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 type
,
313 DriveMediaType media
)
316 const char *file
= NULL
;
318 const char *mediastr
= "";
320 int cyls
, heads
, secs
, translation
;
325 int on_read_error
, on_write_error
;
335 bool has_driver_specific_opts
;
336 BlockDriver
*drv
= NULL
;
338 translation
= BIOS_ATA_TRANSLATION_AUTO
;
340 /* Check common options by copying from bs_opts to opts, all other options
341 * stay in bs_opts for processing by bdrv_open(). */
342 id
= qdict_get_try_str(bs_opts
, "id");
343 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, &error
);
344 if (error_is_set(&error
)) {
345 qerror_report_err(error
);
350 qemu_opts_absorb_qdict(opts
, bs_opts
, &error
);
351 if (error_is_set(&error
)) {
352 qerror_report_err(error
);
358 qdict_del(bs_opts
, "id");
361 has_driver_specific_opts
= !!qdict_size(bs_opts
);
363 /* extract parameters */
364 bus_id
= qemu_opt_get_number(opts
, "bus", 0);
365 unit_id
= qemu_opt_get_number(opts
, "unit", -1);
366 index
= qemu_opt_get_number(opts
, "index", -1);
368 cyls
= qemu_opt_get_number(opts
, "cyls", 0);
369 heads
= qemu_opt_get_number(opts
, "heads", 0);
370 secs
= qemu_opt_get_number(opts
, "secs", 0);
372 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
373 ro
= qemu_opt_get_bool(opts
, "read-only", 0);
374 copy_on_read
= qemu_opt_get_bool(opts
, "copy-on-read", false);
376 file
= qemu_opt_get(opts
, "file");
377 serial
= qemu_opt_get(opts
, "serial");
379 max_devs
= if_max_devs
[type
];
381 if (cyls
|| heads
|| secs
) {
383 error_report("invalid physical cyls number");
387 error_report("invalid physical heads number");
391 error_report("invalid physical secs number");
396 if ((buf
= qemu_opt_get(opts
, "trans")) != NULL
) {
398 error_report("'%s' trans must be used with cyls, heads and secs",
402 if (!strcmp(buf
, "none"))
403 translation
= BIOS_ATA_TRANSLATION_NONE
;
404 else if (!strcmp(buf
, "lba"))
405 translation
= BIOS_ATA_TRANSLATION_LBA
;
406 else if (!strcmp(buf
, "auto"))
407 translation
= BIOS_ATA_TRANSLATION_AUTO
;
409 error_report("'%s' invalid translation type", buf
);
414 if (media
== MEDIA_CDROM
) {
415 if (cyls
|| secs
|| heads
) {
416 error_report("CHS can't be set with media=cdrom");
421 if ((buf
= qemu_opt_get(opts
, "discard")) != NULL
) {
422 if (bdrv_parse_discard_flags(buf
, &bdrv_flags
) != 0) {
423 error_report("invalid discard option");
428 if (qemu_opt_get_bool(opts
, "cache.writeback", true)) {
429 bdrv_flags
|= BDRV_O_CACHE_WB
;
431 if (qemu_opt_get_bool(opts
, "cache.direct", false)) {
432 bdrv_flags
|= BDRV_O_NOCACHE
;
434 if (qemu_opt_get_bool(opts
, "cache.no-flush", false)) {
435 bdrv_flags
|= BDRV_O_NO_FLUSH
;
438 #ifdef CONFIG_LINUX_AIO
439 if ((buf
= qemu_opt_get(opts
, "aio")) != NULL
) {
440 if (!strcmp(buf
, "native")) {
441 bdrv_flags
|= BDRV_O_NATIVE_AIO
;
442 } else if (!strcmp(buf
, "threads")) {
443 /* this is the default */
445 error_report("invalid aio option");
451 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
452 if (is_help_option(buf
)) {
453 error_printf("Supported formats:");
454 bdrv_iterate_format(bdrv_format_print
, NULL
);
459 drv
= bdrv_find_format(buf
);
461 error_report("'%s' invalid format", buf
);
466 /* disk I/O throttling */
467 memset(&cfg
, 0, sizeof(cfg
));
468 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
=
469 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
470 cfg
.buckets
[THROTTLE_BPS_READ
].avg
=
471 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
472 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
=
473 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
474 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
=
475 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
476 cfg
.buckets
[THROTTLE_OPS_READ
].avg
=
477 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
478 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
=
479 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
481 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
=
482 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
483 cfg
.buckets
[THROTTLE_BPS_READ
].max
=
484 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
485 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
=
486 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
487 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
=
488 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
489 cfg
.buckets
[THROTTLE_OPS_READ
].max
=
490 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
491 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
=
492 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
494 cfg
.op_size
= qemu_opt_get_number(opts
, "throttling.iops-size", 0);
496 if (!check_throttle_config(&cfg
, &error
)) {
497 error_report("%s", error_get_pretty(error
));
502 if (qemu_opt_get(opts
, "boot") != NULL
) {
503 fprintf(stderr
, "qemu-kvm: boot=on|off is deprecated and will be "
504 "ignored. Future versions will reject this parameter. Please "
505 "update your scripts.\n");
508 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
509 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
510 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_NONE
) {
511 error_report("werror is not supported by this bus type");
