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 (*action
)(BlkActionState
*common
, Transaction
*tran
, Error
**errp
);
1218 * Describes one Action's state within a Transaction.
1220 * @action: QAPI-defined enum identifying which Action to perform.
1221 * @ops: Table of ActionOps this Action can perform.
1222 * @block_job_txn: Transaction which this action belongs to.
1223 * @entry: List membership for all Actions in this Transaction.
1225 * This structure must be arranged as first member in a subclassed type,
1226 * assuming that the compiler will also arrange it to the same offsets as the
1229 struct BlkActionState
{
1230 TransactionAction
*action
;
1231 const BlkActionOps
*ops
;
1232 JobTxn
*block_job_txn
;
1233 QTAILQ_ENTRY(BlkActionState
) entry
;
1236 /* internal snapshot private data */
1237 typedef struct InternalSnapshotState
{
1238 BlkActionState common
;
1239 BlockDriverState
*bs
;
1240 QEMUSnapshotInfo sn
;
1242 } InternalSnapshotState
;
1244 static void internal_snapshot_abort(void *opaque
);
1245 static void internal_snapshot_clean(void *opaque
);
1246 TransactionActionDrv internal_snapshot_drv
= {
1247 .abort
= internal_snapshot_abort
,
1248 .clean
= internal_snapshot_clean
,
1251 static void internal_snapshot_action(BlkActionState
*common
,
1252 Transaction
*tran
, Error
**errp
)
1254 Error
*local_err
= NULL
;
1257 BlockDriverState
*bs
;
1258 QEMUSnapshotInfo old_sn
, *sn
;
1261 BlockdevSnapshotInternal
*internal
;
1262 InternalSnapshotState
*state
;
1263 AioContext
*aio_context
;
1266 g_assert(common
->action
->type
==
1267 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1268 internal
= common
->action
->u
.blockdev_snapshot_internal_sync
.data
;
1269 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1271 tran_add(tran
, &internal_snapshot_drv
, state
);
1273 device
= internal
->device
;
1274 name
= internal
->name
;
1276 bs
= qmp_get_root_bs(device
, errp
);
1281 aio_context
= bdrv_get_aio_context(bs
);
1282 aio_context_acquire(aio_context
);
1286 /* Paired with .clean() */
1287 bdrv_drained_begin(bs
);
1289 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT
, errp
)) {
1293 if (bdrv_is_read_only(bs
)) {
1294 error_setg(errp
, "Device '%s' is read only", device
);
1298 if (!bdrv_can_snapshot(bs
)) {
1299 error_setg(errp
, "Block format '%s' used by device '%s' "
1300 "does not support internal snapshots",
1301 bs
->drv
->format_name
, device
);
1305 if (!strlen(name
)) {
1306 error_setg(errp
, "Name is empty");
1310 /* check whether a snapshot with name exist */
1311 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
,
1314 error_propagate(errp
, local_err
);
1318 "Snapshot with name '%s' already exists on device '%s'",
1323 /* 3. take the snapshot */
1325 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1326 rt
= g_get_real_time();
1327 sn
->date_sec
= rt
/ G_USEC_PER_SEC
;
1328 sn
->date_nsec
= (rt
% G_USEC_PER_SEC
) * 1000;
1329 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1330 if (replay_mode
!= REPLAY_MODE_NONE
) {
1331 sn
->icount
= replay_get_current_icount();
1336 ret1
= bdrv_snapshot_create(bs
, sn
);
1338 error_setg_errno(errp
, -ret1
,
1339 "Failed to create snapshot '%s' on device '%s'",
1344 /* 4. succeed, mark a snapshot is created */
1345 state
->created
= true;
1348 aio_context_release(aio_context
);
1351 static void internal_snapshot_abort(void *opaque
)
1353 InternalSnapshotState
*state
= opaque
;
1354 BlockDriverState
*bs
= state
->bs
;
1355 QEMUSnapshotInfo
*sn
= &state
->sn
;
1356 AioContext
*aio_context
;
1357 Error
*local_error
= NULL
;
1359 if (!state
->created
) {
1363 aio_context
= bdrv_get_aio_context(state
->bs
);
1364 aio_context_acquire(aio_context
);
1366 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1367 error_reportf_err(local_error
,
1368 "Failed to delete snapshot with id '%s' and "
1369 "name '%s' on device '%s' in abort: ",
1370 sn
->id_str
, sn
->name
,
1371 bdrv_get_device_name(bs
));
1374 aio_context_release(aio_context
);
1377 static void internal_snapshot_clean(void *opaque
)
1379 g_autofree InternalSnapshotState
*state
= opaque
;
1380 AioContext
*aio_context
;
1386 aio_context
= bdrv_get_aio_context(state
->bs
);
1387 aio_context_acquire(aio_context
);
1389 bdrv_drained_end(state
->bs
);
1391 aio_context_release(aio_context
);
1394 /* external snapshot private data */
1395 typedef struct ExternalSnapshotState
{
1396 BlkActionState common
;
1397 BlockDriverState
*old_bs
;
1398 BlockDriverState
*new_bs
;
1399 bool overlay_appended
;
1400 } ExternalSnapshotState
;
1402 static void external_snapshot_commit(void *opaque
);
1403 static void external_snapshot_abort(void *opaque
);
1404 static void external_snapshot_clean(void *opaque
);
1405 TransactionActionDrv external_snapshot_drv
= {
1406 .commit
= external_snapshot_commit
,
1407 .abort
= external_snapshot_abort
,
1408 .clean
= external_snapshot_clean
,
1411 static void external_snapshot_action(BlkActionState
*common
, Transaction
*tran
,
1416 QDict
*options
= NULL
;
1417 Error
*local_err
= NULL
;
1418 /* Device and node name of the image to generate the snapshot from */
1420 const char *node_name
;
1421 /* Reference to the new image (for 'blockdev-snapshot') */
1422 const char *snapshot_ref
;
1423 /* File name of the new image (for 'blockdev-snapshot-sync') */
1424 const char *new_image_file
;
1425 ExternalSnapshotState
*state
=
1426 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1427 TransactionAction
*action
= common
->action
;
1428 AioContext
*aio_context
;
1429 uint64_t perm
, shared
;
1431 tran_add(tran
, &external_snapshot_drv
, state
);
1433 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1434 * purpose but a different set of parameters */
1435 switch (action
->type
) {
1436 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
:
1438 BlockdevSnapshot
*s
= action
->u
.blockdev_snapshot
.data
;
1440 node_name
= s
->node
;
1441 new_image_file
= NULL
;
1442 snapshot_ref
= s
->overlay
;
1445 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
:
1447 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1449 node_name
= s
->node_name
;
1450 new_image_file
= s
->snapshot_file
;
1451 snapshot_ref
= NULL
;
1455 g_assert_not_reached();
1458 /* start processing */
1460 state
->old_bs
= bdrv_lookup_bs(device
, node_name
, errp
);
1461 if (!state
->old_bs
) {
1465 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1466 aio_context_acquire(aio_context
);
1468 /* Paired with .clean() */
1469 bdrv_drained_begin(state
->old_bs
);
1471 if (!bdrv_is_inserted(state
->old_bs
)) {
1472 error_setg(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1476 if (bdrv_op_is_blocked(state
->old_bs
,
1477 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT
, errp
)) {
1481 if (!bdrv_is_read_only(state
->old_bs
)) {
1482 if (bdrv_flush(state
->old_bs
)) {
1483 error_setg(errp
, QERR_IO_ERROR
);
1488 if (action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
) {
1489 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1490 const char *format
= s
->format
?: "qcow2";
1491 enum NewImageMode mode
;
1492 const char *snapshot_node_name
= s
->snapshot_node_name
;
1494 if (node_name
&& !snapshot_node_name
) {
1495 error_setg(errp
, "New overlay node-name missing");
1499 if (snapshot_node_name
&&
1500 bdrv_lookup_bs(snapshot_node_name
, snapshot_node_name
, NULL
)) {
1501 error_setg(errp
, "New overlay node-name already in use");
1505 flags
= state
->old_bs
->open_flags
;
1506 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_COPY_ON_READ
);
1507 flags
|= BDRV_O_NO_BACKING
;
1509 /* create new image w/backing file */
1510 mode
= s
->has_mode
? s
->mode
: NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1511 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1512 int64_t size
= bdrv_getlength(state
->old_bs
);
1514 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1517 bdrv_refresh_filename(state
->old_bs
);
1519 aio_context_release(aio_context
);
1520 bdrv_img_create(new_image_file
, format
,
1521 state
->old_bs
->filename
,
1522 state
->old_bs
->drv
->format_name
,
1523 NULL
, size
, flags
, false, &local_err
);
1524 aio_context_acquire(aio_context
);
1527 error_propagate(errp
, local_err
);
1532 options
= qdict_new();
1533 if (snapshot_node_name
) {
1534 qdict_put_str(options
, "node-name", snapshot_node_name
);
1536 qdict_put_str(options
, "driver", format
);
1539 state
->new_bs
= bdrv_open(new_image_file
, snapshot_ref
, options
, flags
,
1541 /* We will manually add the backing_hd field to the bs later */
1542 if (!state
->new_bs
) {
1547 * Allow attaching a backing file to an overlay that's already in use only
1548 * if the parents don't assume that they are already seeing a valid image.
