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
: {
345 const char *str
= qstring_get_str(qobject_to(QString
,
347 if (parse_uint_full(str
, 10, &length
) == 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
)
665 BlockDriverState
*bs
;
669 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
670 * with other callers) rather than what we want as the real defaults.
671 * Apply the defaults here instead. */
672 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_DIRECT
, "off");
673 qdict_set_default_str(bs_opts
, BDRV_OPT_CACHE_NO_FLUSH
, "off");
674 qdict_set_default_str(bs_opts
, BDRV_OPT_READ_ONLY
, "off");
676 if (runstate_check(RUN_STATE_INMIGRATE
)) {
677 bdrv_flags
|= BDRV_O_INACTIVE
;
680 aio_context_acquire(qemu_get_aio_context());
681 bs
= bdrv_open(NULL
, NULL
, bs_opts
, bdrv_flags
, errp
);
682 aio_context_release(qemu_get_aio_context());
687 void blockdev_close_all_bdrv_states(void)
689 BlockDriverState
*bs
, *next_bs
;
692 QTAILQ_FOREACH_SAFE(bs
, &monitor_bdrv_states
, monitor_list
, next_bs
) {
693 AioContext
*ctx
= bdrv_get_aio_context(bs
);
695 aio_context_acquire(ctx
);
697 aio_context_release(ctx
);
701 /* Iterates over the list of monitor-owned BlockDriverStates */
702 BlockDriverState
*bdrv_next_monitor_owned(BlockDriverState
*bs
)
705 return bs
? QTAILQ_NEXT(bs
, monitor_list
)
706 : QTAILQ_FIRST(&monitor_bdrv_states
);
709 static bool qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
,
714 value
= qemu_opt_get(opts
, from
);
716 if (qemu_opt_find(opts
, to
)) {
717 error_setg(errp
, "'%s' and its alias '%s' can't be used at the "
718 "same time", to
, from
);
723 /* rename all items in opts */
724 while ((value
= qemu_opt_get(opts
, from
))) {
725 qemu_opt_set(opts
, to
, value
, &error_abort
);
726 qemu_opt_unset(opts
, from
);
731 QemuOptsList qemu_legacy_drive_opts
= {
733 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
737 .type
= QEMU_OPT_NUMBER
,
738 .help
= "bus number",
741 .type
= QEMU_OPT_NUMBER
,
742 .help
= "unit number (i.e. lun for scsi)",
745 .type
= QEMU_OPT_NUMBER
,
746 .help
= "index number",
749 .type
= QEMU_OPT_STRING
,
750 .help
= "media type (disk, cdrom)",
753 .type
= QEMU_OPT_STRING
,
754 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
757 .type
= QEMU_OPT_STRING
,
761 /* Options that are passed on, but have special semantics with -drive */
763 .name
= BDRV_OPT_READ_ONLY
,
764 .type
= QEMU_OPT_BOOL
,
765 .help
= "open drive file as read-only",
768 .type
= QEMU_OPT_STRING
,
769 .help
= "read error action",
772 .type
= QEMU_OPT_STRING
,
773 .help
= "write error action",
775 .name
= "copy-on-read",
776 .type
= QEMU_OPT_BOOL
,
777 .help
= "copy read data from backing file into image file",
780 { /* end of list */ }
784 DriveInfo
*drive_new(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
,
789 DriveInfo
*dinfo
= NULL
;
791 QemuOpts
*legacy_opts
;
792 DriveMediaType media
= MEDIA_DISK
;
793 BlockInterfaceType type
;
794 int max_devs
, bus_id
, unit_id
, index
;
795 const char *werror
, *rerror
;
796 bool read_only
= false;
798 const char *filename
;
803 /* Change legacy command line options into QMP ones */
804 static const struct {
808 { "iops", "throttling.iops-total" },
809 { "iops_rd", "throttling.iops-read" },
810 { "iops_wr", "throttling.iops-write" },
812 { "bps", "throttling.bps-total" },
813 { "bps_rd", "throttling.bps-read" },
814 { "bps_wr", "throttling.bps-write" },
816 { "iops_max", "throttling.iops-total-max" },
817 { "iops_rd_max", "throttling.iops-read-max" },
818 { "iops_wr_max", "throttling.iops-write-max" },
820 { "bps_max", "throttling.bps-total-max" },
821 { "bps_rd_max", "throttling.bps-read-max" },
822 { "bps_wr_max", "throttling.bps-write-max" },
824 { "iops_size", "throttling.iops-size" },
826 { "group", "throttling.group" },
828 { "readonly", BDRV_OPT_READ_ONLY
},
831 for (i
= 0; i
< ARRAY_SIZE(opt_renames
); i
++) {
832 if (!qemu_opt_rename(all_opts
, opt_renames
[i
].from
,
833 opt_renames
[i
].to
, errp
)) {
838 value
= qemu_opt_get(all_opts
, "cache");
843 if (bdrv_parse_cache_mode(value
, &flags
, &writethrough
) != 0) {
844 error_setg(errp
, "invalid cache option");
848 /* Specific options take precedence */
849 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_WB
)) {
850 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_WB
,
851 !writethrough
, &error_abort
);
853 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_DIRECT
)) {
854 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_DIRECT
,
855 !!(flags
& BDRV_O_NOCACHE
), &error_abort
);
857 if (!qemu_opt_get(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
)) {
858 qemu_opt_set_bool(all_opts
, BDRV_OPT_CACHE_NO_FLUSH
,
859 !!(flags
& BDRV_O_NO_FLUSH
), &error_abort
);
861 qemu_opt_unset(all_opts
, "cache");
864 /* Get a QDict for processing the options */
865 bs_opts
= qdict_new();
866 qemu_opts_to_qdict(all_opts
, bs_opts
);
868 legacy_opts
= qemu_opts_create(&qemu_legacy_drive_opts
, NULL
, 0,
870 if (!qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, errp
)) {
875 value
= qemu_opt_get(legacy_opts
, "media");
877 if (!strcmp(value
, "disk")) {
879 } else if (!strcmp(value
, "cdrom")) {
883 error_setg(errp
, "'%s' invalid media", value
);
888 /* copy-on-read is disabled with a warning for read-only devices */
889 read_only
|= qemu_opt_get_bool(legacy_opts
, BDRV_OPT_READ_ONLY
, false);
890 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
892 if (read_only
&& copy_on_read
) {
893 warn_report("disabling copy-on-read on read-only drive");
894 copy_on_read
= false;
897 qdict_put_str(bs_opts
, BDRV_OPT_READ_ONLY
, read_only
? "on" : "off");
898 qdict_put_str(bs_opts
, "copy-on-read", copy_on_read
? "on" : "off");
900 /* Controller type */
901 value
= qemu_opt_get(legacy_opts
, "if");
904 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
907 if (type
== IF_COUNT
) {
908 error_setg(errp
, "unsupported bus type '%s'", value
);
912 type
= block_default_type
;
915 /* Device address specified by bus/unit or index.
916 * If none was specified, try to find the first free one. */
917 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
918 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
919 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
921 max_devs
= if_max_devs
[type
];
924 if (bus_id
!= 0 || unit_id
!= -1) {
925 error_setg(errp
, "index cannot be used with bus and unit");
928 bus_id
= drive_index_to_bus_id(type
, index
);
929 unit_id
= drive_index_to_unit_id(type
, index
);
934 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
936 if (max_devs
&& unit_id
>= max_devs
) {
943 if (max_devs
&& unit_id
>= max_devs
) {
944 error_setg(errp
, "unit %d too big (max is %d)", unit_id
, max_devs
- 1);
948 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
949 error_setg(errp
, "drive with bus=%d, unit=%d (index=%d) exists",
950 bus_id
, unit_id
, index
);
954 /* no id supplied -> create one */
955 if (qemu_opts_id(all_opts
) == NULL
) {
957 const char *mediastr
= "";
958 if (type
== IF_IDE
|| type
== IF_SCSI
) {
959 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
962 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
965 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
968 qdict_put_str(bs_opts
, "id", new_id
);
972 /* Add virtio block device */
973 if (type
== IF_VIRTIO
) {
975 devopts
= qemu_opts_create(qemu_find_opts("device"), NULL
, 0,
977 qemu_opt_set(devopts
, "driver", "virtio-blk", &error_abort
);
978 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"),
982 filename
= qemu_opt_get(legacy_opts
, "file");
984 /* Check werror/rerror compatibility with if=... */
985 werror
= qemu_opt_get(legacy_opts
, "werror");
986 if (werror
!= NULL
) {
987 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&&
989 error_setg(errp
, "werror is not supported by this bus type");
992 qdict_put_str(bs_opts
, "werror", werror
);
995 rerror
= qemu_opt_get(legacy_opts
, "rerror");
996 if (rerror
!= NULL
) {
997 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&&
999 error_setg(errp
, "rerror is not supported by this bus type");
1002 qdict_put_str(bs_opts
, "rerror", rerror
);
1005 /* Actual block device init: Functionality shared with blockdev-add */
1006 blk
= blockdev_init(filename
, bs_opts
, errp
);
1012 /* Create legacy DriveInfo */
1013 dinfo
= g_malloc0(sizeof(*dinfo
));
1014 dinfo
->opts
= all_opts
;
1017 dinfo
->bus
= bus_id
;
1018 dinfo
->unit
= unit_id
;
1020 blk_set_legacy_dinfo(blk
, dinfo
);
1027 dinfo
->media_cd
= media
== MEDIA_CDROM
;
1034 qemu_opts_del(legacy_opts
);
1035 qobject_unref(bs_opts
);
1039 static BlockDriverState
*qmp_get_root_bs(const char *name
, Error
**errp
)
1041 BlockDriverState
*bs
;
1042 AioContext
*aio_context
;
1044 GRAPH_RDLOCK_GUARD_MAINLOOP();
1046 bs
= bdrv_lookup_bs(name
, name
, errp
);
1051 if (!bdrv_is_root_node(bs
)) {
1052 error_setg(errp
, "Need a root block node");
1056 aio_context
= bdrv_get_aio_context(bs
);
1057 aio_context_acquire(aio_context
);
1059 if (!bdrv_is_inserted(bs
)) {
1060 error_setg(errp
, "Device has no medium");
1064 aio_context_release(aio_context
);
1069 static void blockdev_do_action(TransactionAction
*action
, Error
**errp
)
1071 TransactionActionList list
;
1073 list
.value
= action
;
1075 qmp_transaction(&list
, NULL
, errp
);
1078 void qmp_blockdev_snapshot_sync(const char *device
, const char *node_name
,
1079 const char *snapshot_file
,
1080 const char *snapshot_node_name
,
1082 bool has_mode
, NewImageMode mode
, Error
**errp
)
1084 BlockdevSnapshotSync snapshot
= {
1085 .device
= (char *) device
,
1086 .node_name
= (char *) node_name
,
1087 .snapshot_file
= (char *) snapshot_file
,
1088 .snapshot_node_name
= (char *) snapshot_node_name
,
1089 .format
= (char *) format
,
1090 .has_mode
= has_mode
,
1093 TransactionAction action
= {
1094 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
1095 .u
.blockdev_snapshot_sync
.data
= &snapshot
,
1097 blockdev_do_action(&action
, errp
);
1100 void qmp_blockdev_snapshot(const char *node
, const char *overlay
,
1103 BlockdevSnapshot snapshot_data
= {
1104 .node
= (char *) node
,
1105 .overlay