515 on_write_error
= parse_block_error_action(buf
, 0);
516 if (on_write_error
< 0) {
521 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
522 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
523 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&& type
!= IF_NONE
) {
524 error_report("rerror is not supported by this bus type");
528 on_read_error
= parse_block_error_action(buf
, 1);
529 if (on_read_error
< 0) {
534 if ((devaddr
= qemu_opt_get(opts
, "addr")) != NULL
) {
535 if (type
!= IF_VIRTIO
) {
536 error_report("addr is not supported by this bus type");
541 /* compute bus and unit according index */
544 if (bus_id
!= 0 || unit_id
!= -1) {
545 error_report("index cannot be used with bus and unit");
548 bus_id
= drive_index_to_bus_id(type
, index
);
549 unit_id
= drive_index_to_unit_id(type
, index
);
552 /* if user doesn't specify a unit_id,
553 * try to find the first free
558 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
560 if (max_devs
&& unit_id
>= max_devs
) {
569 if (max_devs
&& unit_id
>= max_devs
) {
570 error_report("unit %d too big (max is %d)",
571 unit_id
, max_devs
- 1);
576 * catch multiple definitions
579 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
580 error_report("drive with bus=%d, unit=%d (index=%d) exists",
581 bus_id
, unit_id
, index
);
585 /* no id supplied -> create one */
586 if (qemu_opts_id(opts
) == NULL
) {
588 if (type
== IF_IDE
|| type
== IF_SCSI
) {
589 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
592 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
595 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
598 qemu_opts_set_id(opts
, new_id
);
602 dinfo
= g_malloc0(sizeof(*dinfo
));
603 dinfo
->id
= g_strdup(qemu_opts_id(opts
));
604 dinfo
->bdrv
= bdrv_new(dinfo
->id
);
605 dinfo
->bdrv
->open_flags
= snapshot
? BDRV_O_SNAPSHOT
: 0;
606 dinfo
->bdrv
->read_only
= ro
;
607 dinfo
->devaddr
= devaddr
;
610 dinfo
->unit
= unit_id
;
612 dinfo
->heads
= heads
;
614 dinfo
->trans
= translation
;
616 if (serial
!= NULL
) {
617 dinfo
->serial
= g_strdup(serial
);
619 QTAILQ_INSERT_TAIL(&drives
, dinfo
, next
);
621 bdrv_set_on_error(dinfo
->bdrv
, on_read_error
, on_write_error
);
623 /* disk I/O throttling */
624 if (throttle_enabled(&cfg
)) {
625 bdrv_io_limits_enable(dinfo
->bdrv
);
626 bdrv_set_io_limits(dinfo
->bdrv
, &cfg
);
634 dinfo
->media_cd
= media
== MEDIA_CDROM
;
643 /* add virtio block device */
645 devopts
= qemu_opts_create_nofail(qemu_find_opts("device"));
646 if (arch_type
== QEMU_ARCH_S390X
) {
647 qemu_opt_set(devopts
, "driver", "virtio-blk-s390");
649 qemu_opt_set(devopts
, "driver", "virtio-blk-pci");
651 qemu_opt_set(devopts
, "drive", dinfo
->id
);
653 qemu_opt_set(devopts
, "addr", devaddr
);
659 if (!file
|| !*file
) {
660 if (has_driver_specific_opts
) {
667 /* always use cache=unsafe with snapshot */
668 bdrv_flags
&= ~BDRV_O_CACHE_MASK
;
669 bdrv_flags
|= (BDRV_O_SNAPSHOT
|BDRV_O_CACHE_WB
|BDRV_O_NO_FLUSH
);
673 bdrv_flags
|= BDRV_O_COPY_ON_READ
;
676 if (runstate_check(RUN_STATE_INMIGRATE
)) {
677 bdrv_flags
|= BDRV_O_INCOMING
;
680 if (media
== MEDIA_CDROM
) {
681 /* CDROM is fine for any interface, don't check. */
683 } else if (ro
== 1) {
684 if (type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_FLOPPY
&&
685 type
!= IF_NONE
&& type
!= IF_PFLASH
) {
686 error_report("read-only not supported by this bus type");
691 bdrv_flags
|= ro
? 0 : BDRV_O_RDWR
;
693 if (ro
&& copy_on_read
) {
694 error_report("warning: disabling copy_on_read on read-only drive");
698 ret
= bdrv_open(dinfo
->bdrv
, file
, bs_opts
, bdrv_flags
, drv
, &error
);
701 error_report("could not open disk image %s: %s",
702 file
?: dinfo
->id
, error_get_pretty(error
));
706 if (bdrv_key_required(dinfo
->bdrv
))
717 bdrv_unref(dinfo
->bdrv
);
719 QTAILQ_REMOVE(&drives
, dinfo
, next
);
724 static void qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
)
728 value
= qemu_opt_get(opts
, from
);
730 qemu_opt_set(opts
, to
, value
);
731 qemu_opt_unset(opts
, from
);
735 QemuOptsList qemu_legacy_drive_opts
= {
737 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
741 .type
= QEMU_OPT_STRING
,
742 .help
= "media type (disk, cdrom)",
745 .type
= QEMU_OPT_STRING
,
746 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
748 { /* end of list */ }
752 DriveInfo
*drive_init(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
)
755 DriveInfo
*dinfo
= NULL
;
757 QemuOpts
*legacy_opts
;
758 DriveMediaType media
= MEDIA_DISK
;
759 BlockInterfaceType type
;
760 Error
*local_err
= NULL
;
762 /* Change legacy command line options into QMP ones */
763 qemu_opt_rename(all_opts
, "iops", "throttling.iops-total");
764 qemu_opt_rename(all_opts
, "iops_rd", "throttling.iops-read");
765 qemu_opt_rename(all_opts
, "iops_wr", "throttling.iops-write");
767 qemu_opt_rename(all_opts
, "bps", "throttling.bps-total");
768 qemu_opt_rename(all_opts
, "bps_rd", "throttling.bps-read");
769 qemu_opt_rename(all_opts
, "bps_wr", "throttling.bps-write");
771 qemu_opt_rename(all_opts
, "iops_max", "throttling.iops-total-max");
772 qemu_opt_rename(all_opts
, "iops_rd_max", "throttling.iops-read-max");
773 qemu_opt_rename(all_opts
, "iops_wr_max", "throttling.iops-write-max");
775 qemu_opt_rename(all_opts
, "bps_max", "throttling.bps-total-max");