1549 * (Specifically, allow it as a mirror target, which is write-only access.)
1551 bdrv_get_cumulative_perm(state
->new_bs
, &perm
, &shared
);
1552 if (perm
& BLK_PERM_CONSISTENT_READ
) {
1553 error_setg(errp
, "The overlay is already in use");
1557 if (state
->new_bs
->drv
->is_filter
) {
1558 error_setg(errp
, "Filters cannot be used as overlays");
1562 if (bdrv_cow_child(state
->new_bs
)) {
1563 error_setg(errp
, "The overlay already has a backing image");
1567 if (!state
->new_bs
->drv
->supports_backing
) {
1568 error_setg(errp
, "The overlay does not support backing images");
1572 ret
= bdrv_append(state
->new_bs
, state
->old_bs
, errp
);
1576 state
->overlay_appended
= true;
1579 aio_context_release(aio_context
);
1582 static void external_snapshot_commit(void *opaque
)
1584 ExternalSnapshotState
*state
= opaque
;
1585 AioContext
*aio_context
;
1587 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1588 aio_context_acquire(aio_context
);
1590 /* We don't need (or want) to use the transactional
1591 * bdrv_reopen_multiple() across all the entries at once, because we
1592 * don't want to abort all of them if one of them fails the reopen */
1593 if (!qatomic_read(&state
->old_bs
->copy_on_read
)) {
1594 bdrv_reopen_set_read_only(state
->old_bs
, true, NULL
);
1597 aio_context_release(aio_context
);
1600 static void external_snapshot_abort(void *opaque
)
1602 ExternalSnapshotState
*state
= opaque
;
1603 if (state
->new_bs
) {
1604 if (state
->overlay_appended
) {
1605 AioContext
*aio_context
;
1606 AioContext
*tmp_context
;
1609 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1610 aio_context_acquire(aio_context
);
1612 bdrv_ref(state
->old_bs
); /* we can't let bdrv_set_backind_hd()
1613 close state->old_bs; we need it */
1614 bdrv_set_backing_hd(state
->new_bs
, NULL
, &error_abort
);
1617 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1618 * the main AioContext. As we're still going to be using it, return
1619 * it to the AioContext it was before.
1621 tmp_context
= bdrv_get_aio_context(state
->old_bs
);
1622 if (aio_context
!= tmp_context
) {
1623 aio_context_release(aio_context
);
1624 aio_context_acquire(tmp_context
);
1626 ret
= bdrv_try_change_aio_context(state
->old_bs
,
1627 aio_context
, NULL
, NULL
);
1630 aio_context_release(tmp_context
);
1631 aio_context_acquire(aio_context
);
1634 bdrv_replace_node(state
->new_bs
, state
->old_bs
, &error_abort
);
1635 bdrv_unref(state
->old_bs
); /* bdrv_replace_node() ref'ed old_bs */
1637 aio_context_release(aio_context
);
1642 static void external_snapshot_clean(void *opaque
)
1644 g_autofree ExternalSnapshotState
*state
= opaque
;
1645 AioContext
*aio_context
;
1647 if (!state
->old_bs
) {
1651 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1652 aio_context_acquire(aio_context
);
1654 bdrv_drained_end(state
->old_bs
);
1655 bdrv_unref(state
->new_bs
);
1657 aio_context_release(aio_context
);
1660 typedef struct DriveBackupState
{
1661 BlkActionState common
;
1662 BlockDriverState
*bs
;
1666 static BlockJob
*do_backup_common(BackupCommon
*backup
,
1667 BlockDriverState
*bs
,
1668 BlockDriverState
*target_bs
,
1669 AioContext
*aio_context
,
1670 JobTxn
*txn
, Error
**errp
);
1672 static void drive_backup_commit(void *opaque
);
1673 static void drive_backup_abort(void *opaque
);
1674 static void drive_backup_clean(void *opaque
);
1675 TransactionActionDrv drive_backup_drv
= {
1676 .commit
= drive_backup_commit
,
1677 .abort
= drive_backup_abort
,
1678 .clean
= drive_backup_clean
,
1681 static void drive_backup_action(BlkActionState
*common
, Transaction
*tran
,
1684 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1685 DriveBackup
*backup
;
1686 BlockDriverState
*bs
;
1687 BlockDriverState
*target_bs
;
1688 BlockDriverState
*source
= NULL
;
1689 AioContext
*aio_context
;
1690 AioContext
*old_context
;
1693 Error
*local_err
= NULL
;
1696 bool set_backing_hd
= false;
1699 tran_add(tran
, &drive_backup_drv
, state
);
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(void *opaque
)
1833 DriveBackupState
*state
= opaque
;
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(void *opaque
)
1847 DriveBackupState
*state
= opaque
;
1850 job_cancel_sync(&state
->job
->job
, true);
1854 static void drive_backup_clean(void *opaque
)
1856 g_autofree DriveBackupState
*state
= opaque
;
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_commit(void *opaque
);
1878 static void blockdev_backup_abort(void *opaque
);
1879 static void blockdev_backup_clean(void *opaque
);
1880 TransactionActionDrv blockdev_backup_drv
= {
1881 .commit
= blockdev_backup_commit
,
1882 .abort
= blockdev_backup_abort
,
1883 .clean
= blockdev_backup_clean
,
1886 static void blockdev_backup_action(BlkActionState
*common
, Transaction
*tran
,
1889 BlockdevBackupState
*state
= DO_UPCAST(BlockdevBackupState
, common
, common
);
1890 BlockdevBackup
*backup
;
1891 BlockDriverState
*bs
;
1892 BlockDriverState
*target_bs
;
1893 AioContext
*aio_context
;
1894 AioContext
*old_context
;
1897 tran_add(tran
, &blockdev_backup_drv
, state
);
1899 assert(common
->action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
);
1900 backup
= common
->action
->u
.blockdev_backup
.data
;
1902 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1907 target_bs
= bdrv_lookup_bs(backup
->target
, backup
->target
, errp
);
1912 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1913 aio_context
= bdrv_get_aio_context(bs
);
1914 old_context
= bdrv_get_aio_context(target_bs
);
1915 aio_context_acquire(old_context
);
1917 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
1919 aio_context_release(old_context
);
1923 aio_context_release(old_context
);
1924 aio_context_acquire(aio_context
);
1927 /* Paired with .clean() */
1928 bdrv_drained_begin(state
->bs
);
1930 state
->job
= do_backup_common(qapi_BlockdevBackup_base(backup
),
1931 bs
, target_bs
, aio_context
,
1932 common
->block_job_txn
, errp
);
1934 aio_context_release(aio_context
);
1937 static void blockdev_backup_commit(void *opaque
)
1939 BlockdevBackupState
*state
= opaque
;
1940 AioContext
*aio_context
;
1942 aio_context
= bdrv_get_aio_context(state
->bs
);
1943 aio_context_acquire(aio_context
);
1946 job_start(&state
->job
->job
);
1948 aio_context_release(aio_context
);
1951 static void blockdev_backup_abort(void *opaque
)
1953 BlockdevBackupState
*state
= opaque
;
1956 job_cancel_sync(&state
->job
->job
, true);
1960 static void blockdev_backup_clean(void *opaque
)
1962 g_autofree BlockdevBackupState
*state
= opaque
;
1963 AioContext
*aio_context
;
1969 aio_context
= bdrv_get_aio_context(state
->bs
);
1970 aio_context_acquire(aio_context
);
1972 bdrv_drained_end(state
->bs
);
1974 aio_context_release(aio_context
);
1977 typedef struct BlockDirtyBitmapState
{
1978 BlkActionState common
;
1979 BdrvDirtyBitmap
*bitmap
;
1980 BlockDriverState
*bs
;
1984 } BlockDirtyBitmapState
;
1986 static void block_dirty_bitmap_add_abort(void *opaque
);
1987 TransactionActionDrv block_dirty_bitmap_add_drv
= {
1988 .abort
= block_dirty_bitmap_add_abort
,
1992 static void block_dirty_bitmap_add_action(BlkActionState
*common
,
1993 Transaction
*tran
, Error
**errp
)
1995 Error
*local_err
= NULL
;
1996 BlockDirtyBitmapAdd
*action
;
1997 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2000 tran_add(tran
, &block_dirty_bitmap_add_drv
, state
);
2002 action
= common
->action
->u
.block_dirty_bitmap_add
.data
;
2003 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2004 qmp_block_dirty_bitmap_add(action
->node
, action
->name
,
2005 action
->has_granularity
, action
->granularity
,
2006 action
->has_persistent
, action
->persistent
,
2007 action
->has_disabled
, action
->disabled
,
2011 state
->prepared
= true;
2013 error_propagate(errp
, local_err
);
2017 static void block_dirty_bitmap_add_abort(void *opaque
)
2019 BlockDirtyBitmapAdd
*action
;
2020 BlockDirtyBitmapState
*state
= opaque
;
2022 action
= state
->common
.action
->u
.block_dirty_bitmap_add
.data
;
2023 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2024 * then the node reference and bitmap name must have been valid.
