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 "sysemu/blockdev.h"
34 #include "hw/block/block.h"
35 #include "block/blockjob.h"
36 #include "monitor/monitor.h"
37 #include "qapi/qmp/qerror.h"
38 #include "qemu/option.h"
39 #include "qemu/config-file.h"
40 #include "qapi/qmp/types.h"
41 #include "qapi-visit.h"
42 #include "qapi/qmp-output-visitor.h"
43 #include "sysemu/sysemu.h"
44 #include "block/block_int.h"
45 #include "qmp-commands.h"
47 #include "sysemu/arch_init.h"
49 static QTAILQ_HEAD(drivelist
, DriveInfo
) drives
= QTAILQ_HEAD_INITIALIZER(drives
);
50 extern QemuOptsList qemu_common_drive_opts
;
52 static const char *const if_name
[IF_COUNT
] = {
56 [IF_FLOPPY
] = "floppy",
57 [IF_PFLASH
] = "pflash",
60 [IF_VIRTIO
] = "virtio",
64 static const int if_max_devs
[IF_COUNT
] = {
66 * Do not change these numbers! They govern how drive option
67 * index maps to unit and bus. That mapping is ABI.
69 * All controllers used to imlement if=T drives need to support
70 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
71 * Otherwise, some index values map to "impossible" bus, unit
74 * For instance, if you change [IF_SCSI] to 255, -drive
75 * if=scsi,index=12 no longer means bus=1,unit=5, but
76 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
77 * the drive can't be set up. Regression.
84 * We automatically delete the drive when a device using it gets
85 * unplugged. Questionable feature, but we can't just drop it.
86 * Device models call blockdev_mark_auto_del() to schedule the
87 * automatic deletion, and generic qdev code calls blockdev_auto_del()
88 * when deletion is actually safe.
90 void blockdev_mark_auto_del(BlockDriverState
*bs
)
92 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
94 if (dinfo
&& !dinfo
->enable_auto_del
) {
99 block_job_cancel(bs
->job
);
106 void blockdev_auto_del(BlockDriverState
*bs
)
108 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
110 if (dinfo
&& dinfo
->auto_del
) {
111 drive_put_ref(dinfo
);
115 static int drive_index_to_bus_id(BlockInterfaceType type
, int index
)
117 int max_devs
= if_max_devs
[type
];
118 return max_devs
? index
/ max_devs
: 0;
121 static int drive_index_to_unit_id(BlockInterfaceType type
, int index
)
123 int max_devs
= if_max_devs
[type
];
124 return max_devs
? index
% max_devs
: index
;
127 QemuOpts
*drive_def(const char *optstr
)
129 return qemu_opts_parse(qemu_find_opts("drive"), optstr
, 0);
132 QemuOpts
*drive_add(BlockInterfaceType type
, int index
, const char *file
,
138 opts
= drive_def(optstr
);
142 if (type
!= IF_DEFAULT
) {
143 qemu_opt_set(opts
, "if", if_name
[type
]);
146 snprintf(buf
, sizeof(buf
), "%d", index
);
147 qemu_opt_set(opts
, "index", buf
);
150 qemu_opt_set(opts
, "file", file
);
154 DriveInfo
*drive_get(BlockInterfaceType type
, int bus
, int unit
)
158 /* seek interface, bus and unit */
160 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
161 if (dinfo
->type
== type
&&
170 DriveInfo
*drive_get_by_index(BlockInterfaceType type
, int index
)
172 return drive_get(type
,
173 drive_index_to_bus_id(type
, index
),
174 drive_index_to_unit_id(type
, index
));
177 int drive_get_max_bus(BlockInterfaceType type
)
183 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
184 if(dinfo
->type
== type
&&
185 dinfo
->bus
> max_bus
)
186 max_bus
= dinfo
->bus
;
191 /* Get a block device. This should only be used for single-drive devices
192 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
194 DriveInfo
*drive_get_next(BlockInterfaceType type
)
196 static int next_block_unit
[IF_COUNT
];
198 return drive_get(type
, 0, next_block_unit
[type
]++);
201 DriveInfo
*drive_get_by_blockdev(BlockDriverState
*bs
)
205 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
206 if (dinfo
->bdrv
== bs
) {
213 static void bdrv_format_print(void *opaque
, const char *name
)
215 error_printf(" %s", name
);
218 static void drive_uninit(DriveInfo
*dinfo
)
221 qemu_opts_del(dinfo
->opts
);
224 bdrv_unref(dinfo
->bdrv
);
226 QTAILQ_REMOVE(&drives
, dinfo
, next
);
227 g_free(dinfo
->serial
);
231 void drive_put_ref(DriveInfo
*dinfo
)
233 assert(dinfo
->refcount
);
234 if (--dinfo
->refcount
== 0) {
239 void drive_get_ref(DriveInfo
*dinfo
)
246 BlockDriverState
*bs
;
249 static void bdrv_put_ref_bh(void *opaque
)
251 BDRVPutRefBH
*s
= opaque
;
254 qemu_bh_delete(s
->bh
);
259 * Release a BDS reference in a BH
261 * It is not safe to use bdrv_unref() from a callback function when the callers
262 * still need the BlockDriverState. In such cases we schedule a BH to release
265 static void bdrv_put_ref_bh_schedule(BlockDriverState
*bs
)
269 s
= g_new(BDRVPutRefBH
, 1);
270 s
->bh
= qemu_bh_new(bdrv_put_ref_bh
, s
);
272 qemu_bh_schedule(s
->bh
);
275 static int parse_block_error_action(const char *buf
, bool is_read
)
277 if (!strcmp(buf
, "ignore")) {
278 return BLOCKDEV_ON_ERROR_IGNORE
;
279 } else if (!is_read
&& !strcmp(buf
, "enospc")) {
280 return BLOCKDEV_ON_ERROR_ENOSPC
;
281 } else if (!strcmp(buf
, "stop")) {
282 return BLOCKDEV_ON_ERROR_STOP
;
283 } else if (!strcmp(buf
, "report")) {
284 return BLOCKDEV_ON_ERROR_REPORT
;
286 error_report("'%s' invalid %s error action",
287 buf
, is_read
? "read" : "write");
292 static bool check_throttle_config(ThrottleConfig
*cfg
, Error
**errp
)
294 if (throttle_conflicting(cfg
)) {
295 error_setg(errp
, "bps/iops/max total values and read/write values"
296 " cannot be used at the same time");
300 if (!throttle_is_valid(cfg
)) {
301 error_setg(errp
, "bps/iops/maxs values must be 0 or greater");
308 typedef enum { MEDIA_DISK
, MEDIA_CDROM
} DriveMediaType
;
310 /* Takes the ownership of bs_opts */
311 static DriveInfo
*blockdev_init(QDict
*bs_opts
,
312 BlockInterfaceType block_default_type
,
313 DriveMediaType media
)
316 const char *file
= NULL
;
318 const char *mediastr
= "";
319 BlockInterfaceType type
;
321 int cyls
, heads
, secs
, translation
;
326 int on_read_error
, on_write_error
;
336 bool has_driver_specific_opts
;
337 BlockDriver
*drv
= NULL
;
339 translation
= BIOS_ATA_TRANSLATION_AUTO
;
341 /* Check common options by copying from bs_opts to opts, all other options
342 * stay in bs_opts for processing by bdrv_open(). */
343 id
= qdict_get_try_str(bs_opts
, "id");
344 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, &error
);
345 if (error_is_set(&error
)) {
346 qerror_report_err(error
);
351 qemu_opts_absorb_qdict(opts
, bs_opts
, &error
);
352 if (error_is_set(&error
)) {
353 qerror_report_err(error
);
359 qdict_del(bs_opts
, "id");
362 has_driver_specific_opts
= !!qdict_size(bs_opts
);
364 /* extract parameters */
365 bus_id
= qemu_opt_get_number(opts
, "bus", 0);
366 unit_id
= qemu_opt_get_number(opts
, "unit", -1);