= (char *) overlay
1107 TransactionAction action
= {
1108 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
,
1109 .u
.blockdev_snapshot
.data
= &snapshot_data
,
1111 blockdev_do_action(&action
, errp
);
1114 void qmp_blockdev_snapshot_internal_sync(const char *device
,
1118 BlockdevSnapshotInternal snapshot
= {
1119 .device
= (char *) device
,
1120 .name
= (char *) name
1122 TransactionAction action
= {
1123 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
1124 .u
.blockdev_snapshot_internal_sync
.data
= &snapshot
,
1126 blockdev_do_action(&action
, errp
);
1129 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
1134 BlockDriverState
*bs
;
1135 AioContext
*aio_context
;
1136 QEMUSnapshotInfo sn
;
1137 Error
*local_err
= NULL
;
1138 SnapshotInfo
*info
= NULL
;
1141 GLOBAL_STATE_CODE();
1142 GRAPH_RDLOCK_GUARD_MAINLOOP();
1144 bs
= qmp_get_root_bs(device
, errp
);
1148 aio_context
= bdrv_get_aio_context(bs
);
1149 aio_context_acquire(aio_context
);
1152 error_setg(errp
, "Name or id must be provided");
1153 goto out_aio_context
;
1156 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE
, errp
)) {
1157 goto out_aio_context
;
1160 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1162 error_propagate(errp
, local_err
);
1163 goto out_aio_context
;
1167 "Snapshot with id '%s' and name '%s' does not exist on "
1169 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1170 goto out_aio_context
;
1173 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1175 error_propagate(errp
, local_err
);
1176 goto out_aio_context
;
1179 aio_context_release(aio_context
);
1181 info
= g_new0(SnapshotInfo
, 1);
1182 info
->id
= g_strdup(sn
.id_str
);
1183 info
->name
= g_strdup(sn
.name
);
1184 info
->date_nsec
= sn
.date_nsec
;
1185 info
->date_sec
= sn
.date_sec
;
1186 info
->vm_state_size
= sn
.vm_state_size
;
1187 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1188 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1189 if (sn
.icount
!= -1ULL) {
1190 info
->icount
= sn
.icount
;
1191 info
->has_icount
= true;
1197 aio_context_release(aio_context
);
1201 /* internal snapshot private data */
1202 typedef struct InternalSnapshotState
{
1203 BlockDriverState
*bs
;
1204 QEMUSnapshotInfo sn
;
1206 } InternalSnapshotState
;
1208 static void internal_snapshot_abort(void *opaque
);
1209 static void internal_snapshot_clean(void *opaque
);
1210 TransactionActionDrv internal_snapshot_drv
= {
1211 .abort
= internal_snapshot_abort
,
1212 .clean
= internal_snapshot_clean
,
1215 static void internal_snapshot_action(BlockdevSnapshotInternal
*internal
,
1216 Transaction
*tran
, Error
**errp
)
1218 Error
*local_err
= NULL
;
1221 BlockDriverState
*bs
;
1222 QEMUSnapshotInfo old_sn
, *sn
;
1225 InternalSnapshotState
*state
= g_new0(InternalSnapshotState
, 1);
1226 AioContext
*aio_context
;
1229 GLOBAL_STATE_CODE();
1230 GRAPH_RDLOCK_GUARD_MAINLOOP();
1232 tran_add(tran
, &internal_snapshot_drv
, state
);
1234 device
= internal
->device
;
1235 name
= internal
->name
;
1237 bs
= qmp_get_root_bs(device
, errp
);
1242 aio_context
= bdrv_get_aio_context(bs
);
1243 aio_context_acquire(aio_context
);
1247 /* Paired with .clean() */
1248 bdrv_drained_begin(bs
);
1250 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT
, errp
)) {
1254 if (bdrv_is_read_only(bs
)) {
1255 error_setg(errp
, "Device '%s' is read only", device
);
1259 if (!bdrv_can_snapshot(bs
)) {
1260 error_setg(errp
, "Block format '%s' used by device '%s' "
1261 "does not support internal snapshots",
1262 bs
->drv
->format_name
, device
);
1266 if (!strlen(name
)) {
1267 error_setg(errp
, "Name is empty");
1271 /* check whether a snapshot with name exist */
1272 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
,
1275 error_propagate(errp
, local_err
);
1279 "Snapshot with name '%s' already exists on device '%s'",
1284 /* 3. take the snapshot */
1286 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1287 rt
= g_get_real_time();
1288 sn
->date_sec
= rt
/ G_USEC_PER_SEC
;
1289 sn
->date_nsec
= (rt
% G_USEC_PER_SEC
) * 1000;
1290 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1291 if (replay_mode
!= REPLAY_MODE_NONE
) {
1292 sn
->icount
= replay_get_current_icount();
1297 ret1
= bdrv_snapshot_create(bs
, sn
);
1299 error_setg_errno(errp
, -ret1
,
1300 "Failed to create snapshot '%s' on device '%s'",
1305 /* 4. succeed, mark a snapshot is created */
1306 state
->created
= true;
1309 aio_context_release(aio_context
);
1312 static void internal_snapshot_abort(void *opaque
)
1314 InternalSnapshotState
*state
= opaque
;
1315 BlockDriverState
*bs
= state
->bs
;
1316 QEMUSnapshotInfo
*sn
= &state
->sn
;
1317 AioContext
*aio_context
;
1318 Error
*local_error
= NULL
;
1320 GLOBAL_STATE_CODE();
1321 GRAPH_RDLOCK_GUARD_MAINLOOP();
1323 if (!state
->created
) {
1327 aio_context
= bdrv_get_aio_context(state
->bs
);
1328 aio_context_acquire(aio_context
);
1330 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1331 error_reportf_err(local_error
,
1332 "Failed to delete snapshot with id '%s' and "
1333 "name '%s' on device '%s' in abort: ",
1334 sn
->id_str
, sn
->name
,
1335 bdrv_get_device_name(bs
));
1338 aio_context_release(aio_context
);
1341 static void internal_snapshot_clean(void *opaque
)
1343 g_autofree InternalSnapshotState
*state
= opaque
;
1344 AioContext
*aio_context
;
1350 aio_context
= bdrv_get_aio_context(state
->bs
);
1351 aio_context_acquire(aio_context
);
1353 bdrv_drained_end(state
->bs
);
1355 aio_context_release(aio_context
);
1358 /* external snapshot private data */
1359 typedef struct ExternalSnapshotState
{
1360 BlockDriverState
*old_bs
;
1361 BlockDriverState
*new_bs
;
1362 bool overlay_appended
;
1363 } ExternalSnapshotState
;
1365 static void external_snapshot_commit(void *opaque
);
1366 static void external_snapshot_abort(void *opaque
);
1367 static void external_snapshot_clean(void *opaque
);
1368 TransactionActionDrv external_snapshot_drv
= {
1369 .commit
= external_snapshot_commit
,
1370 .abort
= external_snapshot_abort
,
1371 .clean
= external_snapshot_clean
,
1374 static void external_snapshot_action(TransactionAction
*action
,
1375 Transaction
*tran
, Error
**errp
)
1379 QDict
*options
= NULL
;
1380 Error
*local_err
= NULL
;
1381 /* Device and node name of the image to generate the snapshot from */
1383 const char *node_name
;
1384 /* Reference to the new image (for 'blockdev-snapshot') */
1385 const char *snapshot_ref
;
1386 /* File name of the new image (for 'blockdev-snapshot-sync') */
1387 const char *new_image_file
;
1388 ExternalSnapshotState
*state
= g_new0(ExternalSnapshotState
, 1);
1389 AioContext
*aio_context
;
1390 uint64_t perm
, shared
;
1392 /* TODO We'll eventually have to take a writer lock in this function */
1393 GRAPH_RDLOCK_GUARD_MAINLOOP();
1395 tran_add(tran
, &external_snapshot_drv
, state
);
1397 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1398 * purpose but a different set of parameters */
1399 switch (action
->type
) {
1400 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
:
1402 BlockdevSnapshot
*s
= action
->u
.blockdev_snapshot
.data
;
1404 node_name
= s
->node
;
1405 new_image_file
= NULL
;
1406 snapshot_ref
= s
->overlay
;
1409 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
:
1411 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1413 node_name
= s
->node_name
;
1414 new_image_file
= s
->snapshot_file
;
1415 snapshot_ref
= NULL
;
1419 g_assert_not_reached();
1422 /* start processing */
1424 state
->old_bs
= bdrv_lookup_bs(device
, node_name
, errp
);
1425 if (!state
->old_bs
) {
1429 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1430 aio_context_acquire(aio_context
);
1432 /* Paired with .clean() */
1433 bdrv_drained_begin(state
->old_bs
);
1435 if (!bdrv_is_inserted(state
->old_bs
)) {
1436 error_setg(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1440 if (bdrv_op_is_blocked(state
->old_bs
,
1441 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT
, errp
)) {
1445 if (!bdrv_is_read_only(state
->old_bs
)) {
1446 if (bdrv_flush(state
->old_bs
)) {
1447 error_setg(errp
, QERR_IO_ERROR
);
1452 if (action
->type
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
) {
1453 BlockdevSnapshotSync
*s
= action
->u
.blockdev_snapshot_sync
.data
;
1454 const char *format
= s
->format
?: "qcow2";
1455 enum NewImageMode mode
;
1456 const char *snapshot_node_name
= s
->snapshot_node_name
;
1458 if (node_name
&& !snapshot_node_name
) {
1459 error_setg(errp
, "New overlay node-name missing");
1463 if (snapshot_node_name
&&
1464 bdrv_lookup_bs(snapshot_node_name
, snapshot_node_name
, NULL
)) {
1465 error_setg(errp
, "New overlay node-name already in use");
1469 flags
= state
->old_bs
->open_flags
;
1470 flags
&= ~(BDRV_O_SNAPSHOT
| BDRV_O_COPY_ON_READ
);
1471 flags
|= BDRV_O_NO_BACKING
;
1473 /* create new image w/backing file */
1474 mode
= s
->has_mode
? s
->mode
: NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1475 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1476 int64_t size
= bdrv_getlength(state
->old_bs
);
1478 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1481 bdrv_refresh_filename(state
->old_bs
);
1483 aio_context_release(aio_context
);
1484 bdrv_img_create(new_image_file
, format
,
1485 state
->old_bs
->filename
,
1486 state
->old_bs
->drv
->format_name
,
1487 NULL
, size
, flags
, false, &local_err
);
1488 aio_context_acquire(aio_context
);
1491 error_propagate(errp
, local_err
);
1496 options
= qdict_new();
1497 if (snapshot_node_name
) {
1498 qdict_put_str(options
, "node-name", snapshot_node_name
);
1500 qdict_put_str(options
, "driver", format
);
1502 aio_context_release(aio_context
);
1504 aio_context_acquire(qemu_get_aio_context());
1505 state
->new_bs
= bdrv_open(new_image_file
, snapshot_ref
, options
, flags
,
1507 aio_context_release(qemu_get_aio_context());
1509 /* We will manually add the backing_hd field to the bs later */
1510 if (!state
->new_bs
) {
1514 aio_context_acquire(aio_context
);
1517 * Allow attaching a backing file to an overlay that's already in use only
1518 * if the parents don't assume that they are already seeing a valid image.