2026 if (state
->prepared
) {
2027 qmp_block_dirty_bitmap_remove(action
->node
, action
->name
, &error_abort
);
2031 static void block_dirty_bitmap_restore(void *opaque
);
2032 static void block_dirty_bitmap_free_backup(void *opaque
);
2033 TransactionActionDrv block_dirty_bitmap_clear_drv
= {
2034 .abort
= block_dirty_bitmap_restore
,
2035 .commit
= block_dirty_bitmap_free_backup
,
2039 static void block_dirty_bitmap_clear_action(BlkActionState
*common
,
2040 Transaction
*tran
, Error
**errp
)
2042 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2044 BlockDirtyBitmap
*action
;
2046 tran_add(tran
, &block_dirty_bitmap_clear_drv
, state
);
2048 action
= common
->action
->u
.block_dirty_bitmap_clear
.data
;
2049 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2053 if (!state
->bitmap
) {
2057 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_DEFAULT
, errp
)) {
2061 bdrv_clear_dirty_bitmap(state
->bitmap
, &state
->backup
);
2064 static void block_dirty_bitmap_restore(void *opaque
)
2066 BlockDirtyBitmapState
*state
= opaque
;
2068 if (state
->backup
) {
2069 bdrv_restore_dirty_bitmap(state
->bitmap
, state
->backup
);
2073 static void block_dirty_bitmap_free_backup(void *opaque
)
2075 BlockDirtyBitmapState
*state
= opaque
;
2077 hbitmap_free(state
->backup
);
2080 static void block_dirty_bitmap_enable_abort(void *opaque
);
2081 TransactionActionDrv block_dirty_bitmap_enable_drv
= {
2082 .abort
= block_dirty_bitmap_enable_abort
,
2086 static void block_dirty_bitmap_enable_action(BlkActionState
*common
,
2087 Transaction
*tran
, Error
**errp
)
2089 BlockDirtyBitmap
*action
;
2090 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2093 tran_add(tran
, &block_dirty_bitmap_enable_drv
, state
);
2095 action
= common
->action
->u
.block_dirty_bitmap_enable
.data
;
2096 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2100 if (!state
->bitmap
) {
2104 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2108 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2109 bdrv_enable_dirty_bitmap(state
->bitmap
);
2112 static void block_dirty_bitmap_enable_abort(void *opaque
)
2114 BlockDirtyBitmapState
*state
= opaque
;
2116 if (!state
->was_enabled
) {
2117 bdrv_disable_dirty_bitmap(state
->bitmap
);
2121 static void block_dirty_bitmap_disable_abort(void *opaque
);
2122 TransactionActionDrv block_dirty_bitmap_disable_drv
= {
2123 .abort
= block_dirty_bitmap_disable_abort
,
2127 static void block_dirty_bitmap_disable_action(BlkActionState
*common
,
2128 Transaction
*tran
, Error
**errp
)
2130 BlockDirtyBitmap
*action
;
2131 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2134 tran_add(tran
, &block_dirty_bitmap_disable_drv
, state
);
2136 action
= common
->action
->u
.block_dirty_bitmap_disable
.data
;
2137 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2141 if (!state
->bitmap
) {
2145 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2149 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2150 bdrv_disable_dirty_bitmap(state
->bitmap
);
2153 static void block_dirty_bitmap_disable_abort(void *opaque
)
2155 BlockDirtyBitmapState
*state
= opaque
;
2157 if (state
->was_enabled
) {
2158 bdrv_enable_dirty_bitmap(state
->bitmap
);
2162 TransactionActionDrv block_dirty_bitmap_merge_drv
= {
2163 .commit
= block_dirty_bitmap_free_backup
,
2164 .abort
= block_dirty_bitmap_restore
,
2168 static void block_dirty_bitmap_merge_action(BlkActionState
*common
,
2169 Transaction
*tran
, Error
**errp
)
2171 BlockDirtyBitmapMerge
*action
;
2172 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2175 tran_add(tran
, &block_dirty_bitmap_merge_drv
, state
);
2177 action
= common
->action
->u
.block_dirty_bitmap_merge
.data
;
2179 state
->bitmap
= block_dirty_bitmap_merge(action
->node
, action
->target
,
2180 action
->bitmaps
, &state
->backup
,
2184 static void block_dirty_bitmap_remove_commit(void *opaque
);
2185 static void block_dirty_bitmap_remove_abort(void *opaque
);
2186 TransactionActionDrv block_dirty_bitmap_remove_drv
= {
2187 .commit
= block_dirty_bitmap_remove_commit
,
2188 .abort
= block_dirty_bitmap_remove_abort
,
2192 static void block_dirty_bitmap_remove_action(BlkActionState
*common
,
2193 Transaction
*tran
, Error
**errp
)
2195 BlockDirtyBitmap
*action
;
2196 BlockDirtyBitmapState
*state
= DO_UPCAST(BlockDirtyBitmapState
,
2199 tran_add(tran
, &block_dirty_bitmap_remove_drv
, state
);
2201 action
= common
->action
->u
.block_dirty_bitmap_remove
.data
;
2203 state
->bitmap
= block_dirty_bitmap_remove(action
->node
, action
->name
,
2204 false, &state
->bs
, errp
);
2205 if (state
->bitmap
) {
2206 bdrv_dirty_bitmap_skip_store(state
->bitmap
, true);
2207 bdrv_dirty_bitmap_set_busy(state
->bitmap
, true);
2211 static void block_dirty_bitmap_remove_abort(void *opaque
)
2213 BlockDirtyBitmapState
*state
= opaque
;
2215 if (state
->bitmap
) {
2216 bdrv_dirty_bitmap_skip_store(state
->bitmap
, false);
2217 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2221 static void block_dirty_bitmap_remove_commit(void *opaque
)
2223 BlockDirtyBitmapState
*state
= opaque
;
2225 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2226 bdrv_release_dirty_bitmap(state
->bitmap
);
2229 static void abort_commit(void *opaque
);
2230 TransactionActionDrv abort_drv
= {
2231 .commit
= abort_commit
,
2235 static void abort_action(BlkActionState
*common
, Transaction
*tran
,
2238 tran_add(tran
, &abort_drv
, common
);
2239 error_setg(errp
, "Transaction aborted using Abort action");
2242 static void abort_commit(void *opaque
)
2244 g_assert_not_reached(); /* this action never succeeds */
2247 static const BlkActionOps actions_map
[] = {
2248 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
] = {
2249 .instance_size
= sizeof(ExternalSnapshotState
),
2250 .action
= external_snapshot_action
,
2252 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
2253 .instance_size
= sizeof(ExternalSnapshotState
),
2254 .action
= external_snapshot_action
,
2256 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
2257 .instance_size
= sizeof(DriveBackupState
),
2258 .action
= drive_backup_action
,
2260 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
] = {
2261 .instance_size
= sizeof(BlockdevBackupState
),
2262 .action
= blockdev_backup_action
,
2264 [TRANSACTION_ACTION_KIND_ABORT
] = {
2265 .instance_size
= sizeof(BlkActionState
),
2266 .action
= abort_action
,
2268 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
2269 .instance_size
= sizeof(InternalSnapshotState
),
2270 .action
= internal_snapshot_action
,
2272 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD
] = {
2273 .instance_size
= sizeof(BlockDirtyBitmapState
),
2274 .action
= block_dirty_bitmap_add_action
,
2276 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR
] = {
2277 .instance_size
= sizeof(BlockDirtyBitmapState
),
2278 .action
= block_dirty_bitmap_clear_action
,
2280 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE
] = {
2281 .instance_size
= sizeof(BlockDirtyBitmapState
),
2282 .action
= block_dirty_bitmap_enable_action
,
2284 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE
] = {
2285 .instance_size
= sizeof(BlockDirtyBitmapState
),
2286 .action
= block_dirty_bitmap_disable_action
,
2288 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE
] = {
2289 .instance_size
= sizeof(BlockDirtyBitmapState
),
2290 .action
= block_dirty_bitmap_merge_action
,
2292 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE
] = {
2293 .instance_size
= sizeof(BlockDirtyBitmapState
),
2294 .action
= block_dirty_bitmap_remove_action
,
2296 /* Where are transactions for MIRROR, COMMIT and STREAM?