367 index
= qemu_opt_get_number(opts
, "index", -1);
369 cyls
= qemu_opt_get_number(opts
, "cyls", 0);
370 heads
= qemu_opt_get_number(opts
, "heads", 0);
371 secs
= qemu_opt_get_number(opts
, "secs", 0);
373 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
374 ro
= qemu_opt_get_bool(opts
, "read-only", 0);
375 copy_on_read
= qemu_opt_get_bool(opts
, "copy-on-read", false);
377 file
= qemu_opt_get(opts
, "file");
378 serial
= qemu_opt_get(opts
, "serial");
380 if ((buf
= qemu_opt_get(opts
, "if")) != NULL
) {
381 for (type
= 0; type
< IF_COUNT
&& strcmp(buf
, if_name
[type
]); type
++)
383 if (type
== IF_COUNT
) {
384 error_report("unsupported bus type '%s'", buf
);
388 type
= block_default_type
;
391 max_devs
= if_max_devs
[type
];
393 if (cyls
|| heads
|| secs
) {
395 error_report("invalid physical cyls number");
399 error_report("invalid physical heads number");
403 error_report("invalid physical secs number");
408 if ((buf
= qemu_opt_get(opts
, "trans")) != NULL
) {
410 error_report("'%s' trans must be used with cyls, heads and secs",
414 if (!strcmp(buf
, "none"))
415 translation
= BIOS_ATA_TRANSLATION_NONE
;
416 else if (!strcmp(buf
, "lba"))
417 translation
= BIOS_ATA_TRANSLATION_LBA
;
418 else if (!strcmp(buf
, "auto"))
419 translation
= BIOS_ATA_TRANSLATION_AUTO
;
421 error_report("'%s' invalid translation type", buf
);
426 if (media
== MEDIA_CDROM
) {
427 if (cyls
|| secs
|| heads
) {
428 error_report("CHS can't be set with media=cdrom");
433 if ((buf
= qemu_opt_get(opts
, "discard")) != NULL
) {
434 if (bdrv_parse_discard_flags(buf
, &bdrv_flags
) != 0) {
435 error_report("invalid discard option");
440 if (qemu_opt_get_bool(opts
, "cache.writeback", true)) {
441 bdrv_flags
|= BDRV_O_CACHE_WB
;
443 if (qemu_opt_get_bool(opts
, "cache.direct", false)) {
444 bdrv_flags
|= BDRV_O_NOCACHE
;
446 if (qemu_opt_get_bool(opts
, "cache.no-flush", false)) {
447 bdrv_flags
|= BDRV_O_NO_FLUSH
;
450 #ifdef CONFIG_LINUX_AIO
451 if ((buf
= qemu_opt_get(opts
, "aio")) != NULL
) {
452 if (!strcmp(buf
, "native")) {
453 bdrv_flags
|= BDRV_O_NATIVE_AIO
;
454 } else if (!strcmp(buf
, "threads")) {
455 /* this is the default */
457 error_report("invalid aio option");
463 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
464 if (is_help_option(buf
)) {
465 error_printf("Supported formats:");
466 bdrv_iterate_format(bdrv_format_print
, NULL
);
471 drv
= bdrv_find_format(buf
);
473 error_report("'%s' invalid format", buf
);
478 /* disk I/O throttling */
479 memset(&cfg
, 0, sizeof(cfg
));
480 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
=
481 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
482 cfg
.buckets
[THROTTLE_BPS_READ
].avg
=
483 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
484 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
=
485 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
486 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
=
487 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
488 cfg
.buckets
[THROTTLE_OPS_READ
].avg
=
489 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
490 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
=
491 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
493 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
=
494 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
495 cfg
.buckets
[THROTTLE_BPS_READ
].max
=
496 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
497 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
=
498 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
499 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
=
500 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
501 cfg
.buckets
[THROTTLE_OPS_READ
].max
=
502 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
503 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
=
504 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
506 cfg
.op_size
= qemu_opt_get_number(opts
, "throttling.iops-size", 0);
508 if (!check_throttle_config(&cfg
, &error
)) {
509 error_report("%s", error_get_pretty(error
));
514 if (qemu_opt_get(opts
, "boot") != NULL
) {
515 fprintf(stderr
, "qemu-kvm: boot=on|off is deprecated and will be "
516 "ignored. Future versions will reject this parameter. Please "
517 "update your scripts.\n");
520 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
521 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
522 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_NONE
) {
523 error_report("werror is not supported by this bus type");
527 on_write_error
= parse_block_error_action(buf
, 0);
528 if (on_write_error
< 0) {
533 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
534 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
535 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&& type
!= IF_NONE
) {
536 error_report("rerror is not supported by this bus type");
540 on_read_error
= parse_block_error_action(buf
, 1);
541 if (on_read_error
< 0) {
546 if ((devaddr
= qemu_opt_get(opts
, "addr")) != NULL
) {
547 if (type
!= IF_VIRTIO
) {
548 error_report("addr is not supported by this bus type");
553 /* compute bus and unit according index */
556 if (bus_id
!= 0 || unit_id
!= -1) {
557 error_report("index cannot be used with bus and unit");
560 bus_id
= drive_index_to_bus_id(type
, index
);
561 unit_id
= drive_index_to_unit_id(type
, index
);
564 /* if user doesn't specify a unit_id,
565 * try to find the first free
570 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
572 if (max_devs
&& unit_id
>= max_devs
) {
581 if (max_devs
&& unit_id
>= max_devs
) {
582 error_report("unit %d too big (max is %d)",
583 unit_id
, max_devs
- 1);
588 * catch multiple definitions
591 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
592 error_report("drive with bus=%d, unit=%d (index=%d) exists",
593 bus_id
, unit_id
, index
);
597 /* no id supplied -> create one */
598 if (qemu_opts_id(opts
) == NULL
) {
600 if (type
== IF_IDE
|| type
== IF_SCSI
) {
601 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
604 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
607 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
610 qemu_opts_set_id(opts
, new_id
);
614 dinfo
= g_malloc0(sizeof(*dinfo
));
615 dinfo
->id
= g_strdup(qemu_opts_id(opts
));
616 dinfo
->bdrv
= bdrv_new(dinfo
->id
);
617 dinfo
->bdrv
->open_flags
= snapshot
? BDRV_O_SNAPSHOT
: 0;
618 dinfo
->bdrv
->read_only
= ro
;
619 dinfo
->devaddr
= devaddr
;
622 dinfo
->unit
= unit_id
;
624 dinfo
->heads
= heads
;
626 dinfo
->trans
= translation
;
628 if (serial
!= NULL
) {
629 dinfo
->serial
= g_strdup(serial
);
631 QTAILQ_INSERT_TAIL(&drives
, dinfo
, next
);
633 bdrv_set_on_error(dinfo
->bdrv
, on_read_error
