*common
);
1216 void (*clean
)(BlkActionState
*common
);
1221 * Describes one Action's state within a Transaction.
1223 * @action: QAPI-defined enum identifying which Action to perform.
1224 * @ops: Table of ActionOps this Action can perform.
1225 * @block_job_txn: Transaction which this action belongs to.
1226 * @entry: List membership for all Actions in this Transaction.
1228 * This structure must be arranged as first member in a subclassed type,
1229 * assuming that the compiler will also arrange it to the same offsets as the
1232 struct BlkActionState
{
1233 TransactionAction
*action
;
1234 const BlkActionOps
*ops
;
1235 JobTxn
*block_job_txn
;
1236 TransactionProperties
*txn_props
;
1237 QTAILQ_ENTRY(BlkActionState
) entry
;
1240 /* internal snapshot private data */
1241 typedef struct InternalSnapshotState
{
1242 BlkActionState common
;
1243 BlockDriverState
*bs
;
1244 QEMUSnapshotInfo sn
;
1246 } InternalSnapshotState
;
1249 static int action_check_completion_mode(BlkActionState
*s
, Error
**errp
)
1251 if (s
->txn_props
->completion_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
1253 "Action '%s' does not support Transaction property "
1254 "completion-mode = %s",
1255 TransactionActionKind_str(s
->action
->type
),
1256 ActionCompletionMode_str(s
->txn_props
->completion_mode
));
1262 static void internal_snapshot_prepare(BlkActionState
*common
,
1265 Error
*local_err
= NULL
;
1268 BlockDriverState
*bs
;
1269 QEMUSnapshotInfo old_sn
, *sn
;
1272 BlockdevSnapshotInternal
*internal
;
1273 InternalSnapshotState
*state
;
1274 AioContext
*aio_context
;
1277 g_assert(common
->action
->type
==
1278 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1279 internal
= common
->action
->u
.blockdev_snapshot_internal_sync
.data
;
1280 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1282 /* 1. parse input */
1283 device
= internal
->device
;
1284 name
= internal
->name
;
1286 /* 2. check for validation */
1287 if (action_check_completion_mode(common
, errp
) < 0) {
1291 bs
= qmp_get_root_bs(device
, errp
);
1296 aio_context
= bdrv_get_aio_context(bs
);
1297 aio_context_acquire(aio_context
);
1301 /* Paired with .clean() */
1302 bdrv_drained_begin(bs
);
1304 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT
, errp
)) {
1308 if (bdrv_is_read_only(bs
)) {
1309 error_setg(errp
, "Device '%s' is read only", device
);
1313 if (!bdrv_can_snapshot(bs
)) {
1314 error_setg(errp
, "Block format '%s' used by device '%s' "
1315 "does not support internal snapshots",
1316 bs
->drv
->format_name
, device
);
1320 if (!strlen(name
)) {
1321 error_setg(errp
, "Name is empty");
1325 /* check whether a snapshot with name exist */
1326 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
,
1329 error_propagate(errp
, local_err
);
1333 "Snapshot with name '%s' already exists on device '%s'",
1338 /* 3. take the snapshot */
1340 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1341 rt
= g_get_real_time();
1342 sn
->date_sec
= rt
/ G_USEC_PER_SEC
;
1343 sn
->date_nsec
= (rt
% G_USEC_PER_SEC
) * 1000;
1344 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1345 if (replay_mode
!= REPLAY_MODE_NONE
) {
1346 sn
->icount
= replay_get_current_icount();
1351 ret1
= bdrv_snapshot_create(bs
, sn
);
1353 error_setg_errno(errp
, -ret1
,
1354 "Failed to create snapshot '%s' on device '%s'",
1359 /* 4. succeed, mark a snapshot is created */
1360 state
->created
= true;
1363 aio_context_release(aio_context
);
1366 static void internal_snapshot_abort(BlkActionState
*common
)
1368 InternalSnapshotState
*state
=
1369 DO_UPCAST(InternalSnapshotState
, common
, common
);
1370 BlockDriverState
*bs
= state
->bs
;
1371 QEMUSnapshotInfo
*sn
= &state
->sn
;
1372 AioContext
*aio_context
;
1373 Error
*local_error
= NULL
;
1375 if (!state
->created
) {
1379 aio_context
= bdrv_get_aio_context(state
->bs
);
1380 aio_context_acquire(aio_context
);
1382 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1383 error_reportf_err(local_error
,
1384 "Failed to delete snapshot with id '%s' and "
1385 "name '%s' on device '%s' in abort: ",
1386 sn
->id_str
, sn
->name
,
1387 bdrv_get_device_name(bs
));
1390 aio_context_release(aio_context
);
1393 static void internal_snapshot_clean(BlkActionState
*common
)
1395 InternalSnapshotState
*state
= DO_UPCAST(InternalSnapshotState
,
1397 AioContext
*aio_context
;
1403 aio_context
= bdrv_get_aio_context(state
->bs
);
1404 aio_context_acquire(aio_context
);
1406 bdrv_drained_end(state
->bs
);
1408 aio_context_release(aio_context
);
1411 /* external snapshot private data */
1412 typedef struct ExternalSnapshotState
{
1413 BlkActionState common
;
1414 BlockDriverState
*old_bs
;
1415 BlockDriverState
*new_bs
;
1416 bool overlay_appended
;
1417 } ExternalSnapshotState
;
1419 static void external_snapshot_prepare(BlkActionState
*common
,
1424 QDict
*options
= NULL
;
1425 Error
*local_err
= NULL
;
1426 /* Device and node name of the image to generate the snapshot from */
1428 const char *node_name
;
1429 /* Reference to the new image (for 'blockdev-snapshot') */
1430 const char *snapshot_ref
;
1431 /* File name of the new image (for 'blockdev-snapshot-sync') */
1432 const char *new_image_file
;
1433 ExternalSnapshotState
*state
=
1434 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1435 TransactionAction
*action
= common
->action
;
1436 AioContext
*aio_context
;
1437 uint64_t perm
, shared
;
1439 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1440 * purpose but a different set of parameters */
1441 switch (action
->type
) {
1442 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
:
1444 BlockdevSnapshot
*s
= action
->u
.blockdev_snapshot
.data
;
1446 node_name
= s
->node
;
1447 new_image_file
= NULL
;
1448 snapshot_ref
= s
->overlay
;
1451 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
:
1453 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1455 node_name
= s
->node_name
;
1456 new_image_file
= s
->snapshot_file
;
1457 snapshot_ref
= NULL
;
1461 g_assert_not_reached();
1464 /* start processing */
1465 if (action_check_completion_mode(common
, errp
) < 0) {
1469 state
->old_bs
= bdrv_lookup_bs(device
, node_name
, errp
);
1470 if (!state
->old_bs
) {
1474 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1475 aio_context_acquire(aio_context
);
1477 /* Paired with .clean() */
1478 bdrv_drained_begin(state
->old_bs
);
1480 if (!bdrv_is_inserted(state
->old_bs
)) {
1481 error_setg(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1485 if (bdrv_op_is_blocked(state
->old_bs
,
1486 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT
, errp
)) {
1490 if (!bdrv_is_read_only(state
->old_bs
)) {
1491 if (bdrv_flush(state
->old_bs
)) {
1492 error_setg(errp
, QERR_IO_ERROR
);
1497 if (action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
) {
1498 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1499 const char *format
= s
->format
?: "qcow2";
1500 enum NewImageMode mode
;
1501 const char *snapshot_node_name
= s
->snapshot_node_name
;
1503 if (node_name
&& !snapshot_node_name
) {
1504 error_setg(errp
, "New overlay node-name missing");
1508 if (snapshot_node_name
&&
1509 bdrv_lookup_bs(snapshot_node_name
, snapshot_node_name
, NULL
)) {
1510 error_setg(errp
, "New overlay node-name already in use");
1514 flags
= state
->old_bs
->open_flags
;
1515 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_COPY_ON_READ
);
1516 flags
|= BDRV_O_NO_BACKING
;
1518 /* create new image w/backing file */
1519 mode
= s
->has_mode
? s
->mode
: NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1520 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1521 int64_t size
= bdrv_getlength(state
->old_bs
);
1523 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1526 bdrv_refresh_filename(state
->old_bs
);
1528 aio_context_release(aio_context
);
1529 bdrv_img_create(new_image_file
, format
,
1530 state
->old_bs
->filename
,
1531 state
->old_bs
->drv
->format_name
,
1532 NULL
, size
, flags
, false, &local_err
);
1533 aio_context_acquire(aio_context
);
1536 error_propagate(errp
, local_err
);
1541 options
= qdict_new();
1542 if (snapshot_node_name
) {
1543 qdict_put_str(options
, "node-name", snapshot_node_name
);
1545 qdict_put_str(options
, "driver", format
);
1548 state
->new_bs
= bdrv_open(new_image_file
, snapshot_ref
, options
, flags
,
1550 /* We will manually add the backing_hd field to the bs later */
1551 if (!state
->new_bs
) {
1556 * Allow attaching a backing file to an overlay that's already in use only
1557 * if the parents don't assume that they are already seeing a valid image.