2297 * Although these blockjobs use transaction callbacks like the backup job,
2298 * these jobs do not necessarily adhere to transaction semantics.
2299 * These jobs may not fully undo all of their actions on abort, nor do they
2300 * necessarily work in transactions with more than one job in them.
2305 * 'Atomic' group operations. The operations are performed as a set, and if
2306 * any fail then we roll back all operations in the group.
2308 * Always run under BQL.
2310 void qmp_transaction(TransactionActionList
*actions
,
2311 struct TransactionProperties
*properties
,
2314 TransactionActionList
*act
;
2315 JobTxn
*block_job_txn
= NULL
;
2316 Error
*local_err
= NULL
;
2318 ActionCompletionMode comp_mode
=
2319 properties
? properties
->completion_mode
:
2320 ACTION_COMPLETION_MODE_INDIVIDUAL
;
2322 GLOBAL_STATE_CODE();
2324 /* Does this transaction get canceled as a group on failure?
2325 * If not, we don't really need to make a JobTxn.
2327 if (comp_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
2328 for (act
= actions
; act
; act
= act
->next
) {
2329 TransactionActionKind type
= act
->value
->type
;
2331 if (type
!= TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
&&
2332 type
!= TRANSACTION_ACTION_KIND_DRIVE_BACKUP
)
2335 "Action '%s' does not support transaction property "
2336 "completion-mode = %s",
2337 TransactionActionKind_str(type
),
2338 ActionCompletionMode_str(comp_mode
));
2343 block_job_txn
= job_txn_new();
2346 /* drain all i/o before any operations */
2351 /* We don't do anything in this loop that commits us to the operations */
2352 for (act
= actions
; act
; act
= act
->next
) {
2353 TransactionAction
*dev_info
= act
->value
;
2354 const BlkActionOps
*ops
;
2355 BlkActionState
*state
;
2357 assert(dev_info
->type
< ARRAY_SIZE(actions_map
));
2359 ops
= &actions_map
[dev_info
->type
];
2360 assert(ops
->instance_size
> 0);
2362 state
= g_malloc0(ops
->instance_size
);
2364 state
->action
= dev_info
;
2365 state
->block_job_txn
= block_job_txn
;
2367 state
->ops
->action(state
, tran
, &local_err
);
2369 error_propagate(errp
, local_err
);
2370 goto delete_and_fail
;
2380 /* failure, and it is all-or-none; roll back all operations */
2383 job_txn_unref(block_job_txn
);
2386 BlockDirtyBitmapSha256
*qmp_x_debug_block_dirty_bitmap_sha256(const char *node
,
2390 BdrvDirtyBitmap
*bitmap
;
2391 BlockDriverState
*bs
;
2392 BlockDirtyBitmapSha256
*ret
= NULL
;
2395 bitmap
= block_dirty_bitmap_lookup(node
, name
, &bs
, errp
);
2396 if (!bitmap
|| !bs
) {
2400 sha256
= bdrv_dirty_bitmap_sha256(bitmap
, errp
);
2401 if (sha256
== NULL
) {
2405 ret
= g_new(BlockDirtyBitmapSha256
, 1);
2406 ret
->sha256
= sha256
;
2411 void coroutine_fn
qmp_block_resize(const char *device
, const char *node_name
,
2412 int64_t size
, Error
**errp
)
2414 Error
*local_err
= NULL
;
2416 BlockDriverState
*bs
;
2417 AioContext
*old_ctx
;
2419 bs
= bdrv_lookup_bs(device
, node_name
, &local_err
);
2421 error_propagate(errp
, local_err
);
2426 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
2430 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_RESIZE
, NULL
)) {
2431 error_setg(errp
, QERR_DEVICE_IN_USE
, device
);
2435 blk
= blk_co_new_with_bs(bs
, BLK_PERM_RESIZE
, BLK_PERM_ALL
, errp
);
2441 bdrv_drained_begin(bs
);
2444 old_ctx
= bdrv_co_enter(bs
);
2445 blk_co_truncate(blk
, size
, false, PREALLOC_MODE_OFF
, 0, errp
);
2446 bdrv_co_leave(bs
, old_ctx
);
2449 bdrv_drained_end(bs
);
2454 void qmp_block_stream(const char *job_id
, const char *device
,
2456 const char *base_node
,
2457 const char *backing_file
,
2459 bool has_speed
, int64_t speed
,
2460 bool has_on_error
, BlockdevOnError on_error
,
2461 const char *filter_node_name
,
2462 bool has_auto_finalize
, bool auto_finalize
,
2463 bool has_auto_dismiss
, bool auto_dismiss
,
2466 BlockDriverState
*bs
, *iter
, *iter_end
;
2467 BlockDriverState
*base_bs
= NULL
;
2468 BlockDriverState
*bottom_bs
= NULL
;
2469 AioContext
*aio_context
;
2470 Error
*local_err
= NULL
;
2471 int job_flags
= JOB_DEFAULT
;
2473 if (base
&& base_node
) {
2474 error_setg(errp
, "'base' and 'base-node' cannot be specified "
2475 "at the same time");
2479 if (base
&& bottom
) {
2480 error_setg(errp
, "'base' and 'bottom' cannot be specified "
2481 "at the same time");
2485 if (bottom
&& base_node
) {
2486 error_setg(errp
, "'bottom' and 'base-node' cannot be specified "
2487 "at the same time");
2491 if (!has_on_error
) {
2492 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2495 bs
= bdrv_lookup_bs(device
, device
, errp
);
2500 aio_context
= bdrv_get_aio_context(bs
);
2501 aio_context_acquire(aio_context
);
2504 base_bs
= bdrv_find_backing_image(bs
, base
);
2505 if (base_bs
== NULL
) {
2506 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2509 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2513 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2517 if (bs
== base_bs
|| !bdrv_chain_contains(bs
, base_bs
)) {
2518 error_setg(errp
, "Node '%s' is not a backing image of '%s'",
2522 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2523 bdrv_refresh_filename(base_bs
);
2527 bottom_bs
= bdrv_lookup_bs(NULL
, bottom
, errp
);
2531 if (!bottom_bs
->drv
) {
2532 error_setg(errp
, "Node '%s' is not open", bottom
);
2535 if (bottom_bs
->drv
->is_filter
) {
2536 error_setg(errp
, "Node '%s' is a filter, use a non-filter node "
2537 "as 'bottom'", bottom
);
2540 if (!bdrv_chain_contains(bs
, bottom_bs
)) {
2541 error_setg(errp
, "Node '%s' is not in a chain starting from '%s'",
2545 assert(bdrv_get_aio_context(bottom_bs
) == aio_context
);
2549 * Check for op blockers in the whole chain between bs and base (or bottom)
2551 iter_end
= bottom
? bdrv_filter_or_cow_bs(bottom_bs
) : base_bs
;
2552 for (iter
= bs
; iter
&& iter
!= iter_end
;
2553 iter
= bdrv_filter_or_cow_bs(iter
))
2555 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_STREAM
, errp
)) {
2560 /* if we are streaming the entire chain, the result will have no backing
2561 * file, and specifying one is therefore an error */
2562 if (!base_bs
&& backing_file
) {
2563 error_setg(errp
, "backing file specified, but streaming the "
2568 if (has_auto_finalize
&& !auto_finalize
) {
2569 job_flags
|= JOB_MANUAL_FINALIZE
;
2571 if (has_auto_dismiss
&& !auto_dismiss
) {
2572 job_flags
|= JOB_MANUAL_DISMISS
;
2575 stream_start(job_id
, bs
, base_bs
, backing_file
,
2576 bottom_bs
, job_flags
, has_speed
? speed
: 0, on_error
,
2577 filter_node_name
, &local_err
);
2579 error_propagate(errp
, local_err
);
2583 trace_qmp_block_stream(bs
);
2586 aio_context_release(aio_context
);
2589 void qmp_block_commit(const char *job_id
, const char *device
,
2590 const char *base_node
,
2592 const char *top_node
,
2594 const char *backing_file
,
2595 bool has_speed
, int64_t speed
,
2596 bool has_on_error
, BlockdevOnError on_error
,
2597 const char *filter_node_name
,
2598 bool has_auto_finalize
, bool auto_finalize
,
2599 bool has_auto_dismiss