776 qemu_opt_rename(all_opts
, "bps_rd_max", "throttling.bps-read-max");
777 qemu_opt_rename(all_opts
, "bps_wr_max", "throttling.bps-write-max");
779 qemu_opt_rename(all_opts
,
780 "iops_size", "throttling.iops-size");
782 qemu_opt_rename(all_opts
, "readonly", "read-only");
784 value
= qemu_opt_get(all_opts
, "cache");
788 if (bdrv_parse_cache_flags(value
, &flags
) != 0) {
789 error_report("invalid cache option");
793 /* Specific options take precedence */
794 if (!qemu_opt_get(all_opts
, "cache.writeback")) {
795 qemu_opt_set_bool(all_opts
, "cache.writeback",
796 !!(flags
& BDRV_O_CACHE_WB
));
798 if (!qemu_opt_get(all_opts
, "cache.direct")) {
799 qemu_opt_set_bool(all_opts
, "cache.direct",
800 !!(flags
& BDRV_O_NOCACHE
));
802 if (!qemu_opt_get(all_opts
, "cache.no-flush")) {
803 qemu_opt_set_bool(all_opts
, "cache.no-flush",
804 !!(flags
& BDRV_O_NO_FLUSH
));
806 qemu_opt_unset(all_opts
, "cache");
809 /* Get a QDict for processing the options */
810 bs_opts
= qdict_new();
811 qemu_opts_to_qdict(all_opts
, bs_opts
);
813 legacy_opts
= qemu_opts_create_nofail(&qemu_legacy_drive_opts
);
814 qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, &local_err
);
815 if (error_is_set(&local_err
)) {
816 qerror_report_err(local_err
);
817 error_free(local_err
);
822 value
= qemu_opt_get(legacy_opts
, "media");
824 if (!strcmp(value
, "disk")) {
826 } else if (!strcmp(value
, "cdrom")) {
829 error_report("'%s' invalid media", value
);
834 /* Controller type */
835 value
= qemu_opt_get(legacy_opts
, "if");
838 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
841 if (type
== IF_COUNT
) {
842 error_report("unsupported bus type '%s'", value
);
846 type
= block_default_type
;
849 /* Actual block device init: Functionality shared with blockdev-add */
850 dinfo
= blockdev_init(bs_opts
, type
, media
);
855 /* Set legacy DriveInfo fields */
856 dinfo
->enable_auto_del
= true;
857 dinfo
->opts
= all_opts
;
860 qemu_opts_del(legacy_opts
);
864 void do_commit(Monitor
*mon
, const QDict
*qdict
)
866 const char *device
= qdict_get_str(qdict
, "device");
867 BlockDriverState
*bs
;
870 if (!strcmp(device
, "all")) {
871 ret
= bdrv_commit_all();
873 bs
= bdrv_find(device
);
875 monitor_printf(mon
, "Device '%s' not found\n", device
);
878 ret
= bdrv_commit(bs
);
881 monitor_printf(mon
, "'commit' error for '%s': %s\n", device
,
886 static void blockdev_do_action(int kind
, void *data
, Error
**errp
)
888 TransactionAction action
;
889 TransactionActionList list
;
893 list
.value
= &action
;
895 qmp_transaction(&list
, errp
);
898 void qmp_blockdev_snapshot_sync(const char *device
, const char *snapshot_file
,
899 bool has_format
, const char *format
,
900 bool has_mode
, enum NewImageMode mode
,
903 BlockdevSnapshot snapshot
= {
904 .device
= (char *) device
,
905 .snapshot_file
= (char *) snapshot_file
,
906 .has_format
= has_format
,
907 .format
= (char *) format
,
908 .has_mode
= has_mode
,
911 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
915 void qmp_blockdev_snapshot_internal_sync(const char *device
,
919 BlockdevSnapshotInternal snapshot
= {
920 .device
= (char *) device
,
921 .name
= (char *) name
924 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
928 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
935 BlockDriverState
*bs
= bdrv_find(device
);
937 Error
*local_err
= NULL
;
938 SnapshotInfo
*info
= NULL
;
942 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
955 error_setg(errp
, "Name or id must be provided");
959 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
960 if (error_is_set(&local_err
)) {
961 error_propagate(errp
, local_err
);
966 "Snapshot with id '%s' and name '%s' does not exist on "
968 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
972 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
973 if (error_is_set(&local_err
)) {
974 error_propagate(errp
, local_err
);
978 info
= g_malloc0(sizeof(SnapshotInfo
));
979 info
->id
= g_strdup(sn
.id_str
);
980 info
->name
= g_strdup(sn
.name
);
981 info
->date_nsec
= sn
.date_nsec
;
982 info
->date_sec
= sn
.date_sec
;
983 info
->vm_state_size
= sn
.vm_state_size
;
984 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
985 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
990 /* New and old BlockDriverState structs for group snapshots */
992 typedef struct BlkTransactionState BlkTransactionState
;
994 /* Only prepare() may fail. In a single transaction, only one of commit() or
995 abort() will be called, clean() will always be called if it present. */
996 typedef struct BdrvActionOps
{
997 /* Size of state struct, in bytes. */
998 size_t instance_size
;
999 /* Prepare the work, must NOT be NULL. */
1000 void (*prepare
)(BlkTransactionState
*common
, Error
**errp
);
1001 /* Commit the changes, can be NULL. */
1002 void (*commit
)(BlkTransactionState
*common
);
1003 /* Abort the changes on fail, can be NULL. */
1004 void (*abort
)(BlkTransactionState
*common
);
1005 /* Clean up resource in the end, can be NULL. */
1006 void (*clean
)(BlkTransactionState
*common
);
1010 * This structure must be arranged as first member in child type, assuming
1011 * that compiler will also arrange it to the same address with parent instance.
1012 * Later it will be used in free().