1519 * (Specifically, allow it as a mirror target, which is write-only access.)
1521 bdrv_get_cumulative_perm(state
->new_bs
, &perm
, &shared
);
1522 if (perm
& BLK_PERM_CONSISTENT_READ
) {
1523 error_setg(errp
, "The overlay is already in use");
1527 if (state
->new_bs
->drv
->is_filter
) {
1528 error_setg(errp
, "Filters cannot be used as overlays");
1532 if (bdrv_cow_child(state
->new_bs
)) {
1533 error_setg(errp
, "The overlay already has a backing image");
1537 if (!state
->new_bs
->drv
->supports_backing
) {
1538 error_setg(errp
, "The overlay does not support backing images");
1542 ret
= bdrv_append(state
->new_bs
, state
->old_bs
, errp
);
1546 state
->overlay_appended
= true;
1549 aio_context_release(aio_context
);
1552 static void external_snapshot_commit(void *opaque
)
1554 ExternalSnapshotState
*state
= opaque
;
1555 AioContext
*aio_context
;
1557 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1558 aio_context_acquire(aio_context
);
1560 /* We don't need (or want) to use the transactional
1561 * bdrv_reopen_multiple() across all the entries at once, because we
1562 * don't want to abort all of them if one of them fails the reopen */
1563 if (!qatomic_read(&state
->old_bs
->copy_on_read
)) {
1564 bdrv_reopen_set_read_only(state
->old_bs
, true, NULL
);
1567 aio_context_release(aio_context
);
1570 static void external_snapshot_abort(void *opaque
)
1572 ExternalSnapshotState
*state
= opaque
;
1573 if (state
->new_bs
) {
1574 if (state
->overlay_appended
) {
1575 AioContext
*aio_context
;
1576 AioContext
*tmp_context
;
1579 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1580 aio_context_acquire(aio_context
);
1582 bdrv_ref(state
->old_bs
); /* we can't let bdrv_set_backind_hd()
1583 close state->old_bs; we need it */
1584 bdrv_set_backing_hd(state
->new_bs
, NULL
, &error_abort
);
1587 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1588 * the main AioContext. As we're still going to be using it, return
1589 * it to the AioContext it was before.
1591 tmp_context
= bdrv_get_aio_context(state
->old_bs
);
1592 if (aio_context
!= tmp_context
) {
1593 aio_context_release(aio_context
);
1594 aio_context_acquire(tmp_context
);
1596 ret
= bdrv_try_change_aio_context(state
->old_bs
,
1597 aio_context
, NULL
, NULL
);
1600 aio_context_release(tmp_context
);
1601 aio_context_acquire(aio_context
);
1604 bdrv_replace_node(state
->new_bs
, state
->old_bs
, &error_abort
);
1605 bdrv_unref(state
->old_bs
); /* bdrv_replace_node() ref'ed old_bs */
1607 aio_context_release(aio_context
);
1612 static void external_snapshot_clean(void *opaque
)
1614 g_autofree ExternalSnapshotState
*state
= opaque
;
1615 AioContext
*aio_context
;
1617 if (!state
->old_bs
) {
1621 aio_context
= bdrv_get_aio_context(state
->old_bs
);
1622 aio_context_acquire(aio_context
);
1624 bdrv_drained_end(state
->old_bs
);
1625 bdrv_unref(state
->new_bs
);
1627 aio_context_release(aio_context
);
1630 typedef struct DriveBackupState
{
1631 BlockDriverState
*bs
;
1635 static BlockJob
*do_backup_common(BackupCommon
*backup
,
1636 BlockDriverState
*bs
,
1637 BlockDriverState
*target_bs
,
1638 AioContext
*aio_context
,
1639 JobTxn
*txn
, Error
**errp
);
1641 static void drive_backup_commit(void *opaque
);
1642 static void drive_backup_abort(void *opaque
);
1643 static void drive_backup_clean(void *opaque
);
1644 TransactionActionDrv drive_backup_drv
= {
1645 .commit
= drive_backup_commit
,
1646 .abort
= drive_backup_abort
,
1647 .clean
= drive_backup_clean
,
1650 static void drive_backup_action(DriveBackup
*backup
,
1651 JobTxn
*block_job_txn
,
1652 Transaction
*tran
, Error
**errp
)
1654 DriveBackupState
*state
= g_new0(DriveBackupState
, 1);
1655 BlockDriverState
*bs
;
1656 BlockDriverState
*target_bs
;
1657 BlockDriverState
*source
= NULL
;
1658 AioContext
*aio_context
;
1659 AioContext
*old_context
;
1662 Error
*local_err
= NULL
;
1665 bool set_backing_hd
= false;
1668 tran_add(tran
, &drive_backup_drv
, state
);
1670 if (!backup
->has_mode
) {
1671 backup
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1674 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1680 error_setg(errp
, "Device has no medium");
1684 aio_context
= bdrv_get_aio_context(bs
);
1685 aio_context_acquire(aio_context
);
1688 /* Paired with .clean() */
1689 bdrv_drained_begin(bs
);
1691 format
= backup
->format
;
1692 if (!format
&& backup
->mode
!= NEW_IMAGE_MODE_EXISTING
) {
1693 format
= bs
->drv
->format_name
;
1696 /* Early check to avoid creating target */
1697 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_BACKUP_SOURCE
, errp
)) {
1701 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1704 * See if we have a backing HD we can use to create our new image
1707 if (backup
->sync
== MIRROR_SYNC_MODE_TOP
) {
1709 * Backup will not replace the source by the target, so none
1710 * of the filters skipped here will be removed (in contrast to
1711 * mirror). Therefore, we can skip all of them when looking
1712 * for the first COW relationship.
1714 source
= bdrv_cow_bs(bdrv_skip_filters(bs
));
1716 backup
->sync
= MIRROR_SYNC_MODE_FULL
;
1719 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
1721 flags
|= BDRV_O_NO_BACKING
;
1722 set_backing_hd
= true;
1725 size
= bdrv_getlength(bs
);
1727 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1731 if (backup
->mode
!= NEW_IMAGE_MODE_EXISTING
) {
1734 /* Implicit filters should not appear in the filename */
1735 BlockDriverState
*explicit_backing
=
1736 bdrv_skip_implicit_filters(source
);
1738 bdrv_refresh_filename(explicit_backing
);
1739 bdrv_img_create(backup
->target
, format
,
1740 explicit_backing
->filename
,
1741 explicit_backing
->drv
->format_name
, NULL
,
1742 size
, flags
, false, &local_err
);
1744 bdrv_img_create(backup
->target
, format
, NULL
, NULL
, NULL
,
1745 size
, flags
, false, &local_err
);
1750 error_propagate(errp
, local_err
);
1754 options
= qdict_new();
1755 qdict_put_str(options
, "discard", "unmap");
1756 qdict_put_str(options
, "detect-zeroes", "unmap");
1758 qdict_put_str(options
, "driver", format
);
1760 aio_context_release(aio_context
);
1762 aio_context_acquire(qemu_get_aio_context());
1763 target_bs
= bdrv_open(backup
->target
, NULL
, options
, flags
, errp
);
1764 aio_context_release(qemu_get_aio_context());
1770 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1771 old_context
= bdrv_get_aio_context(target_bs
);
1772 aio_context_acquire(old_context
);
1774 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
1776 bdrv_unref(target_bs
);
1777 aio_context_release(old_context
);
1781 aio_context_release(old_context
);
1782 aio_context_acquire(aio_context
);
1784 if (set_backing_hd
) {
1785 if (bdrv_set_backing_hd(target_bs
, source
, errp
) < 0) {
1790 state
->job
= do_backup_common(qapi_DriveBackup_base(backup
),
1791 bs
, target_bs
, aio_context
,
1792 block_job_txn
, errp
);
1795 bdrv_unref(target_bs
);
1797 aio_context_release(aio_context
);
1800 static void drive_backup_commit(void *opaque
)
1802 DriveBackupState
*state
= opaque
;
1803 AioContext
*aio_context
;
1805 aio_context
= bdrv_get_aio_context(state
->bs
);
1806 aio_context_acquire(aio_context
);
1809 job_start(&state
->job
->job
);
1811 aio_context_release(aio_context
);
1814 static void drive_backup_abort(void *opaque
)
1816 DriveBackupState
*state
= opaque
;
1819 job_cancel_sync(&state
->job
->job
, true);
1823 static void drive_backup_clean(void *opaque
)
1825 g_autofree DriveBackupState
*state
= opaque
;
1826 AioContext
*aio_context
;
1832 aio_context
= bdrv_get_aio_context(state
->bs
);
1833 aio_context_acquire(aio_context
);
1835 bdrv_drained_end(state
->bs
);
1837 aio_context_release(aio_context
);
1840 typedef struct BlockdevBackupState
{
1841 BlockDriverState
*bs
;
1843 } BlockdevBackupState
;
1845 static void blockdev_backup_commit(void *opaque
);
1846 static void blockdev_backup_abort(void *opaque
);
1847 static void blockdev_backup_clean(void *opaque
);
1848 TransactionActionDrv blockdev_backup_drv
= {
1849 .commit
= blockdev_backup_commit
,
1850 .abort
= blockdev_backup_abort
,
1851 .clean
= blockdev_backup_clean
,
1854 static void blockdev_backup_action(BlockdevBackup
*backup
,
1855 JobTxn
*block_job_txn
,
1856 Transaction
*tran
, Error
**errp
)
1858 BlockdevBackupState
*state
= g_new0(BlockdevBackupState
, 1);
1859 BlockDriverState
*bs
;
1860 BlockDriverState
*target_bs
;
1861 AioContext
*aio_context
;
1862 AioContext
*old_context
;
1865 tran_add(tran
, &blockdev_backup_drv
, state
);
1867 bs
= bdrv_lookup_bs(backup
->device
, backup
->device
, errp
);
1872 target_bs
= bdrv_lookup_bs(backup
->target
, backup
->target
, errp
);
1877 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1878 aio_context
= bdrv_get_aio_context(bs
);
1879 old_context
= bdrv_get_aio_context(target_bs
);
1880 aio_context_acquire(old_context
);
1882 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
1884 aio_context_release(old_context
);
1888 aio_context_release(old_context
);
1889 aio_context_acquire(aio_context
);
1892 /* Paired with .clean() */