, bool auto_dismiss
,
2602 BlockDriverState
*bs
;
2603 BlockDriverState
*iter
;
2604 BlockDriverState
*base_bs
, *top_bs
;
2605 AioContext
*aio_context
;
2606 Error
*local_err
= NULL
;
2607 int job_flags
= JOB_DEFAULT
;
2608 uint64_t top_perm
, top_shared
;
2613 if (!has_on_error
) {
2614 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2616 if (has_auto_finalize
&& !auto_finalize
) {
2617 job_flags
|= JOB_MANUAL_FINALIZE
;
2619 if (has_auto_dismiss
&& !auto_dismiss
) {
2620 job_flags
|= JOB_MANUAL_DISMISS
;
2624 * libvirt relies on the DeviceNotFound error class in order to probe for
2625 * live commit feature versions; for this to work, we must make sure to
2626 * perform the device lookup before any generic errors that may occur in a
2627 * scenario in which all optional arguments are omitted. */
2628 bs
= qmp_get_root_bs(device
, &local_err
);
2630 bs
= bdrv_lookup_bs(device
, device
, NULL
);
2632 error_free(local_err
);
2633 error_set(errp
, ERROR_CLASS_DEVICE_NOT_FOUND
,
2634 "Device '%s' not found", device
);
2636 error_propagate(errp
, local_err
);
2641 aio_context
= bdrv_get_aio_context(bs
);
2642 aio_context_acquire(aio_context
);
2644 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_COMMIT_SOURCE
, errp
)) {
2648 /* default top_bs is the active layer */
2651 if (top_node
&& top
) {
2652 error_setg(errp
, "'top-node' and 'top' are mutually exclusive");
2654 } else if (top_node
) {
2655 top_bs
= bdrv_lookup_bs(NULL
, top_node
, errp
);
2656 if (top_bs
== NULL
) {
2659 if (!bdrv_chain_contains(bs
, top_bs
)) {
2660 error_setg(errp
, "'%s' is not in this backing file chain",
2665 /* This strcmp() is just a shortcut, there is no need to
2666 * refresh @bs's filename. If it mismatches,
2667 * bdrv_find_backing_image() will do the refresh and may still
2669 if (strcmp(bs
->filename
, top
) != 0) {
2670 top_bs
= bdrv_find_backing_image(bs
, top
);
2674 if (top_bs
== NULL
) {
2675 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
2679 assert(bdrv_get_aio_context(top_bs
) == aio_context
);
2681 if (base_node
&& base
) {
2682 error_setg(errp
, "'base-node' and 'base' are mutually exclusive");
2684 } else if (base_node
) {
2685 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2686 if (base_bs
== NULL
) {
2689 if (!bdrv_chain_contains(top_bs
, base_bs
)) {
2690 error_setg(errp
, "'%s' is not in this backing file chain",
2695 base_bs
= bdrv_find_backing_image(top_bs
, base
);
2696 if (base_bs
== NULL
) {
2697 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2701 base_bs
= bdrv_find_base(top_bs
);
2702 if (base_bs
== NULL
) {
2703 error_setg(errp
, "There is no backimg image");
2708 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2710 for (iter
= top_bs
; iter
!= bdrv_filter_or_cow_bs(base_bs
);
2711 iter
= bdrv_filter_or_cow_bs(iter
))
2713 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2718 /* Do not allow attempts to commit an image into itself */
2719 if (top_bs
== base_bs
) {
2720 error_setg(errp
, "cannot commit an image into itself");
2725 * Active commit is required if and only if someone has taken a
2726 * WRITE permission on the top node. Historically, we have always
2727 * used active commit for top nodes, so continue that practice
2728 * lest we possibly break clients that rely on this behavior, e.g.
2729 * to later attach this node to a writing parent.
2730 * (Active commit is never really wrong.)
2732 bdrv_get_cumulative_perm(top_bs
, &top_perm
, &top_shared
);
2733 if (top_perm
& BLK_PERM_WRITE
||
2734 bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
))
2737 if (bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
)) {
2738 error_setg(errp
, "'backing-file' specified,"
2739 " but 'top' is the active layer");
2741 error_setg(errp
, "'backing-file' specified, but 'top' has a "
2748 * Emulate here what block_job_create() does, because it
2749 * is possible that @bs != @top_bs (the block job should
2750 * be named after @bs, even if @top_bs is the actual
2753 job_id
= bdrv_get_device_name(bs
);
2755 commit_active_start(job_id
, top_bs
, base_bs
, job_flags
, speed
, on_error
,
2756 filter_node_name
, NULL
, NULL
, false, &local_err
);
2758 BlockDriverState
*overlay_bs
= bdrv_find_overlay(bs
, top_bs
);
2759 if (bdrv_op_is_blocked(overlay_bs
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2762 commit_start(job_id
, bs
, base_bs
, top_bs
, job_flags
,
2763 speed
, on_error
, backing_file
,
2764 filter_node_name
, &local_err
);
2766 if (local_err
!= NULL
) {
2767 error_propagate(errp
, local_err
);
2772 aio_context_release(aio_context
);
2775 /* Common QMP interface for drive-backup and blockdev-backup */
2776 static BlockJob
*do_backup_common(BackupCommon
*backup
,
2777 BlockDriverState
*bs
,
2778 BlockDriverState
*target_bs
,
2779 AioContext
*aio_context
,
2780 JobTxn
*txn
, Error
**errp
)
2782 BlockJob
*job
= NULL
;
2783 BdrvDirtyBitmap
*bmap
= NULL
;
2784 BackupPerf perf
= { .max_workers
= 64 };
2785 int job_flags
= JOB_DEFAULT
;
2787 if (!backup
->has_speed
) {
2790 if (!backup
->has_on_source_error
) {
2791 backup
->on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2793 if (!backup
->has_on_target_error
) {
2794 backup
->on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2796 if (!backup
->has_auto_finalize
) {
2797 backup
->auto_finalize
= true;
2799 if (!backup
->has_auto_dismiss
) {
2800 backup
->auto_dismiss
= true;
2802 if (!backup
->has_compress
) {
2803 backup
->compress
= false;
2806 if (backup
->x_perf
) {
2807 if (backup
->x_perf
->has_use_copy_range
) {
2808 perf
.use_copy_range
= backup
->x_perf
->use_copy_range
;
2810 if (backup
->x_perf
->has_max_workers
) {
2811 perf
.max_workers
= backup
->x_perf
->max_workers
;
2813 if (backup
->x_perf
->has_max_chunk
) {
2814 perf
.max_chunk
= backup
->x_perf
->max_chunk
;
2818 if ((backup
->sync
== MIRROR_SYNC_MODE_BITMAP
) ||
2819 (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
)) {
2820 /* done before desugaring 'incremental' to print the right message */
2821 if (!backup
->bitmap
) {
2822 error_setg(errp
, "must provide a valid bitmap name for "
2823 "'%s' sync mode", MirrorSyncMode_str(backup
->sync
));
2828 if (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
) {
2829 if (backup
->has_bitmap_mode
&&
2830 backup
->bitmap_mode
!= BITMAP_SYNC_MODE_ON_SUCCESS
) {
2831 error_setg(errp
, "Bitmap sync mode must be '%s' "
2832 "when using sync mode '%s'",
2833 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS
),
2834 MirrorSyncMode_str(backup
->sync
));
2837 backup
->has_bitmap_mode
= true;
2838 backup
->sync
= MIRROR_SYNC_MODE_BITMAP
;
2839 backup
->bitmap_mode
= BITMAP_SYNC_MODE_ON_SUCCESS
;
2842 if (backup
->bitmap
) {
2843 bmap
= bdrv_find_dirty_bitmap(bs
, backup
->bitmap
);
2845 error_setg(errp
, "Bitmap '%s' could not be found", backup
->bitmap
);
2848 if (!backup
->has_bitmap_mode
) {
2849 error_setg(errp
, "Bitmap sync mode must be given "
2850 "when providing a bitmap");