1014 struct BlkTransactionState
{
1015 TransactionAction
*action
;
1016 const BdrvActionOps
*ops
;
1017 QSIMPLEQ_ENTRY(BlkTransactionState
) entry
;
1020 /* internal snapshot private data */
1021 typedef struct InternalSnapshotState
{
1022 BlkTransactionState common
;
1023 BlockDriverState
*bs
;
1024 QEMUSnapshotInfo sn
;
1025 } InternalSnapshotState
;
1027 static void internal_snapshot_prepare(BlkTransactionState
*common
,
1032 BlockDriverState
*bs
;
1033 QEMUSnapshotInfo old_sn
, *sn
;
1036 BlockdevSnapshotInternal
*internal
;
1037 InternalSnapshotState
*state
;
1040 g_assert(common
->action
->kind
==
1041 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1042 internal
= common
->action
->blockdev_snapshot_internal_sync
;
1043 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1045 /* 1. parse input */
1046 device
= internal
->device
;
1047 name
= internal
->name
;
1049 /* 2. check for validation */
1050 bs
= bdrv_find(device
);
1052 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1056 if (!bdrv_is_inserted(bs
)) {
1057 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1061 if (bdrv_is_read_only(bs
)) {
1062 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1066 if (!bdrv_can_snapshot(bs
)) {
1067 error_set(errp
, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
1068 bs
->drv
->format_name
, device
, "internal snapshot");
1072 if (!strlen(name
)) {
1073 error_setg(errp
, "Name is empty");
1077 /* check whether a snapshot with name exist */
1078 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
, errp
);
1079 if (error_is_set(errp
)) {
1083 "Snapshot with name '%s' already exists on device '%s'",
1088 /* 3. take the snapshot */
1090 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1091 qemu_gettimeofday(&tv
);
1092 sn
->date_sec
= tv
.tv_sec
;
1093 sn
->date_nsec
= tv
.tv_usec
* 1000;
1094 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1096 ret1
= bdrv_snapshot_create(bs
, sn
);
1098 error_setg_errno(errp
, -ret1
,
1099 "Failed to create snapshot '%s' on device '%s'",
1104 /* 4. succeed, mark a snapshot is created */
1108 static void internal_snapshot_abort(BlkTransactionState
*common
)
1110 InternalSnapshotState
*state
=
1111 DO_UPCAST(InternalSnapshotState
, common
, common
);
1112 BlockDriverState
*bs
= state
->bs
;
1113 QEMUSnapshotInfo
*sn
= &state
->sn
;
1114 Error
*local_error
= NULL
;
1120 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1121 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1122 "device '%s' in abort: %s",
1125 bdrv_get_device_name(bs
),
1126 error_get_pretty(local_error
));
1127 error_free(local_error
);
1131 /* external snapshot private data */
1132 typedef struct ExternalSnapshotState
{
1133 BlkTransactionState common
;
1134 BlockDriverState
*old_bs
;
1135 BlockDriverState
*new_bs
;
1136 } ExternalSnapshotState
;
1138 static void external_snapshot_prepare(BlkTransactionState
*common
,
1143 Error
*local_err
= NULL
;
1145 const char *new_image_file
;
1146 const char *format
= "qcow2";
1147 enum NewImageMode mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1148 ExternalSnapshotState
*state
=
1149 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1150 TransactionAction
*action
= common
->action
;
1152 /* get parameters */
1153 g_assert(action
->kind
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
);
1155 device
= action
->blockdev_snapshot_sync
->device
;
1156 new_image_file
= action
->blockdev_snapshot_sync
->snapshot_file
;
1157 if (action
->blockdev_snapshot_sync
->has_format
) {
1158 format
= action
->blockdev_snapshot_sync
->format
;
1160 if (action
->blockdev_snapshot_sync
->has_mode
) {
1161 mode
= action
->blockdev_snapshot_sync
->mode
;
1164 /* start processing */
1165 drv
= bdrv_find_format(format
);
1167 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1171 state
->old_bs
= bdrv_find(device
);
1172 if (!state
->old_bs
) {
1173 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1177 if (!bdrv_is_inserted(state
->old_bs
)) {
1178 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1182 if (bdrv_in_use(state
->old_bs
)) {
1183 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1187 if (!bdrv_is_read_only(state
->old_bs
)) {
1188 if (bdrv_flush(state
->old_bs
)) {
1189 error_set(errp
, QERR_IO_ERROR
);
1194 if (bdrv_check_ext_snapshot(state
->old_bs
) != EXT_SNAPSHOT_ALLOWED
) {
1195 error_set(errp
, QERR_FEATURE_DISABLED
, "snapshot");
1199 flags
= state
->old_bs
->open_flags
;
1201 /* create new image w/backing file */
1202 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1203 bdrv_img_create(new_image_file
, format
,
1204 state
->old_bs
->filename
,
1205 state
->old_bs
->drv
->format_name
,
1206 NULL
, -1, flags
, &local_err
, false);
1207 if (error_is_set(&local_err
)) {
1208 error_propagate(errp
, local_err
);
1213 /* We will manually add the backing_hd field to the bs later */
1214 state
->new_bs
= bdrv_new("");
1215 /* TODO Inherit bs->options or only take explicit options with an
1216 * extended QMP command? */
1217 ret
= bdrv_open(state
->new_bs
, new_image_file
, NULL
,
1218 flags
| BDRV_O_NO_BACKING
, drv
, &local_err
);
1220 error_propagate(errp
, local_err
);
1224 static void external_snapshot_commit(BlkTransactionState
*common
)
1226 ExternalSnapshotState
*state
=
1227 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1229 /* This removes our old bs and adds the new bs */
1230 bdrv_append(state
->new_bs
, state
->old_bs
);
1231 /* We don't need (or want) to use the transactional
1232 * bdrv_reopen_multiple() across all the entries at once, because we