1558 * (Specifically, allow it as a mirror target, which is write-only access.)
1560 bdrv_get_cumulative_perm(state
->new_bs
, &perm
, &shared
);
1561 if (perm
& BLK_PERM_CONSISTENT_READ
) {
1562 error_setg(errp
, "The overlay is already in use");
1566 if (state
->new_bs
->drv
->is_filter
) {
1567 error_setg(errp
, "Filters cannot be used as overlays");
1571 if (bdrv_cow_child(state
->new_bs
)) {
1572 error_setg(errp
, "The overlay already has a backing image");
1576 if (!state
->new_bs
->drv
->supports_backing
) {
1577 error_setg(errp
, "The overlay does not support backing images");
1581 ret
= bdrv_append(state
->new_bs
, state
->old_bs
, errp
);
1585 state
->overlay_appended
= true;
1588 aio_context_release(aio_context
);
1591 static void external_snapshot_commit(BlkActionState
*common
)
1593 ExternalSnapshotState
*state
=
1594 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1595 AioContext
*aio_context
;
1597 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1598 aio_context_acquire(aio_context
);
1600 /* We don't need (or want) to use the transactional
1601 * bdrv_reopen_multiple() across all the entries at once, because we
1602 * don't want to abort all of them if one of them fails the reopen */
1603 if (!qatomic_read(&state
->old_bs
->copy_on_read
)) {
1604 bdrv_reopen_set_read_only(state
->old_bs
, true, NULL
);
1607 aio_context_release(aio_context
);
1610 static void external_snapshot_abort(BlkActionState
*common
)
1612 ExternalSnapshotState
*state
=
1613 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1614 if (state
->new_bs
) {
1615 if (state
->overlay_appended
) {
1616 AioContext
*aio_context
;
1617 AioContext
*tmp_context
;
1620 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1621 aio_context_acquire(aio_context
);
1623 bdrv_ref(state
->old_bs
); /* we can't let bdrv_set_backind_hd()
1624 close state->old_bs; we need it */
1625 bdrv_set_backing_hd(state
->new_bs
, NULL
, &error_abort
);
1628 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1629 * the main AioContext. As we're still going to be using it, return
1630 * it to the AioContext it was before.
1632 tmp_context
= bdrv_get_aio_context(state
->old_bs
);
1633 if (aio_context
!= tmp_context
) {
1634 aio_context_release(aio_context
);
1635 aio_context_acquire(tmp_context
);
1637 ret
= bdrv_try_change_aio_context(state
->old_bs
,
1638 aio_context
, NULL
, NULL
);
1641 aio_context_release(tmp_context
);
1642 aio_context_acquire(aio_context
);
1645 bdrv_replace_node(state
->new_bs
, state
->old_bs
, &error_abort
);
1646 bdrv_unref(state
->old_bs
); /* bdrv_replace_node() ref'ed old_bs */
1648 aio_context_release(aio_context
);
1653 static void external_snapshot_clean(BlkActionState
*common
)
1655 ExternalSnapshotState
*state
=
1656 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1657 AioContext
*aio_context
;
1659 if (!state
->old_bs
) {
1663 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1664 aio_context_acquire(aio_context
);
1666 bdrv_drained_end(state
->old_bs
);
1667 bdrv_unref(state
->new_bs
);
1669 aio_context_release(aio_context
);
1672 typedef struct DriveBackupState
{
1673 BlkActionState common
;
1674 BlockDriverState
*bs
;
1678 static BlockJob
*do_backup_common(BackupCommon
*backup
,
1679 BlockDriverState
*bs
,
1680 BlockDriverState
*target_bs
,
1681 AioContext
*aio_context
,
1682 JobTxn
*txn
, Error
**errp
);
1684 static void drive_backup_prepare(BlkActionState
*common
, Error
**errp
)
1686 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1687 DriveBackup
*backup
;
1688 BlockDriverState
*bs
;
1689 BlockDriverState
*target_bs
;
1690 BlockDriverState
*source
= NULL
;
1691 AioContext
*aio_context
;
1692 AioContext
*old_context
;
1695 Error
*local_err
= NULL
;
1698 bool set_backing_hd
= false;
1701 assert(common
->action
->type
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1702 backup
= common
->action
->u
.drive_backup
.data
;
1704 if (!backup
->has_mode
) {
1705 backup
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1708 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1714 error_setg(errp
, "Device has no medium");
1718 aio_context
= bdrv_get_aio_context(bs
);
1719 aio_context_acquire(aio_context
);
1722 /* Paired with .clean() */
1723 bdrv_drained_begin(bs
);
1725 format
= backup
->format
;
1726 if (!format
&& backup
->mode
!= NEW_IMAGE_MODE_EXISTING
) {
1727 format
= bs
->drv
->format_name
;
1730 /* Early check to avoid creating target */
1731 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_BACKUP_SOURCE
, errp
)) {
1735 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1738 * See if we have a backing HD we can use to create our new image
1741 if (backup
->sync
== MIRROR_SYNC_MODE_TOP
) {
1743 * Backup will not replace the source by the target, so none
1744 * of the filters skipped here will be removed (in contrast to
1745 * mirror). Therefore, we can skip all of them when looking
1746 * for the first COW relationship.
1748 source
= bdrv_cow_bs(bdrv_skip_filters(bs
));
1750 backup
->sync
= MIRROR_SYNC_MODE_FULL
;
1753 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
1755 flags
|= BDRV_O_NO_BACKING
;
1756 set_backing_hd
= true;
1759 size
= bdrv_getlength(bs
);
1761 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1765 if (backup
->mode
!= NEW_IMAGE_MODE_EXISTING
) {
1768 /* Implicit filters should not appear in the filename */
1769 BlockDriverState
*explicit_backing
=
1770 bdrv_skip_implicit_filters(source
);
1772 bdrv_refresh_filename(explicit_backing
);
1773 bdrv_img_create(backup
->target
, format
,
1774 explicit_backing
->filename
,
1775 explicit_backing
->drv
->format_name
, NULL
,
1776 size
, flags
, false, &local_err
);
1778 bdrv_img_create(backup
->target
, format
, NULL
, NULL
, NULL
,
1779 size
, flags
, false, &local_err
);
1784 error_propagate(errp
, local_err
);
1788 options
= qdict_new();
1789 qdict_put_str(options
, "discard", "unmap");
1790 qdict_put_str(options
, "detect-zeroes", "unmap");
1792 qdict_put_str(options
, "driver", format
);
1795 target_bs
= bdrv_open(backup
->target
, NULL
, options
, flags
, errp
);
1800 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1801 old_context
= bdrv_get_aio_context(target_bs
);
1802 aio_context_release(aio_context
);
1803 aio_context_acquire(old_context
);
1805 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
1807 bdrv_unref(target_bs
);
1808 aio_context_release(old_context
);
1812 aio_context_release(old_context
);
1813 aio_context_acquire(aio_context
);
1815 if (set_backing_hd
) {
1816 if (bdrv_set_backing_hd(target_bs
, source
, errp
) < 0) {
1821 state
->job
= do_backup_common(qapi_DriveBackup_base(backup
),
1822 bs
, target_bs
, aio_context
,
1823 common
->block_job_txn
, errp
);
1826 bdrv_unref(target_bs
);
1828 aio_context_release(aio_context
);
1831 static void drive_backup_commit(BlkActionState
*common
)
1833 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1834 AioContext
*aio_context
;
1836 aio_context
= bdrv_get_aio_context(state
->bs
);
1837 aio_context_acquire(aio_context
);
1840 job_start(&state
->job
->job
);
1842 aio_context_release(aio_context
);
1845 static void drive_backup_abort(BlkActionState
*common
)
1847 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1850 job_cancel_sync(&state
->job
->job
, true);
1854 static void drive_backup_clean(BlkActionState
*common
)
1856 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1857 AioContext
*aio_context
;
1863 aio_context
= bdrv_get_aio_context(state
->bs
);
1864 aio_context_acquire(aio_context
);
1866 bdrv_drained_end(state
->bs
);
1868 aio_context_release(aio_context
);
1871 typedef struct BlockdevBackupState
{
1872 BlkActionState common
;
1873 BlockDriverState
*bs
;
1875 } BlockdevBackupState
;
1877 static void blockdev_backup_prepare(BlkActionState
*common
, Error
**errp
)
1879 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1880 BlockdevBackup
*backup
;
1881 BlockDriverState
*bs
;
1882 BlockDriverState
*target_bs
;
1883 AioContext
*aio_context
;
1884 AioContext
*old_context
;
1887 assert(common
->action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
);
1888 backup
= common
->action
->u
.blockdev_backup
.data
;
1890 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1895 target_bs
= bdrv_lookup_bs(backup
->target
, backup
->target
, errp
);
1900 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1901 aio_context
= bdrv_get_aio_context(bs
);
1902 old_context
= bdrv_get_aio_context(target_bs
);
1903 aio_context_acquire(old_context
);
1905 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
1907 aio_context_release(old_context
);
1911 aio_context_release(old_context
);
1912 aio_context_acquire(aio_context
);
1915 /* Paired with .clean() */
1916 bdrv_drained_begin(state
->bs
);
1918 state
->job
= do_backup_common(qapi_BlockdevBackup_base(backup
),
1919 bs
, target_bs
, aio_context
,
1920 common
->block_job_txn
, errp
);
1922 aio_context_release(aio_context
);
1925 static void blockdev_backup_commit(BlkActionState
*common
)
1927 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1928 AioContext
*aio_context
;
1930 aio_context
= bdrv_get_aio_context(state
->bs
);