, on_write_error
);
635 /* disk I/O throttling */
636 if (throttle_enabled(&cfg
)) {
637 bdrv_io_limits_enable(dinfo
->bdrv
);
638 bdrv_set_io_limits(dinfo
->bdrv
, &cfg
);
646 dinfo
->media_cd
= media
== MEDIA_CDROM
;
655 /* add virtio block device */
657 devopts
= qemu_opts_create_nofail(qemu_find_opts("device"));
658 if (arch_type
== QEMU_ARCH_S390X
) {
659 qemu_opt_set(devopts
, "driver", "virtio-blk-s390");
661 qemu_opt_set(devopts
, "driver", "virtio-blk-pci");
663 qemu_opt_set(devopts
, "drive", dinfo
->id
);
665 qemu_opt_set(devopts
, "addr", devaddr
);
671 if (!file
|| !*file
) {
672 if (has_driver_specific_opts
) {
679 /* always use cache=unsafe with snapshot */
680 bdrv_flags
&= ~BDRV_O_CACHE_MASK
;
681 bdrv_flags
|= (BDRV_O_SNAPSHOT
|BDRV_O_CACHE_WB
|BDRV_O_NO_FLUSH
);
685 bdrv_flags
|= BDRV_O_COPY_ON_READ
;
688 if (runstate_check(RUN_STATE_INMIGRATE
)) {
689 bdrv_flags
|= BDRV_O_INCOMING
;
692 if (media
== MEDIA_CDROM
) {
693 /* CDROM is fine for any interface, don't check. */
695 } else if (ro
== 1) {
696 if (type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_FLOPPY
&&
697 type
!= IF_NONE
&& type
!= IF_PFLASH
) {
698 error_report("read-only not supported by this bus type");
703 bdrv_flags
|= ro
? 0 : BDRV_O_RDWR
;
705 if (ro
&& copy_on_read
) {
706 error_report("warning: disabling copy_on_read on read-only drive");
710 ret
= bdrv_open(dinfo
->bdrv
, file
, bs_opts
, bdrv_flags
, drv
, &error
);
713 error_report("could not open disk image %s: %s",
714 file
?: dinfo
->id
, error_get_pretty(error
));
718 if (bdrv_key_required(dinfo
->bdrv
))
729 bdrv_unref(dinfo
->bdrv
);
731 QTAILQ_REMOVE(&drives
, dinfo
, next
);
736 static void qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
)
740 value
= qemu_opt_get(opts
, from
);
742 qemu_opt_set(opts
, to
, value
);
743 qemu_opt_unset(opts
, from
);
747 QemuOptsList qemu_legacy_drive_opts
= {
749 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
753 .type
= QEMU_OPT_STRING
,
754 .help
= "media type (disk, cdrom)",
756 { /* end of list */ }
760 DriveInfo
*drive_init(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
)
763 DriveInfo
*dinfo
= NULL
;
765 QemuOpts
*legacy_opts
;
766 DriveMediaType media
= MEDIA_DISK
;
767 Error
*local_err
= NULL
;
769 /* Change legacy command line options into QMP ones */
770 qemu_opt_rename(all_opts
, "iops", "throttling.iops-total");
771 qemu_opt_rename(all_opts
, "iops_rd", "throttling.iops-read");
772 qemu_opt_rename(all_opts
, "iops_wr", "throttling.iops-write");
774 qemu_opt_rename(all_opts
, "bps", "throttling.bps-total");
775 qemu_opt_rename(all_opts
, "bps_rd", "throttling.bps-read");
776 qemu_opt_rename(all_opts
, "bps_wr", "throttling.bps-write");
778 qemu_opt_rename(all_opts
, "iops_max", "throttling.iops-total-max");
779 qemu_opt_rename(all_opts
, "iops_rd_max", "throttling.iops-read-max");
780 qemu_opt_rename(all_opts
, "iops_wr_max", "throttling.iops-write-max");
782 qemu_opt_rename(all_opts
, "bps_max", "throttling.bps-total-max");
783 qemu_opt_rename(all_opts
, "bps_rd_max", "throttling.bps-read-max");
784 qemu_opt_rename(all_opts
, "bps_wr_max", "throttling.bps-write-max");
786 qemu_opt_rename(all_opts
,
787 "iops_size", "throttling.iops-size");
789 qemu_opt_rename(all_opts
, "readonly", "read-only");
791 value
= qemu_opt_get(all_opts
, "cache");
795 if (bdrv_parse_cache_flags(value
, &flags
) != 0) {
796 error_report("invalid cache option");
800 /* Specific options take precedence */
801 if (!qemu_opt_get(all_opts
, "cache.writeback")) {
802 qemu_opt_set_bool(all_opts
, "cache.writeback",
803 !!(flags
& BDRV_O_CACHE_WB
));
805 if (!qemu_opt_get(all_opts
, "cache.direct")) {
806 qemu_opt_set_bool(all_opts
, "cache.direct",
807 !!(flags
& BDRV_O_NOCACHE
));
809 if (!qemu_opt_get(all_opts
, "cache.no-flush")) {
810 qemu_opt_set_bool(all_opts
, "cache.no-flush",
811 !!(flags
& BDRV_O_NO_FLUSH
));
813 qemu_opt_unset(all_opts
, "cache");
816 /* Get a QDict for processing the options */
817 bs_opts
= qdict_new();
818 qemu_opts_to_qdict(all_opts
, bs_opts
);
820 legacy_opts
= qemu_opts_create_nofail(&qemu_legacy_drive_opts
);
821 qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, &local_err
);
822 if (error_is_set(&local_err
)) {
823 qerror_report_err(local_err
);
824 error_free(local_err
);
829 value
= qemu_opt_get(legacy_opts
, "media");
831 if (!strcmp(value
, "disk")) {
833 } else if (!strcmp(value
, "cdrom")) {
836 error_report("'%s' invalid media", value
);
841 /* Actual block device init: Functionality shared with blockdev-add */
842 dinfo
= blockdev_init(bs_opts
, block_default_type
, media
);
847 /* Set legacy DriveInfo fields */
848 dinfo
->enable_auto_del
= true;
849 dinfo
->opts
= all_opts
;
852 qemu_opts_del(legacy_opts
);
856 void do_commit(Monitor
*mon
, const QDict
*qdict
)
858 const char *device
= qdict_get_str(qdict
, "device");
859 BlockDriverState
*bs
;
862 if (!strcmp(device
, "all")) {
863 ret
= bdrv_commit_all();
865 bs
= bdrv_find(device
);
867 monitor_printf(mon
, "Device '%s' not found\n", device
);
870 ret
= bdrv_commit(bs
);
873 monitor_printf(mon
, "'commit' error for '%s': %s\n", device
,
878 static void blockdev_do_action(int kind
, void *data
, Error
**errp
)
880 TransactionAction action
;
881 TransactionActionList list
;
885 list
.value
= &action
;
887 qmp_transaction(&list
, errp
);
890 void qmp_blockdev_snapshot_sync(const char *device
, const char *snapshot_file
,
891 bool has_format
, const char *format
,
892 bool has_mode
, enum NewImageMode mode
,
895 BlockdevSnapshot snapshot
= {
896 .device
= (char *) device
,
897 .snapshot_file
= (char *) snapshot_file
,
898 .has_format
= has_format
,
899 .format
= (char *) format
,
900 .has_mode
= has_mode
,
903 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
907 void qmp_blockdev_snapshot_internal_sync(const char *device
,
911 BlockdevSnapshotInternal snapshot
= {
912 .device
= (char *) device
,
913 .name
= (char *) name
916 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
920 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
927 BlockDriverState
*bs
= bdrv_find(device
);
929 Error
*local_err
= NULL
;
930 SnapshotInfo
*info
= NULL
;
934 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
947 error_setg(errp
, "Name or id must be provided");
951 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
952 if (error_is_set(&local_err
)) {
953 error_propagate(errp
, local_err
);
958 "Snapshot with id '%s' and name '%s' does not exist on "
960 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
964 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
965 if (error_is_set(&local_err
)) {
966 error_propagate(errp
, local_err
);
970 info
= g_malloc0(sizeof(SnapshotInfo
));