2853 if (bdrv_dirty_bitmap_check(bmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2857 /* This does not produce a useful bitmap artifact: */
2858 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
2859 error_setg(errp
, "sync mode '%s' does not produce meaningful bitmap"
2860 " outputs", MirrorSyncMode_str(backup
->sync
));
2864 /* If the bitmap isn't used for input or output, this is useless: */
2865 if (backup
->bitmap_mode
== BITMAP_SYNC_MODE_NEVER
&&
2866 backup
->sync
!= MIRROR_SYNC_MODE_BITMAP
) {
2867 error_setg(errp
, "Bitmap sync mode '%s' has no meaningful effect"
2868 " when combined with sync mode '%s'",
2869 BitmapSyncMode_str(backup
->bitmap_mode
),
2870 MirrorSyncMode_str(backup
->sync
));
2875 if (!backup
->bitmap
&& backup
->has_bitmap_mode
) {
2876 error_setg(errp
, "Cannot specify bitmap sync mode without a bitmap");
2880 if (!backup
->auto_finalize
) {
2881 job_flags
|= JOB_MANUAL_FINALIZE
;
2883 if (!backup
->auto_dismiss
) {
2884 job_flags
|= JOB_MANUAL_DISMISS
;
2887 job
= backup_job_create(backup
->job_id
, bs
, target_bs
, backup
->speed
,
2888 backup
->sync
, bmap
, backup
->bitmap_mode
,
2890 backup
->filter_node_name
,
2892 backup
->on_source_error
,
2893 backup
->on_target_error
,
2894 job_flags
, NULL
, NULL
, txn
, errp
);
2898 void qmp_drive_backup(DriveBackup
*backup
, Error
**errp
)
2900 TransactionAction action
= {
2901 .type
= TRANSACTION_ACTION_KIND_DRIVE_BACKUP
,
2902 .u
.drive_backup
.data
= backup
,
2904 blockdev_do_action(&action
, errp
);
2907 BlockDeviceInfoList
*qmp_query_named_block_nodes(bool has_flat
,
2911 bool return_flat
= has_flat
&& flat
;
2913 return bdrv_named_nodes_list(return_flat
, errp
);
2916 XDbgBlockGraph
*qmp_x_debug_query_block_graph(Error
**errp
)
2918 return bdrv_get_xdbg_block_graph(errp
);
2921 void qmp_blockdev_backup(BlockdevBackup
*backup
, Error
**errp
)
2923 TransactionAction action
= {
2924 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
,
2925 .u
.blockdev_backup
.data
= backup
,
2927 blockdev_do_action(&action
, errp
);
2930 /* Parameter check and block job starting for drive mirroring.
2931 * Caller should hold @device and @target's aio context (must be the same).
2933 static void blockdev_mirror_common(const char *job_id
, BlockDriverState
*bs
,
2934 BlockDriverState
*target
,
2935 const char *replaces
,
2936 enum MirrorSyncMode sync
,
2937 BlockMirrorBackingMode backing_mode
,
2939 bool has_speed
, int64_t speed
,
2940 bool has_granularity
, uint32_t granularity
,
2941 bool has_buf_size
, int64_t buf_size
,
2942 bool has_on_source_error
,
2943 BlockdevOnError on_source_error
,
2944 bool has_on_target_error
,
2945 BlockdevOnError on_target_error
,
2946 bool has_unmap
, bool unmap
,
2947 const char *filter_node_name
,
2948 bool has_copy_mode
, MirrorCopyMode copy_mode
,
2949 bool has_auto_finalize
, bool auto_finalize
,
2950 bool has_auto_dismiss
, bool auto_dismiss
,
2953 BlockDriverState
*unfiltered_bs
;
2954 int job_flags
= JOB_DEFAULT
;
2956 GLOBAL_STATE_CODE();
2957 GRAPH_RDLOCK_GUARD_MAINLOOP();
2962 if (!has_on_source_error
) {
2963 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2965 if (!has_on_target_error
) {
2966 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2968 if (!has_granularity
) {
2971 if (!has_buf_size
) {
2977 if (!has_copy_mode
) {
2978 copy_mode
= MIRROR_COPY_MODE_BACKGROUND
;
2980 if (has_auto_finalize
&& !auto_finalize
) {
2981 job_flags
|= JOB_MANUAL_FINALIZE
;
2983 if (has_auto_dismiss
&& !auto_dismiss
) {
2984 job_flags
|= JOB_MANUAL_DISMISS
;
2987 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
2988 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2989 "a value in range [512B, 64MB]");
2992 if (granularity
& (granularity
- 1)) {
2993 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2998 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3001 if (bdrv_op_is_blocked(target
, BLOCK_OP_TYPE_MIRROR_TARGET
, errp
)) {
3005 if (!bdrv_backing_chain_next(bs
) && sync
== MIRROR_SYNC_MODE_TOP
) {
3006 sync
= MIRROR_SYNC_MODE_FULL
;
3010 /* We want to mirror from @bs, but keep implicit filters on top */
3011 unfiltered_bs
= bdrv_skip_implicit_filters(bs
);
3012 if (unfiltered_bs
!= bs
) {
3013 replaces
= unfiltered_bs
->node_name
;
3018 BlockDriverState
*to_replace_bs
;
3019 AioContext
*replace_aio_context
;
3020 int64_t bs_size
, replace_size
;
3022 bs_size
= bdrv_getlength(bs
);
3024 error_setg_errno(errp
, -bs_size
, "Failed to query device's size");
3028 to_replace_bs
= check_to_replace_node(bs
, replaces
, errp
);
3029 if (!to_replace_bs
) {
3033 replace_aio_context
= bdrv_get_aio_context(to_replace_bs
);
3034 aio_context_acquire(replace_aio_context
);
3035 replace_size
= bdrv_getlength(to_replace_bs
);
3036 aio_context_release(replace_aio_context
);
3038 if (replace_size
< 0) {
3039 error_setg_errno(errp
, -replace_size
,
3040 "Failed to query the replacement node's size");
3043 if (bs_size
!= replace_size
) {
3044 error_setg(errp
, "cannot replace image with a mirror image of "
3050 /* pass the node name to replace to mirror start since it's loose coupling
3051 * and will allow to check whether the node still exist at mirror completion
3053 mirror_start(job_id
, bs
, target
,
3054 replaces
, job_flags
,
3055 speed
, granularity
, buf_size
, sync
, backing_mode
, zero_target
,
3056 on_source_error
, on_target_error
, unmap
, filter_node_name
,
3060 void qmp_drive_mirror(DriveMirror
*arg
, Error
**errp
)
3062 BlockDriverState
*bs
;
3063 BlockDriverState
*target_backing_bs
, *target_bs
;
3064 AioContext
*aio_context
;
3065 AioContext
*old_context
;
3066 BlockMirrorBackingMode backing_mode
;
3067 Error
*local_err
= NULL
;
3068 QDict
*options
= NULL
;
3071 const char *format
= arg
->format
;
3075 bs
= qmp_get_root_bs(arg
->device
, errp
);
3080 /* Early check to avoid creating target */
3081 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3085 aio_context
= bdrv_get_aio_context(bs
);
3086 aio_context_acquire(aio_context
);
3088 if (!arg
->has_mode
) {
3089 arg
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
3093 format
= (arg
->mode
== NEW_IMAGE_MODE_EXISTING
3094 ? NULL
: bs
->drv
->format_name
);
3097 flags
= bs
->open_flags
| BDRV_O_RDWR
;
3098 target_backing_bs
= bdrv_cow_bs(bdrv_skip_filters(bs
));
3099 if (!target_backing_bs
&& arg
->sync
== MIRROR_SYNC_MODE_TOP
) {
3100 arg
->sync
= MIRROR_SYNC_MODE_FULL
;
3102 if (arg
->sync
== MIRROR_SYNC_MODE_NONE
) {
3103 target_backing_bs
= bs
;
3106 size
= bdrv_getlength(bs
);
3108 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
3112 if (arg
->replaces
) {
3113 if (!arg
->node_name
) {
3114 error_setg(errp
, "a node-name must be provided when replacing a"
3115 " named node of the graph");
3120 if (arg
->mode
== NEW_IMAGE_MODE_ABSOLUTE_PATHS
) {
3121 backing_mode
= MIRROR_SOURCE_BACKING_CHAIN
;
3123 backing_mode
= MIRROR_OPEN_BACKING_CHAIN
;
3126 /* Don't open backing image in create() */
3127 flags
|= BDRV_O_NO_BACKING
;
3129 if ((arg
->sync
== MIRROR_SYNC_MODE_FULL