1931 aio_context_acquire(aio_context
);
1934 job_start(&state
->job
->job
);
1936 aio_context_release(aio_context
);
1939 static void blockdev_backup_abort(BlkActionState
*common
)
1941 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1944 job_cancel_sync(&state
->job
->job
, true);
1948 static void blockdev_backup_clean(BlkActionState
*common
)
1950 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1951 AioContext
*aio_context
;
1957 aio_context
= bdrv_get_aio_context(state
->bs
);
1958 aio_context_acquire(aio_context
);
1960 bdrv_drained_end(state
->bs
);
1962 aio_context_release(aio_context
);
1965 typedef struct BlockDirtyBitmapState
{
1966 BlkActionState common
;
1967 BdrvDirtyBitmap
*bitmap
;
1968 BlockDriverState
*bs
;
1972 } BlockDirtyBitmapState
;
1974 static void block_dirty_bitmap_add_prepare(BlkActionState
*common
,
1977 Error
*local_err
= NULL
;
1978 BlockDirtyBitmapAdd
*action
;
1979 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
1982 if (action_check_completion_mode(common
, errp
) < 0) {
1986 action
= common
->action
->u
.block_dirty_bitmap_add
.data
;
1987 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1988 qmp_block_dirty_bitmap_add(action
->node
, action
->name
,
1989 action
->has_granularity
, action
->granularity
,
1990 action
->has_persistent
, action
->persistent
,
1991 action
->has_disabled
, action
->disabled
,
1995 state
->prepared
= true;
1997 error_propagate(errp
, local_err
);
2001 static void block_dirty_bitmap_add_abort(BlkActionState
*common
)
2003 BlockDirtyBitmapAdd
*action
;
2004 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2007 action
= common
->action
->u
.block_dirty_bitmap_add
.data
;
2008 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2009 * then the node reference and bitmap name must have been valid.
2011 if (state
->prepared
) {
2012 qmp_block_dirty_bitmap_remove(action
->node
, action
->name
, &error_abort
);
2016 static void block_dirty_bitmap_clear_prepare(BlkActionState
*common
,
2019 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2021 BlockDirtyBitmap
*action
;
2023 if (action_check_completion_mode(common
, errp
) < 0) {
2027 action
= common
->action
->u
.block_dirty_bitmap_clear
.data
;
2028 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2032 if (!state
->bitmap
) {
2036 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_DEFAULT
, errp
)) {
2040 bdrv_clear_dirty_bitmap(state
->bitmap
, &state
->backup
);
2043 static void block_dirty_bitmap_restore(BlkActionState
*common
)
2045 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2048 if (state
->backup
) {
2049 bdrv_restore_dirty_bitmap(state
->bitmap
, state
->backup
);
2053 static void block_dirty_bitmap_free_backup(BlkActionState
*common
)
2055 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2058 hbitmap_free(state
->backup
);
2061 static void block_dirty_bitmap_enable_prepare(BlkActionState
*common
,
2064 BlockDirtyBitmap
*action
;
2065 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2068 if (action_check_completion_mode(common
, errp
) < 0) {
2072 action
= common
->action
->u
.block_dirty_bitmap_enable
.data
;
2073 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2077 if (!state
->bitmap
) {
2081 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2085 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2086 bdrv_enable_dirty_bitmap(state
->bitmap
);
2089 static void block_dirty_bitmap_enable_abort(BlkActionState
*common
)
2091 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2094 if (!state
->was_enabled
) {
2095 bdrv_disable_dirty_bitmap(state
->bitmap
);
2099 static void block_dirty_bitmap_disable_prepare(BlkActionState
*common
,
2102 BlockDirtyBitmap
*action
;
2103 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2106 if (action_check_completion_mode(common
, errp
) < 0) {
2110 action
= common
->action
->u
.block_dirty_bitmap_disable
.data
;
2111 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2115 if (!state
->bitmap
) {
2119 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2123 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2124 bdrv_disable_dirty_bitmap(state
->bitmap
);
2127 static void block_dirty_bitmap_disable_abort(BlkActionState
*common
)
2129 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2132 if (state
->was_enabled
) {
2133 bdrv_enable_dirty_bitmap(state
->bitmap
);
2137 static void block_dirty_bitmap_merge_prepare(BlkActionState
*common
,
2140 BlockDirtyBitmapMerge
*action
;
2141 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2144 if (action_check_completion_mode(common
, errp
) < 0) {
2148 action
= common
->action
->u
.block_dirty_bitmap_merge
.data
;
2150 state
->bitmap
= block_dirty_bitmap_merge(action
->node
, action
->target
,
2151 action
->bitmaps
, &state
->backup
,
2155 static void block_dirty_bitmap_remove_prepare(BlkActionState
*common
,
2158 BlockDirtyBitmap
*action
;
2159 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2162 if (action_check_completion_mode(common
, errp
) < 0) {
2166 action
= common
->action
->u
.block_dirty_bitmap_remove
.data
;
2168 state
->bitmap
= block_dirty_bitmap_remove(action
->node
, action
->name
,
2169 false, &state
->bs
, errp
);
2170 if (state
->bitmap
) {
2171 bdrv_dirty_bitmap_skip_store(state
->bitmap
, true);
2172 bdrv_dirty_bitmap_set_busy(state
->bitmap
, true);
2176 static void block_dirty_bitmap_remove_abort(BlkActionState
*common
)
2178 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2181 if (state
->bitmap
) {
2182 bdrv_dirty_bitmap_skip_store(state
->bitmap
, false);
2183 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2187 static void block_dirty_bitmap_remove_commit(BlkActionState
*common
)
2189 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2192 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2193 bdrv_release_dirty_bitmap(state
->bitmap
);
2196 static void abort_prepare(BlkActionState
*common
, Error
**errp
)
2198 error_setg(errp
, "Transaction aborted using Abort action");
2201 static void abort_commit(BlkActionState
*common
)
2203 g_assert_not_reached(); /* this action never succeeds */
2206 static const BlkActionOps actions
[] = {
2207 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
] = {
2208 .instance_size
= sizeof(ExternalSnapshotState
),
2209 .prepare
= external_snapshot_prepare
,
2210 .commit
= external_snapshot_commit
,
2211 .abort
= external_snapshot_abort
,
2212 .clean
= external_snapshot_clean
,
2214 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
2215 .instance_size
= sizeof(ExternalSnapshotState
),
2216 .prepare
= external_snapshot_prepare
,
2217 .commit
= external_snapshot_commit
,
2218 .abort
= external_snapshot_abort
,
2219 .clean
= external_snapshot_clean
,
2221 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
2222 .instance_size
= sizeof(DriveBackupState
),
2223 .prepare
= drive_backup_prepare
,
2224 .commit
= drive_backup_commit
,
2225 .abort
= drive_backup_abort
,
2226 .clean
= drive_backup_clean
,
2228 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
] = {
2229 .instance_size
= sizeof(BlockdevBackupState
),
2230 .prepare
= blockdev_backup_prepare
,
2231 .commit
= blockdev_backup_commit
,
2232 .abort
= blockdev_backup_abort
,
2233 .clean
= blockdev_backup_clean
,
2235 [TRANSACTION_ACTION_KIND_ABORT
] = {
2236 .instance_size
= sizeof(BlkActionState
),
2237 .prepare
= abort_prepare
,
2238 .commit
= abort_commit
,
2240 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
2241 .instance_size
= sizeof(InternalSnapshotState
),
2242 .prepare
= internal_snapshot_prepare
,
2243 .abort
= internal_snapshot_abort
,
2244 .clean
= internal_snapshot_clean
,
2246 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD
] = {
2247 .instance_size
= sizeof(BlockDirtyBitmapState
),
2248 .prepare
= block_dirty_bitmap_add_prepare
,
2249 .abort
= block_dirty_bitmap_add_abort
,
2251 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR
] = {
2252 .instance_size
= sizeof(BlockDirtyBitmapState
),
2253 .prepare
= block_dirty_bitmap_clear_prepare
,
2254 .commit
= block_dirty_bitmap_free_backup
,
2255 .abort
= block_dirty_bitmap_restore
,
2257 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE
] = {
2258 .instance_size
= sizeof(BlockDirtyBitmapState
),
2259 .prepare
= block_dirty_bitmap_enable_prepare
,
2260 .abort
= block_dirty_bitmap_enable_abort
,
2262 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE
] = {
2263 .instance_size
= sizeof(BlockDirtyBitmapState
),
2264 .prepare
= block_dirty_bitmap_disable_prepare
,
2265 .abort
= block_dirty_bitmap_disable_abort
,
2267 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE
] = {
2268 .instance_size
= sizeof(BlockDirtyBitmapState
),
2269 .prepare
= block_dirty_bitmap_merge_prepare
,
2270 .commit
= block_dirty_bitmap_free_backup
,
2271 .abort
= block_dirty_bitmap_restore
,
2273 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE
] = {
2274 .instance_size
= sizeof(BlockDirtyBitmapState
),
2275 .prepare
= block_dirty_bitmap_remove_prepare
,
2276 .commit
= block_dirty_bitmap_remove_commit
,
2277 .abort
= block_dirty_bitmap_remove_abort
,
2279 /* Where are transactions for MIRROR, COMMIT and STREAM?