971 info
->id
= g_strdup(sn
.id_str
);
972 info
->name
= g_strdup(sn
.name
);
973 info
->date_nsec
= sn
.date_nsec
;
974 info
->date_sec
= sn
.date_sec
;
975 info
->vm_state_size
= sn
.vm_state_size
;
976 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
977 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
982 /* New and old BlockDriverState structs for group snapshots */
984 typedef struct BlkTransactionState BlkTransactionState
;
986 /* Only prepare() may fail. In a single transaction, only one of commit() or
987 abort() will be called, clean() will always be called if it present. */
988 typedef struct BdrvActionOps
{
989 /* Size of state struct, in bytes. */
990 size_t instance_size
;
991 /* Prepare the work, must NOT be NULL. */
992 void (*prepare
)(BlkTransactionState
*common
, Error
**errp
);
993 /* Commit the changes, can be NULL. */
994 void (*commit
)(BlkTransactionState
*common
);
995 /* Abort the changes on fail, can be NULL. */
996 void (*abort
)(BlkTransactionState
*common
);
997 /* Clean up resource in the end, can be NULL. */
998 void (*clean
)(BlkTransactionState
*common
);
1002 * This structure must be arranged as first member in child type, assuming
1003 * that compiler will also arrange it to the same address with parent instance.
1004 * Later it will be used in free().
1006 struct BlkTransactionState
{
1007 TransactionAction
*action
;
1008 const BdrvActionOps
*ops
;
1009 QSIMPLEQ_ENTRY(BlkTransactionState
) entry
;
1012 /* internal snapshot private data */
1013 typedef struct InternalSnapshotState
{
1014 BlkTransactionState common
;
1015 BlockDriverState
*bs
;
1016 QEMUSnapshotInfo sn
;
1017 } InternalSnapshotState
;
1019 static void internal_snapshot_prepare(BlkTransactionState
*common
,
1024 BlockDriverState
*bs
;
1025 QEMUSnapshotInfo old_sn
, *sn
;
1028 BlockdevSnapshotInternal
*internal
;
1029 InternalSnapshotState
*state
;
1032 g_assert(common
->action
->kind
==
1033 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1034 internal
= common
->action
->blockdev_snapshot_internal_sync
;
1035 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1037 /* 1. parse input */
1038 device
= internal
->device
;
1039 name
= internal
->name
;
1041 /* 2. check for validation */
1042 bs
= bdrv_find(device
);
1044 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1048 if (!bdrv_is_inserted(bs
)) {
1049 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1053 if (bdrv_is_read_only(bs
)) {
1054 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1058 if (!bdrv_can_snapshot(bs
)) {
1059 error_set(errp
, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
1060 bs
->drv
->format_name
, device
, "internal snapshot");
1064 if (!strlen(name
)) {
1065 error_setg(errp
, "Name is empty");
1069 /* check whether a snapshot with name exist */
1070 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
, errp
);
1071 if (error_is_set(errp
)) {
1075 "Snapshot with name '%s' already exists on device '%s'",
1080 /* 3. take the snapshot */
1082 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1083 qemu_gettimeofday(&tv
);
1084 sn
->date_sec
= tv
.tv_sec
;
1085 sn
->date_nsec
= tv
.tv_usec
* 1000;
1086 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1088 ret1
= bdrv_snapshot_create(bs
, sn
);
1090 error_setg_errno(errp
, -ret1
,
1091 "Failed to create snapshot '%s' on device '%s'",
1096 /* 4. succeed, mark a snapshot is created */
1100 static void internal_snapshot_abort(BlkTransactionState
*common
)
1102 InternalSnapshotState
*state
=
1103 DO_UPCAST(InternalSnapshotState
, common
, common
);
1104 BlockDriverState
*bs
= state
->bs
;
1105 QEMUSnapshotInfo
*sn
= &state
->sn
;
1106 Error
*local_error
= NULL
;
1112 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1113 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1114 "device '%s' in abort: %s",
1117 bdrv_get_device_name(bs
),
1118 error_get_pretty(local_error
));
1119 error_free(local_error
);
1123 /* external snapshot private data */
1124 typedef struct ExternalSnapshotState
{
1125 BlkTransactionState common
;
1126 BlockDriverState
*old_bs
;
1127 BlockDriverState
*new_bs
;
1128 } ExternalSnapshotState
;
1130 static void external_snapshot_prepare(BlkTransactionState
*common
,
1135 Error
*local_err
= NULL
;
1137 const char *new_image_file
;
1138 const char *format
= "qcow2";
1139 enum NewImageMode mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1140 ExternalSnapshotState
*state
=
1141 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1142 TransactionAction
*action
= common
->action
;
1144 /* get parameters */
1145 g_assert(action
->kind
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
);
1147 device
= action
->blockdev_snapshot_sync
->device
;
1148 new_image_file
= action
->blockdev_snapshot_sync
->snapshot_file
;
1149 if (action
->blockdev_snapshot_sync
->has_format
) {
1150 format
= action
->blockdev_snapshot_sync
->format
;
1152 if (action
->blockdev_snapshot_sync
->has_mode
) {
1153 mode
= action
->blockdev_snapshot_sync
->mode
;
1156 /* start processing */
1157 drv
= bdrv_find_format(format
);
1159 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1163 state
->old_bs
= bdrv_find(device
);
1164 if (!state
->old_bs
) {
1165 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1169 if (!bdrv_is_inserted(state
->old_bs
)) {
1170 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1174 if (bdrv_in_use(state
->old_bs
)) {
1175 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1179 if (!bdrv_is_read_only(state
->old_bs
)) {
1180 if (bdrv_flush(state
->old_bs
)) {
1181 error_set(errp
, QERR_IO_ERROR
);
1186 if (bdrv_check_ext_snapshot(state
->old_bs
) != EXT_SNAPSHOT_ALLOWED
) {
1187 error_set(errp
, QERR_FEATURE_DISABLED
, "snapshot");
1191 flags
= state
->old_bs
->open_flags
;
1193 /* create new image w/backing file */
1194 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1195 bdrv_img_create(new_image_file
, format
,
1196 state
->old_bs
->filename
,
1197 state
->old_bs
->drv
->format_name
,
1198 NULL
, -1, flags
, &local_err
, false);
1199 if (error_is_set(&local_err
)) {
1200 error_propagate(errp
, local_err
);
1205 /* We will manually add the backing_hd field to the bs later */
1206 state
->new_bs
= bdrv_new("");
1207 /* TODO Inherit bs->options or only take explicit options with an
1208 * extended QMP command? */
1209 ret
= bdrv_open(state
->new_bs
, new_image_file
, NULL
,
1210 flags
| BDRV_O_NO_BACKING
, drv
, &local_err
);
1212 error_propagate(errp
, local_err
);
1216 static void external_snapshot_commit(BlkTransactionState
*common
)
1218 ExternalSnapshotState
*state
=
1219 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1221 /* This removes our old bs and adds the new bs */