1233 * don't want to abort all of them if one of them fails the reopen */
1234 bdrv_reopen(state
->new_bs
, state
->new_bs
->open_flags
& ~BDRV_O_RDWR
,
1238 static void external_snapshot_abort(BlkTransactionState
*common
)
1240 ExternalSnapshotState
*state
=
1241 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1242 if (state
->new_bs
) {
1243 bdrv_unref(state
->new_bs
);
1247 typedef struct DriveBackupState
{
1248 BlkTransactionState common
;
1249 BlockDriverState
*bs
;
1253 static void drive_backup_prepare(BlkTransactionState
*common
, Error
**errp
)
1255 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1256 DriveBackup
*backup
;
1257 Error
*local_err
= NULL
;
1259 assert(common
->action
->kind
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1260 backup
= common
->action
->drive_backup
;
1262 qmp_drive_backup(backup
->device
, backup
->target
,
1263 backup
->has_format
, backup
->format
,
1265 backup
->has_mode
, backup
->mode
,
1266 backup
->has_speed
, backup
->speed
,
1267 backup
->has_on_source_error
, backup
->on_source_error
,
1268 backup
->has_on_target_error
, backup
->on_target_error
,
1270 if (error_is_set(&local_err
)) {
1271 error_propagate(errp
, local_err
);
1277 state
->bs
= bdrv_find(backup
->device
);
1278 state
->job
= state
->bs
->job
;
1281 static void drive_backup_abort(BlkTransactionState
*common
)
1283 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1284 BlockDriverState
*bs
= state
->bs
;
1286 /* Only cancel if it's the job we started */
1287 if (bs
&& bs
->job
&& bs
->job
== state
->job
) {
1288 block_job_cancel_sync(bs
->job
);
1292 static void abort_prepare(BlkTransactionState
*common
, Error
**errp
)
1294 error_setg(errp
, "Transaction aborted using Abort action");
1297 static void abort_commit(BlkTransactionState
*common
)
1299 g_assert_not_reached(); /* this action never succeeds */
1302 static const BdrvActionOps actions
[] = {
1303 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
1304 .instance_size
= sizeof(ExternalSnapshotState
),
1305 .prepare
= external_snapshot_prepare
,
1306 .commit
= external_snapshot_commit
,
1307 .abort
= external_snapshot_abort
,
1309 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
1310 .instance_size
= sizeof(DriveBackupState
),
1311 .prepare
= drive_backup_prepare
,
1312 .abort
= drive_backup_abort
,
1314 [TRANSACTION_ACTION_KIND_ABORT
] = {
1315 .instance_size
= sizeof(BlkTransactionState
),
1316 .prepare
= abort_prepare
,
1317 .commit
= abort_commit
,
1319 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
1320 .instance_size
= sizeof(InternalSnapshotState
),
1321 .prepare
= internal_snapshot_prepare
,
1322 .abort
= internal_snapshot_abort
,
1327 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1328 * then we do not pivot any of the devices in the group, and abandon the
1331 void qmp_transaction(TransactionActionList
*dev_list
, Error
**errp
)
1333 TransactionActionList
*dev_entry
= dev_list
;
1334 BlkTransactionState
*state
, *next
;
1335 Error
*local_err
= NULL
;
1337 QSIMPLEQ_HEAD(snap_bdrv_states
, BlkTransactionState
) snap_bdrv_states
;
1338 QSIMPLEQ_INIT(&snap_bdrv_states
);
1340 /* drain all i/o before any snapshots */
1343 /* We don't do anything in this loop that commits us to the snapshot */
1344 while (NULL
!= dev_entry
) {
1345 TransactionAction
*dev_info
= NULL
;
1346 const BdrvActionOps
*ops
;
1348 dev_info
= dev_entry
->value
;
1349 dev_entry
= dev_entry
->next
;
1351 assert(dev_info
->kind
< ARRAY_SIZE(actions
));
1353 ops
= &actions
[dev_info
->kind
];
1354 assert(ops
->instance_size
> 0);
1356 state
= g_malloc0(ops
->instance_size
);
1358 state
->action
= dev_info
;
1359 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
1361 state
->ops
->prepare(state
, &local_err
);
1362 if (error_is_set(&local_err
)) {
1363 error_propagate(errp
, local_err
);
1364 goto delete_and_fail
;
1368 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1369 if (state
->ops
->commit
) {
1370 state
->ops
->commit(state
);
1379 * failure, and it is all-or-none; abandon each new bs, and keep using
1380 * the original bs for all images
1382 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1383 if (state
->ops
->abort
) {
1384 state
->ops
->abort(state
);
1388 QSIMPLEQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
1389 if (state
->ops
->clean
) {
1390 state
->ops
->clean(state
);
1397 static void eject_device(BlockDriverState
*bs
, int force
, Error
**errp
)
1399 if (bdrv_in_use(bs
)) {
1400 error_set(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
1403 if (!bdrv_dev_has_removable_media(bs
)) {
1404 error_set(errp
, QERR_DEVICE_NOT_REMOVABLE
, bdrv_get_device_name(bs
));
1408 if (bdrv_dev_is_medium_locked(bs
) && !bdrv_dev_is_tray_open(bs
)) {
1409 bdrv_dev_eject_request(bs
, force
);
1411 error_set(errp
, QERR_DEVICE_LOCKED
, bdrv_get_device_name(bs
));
1419 void qmp_eject(const char *device
, bool has_force
, bool force
, Error
**errp
)
1421 BlockDriverState
*bs
;
1423 bs
= bdrv_find(device
);
1425 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1429 eject_device(bs
, force
, errp
);
1432 void qmp_block_passwd(const char *device
, const char *password
, Error
**errp
)
1434 BlockDriverState
*bs
;
1437 bs
= bdrv_find(device
);
1439 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1443 err
= bdrv_set_key(bs
, password
);
1444 if (err
== -EINVAL
) {
1445 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1447 } else if (err
< 0) {
1448 error_set(errp
, QERR_INVALID_PASSWORD
);
1453 static void qmp_bdrv_open_encrypted(BlockDriverState
*bs
, const char *filename
,
1454 int bdrv_flags
, BlockDriver
*drv
,
1455 const char *password