2280 * Although these blockjobs use transaction callbacks like the backup job,
2281 * these jobs do not necessarily adhere to transaction semantics.
2282 * These jobs may not fully undo all of their actions on abort, nor do they
2283 * necessarily work in transactions with more than one job in them.
2288 * Allocate a TransactionProperties structure if necessary, and fill
2289 * that structure with desired defaults if they are unset.
2291 static TransactionProperties
*get_transaction_properties(
2292 TransactionProperties
*props
)
2295 props
= g_new0(TransactionProperties
, 1);
2298 if (!props
->has_completion_mode
) {
2299 props
->has_completion_mode
= true;
2300 props
->completion_mode
= ACTION_COMPLETION_MODE_INDIVIDUAL
;
2307 * 'Atomic' group operations. The operations are performed as a set, and if
2308 * any fail then we roll back all operations in the group.
2310 * Always run under BQL.
2312 void qmp_transaction(TransactionActionList
*dev_list
,
2313 struct TransactionProperties
*props
,
2316 TransactionActionList
*dev_entry
= dev_list
;
2317 bool has_props
= !!props
;
2318 JobTxn
*block_job_txn
= NULL
;
2319 BlkActionState
*state
, *next
;
2320 Error
*local_err
= NULL
;
2322 GLOBAL_STATE_CODE();
2324 QTAILQ_HEAD(, BlkActionState
) snap_bdrv_states
;
2325 QTAILQ_INIT(&snap_bdrv_states
);
2327 /* Does this transaction get canceled as a group on failure?
2328 * If not, we don't really need to make a JobTxn.
2330 props
= get_transaction_properties(props
);
2331 if (props
->completion_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
2332 block_job_txn
= job_txn_new();
2335 /* drain all i/o before any operations */
2338 /* We don't do anything in this loop that commits us to the operations */
2339 while (NULL
!= dev_entry
) {
2340 TransactionAction
*dev_info
= NULL
;
2341 const BlkActionOps
*ops
;
2343 dev_info
= dev_entry
->value
;
2344 dev_entry
= dev_entry
->next
;
2346 assert(dev_info
->type
< ARRAY_SIZE(actions
));
2348 ops
= &actions
[dev_info
->type
];
2349 assert(ops
->instance_size
> 0);
2351 state
= g_malloc0(ops
->instance_size
);
2353 state
->action
= dev_info
;
2354 state
->block_job_txn
= block_job_txn
;
2355 state
->txn_props
= props
;
2356 QTAILQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
2358 state
->ops
->prepare(state
, &local_err
);
2360 error_propagate(errp
, local_err
);
2361 goto delete_and_fail
;
2365 QTAILQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
2366 if (state
->ops
->commit
) {
2367 state
->ops
->commit(state
);
2375 /* failure, and it is all-or-none; roll back all operations */
2376 QTAILQ_FOREACH_REVERSE(state
, &snap_bdrv_states
, entry
) {
2377 if (state
->ops
->abort
) {
2378 state
->ops
->abort(state
);
2382 QTAILQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
2383 if (state
->ops
->clean
) {
2384 state
->ops
->clean(state
);
2389 qapi_free_TransactionProperties(props
);
2391 job_txn_unref(block_job_txn
);
2394 BlockDirtyBitmapSha256
*qmp_x_debug_block_dirty_bitmap_sha256(const char *node
,
2398 BdrvDirtyBitmap
*bitmap
;
2399 BlockDriverState
*bs
;
2400 BlockDirtyBitmapSha256
*ret
= NULL
;
2403 bitmap
= block_dirty_bitmap_lookup(node
, name
, &bs
, errp
);
2404 if (!bitmap
|| !bs
) {
2408 sha256
= bdrv_dirty_bitmap_sha256(bitmap
, errp
);
2409 if (sha256
== NULL
) {
2413 ret
= g_new(BlockDirtyBitmapSha256
, 1);
2414 ret
->sha256
= sha256
;
2419 void coroutine_fn
qmp_block_resize(const char *device
, const char *node_name
,
2420 int64_t size
, Error
**errp
)
2422 Error
*local_err
= NULL
;
2424 BlockDriverState
*bs
;
2425 AioContext
*old_ctx
;
2427 bs
= bdrv_lookup_bs(device
, node_name
, &local_err
);
2429 error_propagate(errp
, local_err
);
2434 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
2438 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_RESIZE
, NULL
)) {
2439 error_setg(errp
, QERR_DEVICE_IN_USE
, device
);
2443 blk
= blk_co_new_with_bs(bs
, BLK_PERM_RESIZE
, BLK_PERM_ALL
, errp
);
2449 bdrv_drained_begin(bs
);
2452 old_ctx
= bdrv_co_enter(bs
);
2453 blk_co_truncate(blk
, size
, false, PREALLOC_MODE_OFF
, 0, errp
);
2454 bdrv_co_leave(bs
, old_ctx
);
2457 bdrv_drained_end(bs
);
2462 void qmp_block_stream(const char *job_id
, const char *device
,
2464 const char *base_node
,
2465 const char *backing_file
,
2467 bool has_speed
, int64_t speed
,
2468 bool has_on_error
, BlockdevOnError on_error
,
2469 const char *filter_node_name
,
2470 bool has_auto_finalize
, bool auto_finalize
,
2471 bool has_auto_dismiss
, bool auto_dismiss
,
2474 BlockDriverState
*bs
, *iter
, *iter_end
;
2475 BlockDriverState
*base_bs
= NULL
;
2476 BlockDriverState
*bottom_bs
= NULL
;
2477 AioContext
*aio_context
;
2478 Error
*local_err
= NULL
;
2479 int job_flags
= JOB_DEFAULT
;
2481 if (base
&& base_node
) {
2482 error_setg(errp
, "'base' and 'base-node' cannot be specified "
2483 "at the same time");
2487 if (base
&& bottom
) {
2488 error_setg(errp
, "'base' and 'bottom' cannot be specified "
2489 "at the same time");
2493 if (bottom
&& base_node
) {
2494 error_setg(errp
, "'bottom' and 'base-node' cannot be specified "
2495 "at the same time");
2499 if (!has_on_error
) {
2500 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2503 bs
= bdrv_lookup_bs(device
, device
, errp
);
2508 aio_context
= bdrv_get_aio_context(bs
);
2509 aio_context_acquire(aio_context
);
2512 base_bs
= bdrv_find_backing_image(bs
, base
);
2513 if (base_bs
== NULL
) {
2514 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2517 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2521 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2525 if (bs
== base_bs
|| !bdrv_chain_contains(bs
, base_bs
)) {
2526 error_setg(errp
, "Node '%s' is not a backing image of '%s'",
2530 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2531 bdrv_refresh_filename(base_bs
);
2535 bottom_bs
= bdrv_lookup_bs(NULL
, bottom
, errp
);
2539 if (!bottom_bs
->drv
) {
2540 error_setg(errp
, "Node '%s' is not open", bottom
);
2543 if (bottom_bs
->drv
->is_filter
) {
2544 error_setg(errp
, "Node '%s' is a filter, use a non-filter node "
2545 "as 'bottom'", bottom
);
2548 if (!bdrv_chain_contains(bs
, bottom_bs
)) {
2549 error_setg(errp
, "Node '%s' is not in a chain starting from '%s'",
2553 assert(bdrv_get_aio_context(bottom_bs
) == aio_context
);
2557 * Check for op blockers in the whole chain between bs and base (or bottom)
2559 iter_end
= bottom
? bdrv_filter_or_cow_bs(bottom_bs
) : base_bs
;
2560 for (iter
= bs
; iter
&& iter
!= iter_end
;
2561 iter
= bdrv_filter_or_cow_bs(iter
))
2563 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_STREAM
, errp
)) {
2568 /* if we are streaming the entire chain, the result will have no backing
2569 * file, and specifying one is therefore an error */
2570 if (!base_bs
&& backing_file
) {
2571 error_setg(errp
, "backing file specified, but streaming the "
2576 if (has_auto_finalize
&& !auto_finalize
) {
2577 job_flags
|= JOB_MANUAL_FINALIZE
;
2579 if (has_auto_dismiss
&& !auto_dismiss
) {
2580 job_flags
|= JOB_MANUAL_DISMISS
;
2583 stream_start(job_id
, bs
, base_bs
, backing_file
,
2584 bottom_bs
, job_flags