1893 bdrv_drained_begin(state
->bs
);
1895 state
->job
= do_backup_common(qapi_BlockdevBackup_base(backup
),
1896 bs
, target_bs
, aio_context
,
1897 block_job_txn
, errp
);
1899 aio_context_release(aio_context
);
1902 static void blockdev_backup_commit(void *opaque
)
1904 BlockdevBackupState
*state
= opaque
;
1905 AioContext
*aio_context
;
1907 aio_context
= bdrv_get_aio_context(state
->bs
);
1908 aio_context_acquire(aio_context
);
1911 job_start(&state
->job
->job
);
1913 aio_context_release(aio_context
);
1916 static void blockdev_backup_abort(void *opaque
)
1918 BlockdevBackupState
*state
= opaque
;
1921 job_cancel_sync(&state
->job
->job
, true);
1925 static void blockdev_backup_clean(void *opaque
)
1927 g_autofree BlockdevBackupState
*state
= opaque
;
1928 AioContext
*aio_context
;
1934 aio_context
= bdrv_get_aio_context(state
->bs
);
1935 aio_context_acquire(aio_context
);
1937 bdrv_drained_end(state
->bs
);
1939 aio_context_release(aio_context
);
1942 typedef struct BlockDirtyBitmapState
{
1943 BdrvDirtyBitmap
*bitmap
;
1944 BlockDriverState
*bs
;
1947 } BlockDirtyBitmapState
;
1949 static void block_dirty_bitmap_add_abort(void *opaque
);
1950 TransactionActionDrv block_dirty_bitmap_add_drv
= {
1951 .abort
= block_dirty_bitmap_add_abort
,
1955 static void block_dirty_bitmap_add_action(BlockDirtyBitmapAdd
*action
,
1956 Transaction
*tran
, Error
**errp
)
1958 Error
*local_err
= NULL
;
1959 BlockDirtyBitmapState
*state
= g_new0(BlockDirtyBitmapState
, 1);
1961 tran_add(tran
, &block_dirty_bitmap_add_drv
, state
);
1963 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1964 qmp_block_dirty_bitmap_add(action
->node
, action
->name
,
1965 action
->has_granularity
, action
->granularity
,
1966 action
->has_persistent
, action
->persistent
,
1967 action
->has_disabled
, action
->disabled
,
1971 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
, action
->name
,
1972 NULL
, &error_abort
);
1974 error_propagate(errp
, local_err
);
1978 static void block_dirty_bitmap_add_abort(void *opaque
)
1980 BlockDirtyBitmapState
*state
= opaque
;
1982 if (state
->bitmap
) {
1983 bdrv_release_dirty_bitmap(state
->bitmap
);
1987 static void block_dirty_bitmap_restore(void *opaque
);
1988 static void block_dirty_bitmap_free_backup(void *opaque
);
1989 TransactionActionDrv block_dirty_bitmap_clear_drv
= {
1990 .abort
= block_dirty_bitmap_restore
,
1991 .commit
= block_dirty_bitmap_free_backup
,
1995 static void block_dirty_bitmap_clear_action(BlockDirtyBitmap
*action
,
1996 Transaction
*tran
, Error
**errp
)
1998 BlockDirtyBitmapState
*state
= g_new0(BlockDirtyBitmapState
, 1);
2000 tran_add(tran
, &block_dirty_bitmap_clear_drv
, state
);
2002 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2006 if (!state
->bitmap
) {
2010 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_DEFAULT
, errp
)) {
2014 bdrv_clear_dirty_bitmap(state
->bitmap
, &state
->backup
);
2017 static void block_dirty_bitmap_restore(void *opaque
)
2019 BlockDirtyBitmapState
*state
= opaque
;
2021 if (state
->backup
) {
2022 bdrv_restore_dirty_bitmap(state
->bitmap
, state
->backup
);
2026 static void block_dirty_bitmap_free_backup(void *opaque
)
2028 BlockDirtyBitmapState
*state
= opaque
;
2030 hbitmap_free(state
->backup
);
2033 static void block_dirty_bitmap_enable_abort(void *opaque
);
2034 TransactionActionDrv block_dirty_bitmap_enable_drv
= {
2035 .abort
= block_dirty_bitmap_enable_abort
,
2039 static void block_dirty_bitmap_enable_action(BlockDirtyBitmap
*action
,
2040 Transaction
*tran
, Error
**errp
)
2042 BlockDirtyBitmapState
*state
= g_new0(BlockDirtyBitmapState
, 1);
2044 tran_add(tran
, &block_dirty_bitmap_enable_drv
, state
);
2046 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2050 if (!state
->bitmap
) {
2054 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2058 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2059 bdrv_enable_dirty_bitmap(state
->bitmap
);
2062 static void block_dirty_bitmap_enable_abort(void *opaque
)
2064 BlockDirtyBitmapState
*state
= opaque
;
2066 if (!state
->was_enabled
) {
2067 bdrv_disable_dirty_bitmap(state
->bitmap
);
2071 static void block_dirty_bitmap_disable_abort(void *opaque
);
2072 TransactionActionDrv block_dirty_bitmap_disable_drv
= {
2073 .abort
= block_dirty_bitmap_disable_abort
,
2077 static void block_dirty_bitmap_disable_action(BlockDirtyBitmap
*action
,
2078 Transaction
*tran
, Error
**errp
)
2080 BlockDirtyBitmapState
*state
= g_new0(BlockDirtyBitmapState
, 1);
2082 tran_add(tran
, &block_dirty_bitmap_disable_drv
, state
);
2084 state
->bitmap
= block_dirty_bitmap_lookup(action
->node
,
2088 if (!state
->bitmap
) {
2092 if (bdrv_dirty_bitmap_check(state
->bitmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2096 state
->was_enabled
= bdrv_dirty_bitmap_enabled(state
->bitmap
);
2097 bdrv_disable_dirty_bitmap(state
->bitmap
);
2100 static void block_dirty_bitmap_disable_abort(void *opaque
)
2102 BlockDirtyBitmapState
*state
= opaque
;
2104 if (state
->was_enabled
) {
2105 bdrv_enable_dirty_bitmap(state
->bitmap
);
2109 TransactionActionDrv block_dirty_bitmap_merge_drv
= {
2110 .commit
= block_dirty_bitmap_free_backup
,
2111 .abort
= block_dirty_bitmap_restore
,
2115 static void block_dirty_bitmap_merge_action(BlockDirtyBitmapMerge
*action
,
2116 Transaction
*tran
, Error
**errp
)
2118 BlockDirtyBitmapState
*state
= g_new0(BlockDirtyBitmapState
, 1);
2120 tran_add(tran
, &block_dirty_bitmap_merge_drv
, state
);
2122 state
->bitmap
= block_dirty_bitmap_merge(action
->node
, action
->target
,
2123 action
->bitmaps
, &state
->backup
,
2127 static void block_dirty_bitmap_remove_commit(void *opaque
);
2128 static void block_dirty_bitmap_remove_abort(void *opaque
);
2129 TransactionActionDrv block_dirty_bitmap_remove_drv
= {
2130 .commit
= block_dirty_bitmap_remove_commit
,
2131 .abort
= block_dirty_bitmap_remove_abort
,
2135 static void block_dirty_bitmap_remove_action(BlockDirtyBitmap
*action
,
2136 Transaction
*tran
, Error
**errp
)
2138 BlockDirtyBitmapState
*state
= g_new0(BlockDirtyBitmapState
, 1);
2140 tran_add(tran
, &block_dirty_bitmap_remove_drv
, state
);
2143 state
->bitmap
= block_dirty_bitmap_remove(action
->node
, action
->name
,
2144 false, &state
->bs
, errp
);
2145 if (state
->bitmap
) {
2146 bdrv_dirty_bitmap_skip_store(state
->bitmap
, true);
2147 bdrv_dirty_bitmap_set_busy(state
->bitmap
, true);
2151 static void block_dirty_bitmap_remove_abort(void *opaque
)
2153 BlockDirtyBitmapState
*state
= opaque
;
2155 if (state
->bitmap
) {
2156 bdrv_dirty_bitmap_skip_store(state
->bitmap
, false);
2157 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2161 static void block_dirty_bitmap_remove_commit(void *opaque
)
2163 BlockDirtyBitmapState
*state
= opaque
;
2165 bdrv_dirty_bitmap_set_busy(state
->bitmap
, false);
2166 bdrv_release_dirty_bitmap(state
->bitmap
);
2169 static void abort_commit(void *opaque
);
2170 TransactionActionDrv abort_drv
= {
2171 .commit
= abort_commit
,
2174 static void abort_action(Transaction
*tran
, Error
**errp
)
2176 tran_add(tran
, &abort_drv
, NULL
);
2177 error_setg(errp
, "Transaction aborted using Abort action");
2180 static void abort_commit(void *opaque
)
2182 g_assert_not_reached(); /* this action never succeeds */
2185 static void transaction_action(TransactionAction
*act
, JobTxn
*block_job_txn
,
2186 Transaction
*tran
, Error
**errp
)
2188 switch (act
->type
) {
2189 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT
:
2190 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
:
2191 external_snapshot_action(act
, tran
, errp
);
2193 case TRANSACTION_ACTION_KIND_DRIVE_BACKUP
:
2194 drive_backup_action(act
->u
.drive_backup
.data
,
2195 block_job_txn
, tran
, errp
);
2197 case TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
:
2198 blockdev_backup_action(act
->u
.blockdev_backup
.data
,
2199 block_job_txn
, tran
, errp
);
2201 case TRANSACTION_ACTION_KIND_ABORT
:
2202 abort_action(tran
, errp
);
2204 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
:
2205 internal_snapshot_action(act
->u
.blockdev_snapshot_internal_sync
.data
,
2208 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD
:
2209 block_dirty_bitmap_add_action(act
->u
.block_dirty_bitmap_add
.data
,
2212 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR
:
2213 block_dirty_bitmap_clear_action(act
->u
.block_dirty_bitmap_clear
.data
,
2216 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE
:
2217 block_dirty_bitmap_enable_action(act
->u
.block_dirty_bitmap_enable
.data
,
2220 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE
:
2221 block_dirty_bitmap_disable_action(
2222 act
->u
.block_dirty_bitmap_disable
.data
, tran
, errp
);
2224 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE
:
2225 block_dirty_bitmap_merge_action(act
->u
.block_dirty_bitmap_merge
.data
,
2228 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE
:
2229 block_dirty_bitmap_remove_action(act
->u
.block_dirty_bitmap_remove
.data
,
2233 * Where are transactions for MIRROR, COMMIT and STREAM?