, Error
**errp
)
1457 Error
*local_err
= NULL
;
1460 ret
= bdrv_open(bs
, filename
, NULL
, bdrv_flags
, drv
, &local_err
);
1462 error_propagate(errp
, local_err
);
1466 if (bdrv_key_required(bs
)) {
1468 if (bdrv_set_key(bs
, password
) < 0) {
1469 error_set(errp
, QERR_INVALID_PASSWORD
);
1472 error_set(errp
, QERR_DEVICE_ENCRYPTED
, bdrv_get_device_name(bs
),
1473 bdrv_get_encrypted_filename(bs
));
1475 } else if (password
) {
1476 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1480 void qmp_change_blockdev(const char *device
, const char *filename
,
1481 bool has_format
, const char *format
, Error
**errp
)
1483 BlockDriverState
*bs
;
1484 BlockDriver
*drv
= NULL
;
1488 bs
= bdrv_find(device
);
1490 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1495 drv
= bdrv_find_whitelisted_format(format
, bs
->read_only
);
1497 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1502 eject_device(bs
, 0, &err
);
1503 if (error_is_set(&err
)) {
1504 error_propagate(errp
, err
);
1508 bdrv_flags
= bdrv_is_read_only(bs
) ? 0 : BDRV_O_RDWR
;
1509 bdrv_flags
|= bdrv_is_snapshot(bs
) ? BDRV_O_SNAPSHOT
: 0;
1511 qmp_bdrv_open_encrypted(bs
, filename
, bdrv_flags
, drv
, NULL
, errp
);
1514 /* throttling disk I/O limits */
1515 void qmp_block_set_io_throttle(const char *device
, int64_t bps
, int64_t bps_rd
,
1522 bool has_bps_rd_max
,
1524 bool has_bps_wr_max
,
1528 bool has_iops_rd_max
,
1529 int64_t iops_rd_max
,
1530 bool has_iops_wr_max
,
1531 int64_t iops_wr_max
,
1533 int64_t iops_size
, Error
**errp
)
1536 BlockDriverState
*bs
;
1538 bs
= bdrv_find(device
);
1540 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1544 memset(&cfg
, 0, sizeof(cfg
));
1545 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
= bps
;
1546 cfg
.buckets
[THROTTLE_BPS_READ
].avg
= bps_rd
;
1547 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
= bps_wr
;
1549 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
= iops
;
1550 cfg
.buckets
[THROTTLE_OPS_READ
].avg
= iops_rd
;
1551 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
= iops_wr
;
1554 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
= bps_max
;
1556 if (has_bps_rd_max
) {
1557 cfg
.buckets
[THROTTLE_BPS_READ
].max
= bps_rd_max
;
1559 if (has_bps_wr_max
) {
1560 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
= bps_wr_max
;
1563 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
= iops_max
;
1565 if (has_iops_rd_max
) {
1566 cfg
.buckets
[THROTTLE_OPS_READ
].max
= iops_rd_max
;
1568 if (has_iops_wr_max
) {
1569 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
= iops_wr_max
;
1572 if (has_iops_size
) {
1573 cfg
.op_size
= iops_size
;
1576 if (!check_throttle_config(&cfg
, errp
)) {
1580 if (!bs
->io_limits_enabled
&& throttle_enabled(&cfg
)) {
1581 bdrv_io_limits_enable(bs
);
1582 } else if (bs
->io_limits_enabled
&& !throttle_enabled(&cfg
)) {
1583 bdrv_io_limits_disable(bs
);
1586 if (bs
->io_limits_enabled
) {
1587 bdrv_set_io_limits(bs
, &cfg
);
1591 int do_drive_del(Monitor
*mon
, const QDict
*qdict
, QObject
**ret_data
)
1593 const char *id
= qdict_get_str(qdict
, "id");
1594 BlockDriverState
*bs
;
1598 qerror_report(QERR_DEVICE_NOT_FOUND
, id
);
1601 if (bdrv_in_use(bs
)) {
1602 qerror_report(QERR_DEVICE_IN_USE
, id
);
1606 /* quiesce block driver; prevent further io */
1611 /* if we have a device attached to this BlockDriverState
1612 * then we need to make the drive anonymous until the device
1613 * can be removed. If this is a drive with no device backing
1614 * then we can just get rid of the block driver state right here.
1616 if (bdrv_get_attached_dev(bs
)) {
1619 /* Further I/O must not pause the guest */
1620 bdrv_set_on_error(bs
, BLOCKDEV_ON_ERROR_REPORT
,
1621 BLOCKDEV_ON_ERROR_REPORT
);
1623 drive_uninit(drive_get_by_blockdev(bs
));
1629 void qmp_block_resize(const char *device
, int64_t size
, Error
**errp
)
1631 BlockDriverState
*bs
;
1634 bs
= bdrv_find(device
);
1636 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1641 error_set(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
1645 /* complete all in-flight operations before resizing the device */
1648 ret
= bdrv_truncate(bs
, size
);
1653 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1656 error_set(errp
, QERR_UNSUPPORTED
);
1659 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1662 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1665 error_setg_errno(errp
, -ret
, "Could not resize");
1670 static void block_job_cb(void *opaque
, int ret
)
1672 BlockDriverState
*bs
= opaque
;
1675 trace_block_job_cb(bs
, bs
->job
, ret
);
1678 obj
= qobject_from_block_job(bs
->job
);
1680 QDict
*dict
= qobject_to_qdict(obj
);
1681 qdict_put(dict
, "error", qstring_from_str(strerror(-ret
)));
1684 if (block_job_is_cancelled(bs
->job
)) {
1685 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED
, obj
);
1687 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED
, obj
);
1689 qobject_decref(obj
);
1691 bdrv_put_ref_bh_schedule(bs
);
1694 void qmp_block_stream(const char *device
, bool has_base
,
1695 const char *base
, bool has_speed
, int64_t speed
,
1696 bool has_on_error
, BlockdevOnError on_error
,
1699 BlockDriverState
*bs
;
1700 BlockDriverState
*base_bs
= NULL
;
1701 Error
*local_err
= NULL
;
1703 if (!has_on_error
) {
1704 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1707 bs
= bdrv_find(device
);
1709 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1714 base_bs
= bdrv_find_backing_image(bs
, base
);
1715 if (base_bs
== NULL
) {
1716 error_set(errp
, QERR_BASE_NOT_FOUND
, base
);
1721 stream_start(bs
, base_bs
, base