, has_speed
? speed
: 0, on_error
,
2585 filter_node_name
, &local_err
);
2587 error_propagate(errp
, local_err
);
2591 trace_qmp_block_stream(bs
);
2594 aio_context_release(aio_context
);
2597 void qmp_block_commit(const char *job_id
, const char *device
,
2598 const char *base_node
,
2600 const char *top_node
,
2602 const char *backing_file
,
2603 bool has_speed
, int64_t speed
,
2604 bool has_on_error
, BlockdevOnError on_error
,
2605 const char *filter_node_name
,
2606 bool has_auto_finalize
, bool auto_finalize
,
2607 bool has_auto_dismiss
, bool auto_dismiss
,
2610 BlockDriverState
*bs
;
2611 BlockDriverState
*iter
;
2612 BlockDriverState
*base_bs
, *top_bs
;
2613 AioContext
*aio_context
;
2614 Error
*local_err
= NULL
;
2615 int job_flags
= JOB_DEFAULT
;
2616 uint64_t top_perm
, top_shared
;
2621 if (!has_on_error
) {
2622 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2624 if (has_auto_finalize
&& !auto_finalize
) {
2625 job_flags
|= JOB_MANUAL_FINALIZE
;
2627 if (has_auto_dismiss
&& !auto_dismiss
) {
2628 job_flags
|= JOB_MANUAL_DISMISS
;
2632 * libvirt relies on the DeviceNotFound error class in order to probe for
2633 * live commit feature versions; for this to work, we must make sure to
2634 * perform the device lookup before any generic errors that may occur in a
2635 * scenario in which all optional arguments are omitted. */
2636 bs
= qmp_get_root_bs(device
, &local_err
);
2638 bs
= bdrv_lookup_bs(device
, device
, NULL
);
2640 error_free(local_err
);
2641 error_set(errp
, ERROR_CLASS_DEVICE_NOT_FOUND
,
2642 "Device '%s' not found", device
);
2644 error_propagate(errp
, local_err
);
2649 aio_context
= bdrv_get_aio_context(bs
);
2650 aio_context_acquire(aio_context
);
2652 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_COMMIT_SOURCE
, errp
)) {
2656 /* default top_bs is the active layer */
2659 if (top_node
&& top
) {
2660 error_setg(errp
, "'top-node' and 'top' are mutually exclusive");
2662 } else if (top_node
) {
2663 top_bs
= bdrv_lookup_bs(NULL
, top_node
, errp
);
2664 if (top_bs
== NULL
) {
2667 if (!bdrv_chain_contains(bs
, top_bs
)) {
2668 error_setg(errp
, "'%s' is not in this backing file chain",
2673 /* This strcmp() is just a shortcut, there is no need to
2674 * refresh @bs's filename. If it mismatches,
2675 * bdrv_find_backing_image() will do the refresh and may still
2677 if (strcmp(bs
->filename
, top
) != 0) {
2678 top_bs
= bdrv_find_backing_image(bs
, top
);
2682 if (top_bs
== NULL
) {
2683 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
2687 assert(bdrv_get_aio_context(top_bs
) == aio_context
);
2689 if (base_node
&& base
) {
2690 error_setg(errp
, "'base-node' and 'base' are mutually exclusive");
2692 } else if (base_node
) {
2693 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2694 if (base_bs
== NULL
) {
2697 if (!bdrv_chain_contains(top_bs
, base_bs
)) {
2698 error_setg(errp
, "'%s' is not in this backing file chain",
2703 base_bs
= bdrv_find_backing_image(top_bs
, base
);
2704 if (base_bs
== NULL
) {
2705 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2709 base_bs
= bdrv_find_base(top_bs
);
2710 if (base_bs
== NULL
) {
2711 error_setg(errp
, "There is no backimg image");
2716 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2718 for (iter
= top_bs
; iter
!= bdrv_filter_or_cow_bs(base_bs
);
2719 iter
= bdrv_filter_or_cow_bs(iter
))
2721 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2726 /* Do not allow attempts to commit an image into itself */
2727 if (top_bs
== base_bs
) {
2728 error_setg(errp
, "cannot commit an image into itself");
2733 * Active commit is required if and only if someone has taken a
2734 * WRITE permission on the top node. Historically, we have always
2735 * used active commit for top nodes, so continue that practice
2736 * lest we possibly break clients that rely on this behavior, e.g.
2737 * to later attach this node to a writing parent.
2738 * (Active commit is never really wrong.)
2740 bdrv_get_cumulative_perm(top_bs
, &top_perm
, &top_shared
);
2741 if (top_perm
& BLK_PERM_WRITE
||
2742 bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
))
2745 if (bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
)) {
2746 error_setg(errp
, "'backing-file' specified,"
2747 " but 'top' is the active layer");
2749 error_setg(errp
, "'backing-file' specified, but 'top' has a "
2756 * Emulate here what block_job_create() does, because it
2757 * is possible that @bs != @top_bs (the block job should
2758 * be named after @bs, even if @top_bs is the actual
2761 job_id
= bdrv_get_device_name(bs
);
2763 commit_active_start(job_id
, top_bs
, base_bs
, job_flags
, speed
, on_error
,
2764 filter_node_name
, NULL
, NULL
, false, &local_err
);
2766 BlockDriverState
*overlay_bs
= bdrv_find_overlay(bs
, top_bs
);
2767 if (bdrv_op_is_blocked(overlay_bs
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2770 commit_start(job_id
, bs
, base_bs
, top_bs
, job_flags
,
2771 speed
, on_error
, backing_file
,
2772 filter_node_name
, &local_err
);
2774 if (local_err
!= NULL
) {
2775 error_propagate(errp
, local_err
);
2780 aio_context_release(aio_context
);
2783 /* Common QMP interface for drive-backup and blockdev-backup */
2784 static BlockJob
*do_backup_common(BackupCommon
*backup
,
2785 BlockDriverState
*bs
,
2786 BlockDriverState
*target_bs
,
2787 AioContext
*aio_context
,
2788 JobTxn
*txn
, Error
**errp
)
2790 BlockJob
*job
= NULL
;
2791 BdrvDirtyBitmap
*bmap
= NULL
;
2792 BackupPerf perf
= { .max_workers
= 64 };
2793 int job_flags
= JOB_DEFAULT
;
2795 if (!backup
->has_speed
) {
2798 if (!backup
->has_on_source_error
) {
2799 backup
->on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2801 if (!backup
->has_on_target_error
) {
2802 backup
->on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2804 if (!backup
->has_auto_finalize
) {
2805 backup
->auto_finalize
= true;
2807 if (!backup
->has_auto_dismiss
) {
2808 backup
->auto_dismiss
= true;
2810 if (!backup
->has_compress
) {
2811 backup
->compress
= false;
2814 if (backup
->x_perf
) {
2815 if (backup
->x_perf
->has_use_copy_range
) {
2816 perf
.use_copy_range
= backup
->x_perf
->use_copy_range
;
2818 if (backup
->x_perf
->has_max_workers
) {
2819 perf
.max_workers
= backup
->x_perf
->max_workers
;
2821 if (backup
->x_perf
->has_max_chunk
) {
2822 perf
.max_chunk
= backup
->x_perf
->max_chunk
;
2826 if ((backup
->sync
== MIRROR_SYNC_MODE_BITMAP
) ||
2827 (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
)) {
2828 /* done before desugaring 'incremental' to print the right message */
2829 if (!backup
->bitmap
) {
2830 error_setg(errp
, "must provide a valid bitmap name for "
2831 "'%s' sync mode", MirrorSyncMode_str(backup
->sync
));
2836 if (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
) {
2837 if (backup
->has_bitmap_mode
&&
2838 backup
->bitmap_mode
!= BITMAP_SYNC_MODE_ON_SUCCESS
) {
2839 error_setg(errp
, "Bitmap sync mode must be '%s' "
2840 "when using sync mode '%s'",
2841 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS
),
2842 MirrorSyncMode_str(backup
->sync
));
2845 backup
->has_bitmap_mode
= true;
2846 backup