|| !target_backing_bs
)
3130 && arg
->mode
!= NEW_IMAGE_MODE_EXISTING
)
3132 /* create new image w/o backing file */
3134 bdrv_img_create(arg
->target
, format
,
3135 NULL
, NULL
, NULL
, size
, flags
, false, &local_err
);
3137 /* Implicit filters should not appear in the filename */
3138 BlockDriverState
*explicit_backing
=
3139 bdrv_skip_implicit_filters(target_backing_bs
);
3141 switch (arg
->mode
) {
3142 case NEW_IMAGE_MODE_EXISTING
:
3144 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
3145 /* create new image with backing file */
3146 bdrv_refresh_filename(explicit_backing
);
3147 bdrv_img_create(arg
->target
, format
,
3148 explicit_backing
->filename
,
3149 explicit_backing
->drv
->format_name
,
3150 NULL
, size
, flags
, false, &local_err
);
3158 error_propagate(errp
, local_err
);
3162 options
= qdict_new();
3163 if (arg
->node_name
) {
3164 qdict_put_str(options
, "node-name", arg
->node_name
);
3167 qdict_put_str(options
, "driver", format
);
3170 /* Mirroring takes care of copy-on-write using the source's backing
3173 target_bs
= bdrv_open(arg
->target
, NULL
, options
, flags
, errp
);
3178 zero_target
= (arg
->sync
== MIRROR_SYNC_MODE_FULL
&&
3179 (arg
->mode
== NEW_IMAGE_MODE_EXISTING
||
3180 !bdrv_has_zero_init(target_bs
)));
3183 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3184 old_context
= bdrv_get_aio_context(target_bs
);
3185 aio_context_release(aio_context
);
3186 aio_context_acquire(old_context
);
3188 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
3190 bdrv_unref(target_bs
);
3191 aio_context_release(old_context
);
3195 aio_context_release(old_context
);
3196 aio_context_acquire(aio_context
);
3198 blockdev_mirror_common(arg
->job_id
, bs
, target_bs
,
3199 arg
->replaces
, arg
->sync
,
3200 backing_mode
, zero_target
,
3201 arg
->has_speed
, arg
->speed
,
3202 arg
->has_granularity
, arg
->granularity
,
3203 arg
->has_buf_size
, arg
->buf_size
,
3204 arg
->has_on_source_error
, arg
->on_source_error
,
3205 arg
->has_on_target_error
, arg
->on_target_error
,
3206 arg
->has_unmap
, arg
->unmap
,
3208 arg
->has_copy_mode
, arg
->copy_mode
,
3209 arg
->has_auto_finalize
, arg
->auto_finalize
,
3210 arg
->has_auto_dismiss
, arg
->auto_dismiss
,
3212 bdrv_unref(target_bs
);
3214 aio_context_release(aio_context
);
3217 void qmp_blockdev_mirror(const char *job_id
,
3218 const char *device
, const char *target
,
3219 const char *replaces
,
3220 MirrorSyncMode sync
,
3221 bool has_speed
, int64_t speed
,
3222 bool has_granularity
, uint32_t granularity
,
3223 bool has_buf_size
, int64_t buf_size
,
3224 bool has_on_source_error
,
3225 BlockdevOnError on_source_error
,
3226 bool has_on_target_error
,
3227 BlockdevOnError on_target_error
,
3228 const char *filter_node_name
,
3229 bool has_copy_mode
, MirrorCopyMode copy_mode
,
3230 bool has_auto_finalize
, bool auto_finalize
,
3231 bool has_auto_dismiss
, bool auto_dismiss
,
3234 BlockDriverState
*bs
;
3235 BlockDriverState
*target_bs
;
3236 AioContext
*aio_context
;
3237 AioContext
*old_context
;
3238 BlockMirrorBackingMode backing_mode
= MIRROR_LEAVE_BACKING_CHAIN
;
3242 bs
= qmp_get_root_bs(device
, errp
);
3247 target_bs
= bdrv_lookup_bs(target
, target
, errp
);
3252 zero_target
= (sync
== MIRROR_SYNC_MODE_FULL
);
3254 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3255 old_context
= bdrv_get_aio_context(target_bs
);
3256 aio_context
= bdrv_get_aio_context(bs
);
3257 aio_context_acquire(old_context
);
3259 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
3261 aio_context_release(old_context
);
3262 aio_context_acquire(aio_context
);
3268 blockdev_mirror_common(job_id
, bs
, target_bs
,
3269 replaces
, sync
, backing_mode
,
3270 zero_target
, has_speed
, speed
,
3271 has_granularity
, granularity
,
3272 has_buf_size
, buf_size
,
3273 has_on_source_error
, on_source_error
,
3274 has_on_target_error
, on_target_error
,
3275 true, true, filter_node_name
,
3276 has_copy_mode
, copy_mode
,
3277 has_auto_finalize
, auto_finalize
,
3278 has_auto_dismiss
, auto_dismiss
,
3281 aio_context_release(aio_context
);
3285 * Get a block job using its ID. Called with job_mutex held.
3287 static BlockJob
*find_block_job_locked(const char *id
, Error
**errp
)
3293 job
= block_job_get_locked(id
);
3296 error_set(errp
, ERROR_CLASS_DEVICE_NOT_ACTIVE
,
3297 "Block job '%s' not found", id
);
3304 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
3309 job
= find_block_job_locked(device
, errp
);
3315 block_job_set_speed_locked(job
, speed
, errp
);
3318 void qmp_block_job_cancel(const char *device
,
3319 bool has_force
, bool force
, Error
**errp
)
3324 job
= find_block_job_locked(device
, errp
);
3334 if (job_user_paused_locked(&job
->job
) && !force
) {
3335 error_setg(errp
, "The block job for device '%s' is currently paused",
3340 trace_qmp_block_job_cancel(job
);
3341 job_user_cancel_locked(&job
->job
, force
, errp
);
3344 void qmp_block_job_pause(const char *device
, Error
**errp
)
3349 job
= find_block_job_locked(device
, errp
);
3355 trace_qmp_block_job_pause(job
);
3356 job_user_pause_locked(&job
->job
, errp
);
3359 void qmp_block_job_resume(const char *device
, Error
**errp
)
3364 job
= find_block_job_locked(device
, errp
);
3370 trace_qmp_block_job_resume(job
);
3371 job_user_resume_locked(&job
->job
, errp
);
3374 void qmp_block_job_complete(const char *device
, Error
**errp
)
3379 job
= find_block_job_locked(device
, errp
);
3385 trace_qmp_block_job_complete(job
);
3386 job_complete_locked(&job
->job
, errp
);
3389 void qmp_block_job_finalize(const char *id
, Error
**errp
)
3394 job
= find_block_job_locked(id
, errp
);
3400 trace_qmp_block_job_finalize(job
);
3401 job_ref_locked(&job
->job
);
3402 job_finalize_locked(&job
->job
, errp
);
3404 job_unref_locked(&job
->job
);
3407 void qmp_block_job_dismiss(const char *id
, Error
**errp
)
3413 bjob
= find_block_job_locked(id
, errp
);
3419 trace_qmp_block_job_dismiss(bjob
);
3421 job_dismiss_locked(&job
, errp
);
3424 void qmp_change_backing_file(const char *device
,
3425 const char *image_node_name
,
3426 const char *backing_file
,
3429 BlockDriverState
*bs
= NULL
;
3430 AioContext
*aio_context
;
3431 BlockDriverState
*image_bs
= NULL
;
3432 Error
*local_err
= NULL
;
3436 bs
= qmp_get_root_bs(device
, errp
);
3441 aio_context
= bdrv_get_aio_context(bs
);
3442 aio_context_acquire(aio_context
);
3444 image_bs
= bdrv_lookup_bs(NULL
, image_node_name
, &local_err
);
3446 error_propagate(errp
, local_err
);
3451 error_setg(errp
, "image file not found");
3455 if (bdrv_find_base(image_bs
) == image_bs
) {
3456 error_setg(errp
, "not allowing backing file change on an image "
3457 "without a backing file");
3461 /* even though we are not necessarily operating on bs, we need it to
3462 * determine if block ops are currently prohibited on the chain */
3463 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_CHANGE
, errp
)) {
3467 /* final sanity check */
3468 if (!bdrv_chain_contains(bs
, image_bs