1222 bdrv_append(state
->new_bs
, state
->old_bs
);
1223 /* We don't need (or want) to use the transactional
1224 * bdrv_reopen_multiple() across all the entries at once, because we
1225 * don't want to abort all of them if one of them fails the reopen */
1226 bdrv_reopen(state
->new_bs
, state
->new_bs
->open_flags
& ~BDRV_O_RDWR
,
1230 static void external_snapshot_abort(BlkTransactionState
*common
)
1232 ExternalSnapshotState
*state
=
1233 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1234 if (state
->new_bs
) {
1235 bdrv_unref(state
->new_bs
);
1239 typedef struct DriveBackupState
{
1240 BlkTransactionState common
;
1241 BlockDriverState
*bs
;
1245 static void drive_backup_prepare(BlkTransactionState
*common
, Error
**errp
)
1247 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1248 DriveBackup
*backup
;
1249 Error
*local_err
= NULL
;
1251 assert(common
->action
->kind
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1252 backup
= common
->action
->drive_backup
;
1254 qmp_drive_backup(backup
->device
, backup
->target
,
1255 backup
->has_format
, backup
->format
,
1257 backup
->has_mode
, backup
->mode
,
1258 backup
->has_speed
, backup
->speed
,
1259 backup
->has_on_source_error
, backup
->on_source_error
,
1260 backup
->has_on_target_error
, backup
->on_target_error
,
1262 if (error_is_set(&local_err
)) {
1263 error_propagate(errp
, local_err
);
1269 state
->bs
= bdrv_find(backup
->device
);
1270 state
->job
= state
->bs
->job
;
1273 static void drive_backup_abort(BlkTransactionState
*common
)
1275 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1276 BlockDriverState
*bs
= state
->bs
;
1278 /* Only cancel if it's the job we started */
1279 if (bs
&& bs
->job
&& bs
->job
== state
->job
) {
1280 block_job_cancel_sync(bs
->job
);
1284 static void abort_prepare(BlkTransactionState
*common
, Error
**errp
)
1286 error_setg(errp
, "Transaction aborted using Abort action");
1289 static void abort_commit(BlkTransactionState
*common
)
1291 g_assert_not_reached(); /* this action never succeeds */
1294 static const BdrvActionOps actions
[] = {
1295 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
1296 .instance_size
= sizeof(ExternalSnapshotState
),
1297 .prepare
= external_snapshot_prepare
,
1298 .commit
= external_snapshot_commit
,
1299 .abort
= external_snapshot_abort
,
1301 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
1302 .instance_size
= sizeof(DriveBackupState
),
1303 .prepare
= drive_backup_prepare
,
1304 .abort
= drive_backup_abort
,
1306 [TRANSACTION_ACTION_KIND_ABORT
] = {
1307 .instance_size
= sizeof(BlkTransactionState
),
1308 .prepare
= abort_prepare
,
1309 .commit
= abort_commit
,
1311 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
1312 .instance_size
= sizeof(InternalSnapshotState
),
1313 .prepare
= internal_snapshot_prepare
,
1314 .abort
= internal_snapshot_abort
,
1319 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1320 * then we do not pivot any of the devices in the group, and abandon the
1323 void qmp_transaction(TransactionActionList
*dev_list
, Error
**errp
)
1325 TransactionActionList
*dev_entry
= dev_list
;
1326 BlkTransactionState
*state
, *next
;
1327 Error
*local_err
= NULL
;
1329 QSIMPLEQ_HEAD(snap_bdrv_states
, BlkTransactionState
) snap_bdrv_states
;
1330 QSIMPLEQ_INIT(&snap_bdrv_states
);
1332 /* drain all i/o before any snapshots */
1335 /* We don't do anything in this loop that commits us to the snapshot */
1336 while (NULL
!= dev_entry
) {
1337 TransactionAction
*dev_info
= NULL
;
1338 const BdrvActionOps
*ops
;
1340 dev_info
= dev_entry
->value
;
1341 dev_entry
= dev_entry
->next
;
1343 assert(dev_info
->kind
< ARRAY_SIZE(actions
));
1345 ops
= &actions
[dev_info
->kind
];
1346 assert(ops
->instance_size
> 0);
1348 state
= g_malloc0(ops
->instance_size
);
1350 state
->action
= dev_info
;
1351 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
1353 state
->ops
->prepare(state
, &local_err
);
1354 if (error_is_set(&local_err
)) {
1355 error_propagate(errp
, local_err
);
1356 goto delete_and_fail
;
1360 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1361 if (state
->ops
->commit
) {
1362 state
->ops
->commit(state
);
1371 * failure, and it is all-or-none; abandon each new bs, and keep using
1372 * the original bs for all images
1374 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1375 if (state
->ops
->abort
) {
1376 state
->ops
->abort(state
);
1380 QSIMPLEQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
1381 if (state
->ops
->clean
) {
1382 state
->ops
->clean(state
);
1389 static void eject_device(BlockDriverState
*bs
, int force
, Error
**errp
)
1391 if (bdrv_in_use(bs
)) {
1392 error_set(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
1395 if (!bdrv_dev_has_removable_media(bs
)) {
1396 error_set(errp
, QERR_DEVICE_NOT_REMOVABLE
, bdrv_get_device_name(bs
));
1400 if (bdrv_dev_is_medium_locked(bs
) && !bdrv_dev_is_tray_open(bs
)) {
1401 bdrv_dev_eject_request(bs
, force
);
1403 error_set(errp
, QERR_DEVICE_LOCKED
, bdrv_get_device_name(bs
));
1411 void qmp_eject(const char *device
, bool has_force
, bool force
, Error
**errp
)
1413 BlockDriverState
*bs
;
1415 bs
= bdrv_find(device
);
1417 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1421 eject_device(bs
, force
, errp
);
1424 void qmp_block_passwd(const char *device
, const char *password
, Error
**errp
)
1426 BlockDriverState
*bs
;
1429 bs
= bdrv_find(device
);
1431 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1435 err
= bdrv_set_key(bs
, password
);
1436 if (err
== -EINVAL
) {
1437 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1439 } else if (err
< 0) {
1440 error_set(errp
, QERR_INVALID_PASSWORD
);
1445 static void qmp_bdrv_open_encrypted(BlockDriverState
*bs
, const char *filename
,
1446 int bdrv_flags
, BlockDriver
*drv
,
1447 const char *password
, Error
**errp
)
1449 Error
*local_err
= NULL
;
1452 ret
= bdrv_open(bs
, filename
, NULL
, bdrv_flags
, drv
, &local_err
);
1454 error_propagate(errp
, local_err
);
1458 if (bdrv_key_required(bs
)) {
1460 if (bdrv_set_key(bs
, password
) < 0) {
1461 error_set(errp
, QERR_INVALID_PASSWORD
);
1464 error_set(errp
, QERR_DEVICE_ENCRYPTED
, bdrv_get_device_name(bs
),
1465 bdrv_get_encrypted_filename(bs
));
1467 } else if (password
) {
1468 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1472 void qmp_change_blockdev(const char *device
, const char *filename
,
1473 bool has_format
, const char *format
, Error
**errp
)
1475 BlockDriverState
*bs
;
1476 BlockDriver
*drv
= NULL
;
1480 bs
= bdrv_find(device
);
1482 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1487 drv
= bdrv_find_whitelisted_format(format
, bs
->read_only
);