->sync
= MIRROR_SYNC_MODE_BITMAP
;
2847 backup
->bitmap_mode
= BITMAP_SYNC_MODE_ON_SUCCESS
;
2850 if (backup
->bitmap
) {
2851 bmap
= bdrv_find_dirty_bitmap(bs
, backup
->bitmap
);
2853 error_setg(errp
, "Bitmap '%s' could not be found", backup
->bitmap
);
2856 if (!backup
->has_bitmap_mode
) {
2857 error_setg(errp
, "Bitmap sync mode must be given "
2858 "when providing a bitmap");
2861 if (bdrv_dirty_bitmap_check(bmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2865 /* This does not produce a useful bitmap artifact: */
2866 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
2867 error_setg(errp
, "sync mode '%s' does not produce meaningful bitmap"
2868 " outputs", MirrorSyncMode_str(backup
->sync
));
2872 /* If the bitmap isn't used for input or output, this is useless: */
2873 if (backup
->bitmap_mode
== BITMAP_SYNC_MODE_NEVER
&&
2874 backup
->sync
!= MIRROR_SYNC_MODE_BITMAP
) {
2875 error_setg(errp
, "Bitmap sync mode '%s' has no meaningful effect"
2876 " when combined with sync mode '%s'",
2877 BitmapSyncMode_str(backup
->bitmap_mode
),
2878 MirrorSyncMode_str(backup
->sync
));
2883 if (!backup
->bitmap
&& backup
->has_bitmap_mode
) {
2884 error_setg(errp
, "Cannot specify bitmap sync mode without a bitmap");
2888 if (!backup
->auto_finalize
) {
2889 job_flags
|= JOB_MANUAL_FINALIZE
;
2891 if (!backup
->auto_dismiss
) {
2892 job_flags
|= JOB_MANUAL_DISMISS
;
2895 job
= backup_job_create(backup
->job_id
, bs
, target_bs
, backup
->speed
,
2896 backup
->sync
, bmap
, backup
->bitmap_mode
,
2898 backup
->filter_node_name
,
2900 backup
->on_source_error
,
2901 backup
->on_target_error
,
2902 job_flags
, NULL
, NULL
, txn
, errp
);
2906 void qmp_drive_backup(DriveBackup
*backup
, Error
**errp
)
2908 TransactionAction action
= {
2909 .type
= TRANSACTION_ACTION_KIND_DRIVE_BACKUP
,
2910 .u
.drive_backup
.data
= backup
,
2912 blockdev_do_action(&action
, errp
);
2915 BlockDeviceInfoList
*qmp_query_named_block_nodes(bool has_flat
,
2919 bool return_flat
= has_flat
&& flat
;
2921 return bdrv_named_nodes_list(return_flat
, errp
);
2924 XDbgBlockGraph
*qmp_x_debug_query_block_graph(Error
**errp
)
2926 return bdrv_get_xdbg_block_graph(errp
);
2929 void qmp_blockdev_backup(BlockdevBackup
*backup
, Error
**errp
)
2931 TransactionAction action
= {
2932 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
,
2933 .u
.blockdev_backup
.data
= backup
,
2935 blockdev_do_action(&action
, errp
);
2938 /* Parameter check and block job starting for drive mirroring.
2939 * Caller should hold @device and @target's aio context (must be the same).
2941 static void blockdev_mirror_common(const char *job_id
, BlockDriverState
*bs
,
2942 BlockDriverState
*target
,
2943 const char *replaces
,
2944 enum MirrorSyncMode sync
,
2945 BlockMirrorBackingMode backing_mode
,
2947 bool has_speed
, int64_t speed
,
2948 bool has_granularity
, uint32_t granularity
,
2949 bool has_buf_size
, int64_t buf_size
,
2950 bool has_on_source_error
,
2951 BlockdevOnError on_source_error
,
2952 bool has_on_target_error
,
2953 BlockdevOnError on_target_error
,
2954 bool has_unmap
, bool unmap
,
2955 const char *filter_node_name
,
2956 bool has_copy_mode
, MirrorCopyMode copy_mode
,
2957 bool has_auto_finalize
, bool auto_finalize
,
2958 bool has_auto_dismiss
, bool auto_dismiss
,
2961 BlockDriverState
*unfiltered_bs
;
2962 int job_flags
= JOB_DEFAULT
;
2964 GLOBAL_STATE_CODE();
2965 GRAPH_RDLOCK_GUARD_MAINLOOP();
2970 if (!has_on_source_error
) {
2971 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2973 if (!has_on_target_error
) {
2974 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2976 if (!has_granularity
) {
2979 if (!has_buf_size
) {
2985 if (!has_copy_mode
) {
2986 copy_mode
= MIRROR_COPY_MODE_BACKGROUND
;
2988 if (has_auto_finalize
&& !auto_finalize
) {
2989 job_flags
|= JOB_MANUAL_FINALIZE
;
2991 if (has_auto_dismiss
&& !auto_dismiss
) {
2992 job_flags
|= JOB_MANUAL_DISMISS
;
2995 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
2996 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2997 "a value in range [512B, 64MB]");
3000 if (granularity
& (granularity
- 1)) {
3001 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
3006 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3009 if (bdrv_op_is_blocked(target
, BLOCK_OP_TYPE_MIRROR_TARGET
, errp
)) {
3013 if (!bdrv_backing_chain_next(bs
) && sync
== MIRROR_SYNC_MODE_TOP
) {
3014 sync
= MIRROR_SYNC_MODE_FULL
;
3018 /* We want to mirror from @bs, but keep implicit filters on top */
3019 unfiltered_bs
= bdrv_skip_implicit_filters(bs
);
3020 if (unfiltered_bs
!= bs
) {
3021 replaces
= unfiltered_bs
->node_name
;
3026 BlockDriverState
*to_replace_bs
;
3027 AioContext
*replace_aio_context
;
3028 int64_t bs_size
, replace_size
;
3030 bs_size
= bdrv_getlength(bs
);
3032 error_setg_errno(errp
, -bs_size
, "Failed to query device's size");
3036 to_replace_bs
= check_to_replace_node(bs
, replaces
, errp
);
3037 if (!to_replace_bs
) {
3041 replace_aio_context
= bdrv_get_aio_context(to_replace_bs
);
3042 aio_context_acquire(replace_aio_context
);
3043 replace_size
= bdrv_getlength(to_replace_bs
);
3044 aio_context_release(replace_aio_context
);
3046 if (replace_size
< 0) {
3047 error_setg_errno(errp
, -replace_size
,
3048 "Failed to query the replacement node's size");
3051 if (bs_size
!= replace_size
) {
3052 error_setg(errp
, "cannot replace image with a mirror image of "
3058 /* pass the node name to replace to mirror start since it's loose coupling
3059 * and will allow to check whether the node still exist at mirror completion
3061 mirror_start(job_id
, bs
, target
,
3062 replaces
, job_flags
,
3063 speed
, granularity
, buf_size
, sync
, backing_mode
, zero_target
,
3064 on_source_error
, on_target_error
, unmap
, filter_node_name
,
3068 void qmp_drive_mirror(DriveMirror
*arg
, Error
**errp
)
3070 BlockDriverState
*bs
;
3071 BlockDriverState
*target_backing_bs
, *target_bs
;
3072 AioContext
*aio_context
;
3073 AioContext
*old_context
;
3074 BlockMirrorBackingMode backing_mode
;
3075 Error
*local_err
= NULL
;
3076 QDict
*options
= NULL
;
3079 const char *format
= arg
->format
;
3083 bs
= qmp_get_root_bs(arg
->device
, errp
);
3088 /* Early check to avoid creating target */
3089 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3093 aio_context
= bdrv_get_aio_context(bs
);
3094 aio_context_acquire(aio_context
);
3096 if (!arg
->has_mode
) {
3097 arg
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
3101 format
= (arg
->mode
== NEW_IMAGE_MODE_EXISTING
3102 ? NULL
: bs
->drv
->format_name
);
3105 flags
= bs
->open_flags
| BDRV_O_RDWR
;
3106 target_backing_bs
= bdrv_cow_bs(bdrv_skip_filters(bs
));
3107 if (!target_backing_bs
&& arg
->sync
== MIRROR_SYNC_MODE_TOP
) {
3108 arg
->sync
= MIRROR_SYNC_MODE_FULL
;
3110 if (arg
->sync
== MIRROR_SYNC_MODE_NONE
) {
3111 target_backing_bs
= bs
;
3114 size
= bdrv_getlength(bs
);
3116 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
3120 if (arg
->replaces
) {
3121 if (!arg
->node_name
) {
3122 error_setg(errp
, "a node-name must be provided when replacing a"
3123 " named node of the graph");