)) {
3469 error_setg(errp
, "'%s' and image file are not in the same chain",
3474 /* if not r/w, reopen to make r/w */
3475 ro
= bdrv_is_read_only(image_bs
);
3478 if (bdrv_reopen_set_read_only(image_bs
, false, errp
) != 0) {
3483 ret
= bdrv_change_backing_file(image_bs
, backing_file
,
3484 image_bs
->drv
? image_bs
->drv
->format_name
: "",
3488 error_setg_errno(errp
, -ret
, "Could not change backing file to '%s'",
3490 /* don't exit here, so we can try to restore open flags if
3495 bdrv_reopen_set_read_only(image_bs
, true, errp
);
3499 aio_context_release(aio_context
);
3502 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
3504 BlockDriverState
*bs
;
3506 Visitor
*v
= qobject_output_visitor_new(&obj
);
3509 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3510 visit_complete(v
, &obj
);
3511 qdict
= qobject_to(QDict
, obj
);
3513 qdict_flatten(qdict
);
3515 if (!qdict_get_try_str(qdict
, "node-name")) {
3516 error_setg(errp
, "'node-name' must be specified for the root node");
3520 bs
= bds_tree_init(qdict
, errp
);
3525 bdrv_set_monitor_owned(bs
);
3531 void qmp_blockdev_reopen(BlockdevOptionsList
*reopen_list
, Error
**errp
)
3533 BlockReopenQueue
*queue
= NULL
;
3535 /* Add each one of the BDS that we want to reopen to the queue */
3536 for (; reopen_list
!= NULL
; reopen_list
= reopen_list
->next
) {
3537 BlockdevOptions
*options
= reopen_list
->value
;
3538 BlockDriverState
*bs
;
3544 /* Check for the selected node name */
3545 if (!options
->node_name
) {
3546 error_setg(errp
, "node-name not specified");
3550 bs
= bdrv_find_node(options
->node_name
);
3552 error_setg(errp
, "Failed to find node with node-name='%s'",
3553 options
->node_name
);
3557 /* Put all options in a QDict and flatten it */
3558 v
= qobject_output_visitor_new(&obj
);
3559 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3560 visit_complete(v
, &obj
);
3563 qdict
= qobject_to(QDict
, obj
);
3565 qdict_flatten(qdict
);
3567 ctx
= bdrv_get_aio_context(bs
);
3568 aio_context_acquire(ctx
);
3570 queue
= bdrv_reopen_queue(queue
, bs
, qdict
, false);
3572 aio_context_release(ctx
);
3575 /* Perform the reopen operation */
3576 bdrv_reopen_multiple(queue
, errp
);
3580 bdrv_reopen_queue_free(queue
);
3583 void qmp_blockdev_del(const char *node_name
, Error
**errp
)
3585 AioContext
*aio_context
;
3586 BlockDriverState
*bs
;
3588 GLOBAL_STATE_CODE();
3590 bs
= bdrv_find_node(node_name
);
3592 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3595 if (bdrv_has_blk(bs
)) {
3596 error_setg(errp
, "Node %s is in use", node_name
);
3599 aio_context
= bdrv_get_aio_context(bs
);
3600 aio_context_acquire(aio_context
);
3602 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_DRIVE_DEL
, errp
)) {
3606 if (!QTAILQ_IN_USE(bs
, monitor_list
)) {
3607 error_setg(errp
, "Node %s is not owned by the monitor",
3612 if (bs
->refcnt
> 1) {
3613 error_setg(errp
, "Block device %s is in use",
3614 bdrv_get_device_or_node_name(bs
));
3618 QTAILQ_REMOVE(&monitor_bdrv_states
, bs
, monitor_list
);
3622 aio_context_release(aio_context
);
3625 static BdrvChild
*bdrv_find_child(BlockDriverState
*parent_bs
,
3626 const char *child_name
)
3630 QLIST_FOREACH(child
, &parent_bs
->children
, next
) {
3631 if (strcmp(child
->name
, child_name
) == 0) {
3639 void qmp_x_blockdev_change(const char *parent
, const char *child
,
3640 const char *node
, Error
**errp
)
3642 BlockDriverState
*parent_bs
, *new_bs
= NULL
;
3645 parent_bs
= bdrv_lookup_bs(parent
, parent
, errp
);
3650 if (!child
== !node
) {
3652 error_setg(errp
, "The parameters child and node are in conflict");
3654 error_setg(errp
, "Either child or node must be specified");
3660 p_child
= bdrv_find_child(parent_bs
, child
);
3662 error_setg(errp
, "Node '%s' does not have child '%s'",
3666 bdrv_del_child(parent_bs
, p_child
, errp
);
3670 new_bs
= bdrv_find_node(node
);
3672 error_setg(errp
, "Node '%s' not found", node
);
3675 bdrv_add_child(parent_bs
, new_bs
, errp
);
3679 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
3681 BlockJobInfoList
*head
= NULL
, **tail
= &head
;
3686 for (job
= block_job_next_locked(NULL
); job
;
3687 job
= block_job_next_locked(job
)) {
3688 BlockJobInfo
*value
;
3690 if (block_job_is_internal(job
)) {
3693 value
= block_job_query_locked(job
, errp
);
3695 qapi_free_BlockJobInfoList(head
);
3698 QAPI_LIST_APPEND(tail
, value
);
3704 void qmp_x_blockdev_set_iothread(const char *node_name
, StrOrNull
*iothread
,
3705 bool has_force
, bool force
, Error
**errp
)
3707 AioContext
*old_context
;
3708 AioContext
*new_context
;
3709 BlockDriverState
*bs
;
3711 bs
= bdrv_find_node(node_name
);
3713 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3717 /* Protects against accidents. */
3718 if (!(has_force
&& force
) && bdrv_has_blk(bs
)) {
3719 error_setg(errp
, "Node %s is associated with a BlockBackend and could "
3720 "be in use (use force=true to override this check)",
3725 if (iothread
->type
== QTYPE_QSTRING
) {
3726 IOThread
*obj
= iothread_by_id(iothread
->u
.s
);
3728 error_setg(errp
, "Cannot find iothread %s", iothread
->u
.s
);
3732 new_context
= iothread_get_aio_context(obj
);
3734 new_context
= qemu_get_aio_context();
3737 old_context
= bdrv_get_aio_context(bs
);
3738 aio_context_acquire(old_context
);
3740 bdrv_try_change_aio_context(bs
, new_context
, NULL
, errp
);
3742 aio_context_release(old_context
);
3745 QemuOptsList qemu_common_drive_opts
= {
3747 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
3751 .type
= QEMU_OPT_BOOL
,
3752 .help
= "enable/disable snapshot mode",
3755 .type
= QEMU_OPT_STRING
,
3756 .help
= "host AIO implementation (threads, native, io_uring)",
3758 .name
= BDRV_OPT_CACHE_WB
,
3759 .type
= QEMU_OPT_BOOL
,
3760 .help
= "Enable writeback mode",
3763 .type
= QEMU_OPT_STRING
,
3764 .help
= "disk format (raw, qcow2, ...)",
3767 .type
= QEMU_OPT_STRING
,
3768 .help
= "read error action",
3771 .type
= QEMU_OPT_STRING
,
3772 .help
= "write error action",
3774 .name
= BDRV_OPT_READ_ONLY
,
3775 .type
= QEMU_OPT_BOOL
,
3776 .help
= "open drive file as read-only",
3782 .name
= "throttling.group",
3783 .type
= QEMU_OPT_STRING
,
3784 .help
= "name of the block throttling group",
3786 .name
= "copy-on-read",
3787 .type
= QEMU_OPT_BOOL
,
3788 .help
= "copy read data from backing file into image file",
3790 .name
= "detect-zeroes",
3791 .type
= QEMU_OPT_STRING
,
3792 .help
= "try to optimize zero writes (off, on, unmap)",
3794 .name
= "stats-account-invalid",
3795 .type
= QEMU_OPT_BOOL
,
3796 .help
= "whether to account for invalid I/O operations "
3797 "in the statistics",
3799 .name
= "stats-account-failed",
3800 .type
= QEMU_OPT_BOOL
,
3801 .help
= "whether to account for failed I/O operations "
3802 "in the statistics",
3804 { /* end of list */ }
3808 QemuOptsList qemu_drive_opts
= {
3810 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
3813 * no elements => accept any params
3814 * validation will happen later
3816 { /* end of list */ }