1489 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1494 eject_device(bs
, 0, &err
);
1495 if (error_is_set(&err
)) {
1496 error_propagate(errp
, err
);
1500 bdrv_flags
= bdrv_is_read_only(bs
) ? 0 : BDRV_O_RDWR
;
1501 bdrv_flags
|= bdrv_is_snapshot(bs
) ? BDRV_O_SNAPSHOT
: 0;
1503 qmp_bdrv_open_encrypted(bs
, filename
, bdrv_flags
, drv
, NULL
, errp
);
1506 /* throttling disk I/O limits */
1507 void qmp_block_set_io_throttle(const char *device
, int64_t bps
, int64_t bps_rd
,
1514 bool has_bps_rd_max
,
1516 bool has_bps_wr_max
,
1520 bool has_iops_rd_max
,
1521 int64_t iops_rd_max
,
1522 bool has_iops_wr_max
,
1523 int64_t iops_wr_max
,
1525 int64_t iops_size
, Error
**errp
)
1528 BlockDriverState
*bs
;
1530 bs
= bdrv_find(device
);
1532 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1536 memset(&cfg
, 0, sizeof(cfg
));
1537 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
= bps
;
1538 cfg
.buckets
[THROTTLE_BPS_READ
].avg
= bps_rd
;
1539 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
= bps_wr
;
1541 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
= iops
;
1542 cfg
.buckets
[THROTTLE_OPS_READ
].avg
= iops_rd
;
1543 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
= iops_wr
;
1546 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
= bps_max
;
1548 if (has_bps_rd_max
) {
1549 cfg
.buckets
[THROTTLE_BPS_READ
].max
= bps_rd_max
;
1551 if (has_bps_wr_max
) {
1552 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
= bps_wr_max
;
1555 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
= iops_max
;
1557 if (has_iops_rd_max
) {
1558 cfg
.buckets
[THROTTLE_OPS_READ
].max
= iops_rd_max
;
1560 if (has_iops_wr_max
) {
1561 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
= iops_wr_max
;
1564 if (has_iops_size
) {
1565 cfg
.op_size
= iops_size
;
1568 if (!check_throttle_config(&cfg
, errp
)) {
1572 if (!bs
->io_limits_enabled
&& throttle_enabled(&cfg
)) {
1573 bdrv_io_limits_enable(bs
);
1574 } else if (bs
->io_limits_enabled
&& !throttle_enabled(&cfg
)) {
1575 bdrv_io_limits_disable(bs
);
1578 if (bs
->io_limits_enabled
) {
1579 bdrv_set_io_limits(bs
, &cfg
);
1583 int do_drive_del(Monitor
*mon
, const QDict
*qdict
, QObject
**ret_data
)
1585 const char *id
= qdict_get_str(qdict
, "id");
1586 BlockDriverState
*bs
;
1590 qerror_report(QERR_DEVICE_NOT_FOUND
, id
);
1593 if (bdrv_in_use(bs
)) {
1594 qerror_report(QERR_DEVICE_IN_USE
, id
);
1598 /* quiesce block driver; prevent further io */
1603 /* if we have a device attached to this BlockDriverState
1604 * then we need to make the drive anonymous until the device
1605 * can be removed. If this is a drive with no device backing
1606 * then we can just get rid of the block driver state right here.
1608 if (bdrv_get_attached_dev(bs
)) {
1611 /* Further I/O must not pause the guest */
1612 bdrv_set_on_error(bs
, BLOCKDEV_ON_ERROR_REPORT
,
1613 BLOCKDEV_ON_ERROR_REPORT
);
1615 drive_uninit(drive_get_by_blockdev(bs
));
1621 void qmp_block_resize(const char *device
, int64_t size
, Error
**errp
)
1623 BlockDriverState
*bs
;
1626 bs
= bdrv_find(device
);
1628 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1633 error_set(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
1637 /* complete all in-flight operations before resizing the device */
1640 ret
= bdrv_truncate(bs
, size
);
1645 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1648 error_set(errp
, QERR_UNSUPPORTED
);
1651 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1654 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1657 error_setg_errno(errp
, -ret
, "Could not resize");
1662 static void block_job_cb(void *opaque
, int ret
)
1664 BlockDriverState
*bs
= opaque
;
1667 trace_block_job_cb(bs
, bs
->job
, ret
);
1670 obj
= qobject_from_block_job(bs
->job
);
1672 QDict
*dict
= qobject_to_qdict(obj
);
1673 qdict_put(dict
, "error", qstring_from_str(strerror(-ret
)));
1676 if (block_job_is_cancelled(bs
->job
)) {
1677 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED
, obj
);
1679 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED
, obj
);
1681 qobject_decref(obj
);
1683 bdrv_put_ref_bh_schedule(bs
);
1686 void qmp_block_stream(const char *device
, bool has_base
,
1687 const char *base
, bool has_speed
, int64_t speed
,
1688 bool has_on_error
, BlockdevOnError on_error
,
1691 BlockDriverState
*bs
;
1692 BlockDriverState
*base_bs
= NULL
;
1693 Error
*local_err
= NULL
;
1695 if (!has_on_error
) {
1696 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1699 bs
= bdrv_find(device
);
1701 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1706 base_bs
= bdrv_find_backing_image(bs
, base
);
1707 if (base_bs
== NULL
) {
1708 error_set(errp
, QERR_BASE_NOT_FOUND
, base
);
1713 stream_start(bs
, base_bs
, base
, has_speed
? speed
: 0,
1714 on_error
, block_job_cb
, bs
, &local_err
);
1715 if (error_is_set(&local_err
)) {
1716 error_propagate(errp
, local_err
);
1720 trace_qmp_block_stream(bs
, bs
->job
);
1723 void qmp_block_commit(const char *device
,
1724 bool has_base
, const char *base
, const char *top
,
1725 bool has_speed
, int64_t speed
,
1728 BlockDriverState
*bs
;
1729 BlockDriverState
*base_bs
, *top_bs
;
1730 Error
*local_err
= NULL
;
1731 /* This will be part of the QMP command, if/when the
1732 * BlockdevOnError change for blkmirror makes it in
1734 BlockdevOnError on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1736 /* drain all i/o before commits */
1739 bs
= bdrv_find(device
);
1741 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1745 /* default top_bs is the active layer */
1749 if (strcmp(bs
->filename
, top
) != 0) {
1750 top_bs
= bdrv_find_backing_image(bs
, top
);
1754 if (top_bs
== NULL
) {
1755 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
1759 if (has_base
&& base
) {
1760 base_bs
= bdrv_find_backing_image(top_bs
, base
);
1762 base_bs
= bdrv_find_base(top_bs
);
1765 if (base_bs
== NULL
) {
1766 error_set(errp
, QERR_BASE_NOT_FOUND
, base
? base
: "NULL");
1770 commit_start(bs
, base_bs
, top_bs
, speed
, on_error
, block_job_cb
, bs
,
1772 if (local_err
!= NULL
) {
1773 error_propagate(errp
, local_err
);
1778 void qmp_drive_backup(const char *device
, const char *target
,
1779 bool has_format
, const char *format
,
1780 enum MirrorSyncMode sync
,
1781 bool has_mode
, enum NewImageMode mode
,
1782 bool has_speed
, int64_t speed
,
1783 bool has_on_source_error
, BlockdevOnError on_source_error
,
1784 bool has_on_target_error
, BlockdevOnError on_target_error
,
1787 BlockDriverState
*bs
;
1788 BlockDriverState
*target_bs
;
1789 BlockDriverState
*source
= NULL
;
1790 BlockDriver
*drv
= NULL
;
1791 Error
*local_err
= NULL
;
1799 if (!has_on_source_error
) {