3128 if (arg
->mode
== NEW_IMAGE_MODE_ABSOLUTE_PATHS
) {
3129 backing_mode
= MIRROR_SOURCE_BACKING_CHAIN
;
3131 backing_mode
= MIRROR_OPEN_BACKING_CHAIN
;
3134 /* Don't open backing image in create() */
3135 flags
|= BDRV_O_NO_BACKING
;
3137 if ((arg
->sync
== MIRROR_SYNC_MODE_FULL
|| !target_backing_bs
)
3138 && arg
->mode
!= NEW_IMAGE_MODE_EXISTING
)
3140 /* create new image w/o backing file */
3142 bdrv_img_create(arg
->target
, format
,
3143 NULL
, NULL
, NULL
, size
, flags
, false, &local_err
);
3145 /* Implicit filters should not appear in the filename */
3146 BlockDriverState
*explicit_backing
=
3147 bdrv_skip_implicit_filters(target_backing_bs
);
3149 switch (arg
->mode
) {
3150 case NEW_IMAGE_MODE_EXISTING
:
3152 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
3153 /* create new image with backing file */
3154 bdrv_refresh_filename(explicit_backing
);
3155 bdrv_img_create(arg
->target
, format
,
3156 explicit_backing
->filename
,
3157 explicit_backing
->drv
->format_name
,
3158 NULL
, size
, flags
, false, &local_err
);
3166 error_propagate(errp
, local_err
);
3170 options
= qdict_new();
3171 if (arg
->node_name
) {
3172 qdict_put_str(options
, "node-name", arg
->node_name
);
3175 qdict_put_str(options
, "driver", format
);
3178 /* Mirroring takes care of copy-on-write using the source's backing
3181 target_bs
= bdrv_open(arg
->target
, NULL
, options
, flags
, errp
);
3186 zero_target
= (arg
->sync
== MIRROR_SYNC_MODE_FULL
&&
3187 (arg
->mode
== NEW_IMAGE_MODE_EXISTING
||
3188 !bdrv_has_zero_init(target_bs
)));
3191 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3192 old_context
= bdrv_get_aio_context(target_bs
);
3193 aio_context_release(aio_context
);
3194 aio_context_acquire(old_context
);
3196 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
3198 bdrv_unref(target_bs
);
3199 aio_context_release(old_context
);
3203 aio_context_release(old_context
);
3204 aio_context_acquire(aio_context
);
3206 blockdev_mirror_common(arg
->job_id
, bs
, target_bs
,
3207 arg
->replaces
, arg
->sync
,
3208 backing_mode
, zero_target
,
3209 arg
->has_speed
, arg
->speed
,
3210 arg
->has_granularity
, arg
->granularity
,
3211 arg
->has_buf_size
, arg
->buf_size
,
3212 arg
->has_on_source_error
, arg
->on_source_error
,
3213 arg
->has_on_target_error
, arg
->on_target_error
,
3214 arg
->has_unmap
, arg
->unmap
,
3216 arg
->has_copy_mode
, arg
->copy_mode
,
3217 arg
->has_auto_finalize
, arg
->auto_finalize
,
3218 arg
->has_auto_dismiss
, arg
->auto_dismiss
,
3220 bdrv_unref(target_bs
);
3222 aio_context_release(aio_context
);
3225 void qmp_blockdev_mirror(const char *job_id
,
3226 const char *device
, const char *target
,
3227 const char *replaces
,
3228 MirrorSyncMode sync
,
3229 bool has_speed
, int64_t speed
,
3230 bool has_granularity
, uint32_t granularity
,
3231 bool has_buf_size
, int64_t buf_size
,
3232 bool has_on_source_error
,
3233 BlockdevOnError on_source_error
,
3234 bool has_on_target_error
,
3235 BlockdevOnError on_target_error
,
3236 const char *filter_node_name
,
3237 bool has_copy_mode
, MirrorCopyMode copy_mode
,
3238 bool has_auto_finalize
, bool auto_finalize
,
3239 bool has_auto_dismiss
, bool auto_dismiss
,
3242 BlockDriverState
*bs
;
3243 BlockDriverState
*target_bs
;
3244 AioContext
*aio_context
;
3245 AioContext
*old_context
;
3246 BlockMirrorBackingMode backing_mode
= MIRROR_LEAVE_BACKING_CHAIN
;
3250 bs
= qmp_get_root_bs(device
, errp
);
3255 target_bs
= bdrv_lookup_bs(target
, target
, errp
);
3260 zero_target
= (sync
== MIRROR_SYNC_MODE_FULL
);
3262 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3263 old_context
= bdrv_get_aio_context(target_bs
);
3264 aio_context
= bdrv_get_aio_context(bs
);
3265 aio_context_acquire(old_context
);
3267 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
3269 aio_context_release(old_context
);
3270 aio_context_acquire(aio_context
);
3276 blockdev_mirror_common(job_id
, bs
, target_bs
,
3277 replaces
, sync
, backing_mode
,
3278 zero_target
, has_speed
, speed
,
3279 has_granularity
, granularity
,
3280 has_buf_size
, buf_size
,
3281 has_on_source_error
, on_source_error
,
3282 has_on_target_error
, on_target_error
,
3283 true, true, filter_node_name
,
3284 has_copy_mode
, copy_mode
,
3285 has_auto_finalize
, auto_finalize
,
3286 has_auto_dismiss
, auto_dismiss
,
3289 aio_context_release(aio_context
);
3293 * Get a block job using its ID. Called with job_mutex held.
3295 static BlockJob
*find_block_job_locked(const char *id
, Error
**errp
)
3301 job
= block_job_get_locked(id
);
3304 error_set(errp
, ERROR_CLASS_DEVICE_NOT_ACTIVE
,
3305 "Block job '%s' not found", id
);
3312 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
3317 job
= find_block_job_locked(device
, errp
);
3323 block_job_set_speed_locked(job
, speed
, errp
);
3326 void qmp_block_job_cancel(const char *device
,
3327 bool has_force
, bool force
, Error
**errp
)
3332 job
= find_block_job_locked(device
, errp
);
3342 if (job_user_paused_locked(&job
->job
) && !force
) {
3343 error_setg(errp
, "The block job for device '%s' is currently paused",
3348 trace_qmp_block_job_cancel(job
);
3349 job_user_cancel_locked(&job
->job
, force
, errp
);
3352 void qmp_block_job_pause(const char *device
, Error
**errp
)
3357 job
= find_block_job_locked(device
, errp
);
3363 trace_qmp_block_job_pause(job
);
3364 job_user_pause_locked(&job
->job
, errp
);
3367 void qmp_block_job_resume(const char *device
, Error
**errp
)
3372 job
= find_block_job_locked(device
, errp
);
3378 trace_qmp_block_job_resume(job
);
3379 job_user_resume_locked(&job
->job
, errp
);
3382 void qmp_block_job_complete(const char *device
, Error
**errp
)
3387 job
= find_block_job_locked(device
, errp
);
3393 trace_qmp_block_job_complete(job
);
3394 job_complete_locked(&job
->job
, errp
);
3397 void qmp_block_job_finalize(const char *id
, Error
**errp
)
3402 job
= find_block_job_locked(id
, errp
);
3408 trace_qmp_block_job_finalize(job
);
3409 job_ref_locked(&job
->job
);
3410 job_finalize_locked(&job
->job
, errp
);
3412 job_unref_locked(&job
->job
);
3415 void qmp_block_job_dismiss(const char *id
, Error
**errp
)
3421 bjob
= find_block_job_locked(id
, errp
);
3427 trace_qmp_block_job_dismiss(bjob
);
3429 job_dismiss_locked(&job
, errp
);
3432 void qmp_change_backing_file(const char *device
,
3433 const char *image_node_name
,
3434 const char *backing_file
,
3437 BlockDriverState
*bs
= NULL
;
3438 AioContext
*aio_context
;
3439 BlockDriverState
*image_bs
= NULL
;
3440 Error
*local_err
= NULL
;
3444 bs
= qmp_get_root_bs(device
, errp
);
3449 aio_context
= bdrv_get_aio_context(bs
);
3450 aio_context_acquire(aio_context
);
3452 image_bs
= bdrv_lookup_bs(NULL
, image_node_name
, &local_err
);
3454 error_propagate(errp
, local_err
);
3459 error_setg(errp
, "image file not found");
3463 if (bdrv_find_base(image_bs
) == image_bs
) {
3464 error_setg(errp
, "not allowing backing file change on an image "
3465 "without a backing file");
3469 /* even though we are not necessarily operating on bs, we need it to
3470 * determine if block ops are currently prohibited on the chain */