, has_speed
? speed
: 0,
1722 on_error
, block_job_cb
, bs
, &local_err
);
1723 if (error_is_set(&local_err
)) {
1724 error_propagate(errp
, local_err
);
1728 trace_qmp_block_stream(bs
, bs
->job
);
1731 void qmp_block_commit(const char *device
,
1732 bool has_base
, const char *base
, const char *top
,
1733 bool has_speed
, int64_t speed
,
1736 BlockDriverState
*bs
;
1737 BlockDriverState
*base_bs
, *top_bs
;
1738 Error
*local_err
= NULL
;
1739 /* This will be part of the QMP command, if/when the
1740 * BlockdevOnError change for blkmirror makes it in
1742 BlockdevOnError on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1744 /* drain all i/o before commits */
1747 bs
= bdrv_find(device
);
1749 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1753 /* default top_bs is the active layer */
1757 if (strcmp(bs
->filename
, top
) != 0) {
1758 top_bs
= bdrv_find_backing_image(bs
, top
);
1762 if (top_bs
== NULL
) {
1763 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
1767 if (has_base
&& base
) {
1768 base_bs
= bdrv_find_backing_image(top_bs
, base
);
1770 base_bs
= bdrv_find_base(top_bs
);
1773 if (base_bs
== NULL
) {
1774 error_set(errp
, QERR_BASE_NOT_FOUND
, base
? base
: "NULL");
1778 commit_start(bs
, base_bs
, top_bs
, speed
, on_error
, block_job_cb
, bs
,
1780 if (local_err
!= NULL
) {
1781 error_propagate(errp
, local_err
);
1786 void qmp_drive_backup(const char *device
, const char *target
,
1787 bool has_format
, const char *format
,
1788 enum MirrorSyncMode sync
,
1789 bool has_mode
, enum NewImageMode mode
,
1790 bool has_speed
, int64_t speed
,
1791 bool has_on_source_error
, BlockdevOnError on_source_error
,
1792 bool has_on_target_error
, BlockdevOnError on_target_error
,
1795 BlockDriverState
*bs
;
1796 BlockDriverState
*target_bs
;
1797 BlockDriverState
*source
= NULL
;
1798 BlockDriver
*drv
= NULL
;
1799 Error
*local_err
= NULL
;
1807 if (!has_on_source_error
) {
1808 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1810 if (!has_on_target_error
) {
1811 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1814 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1817 bs
= bdrv_find(device
);
1819 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1823 if (!bdrv_is_inserted(bs
)) {
1824 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1829 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1832 drv
= bdrv_find_format(format
);
1834 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1839 if (bdrv_in_use(bs
)) {
1840 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1844 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1846 /* See if we have a backing HD we can use to create our new image
1848 if (sync
== MIRROR_SYNC_MODE_TOP
) {
1849 source
= bs
->backing_hd
;
1851 sync
= MIRROR_SYNC_MODE_FULL
;
1854 if (sync
== MIRROR_SYNC_MODE_NONE
) {
1858 size
= bdrv_getlength(bs
);
1860 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1864 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1865 assert(format
&& drv
);
1867 bdrv_img_create(target
, format
, source
->filename
,
1868 source
->drv
->format_name
, NULL
,
1869 size
, flags
, &local_err
, false);
1871 bdrv_img_create(target
, format
, NULL
, NULL
, NULL
,
1872 size
, flags
, &local_err
, false);
1876 if (error_is_set(&local_err
)) {
1877 error_propagate(errp
, local_err
);
1881 target_bs
= bdrv_new("");
1882 ret
= bdrv_open(target_bs
, target
, NULL
, flags
, drv
, &local_err
);
1884 bdrv_unref(target_bs
);
1885 error_propagate(errp
, local_err
);
1889 backup_start(bs
, target_bs
, speed
, sync
, on_source_error
, on_target_error
,
1890 block_job_cb
, bs
, &local_err
);
1891 if (local_err
!= NULL
) {
1892 bdrv_unref(target_bs
);
1893 error_propagate(errp
, local_err
);
1898 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1900 void qmp_drive_mirror(const char *device
, const char *target
,
1901 bool has_format
, const char *format
,
1902 enum MirrorSyncMode sync
,
1903 bool has_mode
, enum NewImageMode mode
,
1904 bool has_speed
, int64_t speed
,
1905 bool has_granularity
, uint32_t granularity
,
1906 bool has_buf_size
, int64_t buf_size
,
1907 bool has_on_source_error
, BlockdevOnError on_source_error
,
1908 bool has_on_target_error
, BlockdevOnError on_target_error
,
1911 BlockDriverState
*bs
;
1912 BlockDriverState
*source
, *target_bs
;
1913 BlockDriver
*drv
= NULL
;
1914 Error
*local_err
= NULL
;
1922 if (!has_on_source_error
) {
1923 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1925 if (!has_on_target_error
) {
1926 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1929 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1931 if (!has_granularity
) {
1934 if (!has_buf_size
) {
1935 buf_size
= DEFAULT_MIRROR_BUF_SIZE
;
1938 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
1939 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1942 if (granularity
& (granularity
- 1)) {
1943 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1947 bs
= bdrv_find(device
);
1949 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1953 if (!bdrv_is_inserted(bs
)) {
1954 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1959 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1962 drv
= bdrv_find_format(format
);
1964 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1969 if (bdrv_in_use(bs
)) {
1970 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1974 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1975 source
= bs
->backing_hd
;
1976 if (!source
&& sync
== MIRROR_SYNC_MODE_TOP
) {
1977 sync
= MIRROR_SYNC_MODE_FULL
;
1980 size
= bdrv_getlength(bs
);