2234 * Although these blockjobs use transaction callbacks like the backup job,
2235 * these jobs do not necessarily adhere to transaction semantics.
2236 * These jobs may not fully undo all of their actions on abort, nor do they
2237 * necessarily work in transactions with more than one job in them.
2239 case TRANSACTION_ACTION_KIND__MAX
:
2241 g_assert_not_reached();
2247 * 'Atomic' group operations. The operations are performed as a set, and if
2248 * any fail then we roll back all operations in the group.
2250 * Always run under BQL.
2252 void qmp_transaction(TransactionActionList
*actions
,
2253 struct TransactionProperties
*properties
,
2256 TransactionActionList
*act
;
2257 JobTxn
*block_job_txn
= NULL
;
2258 Error
*local_err
= NULL
;
2260 ActionCompletionMode comp_mode
=
2261 properties
? properties
->completion_mode
:
2262 ACTION_COMPLETION_MODE_INDIVIDUAL
;
2264 GLOBAL_STATE_CODE();
2266 /* Does this transaction get canceled as a group on failure?
2267 * If not, we don't really need to make a JobTxn.
2269 if (comp_mode
!= ACTION_COMPLETION_MODE_INDIVIDUAL
) {
2270 for (act
= actions
; act
; act
= act
->next
) {
2271 TransactionActionKind type
= act
->value
->type
;
2273 if (type
!= TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
&&
2274 type
!= TRANSACTION_ACTION_KIND_DRIVE_BACKUP
)
2277 "Action '%s' does not support transaction property "
2278 "completion-mode = %s",
2279 TransactionActionKind_str(type
),
2280 ActionCompletionMode_str(comp_mode
));
2285 block_job_txn
= job_txn_new();
2288 /* drain all i/o before any operations */
2293 /* We don't do anything in this loop that commits us to the operations */
2294 for (act
= actions
; act
; act
= act
->next
) {
2295 transaction_action(act
->value
, block_job_txn
, tran
, &local_err
);
2297 error_propagate(errp
, local_err
);
2298 goto delete_and_fail
;
2308 /* failure, and it is all-or-none; roll back all operations */
2311 job_txn_unref(block_job_txn
);
2314 BlockDirtyBitmapSha256
*qmp_x_debug_block_dirty_bitmap_sha256(const char *node
,
2318 BdrvDirtyBitmap
*bitmap
;
2319 BlockDriverState
*bs
;
2320 BlockDirtyBitmapSha256
*ret
= NULL
;
2323 bitmap
= block_dirty_bitmap_lookup(node
, name
, &bs
, errp
);
2324 if (!bitmap
|| !bs
) {
2328 sha256
= bdrv_dirty_bitmap_sha256(bitmap
, errp
);
2329 if (sha256
== NULL
) {
2333 ret
= g_new(BlockDirtyBitmapSha256
, 1);
2334 ret
->sha256
= sha256
;
2339 void coroutine_fn
qmp_block_resize(const char *device
, const char *node_name
,
2340 int64_t size
, Error
**errp
)
2342 Error
*local_err
= NULL
;
2344 BlockDriverState
*bs
;
2345 AioContext
*old_ctx
;
2347 bs
= bdrv_lookup_bs(device
, node_name
, &local_err
);
2349 error_propagate(errp
, local_err
);
2354 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
2358 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_RESIZE
, NULL
)) {
2359 error_setg(errp
, QERR_DEVICE_IN_USE
, device
);
2363 blk
= blk_co_new_with_bs(bs
, BLK_PERM_RESIZE
, BLK_PERM_ALL
, errp
);
2369 bdrv_drained_begin(bs
);
2372 old_ctx
= bdrv_co_enter(bs
);
2373 blk_co_truncate(blk
, size
, false, PREALLOC_MODE_OFF
, 0, errp
);
2374 bdrv_co_leave(bs
, old_ctx
);
2377 bdrv_drained_end(bs
);
2382 void qmp_block_stream(const char *job_id
, const char *device
,
2384 const char *base_node
,
2385 const char *backing_file
,
2387 bool has_speed
, int64_t speed
,
2388 bool has_on_error
, BlockdevOnError on_error
,
2389 const char *filter_node_name
,
2390 bool has_auto_finalize
, bool auto_finalize
,
2391 bool has_auto_dismiss
, bool auto_dismiss
,
2394 BlockDriverState
*bs
, *iter
, *iter_end
;
2395 BlockDriverState
*base_bs
= NULL
;
2396 BlockDriverState
*bottom_bs
= NULL
;
2397 AioContext
*aio_context
;
2398 Error
*local_err
= NULL
;
2399 int job_flags
= JOB_DEFAULT
;
2401 if (base
&& base_node
) {
2402 error_setg(errp
, "'base' and 'base-node' cannot be specified "
2403 "at the same time");
2407 if (base
&& bottom
) {
2408 error_setg(errp
, "'base' and 'bottom' cannot be specified "
2409 "at the same time");
2413 if (bottom
&& base_node
) {
2414 error_setg(errp
, "'bottom' and 'base-node' cannot be specified "
2415 "at the same time");
2419 if (!has_on_error
) {
2420 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2423 bs
= bdrv_lookup_bs(device
, device
, errp
);
2428 aio_context
= bdrv_get_aio_context(bs
);
2429 aio_context_acquire(aio_context
);
2432 base_bs
= bdrv_find_backing_image(bs
, base
);
2433 if (base_bs
== NULL
) {
2434 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2437 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2441 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2445 if (bs
== base_bs
|| !bdrv_chain_contains(bs
, base_bs
)) {
2446 error_setg(errp
, "Node '%s' is not a backing image of '%s'",
2450 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2451 bdrv_refresh_filename(base_bs
);
2455 bottom_bs
= bdrv_lookup_bs(NULL
, bottom
, errp
);
2459 if (!bottom_bs
->drv
) {
2460 error_setg(errp
, "Node '%s' is not open", bottom
);
2463 if (bottom_bs
->drv
->is_filter
) {
2464 error_setg(errp
, "Node '%s' is a filter, use a non-filter node "
2465 "as 'bottom'", bottom
);
2468 if (!bdrv_chain_contains(bs
, bottom_bs
)) {
2469 error_setg(errp
, "Node '%s' is not in a chain starting from '%s'",
2473 assert(bdrv_get_aio_context(bottom_bs
) == aio_context
);
2477 * Check for op blockers in the whole chain between bs and base (or bottom)
2479 iter_end
= bottom
? bdrv_filter_or_cow_bs(bottom_bs
) : base_bs
;
2480 for (iter
= bs
; iter
&& iter
!= iter_end
;
2481 iter
= bdrv_filter_or_cow_bs(iter
))
2483 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_STREAM
, errp
)) {
2488 /* if we are streaming the entire chain, the result will have no backing
2489 * file, and specifying one is therefore an error */
2490 if (!base_bs
&& backing_file
) {
2491 error_setg(errp
, "backing file specified, but streaming the "
2496 if (has_auto_finalize
&& !auto_finalize
) {
2497 job_flags
|= JOB_MANUAL_FINALIZE
;
2499 if (has_auto_dismiss
&& !auto_dismiss
) {
2500 job_flags
|= JOB_MANUAL_DISMISS
;
2503 stream_start(job_id
, bs
, base_bs
, backing_file
,
2504 bottom_bs
, job_flags
, has_speed
? speed
: 0, on_error
,
2505 filter_node_name
, &local_err
);
2507 error_propagate(errp
, local_err
);
2511 trace_qmp_block_stream(bs
);
2514 aio_context_release(aio_context
);
2517 void qmp_block_commit(const char *job_id
, const char *device
,
2518 const char *base_node
,
2520 const char *top_node
,
2522 const char *backing_file
,
2523 bool has_speed
, int64_t speed
,
2524 bool has_on_error
, BlockdevOnError on_error
,
2525 const char *filter_node_name
,
2526 bool has_auto_finalize
, bool auto_finalize
,
2527 bool has_auto_dismiss
, bool auto_dismiss
,
2530 BlockDriverState
*bs
;
2531 BlockDriverState
*iter
;
2532 BlockDriverState
*base_bs
, *top_bs
;
2533 AioContext
*aio_context
;
2534 Error
*local_err
= NULL
;
2535 int job_flags
= JOB_DEFAULT
;
2536 uint64_t top_perm
, top_shared
;
2538 /* TODO We'll eventually have to take a writer lock in this function */
2539 GRAPH_RDLOCK_GUARD_MAINLOOP();
2544 if (!has_on_error
) {
2545 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
2547 if (has_auto_finalize
&& !auto_finalize
) {
2548 job_flags
|= JOB_MANUAL_FINALIZE
;
2550 if (has_auto_dismiss
&& !auto_dismiss
) {
2551 job_flags
|= JOB_MANUAL_DISMISS
;
2555 * libvirt relies on the DeviceNotFound error class in order to probe for
2556 * live commit feature versions; for this to work, we must make sure to
2557 * perform the device lookup before any generic errors that may occur in a
2558 * scenario in which all optional arguments are omitted. */
2559 bs
= qmp_get_root_bs(device
, &local_err
);
2561 bs
= bdrv_lookup_bs(device
, device
, NULL
);
2563 error_free(local_err
);
2564 error_set(errp
, ERROR_CLASS_DEVICE_NOT_FOUND
,
2565 "Device '%s' not found", device
);
2567 error_propagate(errp
, local_err
);
2572 aio_context
= bdrv_get_aio_context(bs
);
2573 aio_context_acquire(aio_context
);
2575 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_COMMIT_SOURCE
, errp
)) {
2579 /* default top_bs is the active layer */
2582 if (top_node
&& top
) {
2583 error_setg(errp
, "'top-node' and 'top' are mutually exclusive");
2585 } else if (top_node
) {
2586 top_bs
= bdrv_lookup_bs(NULL
, top_node
, errp
);
2587 if (top_bs
== NULL
) {
2590 if (!bdrv_chain_contains(bs
, top_bs
)) {
2591 error_setg(errp
, "'%s' is not in this backing file chain",
2596 /* This strcmp() is just a shortcut, there is no need to
2597 * refresh @bs's filename. If it mismatches,
2598 * bdrv_find_backing_image() will do the refresh and may still
2600 if (strcmp(bs
->filename
, top
) != 0) {
2601 top_bs
= bdrv_find_backing_image(bs
, top
);
2605 if (top_bs
== NULL
) {
2606 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
2610 assert(bdrv_get_aio_context(top_bs
) == aio_context
);
2612 if (base_node
&& base
) {
2613 error_setg(errp
, "'base-node' and 'base' are mutually exclusive");
2615 } else if (base_node
) {
2616 base_bs
= bdrv_lookup_bs(NULL
, base_node
, errp
);
2617 if (base_bs
== NULL
) {
2620 if (!bdrv_chain_contains(top_bs
, base_bs
)) {
2621 error_setg(errp
, "'%s' is not in this backing file chain",
2626 base_bs
= bdrv_find_backing_image(top_bs
, base
);
2627 if (base_bs
== NULL
) {
2628 error_setg(errp
, "Can't find '%s' in the backing chain", base
);
2632 base_bs
= bdrv_find_base(top_bs
);
2633 if (base_bs
== NULL
) {
2634 error_setg(errp
, "There is no backimg image");
2639 assert(bdrv_get_aio_context(base_bs
) == aio_context
);
2641 for (iter
= top_bs
; iter
!= bdrv_filter_or_cow_bs(base_bs
);
2642 iter
= bdrv_filter_or_cow_bs(iter
))
2644 if (bdrv_op_is_blocked(iter
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2649 /* Do not allow attempts to commit an image into itself */
2650 if (top_bs
== base_bs
) {
2651 error_setg(errp
, "cannot commit an image into itself");
2656 * Active commit is required if and only if someone has taken a
2657 * WRITE permission on the top node. Historically, we have always
2658 * used active commit for top nodes, so continue that practice
2659 * lest we possibly break clients that rely on this behavior, e.g.