3471 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_CHANGE
, errp
)) {
3475 /* final sanity check */
3476 if (!bdrv_chain_contains(bs
, image_bs
)) {
3477 error_setg(errp
, "'%s' and image file are not in the same chain",
3482 /* if not r/w, reopen to make r/w */
3483 ro
= bdrv_is_read_only(image_bs
);
3486 if (bdrv_reopen_set_read_only(image_bs
, false, errp
) != 0) {
3491 ret
= bdrv_change_backing_file(image_bs
, backing_file
,
3492 image_bs
->drv
? image_bs
->drv
->format_name
: "",
3496 error_setg_errno(errp
, -ret
, "Could not change backing file to '%s'",
3498 /* don't exit here, so we can try to restore open flags if
3503 bdrv_reopen_set_read_only(image_bs
, true, errp
);
3507 aio_context_release(aio_context
);
3510 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
3512 BlockDriverState
*bs
;
3514 Visitor
*v
= qobject_output_visitor_new(&obj
);
3517 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3518 visit_complete(v
, &obj
);
3519 qdict
= qobject_to(QDict
, obj
);
3521 qdict_flatten(qdict
);
3523 if (!qdict_get_try_str(qdict
, "node-name")) {
3524 error_setg(errp
, "'node-name' must be specified for the root node");
3528 bs
= bds_tree_init(qdict
, errp
);
3533 bdrv_set_monitor_owned(bs
);
3539 void qmp_blockdev_reopen(BlockdevOptionsList
*reopen_list
, Error
**errp
)
3541 BlockReopenQueue
*queue
= NULL
;
3543 /* Add each one of the BDS that we want to reopen to the queue */
3544 for (; reopen_list
!= NULL
; reopen_list
= reopen_list
->next
) {
3545 BlockdevOptions
*options
= reopen_list
->value
;
3546 BlockDriverState
*bs
;
3552 /* Check for the selected node name */
3553 if (!options
->node_name
) {
3554 error_setg(errp
, "node-name not specified");
3558 bs
= bdrv_find_node(options
->node_name
);
3560 error_setg(errp
, "Failed to find node with node-name='%s'",
3561 options
->node_name
);
3565 /* Put all options in a QDict and flatten it */
3566 v
= qobject_output_visitor_new(&obj
);
3567 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3568 visit_complete(v
, &obj
);
3571 qdict
= qobject_to(QDict
, obj
);
3573 qdict_flatten(qdict
);
3575 ctx
= bdrv_get_aio_context(bs
);
3576 aio_context_acquire(ctx
);
3578 queue
= bdrv_reopen_queue(queue
, bs
, qdict
, false);
3580 aio_context_release(ctx
);
3583 /* Perform the reopen operation */
3584 bdrv_reopen_multiple(queue
, errp
);
3588 bdrv_reopen_queue_free(queue
);
3591 void qmp_blockdev_del(const char *node_name
, Error
**errp
)
3593 AioContext
*aio_context
;
3594 BlockDriverState
*bs
;
3596 GLOBAL_STATE_CODE();
3598 bs
= bdrv_find_node(node_name
);
3600 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3603 if (bdrv_has_blk(bs
)) {
3604 error_setg(errp
, "Node %s is in use", node_name
);
3607 aio_context
= bdrv_get_aio_context(bs
);
3608 aio_context_acquire(aio_context
);
3610 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_DRIVE_DEL
, errp
)) {
3614 if (!QTAILQ_IN_USE(bs
, monitor_list
)) {
3615 error_setg(errp
, "Node %s is not owned by the monitor",
3620 if (bs
->refcnt
> 1) {
3621 error_setg(errp
, "Block device %s is in use",
3622 bdrv_get_device_or_node_name(bs
));
3626 QTAILQ_REMOVE(&monitor_bdrv_states
, bs
, monitor_list
);
3630 aio_context_release(aio_context
);
3633 static BdrvChild
*bdrv_find_child(BlockDriverState
*parent_bs
,
3634 const char *child_name
)
3638 QLIST_FOREACH(child
, &parent_bs
->children
, next
) {
3639 if (strcmp(child
->name
, child_name
) == 0) {
3647 void qmp_x_blockdev_change(const char *parent
, const char *child
,
3648 const char *node
, Error
**errp
)
3650 BlockDriverState
*parent_bs
, *new_bs
= NULL
;
3653 parent_bs
= bdrv_lookup_bs(parent
, parent
, errp
);
3658 if (!child
== !node
) {
3660 error_setg(errp
, "The parameters child and node are in conflict");
3662 error_setg(errp
, "Either child or node must be specified");
3668 p_child
= bdrv_find_child(parent_bs
, child
);
3670 error_setg(errp
, "Node '%s' does not have child '%s'",
3674 bdrv_del_child(parent_bs
, p_child
, errp
);
3678 new_bs
= bdrv_find_node(node
);
3680 error_setg(errp
, "Node '%s' not found", node
);
3683 bdrv_add_child(parent_bs
, new_bs
, errp
);
3687 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
3689 BlockJobInfoList
*head
= NULL
, **tail
= &head
;
3694 for (job
= block_job_next_locked(NULL
); job
;
3695 job
= block_job_next_locked(job
)) {
3696 BlockJobInfo
*value
;
3698 if (block_job_is_internal(job
)) {
3701 value
= block_job_query_locked(job
, errp
);
3703 qapi_free_BlockJobInfoList(head
);
3706 QAPI_LIST_APPEND(tail
, value
);
3712 void qmp_x_blockdev_set_iothread(const char *node_name
, StrOrNull
*iothread
,
3713 bool has_force
, bool force
, Error
**errp
)
3715 AioContext
*old_context
;
3716 AioContext
*new_context
;
3717 BlockDriverState
*bs
;
3719 bs
= bdrv_find_node(node_name
);
3721 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3725 /* Protects against accidents. */
3726 if (!(has_force
&& force
) && bdrv_has_blk(bs
)) {
3727 error_setg(errp
, "Node %s is associated with a BlockBackend and could "
3728 "be in use (use force=true to override this check)",
3733 if (iothread
->type
== QTYPE_QSTRING
) {
3734 IOThread
*obj
= iothread_by_id(iothread
->u
.s
);
3736 error_setg(errp
, "Cannot find iothread %s", iothread
->u
.s
);
3740 new_context
= iothread_get_aio_context(obj
);
3742 new_context
= qemu_get_aio_context();
3745 old_context
= bdrv_get_aio_context(bs
);
3746 aio_context_acquire(old_context
);
3748 bdrv_try_change_aio_context(bs
, new_context
, NULL
, errp
);
3750 aio_context_release(old_context
);
3753 QemuOptsList qemu_common_drive_opts
= {
3755 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
3759 .type
= QEMU_OPT_BOOL
,
3760 .help
= "enable/disable snapshot mode",
3763 .type
= QEMU_OPT_STRING
,
3764 .help
= "host AIO implementation (threads, native, io_uring)",
3766 .name
= BDRV_OPT_CACHE_WB
,
3767 .type
= QEMU_OPT_BOOL
,
3768 .help
= "Enable writeback mode",
3771 .type
= QEMU_OPT_STRING
,
3772 .help
= "disk format (raw, qcow2, ...)",
3775 .type
= QEMU_OPT_STRING
,
3776 .help
= "read error action",
3779 .type
= QEMU_OPT_STRING
,
3780 .help
= "write error action",
3782 .name
= BDRV_OPT_READ_ONLY
,
3783 .type
= QEMU_OPT_BOOL
,
3784 .help
= "open drive file as read-only",
3790 .name
= "throttling.group",
3791 .type
= QEMU_OPT_STRING
,
3792 .help
= "name of the block throttling group",
3794 .name
= "copy-on-read",
3795 .type
= QEMU_OPT_BOOL
,
3796 .help
= "copy read data from backing file into image file",
3798 .name
= "detect-zeroes",
3799 .type
= QEMU_OPT_STRING
,
3800 .help
= "try to optimize zero writes (off, on, unmap)",
3802 .name
= "stats-account-invalid",
3803 .type
= QEMU_OPT_BOOL
,
3804 .help
= "whether to account for invalid I/O operations "
3805 "in the statistics",
3807 .name
= "stats-account-failed",
3808 .type
= QEMU_OPT_BOOL
,
3809 .help
= "whether to account for failed I/O operations "
3810 "in the statistics",
3812 { /* end of list */ }
3816 QemuOptsList qemu_drive_opts
= {
3818 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
3821 * no elements => accept any params
3822 * validation will happen later
3824 { /* end of list */ }