1800 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1802 if (!has_on_target_error
) {
1803 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1806 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1809 bs
= bdrv_find(device
);
1811 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1815 if (!bdrv_is_inserted(bs
)) {
1816 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1821 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1824 drv
= bdrv_find_format(format
);
1826 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1831 if (bdrv_in_use(bs
)) {
1832 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1836 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1838 /* See if we have a backing HD we can use to create our new image
1840 if (sync
== MIRROR_SYNC_MODE_TOP
) {
1841 source
= bs
->backing_hd
;
1843 sync
= MIRROR_SYNC_MODE_FULL
;
1846 if (sync
== MIRROR_SYNC_MODE_NONE
) {
1850 size
= bdrv_getlength(bs
);
1852 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1856 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1857 assert(format
&& drv
);
1859 bdrv_img_create(target
, format
, source
->filename
,
1860 source
->drv
->format_name
, NULL
,
1861 size
, flags
, &local_err
, false);
1863 bdrv_img_create(target
, format
, NULL
, NULL
, NULL
,
1864 size
, flags
, &local_err
, false);
1868 if (error_is_set(&local_err
)) {
1869 error_propagate(errp
, local_err
);
1873 target_bs
= bdrv_new("");
1874 ret
= bdrv_open(target_bs
, target
, NULL
, flags
, drv
, &local_err
);
1876 bdrv_unref(target_bs
);
1877 error_propagate(errp
, local_err
);
1881 backup_start(bs
, target_bs
, speed
, sync
, on_source_error
, on_target_error
,
1882 block_job_cb
, bs
, &local_err
);
1883 if (local_err
!= NULL
) {
1884 bdrv_unref(target_bs
);
1885 error_propagate(errp
, local_err
);
1890 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1892 void qmp_drive_mirror(const char *device
, const char *target
,
1893 bool has_format
, const char *format
,
1894 enum MirrorSyncMode sync
,
1895 bool has_mode
, enum NewImageMode mode
,
1896 bool has_speed
, int64_t speed
,
1897 bool has_granularity
, uint32_t granularity
,
1898 bool has_buf_size
, int64_t buf_size
,
1899 bool has_on_source_error
, BlockdevOnError on_source_error
,
1900 bool has_on_target_error
, BlockdevOnError on_target_error
,
1903 BlockDriverState
*bs
;
1904 BlockDriverState
*source
, *target_bs
;
1905 BlockDriver
*drv
= NULL
;
1906 Error
*local_err
= NULL
;
1914 if (!has_on_source_error
) {
1915 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1917 if (!has_on_target_error
) {
1918 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1921 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1923 if (!has_granularity
) {
1926 if (!has_buf_size
) {
1927 buf_size
= DEFAULT_MIRROR_BUF_SIZE
;
1930 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
1931 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1934 if (granularity
& (granularity
- 1)) {
1935 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1939 bs
= bdrv_find(device
);
1941 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1945 if (!bdrv_is_inserted(bs
)) {
1946 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1951 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1954 drv
= bdrv_find_format(format
);
1956 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1961 if (bdrv_in_use(bs
)) {
1962 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1966 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1967 source
= bs
->backing_hd
;
1968 if (!source
&& sync
== MIRROR_SYNC_MODE_TOP
) {
1969 sync
= MIRROR_SYNC_MODE_FULL
;
1972 size
= bdrv_getlength(bs
);
1974 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1978 if (sync
== MIRROR_SYNC_MODE_FULL
&& mode
!= NEW_IMAGE_MODE_EXISTING
) {
1979 /* create new image w/o backing file */
1980 assert(format
&& drv
);
1981 bdrv_img_create(target
, format
,
1982 NULL
, NULL
, NULL
, size
, flags
, &local_err
, false);
1985 case NEW_IMAGE_MODE_EXISTING
:
1987 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
1988 /* create new image with backing file */
1989 bdrv_img_create(target
, format
,
1991 source
->drv
->format_name
,
1992 NULL
, size
, flags
, &local_err
, false);
1999 if (error_is_set(&local_err
)) {
2000 error_propagate(errp
, local_err
);
2004 /* Mirroring takes care of copy-on-write using the source's backing
2007 target_bs
= bdrv_new("");
2008 ret
= bdrv_open(target_bs
, target
, NULL
, flags
| BDRV_O_NO_BACKING
, drv
,
2011 bdrv_unref(target_bs
);
2012 error_propagate(errp
, local_err
);
2016 mirror_start(bs
, target_bs
, speed
, granularity
, buf_size
, sync
,
2017 on_source_error
, on_target_error
,
2018 block_job_cb
, bs
, &local_err
);
2019 if (local_err
!= NULL
) {
2020 bdrv_unref(target_bs
);
2021 error_propagate(errp
, local_err
);
2026 static BlockJob
*find_block_job(const char *device
)
2028 BlockDriverState
*bs
;
2030 bs
= bdrv_find(device
);
2031 if (!bs
|| !bs
->job
) {
2037 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
2039 BlockJob
*job
= find_block_job(device
);
2042 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2046 block_job_set_speed(job
, speed
, errp
);
2049 void qmp_block_job_cancel(const char *device
,
2050 bool has_force
, bool force
, Error
**errp
)
2052 BlockJob
*job
= find_block_job(device
);
2059 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2062 if (job
->paused
&& !force
) {
2063 error_set(errp
, QERR_BLOCK_JOB_PAUSED
, device
);
2067 trace_qmp_block_job_cancel(job
);
2068 block_job_cancel(job
);
2071 void qmp_block_job_pause(const char *device
, Error
**errp
)
2073 BlockJob
*job
= find_block_job(device
);
2076 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2080 trace_qmp_block_job_pause(job
);
2081 block_job_pause(job
);
2084 void qmp_block_job_resume(const char *device
, Error
**errp
)
2086 BlockJob
*job
= find_block_job(device
);
2089 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2093 trace_qmp_block_job_resume(job
);
2094 block_job_resume(job
);
2097 void qmp_block_job_complete(const char *device
, Error
**errp
)
2099 BlockJob
*job
= find_block_job(device
);
2102 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2106 trace_qmp_block_job_complete(job
);
2107 block_job_complete(job
, errp
);
2110 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
2112 QmpOutputVisitor
*ov
= qmp_output_visitor_new();
2116 Error
*local_err
= NULL
;
2118 /* Require an ID in the top level */
2119 if (!options
->has_id
) {
2120 error_setg(errp
, "Block device needs an ID");
2124 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2125 * cache.direct=false instead of silently switching to aio=threads, except
2126 * if called from drive_init.