2660 * to later attach this node to a writing parent.
2661 * (Active commit is never really wrong.)
2663 bdrv_get_cumulative_perm(top_bs
, &top_perm
, &top_shared
);
2664 if (top_perm
& BLK_PERM_WRITE
||
2665 bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
))
2668 if (bdrv_skip_filters(top_bs
) == bdrv_skip_filters(bs
)) {
2669 error_setg(errp
, "'backing-file' specified,"
2670 " but 'top' is the active layer");
2672 error_setg(errp
, "'backing-file' specified, but 'top' has a "
2679 * Emulate here what block_job_create() does, because it
2680 * is possible that @bs != @top_bs (the block job should
2681 * be named after @bs, even if @top_bs is the actual
2684 job_id
= bdrv_get_device_name(bs
);
2686 commit_active_start(job_id
, top_bs
, base_bs
, job_flags
, speed
, on_error
,
2687 filter_node_name
, NULL
, NULL
, false, &local_err
);
2689 BlockDriverState
*overlay_bs
= bdrv_find_overlay(bs
, top_bs
);
2690 if (bdrv_op_is_blocked(overlay_bs
, BLOCK_OP_TYPE_COMMIT_TARGET
, errp
)) {
2693 commit_start(job_id
, bs
, base_bs
, top_bs
, job_flags
,
2694 speed
, on_error
, backing_file
,
2695 filter_node_name
, &local_err
);
2697 if (local_err
!= NULL
) {
2698 error_propagate(errp
, local_err
);
2703 aio_context_release(aio_context
);
2706 /* Common QMP interface for drive-backup and blockdev-backup */
2707 static BlockJob
*do_backup_common(BackupCommon
*backup
,
2708 BlockDriverState
*bs
,
2709 BlockDriverState
*target_bs
,
2710 AioContext
*aio_context
,
2711 JobTxn
*txn
, Error
**errp
)
2713 BlockJob
*job
= NULL
;
2714 BdrvDirtyBitmap
*bmap
= NULL
;
2715 BackupPerf perf
= { .max_workers
= 64 };
2716 int job_flags
= JOB_DEFAULT
;
2718 if (!backup
->has_speed
) {
2721 if (!backup
->has_on_source_error
) {
2722 backup
->on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2724 if (!backup
->has_on_target_error
) {
2725 backup
->on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2727 if (!backup
->has_auto_finalize
) {
2728 backup
->auto_finalize
= true;
2730 if (!backup
->has_auto_dismiss
) {
2731 backup
->auto_dismiss
= true;
2733 if (!backup
->has_compress
) {
2734 backup
->compress
= false;
2737 if (backup
->x_perf
) {
2738 if (backup
->x_perf
->has_use_copy_range
) {
2739 perf
.use_copy_range
= backup
->x_perf
->use_copy_range
;
2741 if (backup
->x_perf
->has_max_workers
) {
2742 perf
.max_workers
= backup
->x_perf
->max_workers
;
2744 if (backup
->x_perf
->has_max_chunk
) {
2745 perf
.max_chunk
= backup
->x_perf
->max_chunk
;
2749 if ((backup
->sync
== MIRROR_SYNC_MODE_BITMAP
) ||
2750 (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
)) {
2751 /* done before desugaring 'incremental' to print the right message */
2752 if (!backup
->bitmap
) {
2753 error_setg(errp
, "must provide a valid bitmap name for "
2754 "'%s' sync mode", MirrorSyncMode_str(backup
->sync
));
2759 if (backup
->sync
== MIRROR_SYNC_MODE_INCREMENTAL
) {
2760 if (backup
->has_bitmap_mode
&&
2761 backup
->bitmap_mode
!= BITMAP_SYNC_MODE_ON_SUCCESS
) {
2762 error_setg(errp
, "Bitmap sync mode must be '%s' "
2763 "when using sync mode '%s'",
2764 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS
),
2765 MirrorSyncMode_str(backup
->sync
));
2768 backup
->has_bitmap_mode
= true;
2769 backup
->sync
= MIRROR_SYNC_MODE_BITMAP
;
2770 backup
->bitmap_mode
= BITMAP_SYNC_MODE_ON_SUCCESS
;
2773 if (backup
->bitmap
) {
2774 bmap
= bdrv_find_dirty_bitmap(bs
, backup
->bitmap
);
2776 error_setg(errp
, "Bitmap '%s' could not be found", backup
->bitmap
);
2779 if (!backup
->has_bitmap_mode
) {
2780 error_setg(errp
, "Bitmap sync mode must be given "
2781 "when providing a bitmap");
2784 if (bdrv_dirty_bitmap_check(bmap
, BDRV_BITMAP_ALLOW_RO
, errp
)) {
2788 /* This does not produce a useful bitmap artifact: */
2789 if (backup
->sync
== MIRROR_SYNC_MODE_NONE
) {
2790 error_setg(errp
, "sync mode '%s' does not produce meaningful bitmap"
2791 " outputs", MirrorSyncMode_str(backup
->sync
));
2795 /* If the bitmap isn't used for input or output, this is useless: */
2796 if (backup
->bitmap_mode
== BITMAP_SYNC_MODE_NEVER
&&
2797 backup
->sync
!= MIRROR_SYNC_MODE_BITMAP
) {
2798 error_setg(errp
, "Bitmap sync mode '%s' has no meaningful effect"
2799 " when combined with sync mode '%s'",
2800 BitmapSyncMode_str(backup
->bitmap_mode
),
2801 MirrorSyncMode_str(backup
->sync
));
2806 if (!backup
->bitmap
&& backup
->has_bitmap_mode
) {
2807 error_setg(errp
, "Cannot specify bitmap sync mode without a bitmap");
2811 if (!backup
->auto_finalize
) {
2812 job_flags
|= JOB_MANUAL_FINALIZE
;
2814 if (!backup
->auto_dismiss
) {
2815 job_flags
|= JOB_MANUAL_DISMISS
;
2818 job
= backup_job_create(backup
->job_id
, bs
, target_bs
, backup
->speed
,
2819 backup
->sync
, bmap
, backup
->bitmap_mode
,
2821 backup
->filter_node_name
,
2823 backup
->on_source_error
,
2824 backup
->on_target_error
,
2825 job_flags
, NULL
, NULL
, txn
, errp
);
2829 void qmp_drive_backup(DriveBackup
*backup
, Error
**errp
)
2831 TransactionAction action
= {
2832 .type
= TRANSACTION_ACTION_KIND_DRIVE_BACKUP
,
2833 .u
.drive_backup
.data
= backup
,
2835 blockdev_do_action(&action
, errp
);
2838 BlockDeviceInfoList
*qmp_query_named_block_nodes(bool has_flat
,
2842 bool return_flat
= has_flat
&& flat
;
2844 return bdrv_named_nodes_list(return_flat
, errp
);
2847 XDbgBlockGraph
*qmp_x_debug_query_block_graph(Error
**errp
)
2849 return bdrv_get_xdbg_block_graph(errp
);
2852 void qmp_blockdev_backup(BlockdevBackup
*backup
, Error
**errp
)
2854 TransactionAction action
= {
2855 .type
= TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP
,
2856 .u
.blockdev_backup
.data
= backup
,
2858 blockdev_do_action(&action
, errp
);
2861 /* Parameter check and block job starting for drive mirroring.
2862 * Caller should hold @device and @target's aio context (must be the same).
2864 static void blockdev_mirror_common(const char *job_id
, BlockDriverState
*bs
,
2865 BlockDriverState
*target
,
2866 const char *replaces
,
2867 enum MirrorSyncMode sync
,
2868 BlockMirrorBackingMode backing_mode
,
2870 bool has_speed
, int64_t speed
,
2871 bool has_granularity
, uint32_t granularity
,
2872 bool has_buf_size
, int64_t buf_size
,
2873 bool has_on_source_error
,
2874 BlockdevOnError on_source_error
,
2875 bool has_on_target_error
,
2876 BlockdevOnError on_target_error
,
2877 bool has_unmap
, bool unmap
,
2878 const char *filter_node_name
,
2879 bool has_copy_mode
, MirrorCopyMode copy_mode
,
2880 bool has_auto_finalize
, bool auto_finalize
,
2881 bool has_auto_dismiss
, bool auto_dismiss
,
2884 BlockDriverState
*unfiltered_bs
;
2885 int job_flags
= JOB_DEFAULT
;
2887 GLOBAL_STATE_CODE();
2888 GRAPH_RDLOCK_GUARD_MAINLOOP();
2893 if (!has_on_source_error
) {
2894 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2896 if (!has_on_target_error
) {
2897 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2899 if (!has_granularity
) {
2902 if (!has_buf_size
) {
2908 if (!has_copy_mode
) {
2909 copy_mode
= MIRROR_COPY_MODE_BACKGROUND
;
2911 if (has_auto_finalize
&& !auto_finalize
) {
2912 job_flags
|= JOB_MANUAL_FINALIZE
;
2914 if (has_auto_dismiss
&& !auto_dismiss
) {
2915 job_flags
|= JOB_MANUAL_DISMISS
;
2918 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
2919 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2920 "a value in range [512B, 64MB]");
2923 if (granularity
& (granularity
- 1)) {
2924 error_setg(errp
, QERR_INVALID_PARAMETER_VALUE
, "granularity",
2929 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
2932 if (bdrv_op_is_blocked(target
, BLOCK_OP_TYPE_MIRROR_TARGET
, errp
)) {
2936 if (!bdrv_backing_chain_next(bs
) && sync
== MIRROR_SYNC_MODE_TOP
) {
2937 sync
= MIRROR_SYNC_MODE_FULL
;
2941 /* We want to mirror from @bs, but keep implicit filters on top */
2942 unfiltered_bs
= bdrv_skip_implicit_filters(bs
);
2943 if (unfiltered_bs
!= bs
) {
2944 replaces
= unfiltered_bs
->node_name
;
2949 BlockDriverState
*to_replace_bs
;
2950 AioContext
*replace_aio_context
;
2951 int64_t bs_size
, replace_size
;
2953 bs_size
= bdrv_getlength(bs
);
2955 error_setg_errno(errp
, -bs_size
, "Failed to query device's size");
2959 to_replace_bs
= check_to_replace_node(bs
, replaces
, errp
);
2960 if (!to_replace_bs
) {
2964 replace_aio_context
= bdrv_get_aio_context(to_replace_bs
);
2965 aio_context_acquire(replace_aio_context
);
2966 replace_size
= bdrv_getlength(to_replace_bs
);
2967 aio_context_release(replace_aio_context
);
2969 if (replace_size
< 0) {
2970 error_setg_errno(errp
, -replace_size
,
2971 "Failed to query the replacement node's size");
2974 if (bs_size
!= replace_size
) {
2975 error_setg(errp
, "cannot replace image with a mirror image of "
2981 /* pass the node name to replace to mirror start since it's loose coupling
2982 * and will allow to check whether the node still exist at mirror completion
2984 mirror_start(job_id
, bs
, target
,
2985 replaces
, job_flags
,
2986 speed
, granularity
, buf_size
, sync
, backing_mode
, zero_target
,
2987 on_source_error
, on_target_error
, unmap
, filter_node_name
,
2991 void qmp_drive_mirror(DriveMirror
*arg
, Error
**errp
)
2993 BlockDriverState
*bs
;
2994 BlockDriverState
*target_backing_bs
, *target_bs
;
2995 AioContext
*aio_context
;
2996 AioContext
*old_context
;
2997 BlockMirrorBackingMode backing_mode
;
2998 Error
*local_err
= NULL
;
2999 QDict
*options
= NULL
;
3002 const char *format
= arg
->format
;
3006 bs
= qmp_get_root_bs(arg
->device
, errp
);
3011 /* Early check to avoid creating target */
3012 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_MIRROR_SOURCE
, errp
)) {
3016 aio_context
= bdrv_get_aio_context(bs
);
3017 aio_context_acquire(aio_context
);
3019 if (!arg
->has_mode
) {
3020 arg
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
3024 format
= (arg
->mode
== NEW_IMAGE_MODE_EXISTING
3025 ? NULL
: bs
->drv
->format_name
);
3028 flags
= bs
->open_flags
| BDRV_O_RDWR
;
3029 target_backing_bs
= bdrv_cow_bs(bdrv_skip_filters(bs
));
3030 if (!target_backing_bs
&& arg
->sync
== MIRROR_SYNC_MODE_TOP
) {
3031 arg
->sync
= MIRROR_SYNC_MODE_FULL
;
3033 if (arg
->sync
== MIRROR_SYNC_MODE_NONE
) {
3034 target_backing_bs
= bs
;
3037 size
= bdrv_getlength(bs
);
3039 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
3043 if (arg
->replaces
) {
3044 if (!arg
->node_name
) {
3045 error_setg(errp
, "a node-name must be provided when replacing a"
3046 " named node of the graph");
3051 if (arg
->mode
== NEW_IMAGE_MODE_ABSOLUTE_PATHS
) {
3052 backing_mode
= MIRROR_SOURCE_BACKING_CHAIN
;
3054 backing_mode
= MIRROR_OPEN_BACKING_CHAIN
;
3057 /* Don't open backing image in create() */
3058 flags
|= BDRV_O_NO_BACKING
;
3060 if ((arg
->sync
== MIRROR_SYNC_MODE_FULL
|| !target_backing_bs
)
3061 && arg
->mode
!= NEW_IMAGE_MODE_EXISTING
)
3063 /* create new image w/o backing file */
3065 bdrv_img_create(arg
->target
, format
,
3066 NULL
, NULL
, NULL
, size
, flags
, false, &local_err
);
3068 /* Implicit filters should not appear in the filename */
3069 BlockDriverState
*explicit_backing
=
3070 bdrv_skip_implicit_filters(target_backing_bs
);
3072 switch (arg
->mode
) {
3073 case NEW_IMAGE_MODE_EXISTING
:
3075 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
3076 /* create new image with backing file */
3077 bdrv_refresh_filename(explicit_backing
);
3078 bdrv_img_create(arg
->target
, format
,
3079 explicit_backing
->filename
,
3080 explicit_backing
->drv
->format_name
,
3081 NULL
, size
, flags
, false, &local_err
);
3089 error_propagate(errp
, local_err
);
3093 options
= qdict_new();
3094 if (arg
->node_name
) {
3095 qdict_put_str(options
, "node-name", arg
->node_name
);
3098 qdict_put_str(options
, "driver", format
);
3100 aio_context_release(aio_context
);
3102 /* Mirroring takes care of copy-on-write using the source's backing
3105 aio_context_acquire(qemu_get_aio_context());
3106 target_bs
= bdrv_open(arg
->target
, NULL
, options
, flags
, errp
);
3107 aio_context_release(qemu_get_aio_context());
3113 zero_target
= (arg
->sync
== MIRROR_SYNC_MODE_FULL
&&
3114 (arg
->mode
== NEW_IMAGE_MODE_EXISTING
||
3115 !bdrv_has_zero_init(target_bs
)));
3118 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3119 old_context
= bdrv_get_aio_context(target_bs
);
3120 aio_context_acquire(old_context
);
3122 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
3124 bdrv_unref(target_bs
);
3125 aio_context_release(old_context
);
3129 aio_context_release(old_context
);
3130 aio_context_acquire(aio_context
);
3132 blockdev_mirror_common(arg
->job_id
, bs
, target_bs
,
3133 arg
->replaces
, arg
->sync
,
3134 backing_mode
, zero_target
,
3135 arg
->has_speed
, arg
->speed
,
3136 arg
->has_granularity
, arg
->granularity
,
3137 arg
->has_buf_size
, arg
->buf_size
,
3138 arg
->has_on_source_error
, arg
->on_source_error
,
3139 arg
->has_on_target_error
, arg
->on_target_error
,
3140 arg
->has_unmap
, arg
->unmap
,
3142 arg
->has_copy_mode
, arg
->copy_mode
,
3143 arg
->has_auto_finalize
, arg
->auto_finalize
,
3144 arg
->has_auto_dismiss
, arg
->auto_dismiss
,
3146 bdrv_unref(target_bs
);
3148 aio_context_release(aio_context
);
3151 void qmp_blockdev_mirror(const char *job_id
,
3152 const char *device
, const char *target
,
3153 const char *replaces
,
3154 MirrorSyncMode sync
,
3155 bool has_speed
, int64_t speed
,
3156 bool has_granularity
, uint32_t granularity
,
3157 bool has_buf_size
, int64_t buf_size
,
3158 bool has_on_source_error
,
3159 BlockdevOnError on_source_error
,
3160 bool has_on_target_error
,
3161 BlockdevOnError on_target_error
,
3162 const char *filter_node_name
,
3163 bool has_copy_mode
, MirrorCopyMode copy_mode
,
3164 bool has_auto_finalize
, bool auto_finalize
,
3165 bool has_auto_dismiss
, bool auto_dismiss
,
3168 BlockDriverState
*bs
;
3169 BlockDriverState
*target_bs
;
3170 AioContext
*aio_context
;
3171 AioContext
*old_context
;
3172 BlockMirrorBackingMode backing_mode
= MIRROR_LEAVE_BACKING_CHAIN
;
3176 bs
= qmp_get_root_bs(device
, errp
);
3181 target_bs
= bdrv_lookup_bs(target
, target
, errp
);
3186 zero_target
= (sync
== MIRROR_SYNC_MODE_FULL
);
3188 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3189 old_context
= bdrv_get_aio_context(target_bs
);
3190 aio_context
= bdrv_get_aio_context(bs
);
3191 aio_context_acquire(old_context
);
3193 ret
= bdrv_try_change_aio_context(target_bs
, aio_context
, NULL
, errp
);
3195 aio_context_release(old_context
);
3196 aio_context_acquire(aio_context
);
3202 blockdev_mirror_common(job_id
, bs
, target_bs
,
3203 replaces
, sync
, backing_mode
,
3204 zero_target
, has_speed
, speed
,
3205 has_granularity
, granularity
,
3206 has_buf_size
, buf_size
,
3207 has_on_source_error
, on_source_error
,
3208 has_on_target_error
, on_target_error
,
3209 true, true, filter_node_name
,
3210 has_copy_mode
, copy_mode
,
3211 has_auto_finalize
, auto_finalize
,
3212 has_auto_dismiss
, auto_dismiss
,
3215 aio_context_release(aio_context
);
3219 * Get a block job using its ID. Called with job_mutex held.