2128 * For now, simply forbidding the combination for all drivers will do. */
2129 if (options
->has_aio
&& options
->aio
== BLOCKDEV_AIO_OPTIONS_NATIVE
) {
2130 bool direct
= options
->cache
->has_direct
&& options
->cache
->direct
;
2131 if (!options
->has_cache
&& !direct
) {
2132 error_setg(errp
, "aio=native requires cache.direct=true");
2137 visit_type_BlockdevOptions(qmp_output_get_visitor(ov
),
2138 &options
, NULL
, &local_err
);
2139 if (error_is_set(&local_err
)) {
2140 error_propagate(errp
, local_err
);
2144 obj
= qmp_output_get_qobject(ov
);
2145 qdict
= qobject_to_qdict(obj
);
2147 qdict_flatten(qdict
);
2149 dinfo
= blockdev_init(qdict
, IF_NONE
, MEDIA_DISK
);
2151 error_setg(errp
, "Could not open image");
2156 qmp_output_visitor_cleanup(ov
);
2159 static void do_qmp_query_block_jobs_one(void *opaque
, BlockDriverState
*bs
)
2161 BlockJobInfoList
**prev
= opaque
;
2162 BlockJob
*job
= bs
->job
;
2165 BlockJobInfoList
*elem
= g_new0(BlockJobInfoList
, 1);
2166 elem
->value
= block_job_query(bs
->job
);
2167 (*prev
)->next
= elem
;
2172 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
2174 /* Dummy is a fake list element for holding the head pointer */
2175 BlockJobInfoList dummy
= {};
2176 BlockJobInfoList
*prev
= &dummy
;
2177 bdrv_iterate(do_qmp_query_block_jobs_one
, &prev
);
2181 QemuOptsList qemu_common_drive_opts
= {
2183 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
2187 .type
= QEMU_OPT_NUMBER
,
2188 .help
= "bus number",
2191 .type
= QEMU_OPT_NUMBER
,
2192 .help
= "unit number (i.e. lun for scsi)",
2195 .type
= QEMU_OPT_STRING
,
2196 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
2199 .type
= QEMU_OPT_NUMBER
,
2200 .help
= "index number",
2203 .type
= QEMU_OPT_NUMBER
,
2204 .help
= "number of cylinders (ide disk geometry)",
2207 .type
= QEMU_OPT_NUMBER
,
2208 .help
= "number of heads (ide disk geometry)",
2211 .type
= QEMU_OPT_NUMBER
,
2212 .help
= "number of sectors (ide disk geometry)",
2215 .type
= QEMU_OPT_STRING
,
2216 .help
= "chs translation (auto, lba. none)",
2219 .type
= QEMU_OPT_BOOL
,
2220 .help
= "enable/disable snapshot mode",
2223 .type
= QEMU_OPT_STRING
,
2224 .help
= "disk image",
2227 .type
= QEMU_OPT_STRING
,
2228 .help
= "discard operation (ignore/off, unmap/on)",
2230 .name
= "cache.writeback",
2231 .type
= QEMU_OPT_BOOL
,
2232 .help
= "enables writeback mode for any caches",
2234 .name
= "cache.direct",
2235 .type
= QEMU_OPT_BOOL
,
2236 .help
= "enables use of O_DIRECT (bypass the host page cache)",
2238 .name
= "cache.no-flush",
2239 .type
= QEMU_OPT_BOOL
,
2240 .help
= "ignore any flush requests for the device",
2243 .type
= QEMU_OPT_STRING
,
2244 .help
= "host AIO implementation (threads, native)",
2247 .type
= QEMU_OPT_STRING
,
2248 .help
= "disk format (raw, qcow2, ...)",
2251 .type
= QEMU_OPT_STRING
,
2252 .help
= "disk serial number",
2255 .type
= QEMU_OPT_STRING
,
2256 .help
= "read error action",
2259 .type
= QEMU_OPT_STRING
,
2260 .help
= "write error action",
2263 .type
= QEMU_OPT_STRING
,
2264 .help
= "pci address (virtio only)",
2266 .name
= "read-only",
2267 .type
= QEMU_OPT_BOOL
,
2268 .help
= "open drive file as read-only",
2270 .name
= "throttling.iops-total",
2271 .type
= QEMU_OPT_NUMBER
,
2272 .help
= "limit total I/O operations per second",
2274 .name
= "throttling.iops-read",
2275 .type
= QEMU_OPT_NUMBER
,
2276 .help
= "limit read operations per second",
2278 .name
= "throttling.iops-write",
2279 .type
= QEMU_OPT_NUMBER
,
2280 .help
= "limit write operations per second",
2282 .name
= "throttling.bps-total",
2283 .type
= QEMU_OPT_NUMBER
,
2284 .help
= "limit total bytes per second",
2286 .name
= "throttling.bps-read",
2287 .type
= QEMU_OPT_NUMBER
,
2288 .help
= "limit read bytes per second",
2290 .name
= "throttling.bps-write",
2291 .type
= QEMU_OPT_NUMBER
,
2292 .help
= "limit write bytes per second",
2294 .name
= "throttling.iops-total-max",
2295 .type
= QEMU_OPT_NUMBER
,
2296 .help
= "I/O operations burst",
2298 .name
= "throttling.iops-read-max",
2299 .type
= QEMU_OPT_NUMBER
,
2300 .help
= "I/O operations read burst",
2302 .name
= "throttling.iops-write-max",
2303 .type
= QEMU_OPT_NUMBER
,
2304 .help
= "I/O operations write burst",
2306 .name
= "throttling.bps-total-max",
2307 .type
= QEMU_OPT_NUMBER
,
2308 .help
= "total bytes burst",
2310 .name
= "throttling.bps-read-max",
2311 .type
= QEMU_OPT_NUMBER
,
2312 .help
= "total bytes read burst",
2314 .name
= "throttling.bps-write-max",
2315 .type
= QEMU_OPT_NUMBER
,
2316 .help
= "total bytes write burst",
2318 .name
= "throttling.iops-size",
2319 .type
= QEMU_OPT_NUMBER
,
2320 .help
= "when limiting by iops max size of an I/O in bytes",
2322 .name
= "copy-on-read",
2323 .type
= QEMU_OPT_BOOL
,
2324 .help
= "copy read data from backing file into image file",
2327 .type
= QEMU_OPT_BOOL
,
2328 .help
= "(deprecated, ignored)",
2330 { /* end of list */ }
2334 QemuOptsList qemu_drive_opts
= {
2336 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
2339 * no elements => accept any params
2340 * validation will happen later
2342 { /* end of list */ }