3221 static BlockJob
*find_block_job_locked(const char *id
, Error
**errp
)
3227 job
= block_job_get_locked(id
);
3230 error_set(errp
, ERROR_CLASS_DEVICE_NOT_ACTIVE
,
3231 "Block job '%s' not found", id
);
3238 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
3243 job
= find_block_job_locked(device
, errp
);
3249 block_job_set_speed_locked(job
, speed
, errp
);
3252 void qmp_block_job_cancel(const char *device
,
3253 bool has_force
, bool force
, Error
**errp
)
3258 job
= find_block_job_locked(device
, errp
);
3268 if (job_user_paused_locked(&job
->job
) && !force
) {
3269 error_setg(errp
, "The block job for device '%s' is currently paused",
3274 trace_qmp_block_job_cancel(job
);
3275 job_user_cancel_locked(&job
->job
, force
, errp
);
3278 void qmp_block_job_pause(const char *device
, Error
**errp
)
3283 job
= find_block_job_locked(device
, errp
);
3289 trace_qmp_block_job_pause(job
);
3290 job_user_pause_locked(&job
->job
, errp
);
3293 void qmp_block_job_resume(const char *device
, Error
**errp
)
3298 job
= find_block_job_locked(device
, errp
);
3304 trace_qmp_block_job_resume(job
);
3305 job_user_resume_locked(&job
->job
, errp
);
3308 void qmp_block_job_complete(const char *device
, Error
**errp
)
3313 job
= find_block_job_locked(device
, errp
);
3319 trace_qmp_block_job_complete(job
);
3320 job_complete_locked(&job
->job
, errp
);
3323 void qmp_block_job_finalize(const char *id
, Error
**errp
)
3328 job
= find_block_job_locked(id
, errp
);
3334 trace_qmp_block_job_finalize(job
);
3335 job_ref_locked(&job
->job
);
3336 job_finalize_locked(&job
->job
, errp
);
3338 job_unref_locked(&job
->job
);
3341 void qmp_block_job_dismiss(const char *id
, Error
**errp
)
3347 bjob
= find_block_job_locked(id
, errp
);
3353 trace_qmp_block_job_dismiss(bjob
);
3355 job_dismiss_locked(&job
, errp
);
3358 void qmp_change_backing_file(const char *device
,
3359 const char *image_node_name
,
3360 const char *backing_file
,
3363 BlockDriverState
*bs
= NULL
;
3364 AioContext
*aio_context
;
3365 BlockDriverState
*image_bs
= NULL
;
3366 Error
*local_err
= NULL
;
3370 bs
= qmp_get_root_bs(device
, errp
);
3375 aio_context
= bdrv_get_aio_context(bs
);
3376 aio_context_acquire(aio_context
);
3378 image_bs
= bdrv_lookup_bs(NULL
, image_node_name
, &local_err
);
3380 error_propagate(errp
, local_err
);
3385 error_setg(errp
, "image file not found");
3389 if (bdrv_find_base(image_bs
) == image_bs
) {
3390 error_setg(errp
, "not allowing backing file change on an image "
3391 "without a backing file");
3395 /* even though we are not necessarily operating on bs, we need it to
3396 * determine if block ops are currently prohibited on the chain */
3397 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_CHANGE
, errp
)) {
3401 /* final sanity check */
3402 if (!bdrv_chain_contains(bs
, image_bs
)) {
3403 error_setg(errp
, "'%s' and image file are not in the same chain",
3408 /* if not r/w, reopen to make r/w */
3409 ro
= bdrv_is_read_only(image_bs
);
3412 if (bdrv_reopen_set_read_only(image_bs
, false, errp
) != 0) {
3417 ret
= bdrv_change_backing_file(image_bs
, backing_file
,
3418 image_bs
->drv
? image_bs
->drv
->format_name
: "",
3422 error_setg_errno(errp
, -ret
, "Could not change backing file to '%s'",
3424 /* don't exit here, so we can try to restore open flags if
3429 bdrv_reopen_set_read_only(image_bs
, true, errp
);
3433 aio_context_release(aio_context
);
3436 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
3438 BlockDriverState
*bs
;
3440 Visitor
*v
= qobject_output_visitor_new(&obj
);
3443 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3444 visit_complete(v
, &obj
);
3445 qdict
= qobject_to(QDict
, obj
);
3447 qdict_flatten(qdict
);
3449 if (!qdict_get_try_str(qdict
, "node-name")) {
3450 error_setg(errp
, "'node-name' must be specified for the root node");
3454 bs
= bds_tree_init(qdict
, errp
);
3459 bdrv_set_monitor_owned(bs
);
3465 void qmp_blockdev_reopen(BlockdevOptionsList
*reopen_list
, Error
**errp
)
3467 BlockReopenQueue
*queue
= NULL
;
3469 /* Add each one of the BDS that we want to reopen to the queue */
3470 for (; reopen_list
!= NULL
; reopen_list
= reopen_list
->next
) {
3471 BlockdevOptions
*options
= reopen_list
->value
;
3472 BlockDriverState
*bs
;
3478 /* Check for the selected node name */
3479 if (!options
->node_name
) {
3480 error_setg(errp
, "node-name not specified");
3484 bs
= bdrv_find_node(options
->node_name
);
3486 error_setg(errp
, "Failed to find node with node-name='%s'",
3487 options
->node_name
);
3491 /* Put all options in a QDict and flatten it */
3492 v
= qobject_output_visitor_new(&obj
);
3493 visit_type_BlockdevOptions(v
, NULL
, &options
, &error_abort
);
3494 visit_complete(v
, &obj
);
3497 qdict
= qobject_to(QDict
, obj
);
3499 qdict_flatten(qdict
);
3501 ctx
= bdrv_get_aio_context(bs
);
3502 aio_context_acquire(ctx
);
3504 queue
= bdrv_reopen_queue(queue
, bs
, qdict
, false);
3506 aio_context_release(ctx
);
3509 /* Perform the reopen operation */
3510 bdrv_reopen_multiple(queue
, errp
);
3514 bdrv_reopen_queue_free(queue
);
3517 void qmp_blockdev_del(const char *node_name
, Error
**errp
)
3519 AioContext
*aio_context
;
3520 BlockDriverState
*bs
;
3522 GLOBAL_STATE_CODE();
3523 GRAPH_RDLOCK_GUARD_MAINLOOP();
3525 bs
= bdrv_find_node(node_name
);
3527 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3530 if (bdrv_has_blk(bs
)) {
3531 error_setg(errp
, "Node %s is in use", node_name
);
3534 aio_context
= bdrv_get_aio_context(bs
);
3535 aio_context_acquire(aio_context
);
3537 if (bdrv_op_is_blocked(bs
, BLOCK_OP_TYPE_DRIVE_DEL
, errp
)) {
3541 if (!QTAILQ_IN_USE(bs
, monitor_list
)) {
3542 error_setg(errp
, "Node %s is not owned by the monitor",
3547 if (bs
->refcnt
> 1) {
3548 error_setg(errp
, "Block device %s is in use",
3549 bdrv_get_device_or_node_name(bs
));
3553 QTAILQ_REMOVE(&monitor_bdrv_states
, bs
, monitor_list
);
3557 aio_context_release(aio_context
);
3560 static BdrvChild
* GRAPH_RDLOCK
3561 bdrv_find_child(BlockDriverState
*parent_bs
, const char *child_name
)
3565 QLIST_FOREACH(child
, &parent_bs
->children
, next
) {
3566 if (strcmp(child
->name
, child_name
) == 0) {
3574 void qmp_x_blockdev_change(const char *parent
, const char *child
,
3575 const char *node
, Error
**errp
)
3577 BlockDriverState
*parent_bs
, *new_bs
= NULL
;
3580 bdrv_graph_wrlock(NULL
);
3582 parent_bs
= bdrv_lookup_bs(parent
, parent
, errp
);
3587 if (!child
== !node
) {
3589 error_setg(errp
, "The parameters child and node are in conflict");
3591 error_setg(errp
, "Either child or node must be specified");
3597 p_child
= bdrv_find_child(parent_bs
, child
);
3599 error_setg(errp
, "Node '%s' does not have child '%s'",
3603 bdrv_del_child(parent_bs
, p_child
, errp
);
3607 new_bs
= bdrv_find_node(node
);
3609 error_setg(errp
, "Node '%s' not found", node
);
3612 bdrv_add_child(parent_bs
, new_bs
, errp
);
3616 bdrv_graph_wrunlock();
3619 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
3621 BlockJobInfoList
*head
= NULL
, **tail
= &head
;
3626 for (job
= block_job_next_locked(NULL
); job
;
3627 job
= block_job_next_locked(job
)) {
3628 BlockJobInfo
*value
;
3630 if (block_job_is_internal(job
)) {
3633 value
= block_job_query_locked(job
, errp
);
3635 qapi_free_BlockJobInfoList(head
);
3638 QAPI_LIST_APPEND(tail
, value
);
3644 void qmp_x_blockdev_set_iothread(const char *node_name
, StrOrNull
*iothread
,
3645 bool has_force
, bool force
, Error
**errp
)
3647 AioContext
*old_context
;
3648 AioContext
*new_context
;
3649 BlockDriverState
*bs
;
3651 GRAPH_RDLOCK_GUARD_MAINLOOP();
3653 bs
= bdrv_find_node(node_name
);
3655 error_setg(errp
, "Failed to find node with node-name='%s'", node_name
);
3659 /* Protects against accidents. */
3660 if (!(has_force
&& force
) && bdrv_has_blk(bs
)) {
3661 error_setg(errp
, "Node %s is associated with a BlockBackend and could "
3662 "be in use (use force=true to override this check)",
3667 if (iothread
->type
== QTYPE_QSTRING
) {
3668 IOThread
*obj
= iothread_by_id(iothread
->u
.s
);
3670 error_setg(errp
, "Cannot find iothread %s", iothread
->u
.s
);
3674 new_context
= iothread_get_aio_context(obj
);
3676 new_context
= qemu_get_aio_context();
3679 old_context
= bdrv_get_aio_context(bs
);
3680 aio_context_acquire(old_context
);
3682 bdrv_try_change_aio_context(bs
, new_context
, NULL
, errp
);
3684 aio_context_release(old_context
);
3687 QemuOptsList qemu_common_drive_opts
= {
3689 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
3693 .type
= QEMU_OPT_BOOL
,
3694 .help
= "enable/disable snapshot mode",
3697 .type
= QEMU_OPT_STRING
,
3698 .help
= "host AIO implementation (threads, native, io_uring)",
3700 .name
= BDRV_OPT_CACHE_WB
,
3701 .type
= QEMU_OPT_BOOL
,
3702 .help
= "Enable writeback mode",
3705 .type
= QEMU_OPT_STRING
,
3706 .help
= "disk format (raw, qcow2, ...)",
3709 .type
= QEMU_OPT_STRING
,
3710 .help
= "read error action",
3713 .type
= QEMU_OPT_STRING
,
3714 .help
= "write error action",
3716 .name
= BDRV_OPT_READ_ONLY
,
3717 .type
= QEMU_OPT_BOOL
,
3718 .help
= "open drive file as read-only",
3724 .name
= "throttling.group",
3725 .type
= QEMU_OPT_STRING
,
3726 .help
= "name of the block throttling group",
3728 .name
= "copy-on-read",
3729 .type
= QEMU_OPT_BOOL
,
3730 .help
= "copy read data from backing file into image file",
3732 .name
= "detect-zeroes",
3733 .type
= QEMU_OPT_STRING
,
3734 .help
= "try to optimize zero writes (off, on, unmap)",
3736 .name
= "stats-account-invalid",
3737 .type
= QEMU_OPT_BOOL
,
3738 .help
= "whether to account for invalid I/O operations "
3739 "in the statistics",
3741 .name
= "stats-account-failed",
3742 .type
= QEMU_OPT_BOOL
,
3743 .help
= "whether to account for failed I/O operations "
3744 "in the statistics",
3746 { /* end of list */ }
3750 QemuOptsList qemu_drive_opts
= {
3752 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
3755 * no elements => accept any params
3756 * validation will happen later
3758 { /* end of list */ }