1982 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1986 if (sync
== MIRROR_SYNC_MODE_FULL
&& mode
!= NEW_IMAGE_MODE_EXISTING
) {
1987 /* create new image w/o backing file */
1988 assert(format
&& drv
);
1989 bdrv_img_create(target
, format
,
1990 NULL
, NULL
, NULL
, size
, flags
, &local_err
, false);
1993 case NEW_IMAGE_MODE_EXISTING
:
1995 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
1996 /* create new image with backing file */
1997 bdrv_img_create(target
, format
,
1999 source
->drv
->format_name
,
2000 NULL
, size
, flags
, &local_err
, false);
2007 if (error_is_set(&local_err
)) {
2008 error_propagate(errp
, local_err
);
2012 /* Mirroring takes care of copy-on-write using the source's backing
2015 target_bs
= bdrv_new("");
2016 ret
= bdrv_open(target_bs
, target
, NULL
, flags
| BDRV_O_NO_BACKING
, drv
,
2019 bdrv_unref(target_bs
);
2020 error_propagate(errp
, local_err
);
2024 mirror_start(bs
, target_bs
, speed
, granularity
, buf_size
, sync
,
2025 on_source_error
, on_target_error
,
2026 block_job_cb
, bs
, &local_err
);
2027 if (local_err
!= NULL
) {
2028 bdrv_unref(target_bs
);
2029 error_propagate(errp
, local_err
);
2034 static BlockJob
*find_block_job(const char *device
)
2036 BlockDriverState
*bs
;
2038 bs
= bdrv_find(device
);
2039 if (!bs
|| !bs
->job
) {
2045 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
2047 BlockJob
*job
= find_block_job(device
);
2050 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2054 block_job_set_speed(job
, speed
, errp
);
2057 void qmp_block_job_cancel(const char *device
,
2058 bool has_force
, bool force
, Error
**errp
)
2060 BlockJob
*job
= find_block_job(device
);
2067 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2070 if (job
->paused
&& !force
) {
2071 error_set(errp
, QERR_BLOCK_JOB_PAUSED
, device
);
2075 trace_qmp_block_job_cancel(job
);
2076 block_job_cancel(job
);
2079 void qmp_block_job_pause(const char *device
, Error
**errp
)
2081 BlockJob
*job
= find_block_job(device
);
2084 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2088 trace_qmp_block_job_pause(job
);
2089 block_job_pause(job
);
2092 void qmp_block_job_resume(const char *device
, Error
**errp
)
2094 BlockJob
*job
= find_block_job(device
);
2097 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2101 trace_qmp_block_job_resume(job
);
2102 block_job_resume(job
);
2105 void qmp_block_job_complete(const char *device
, Error
**errp
)
2107 BlockJob
*job
= find_block_job(device
);
2110 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2114 trace_qmp_block_job_complete(job
);
2115 block_job_complete(job
, errp
);
2118 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
2120 QmpOutputVisitor
*ov
= qmp_output_visitor_new();
2124 Error
*local_err
= NULL
;
2126 /* Require an ID in the top level */
2127 if (!options
->has_id
) {
2128 error_setg(errp
, "Block device needs an ID");
2132 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2133 * cache.direct=false instead of silently switching to aio=threads, except
2134 * if called from drive_init.
2136 * For now, simply forbidding the combination for all drivers will do. */
2137 if (options
->has_aio
&& options
->aio
== BLOCKDEV_AIO_OPTIONS_NATIVE
) {
2138 bool direct
= options
->cache
->has_direct
&& options
->cache
->direct
;
2139 if (!options
->has_cache
&& !direct
) {
2140 error_setg(errp
, "aio=native requires cache.direct=true");
2145 visit_type_BlockdevOptions(qmp_output_get_visitor(ov
),
2146 &options
, NULL
, &local_err
);
2147 if (error_is_set(&local_err
)) {
2148 error_propagate(errp
, local_err
);
2152 obj
= qmp_output_get_qobject(ov
);
2153 qdict
= qobject_to_qdict(obj
);
2155 qdict_flatten(qdict
);
2157 dinfo
= blockdev_init(qdict
, IF_NONE
, MEDIA_DISK
);
2159 error_setg(errp
, "Could not open image");
2164 qmp_output_visitor_cleanup(ov
);
2167 static void do_qmp_query_block_jobs_one(void *opaque
, BlockDriverState
*bs
)
2169 BlockJobInfoList
**prev
= opaque
;
2170 BlockJob
*job
= bs
->job
;
2173 BlockJobInfoList
*elem
= g_new0(BlockJobInfoList
, 1);
2174 elem
->value
= block_job_query(bs
->job
);
2175 (*prev
)->next
= elem
;
2180 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
2182 /* Dummy is a fake list element for holding the head pointer */
2183 BlockJobInfoList dummy
= {};
2184 BlockJobInfoList
*prev
= &dummy
;
2185 bdrv_iterate(do_qmp_query_block_jobs_one
, &prev
);
2189 QemuOptsList qemu_common_drive_opts
= {
2191 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
2195 .type
= QEMU_OPT_NUMBER
,
2196 .help
= "bus number",
2199 .type
= QEMU_OPT_NUMBER
,
2200 .help
= "unit number (i.e. lun for scsi)",
2203 .type
= QEMU_OPT_NUMBER
,
2204 .help
= "index number",
2207 .type
= QEMU_OPT_NUMBER
,
2208 .help
= "number of cylinders (ide disk geometry)",
2211 .type
= QEMU_OPT_NUMBER
,
2212 .help
= "number of heads (ide disk geometry)",
2215 .type
= QEMU_OPT_NUMBER
,
2216 .help
= "number of sectors (ide disk geometry)",
2219 .type
= QEMU_OPT_STRING
,
2220 .help
= "chs translation (auto, lba. none)",
2223 .type
= QEMU_OPT_BOOL
,
2224 .help
= "enable/disable snapshot mode",
2227 .type
= QEMU_OPT_STRING
,
2228 .help
= "disk image",
2231 .type
= QEMU_OPT_STRING
,
2232 .help
= "discard operation (ignore/off, unmap/on)",
2234 .name
= "cache.writeback",
2235 .type
= QEMU_OPT_BOOL
,
2236 .help
= "enables writeback mode for any caches",
2238 .name
= "cache.direct",
2239 .type
= QEMU_OPT_BOOL
,
2240 .help
= "enables use of O_DIRECT (bypass the host page cache)",
2242 .name
= "cache.no-flush",
2243 .type
= QEMU_OPT_BOOL
,
2244 .help
= "ignore any flush requests for the device",
2247 .type
= QEMU_OPT_STRING
,
2248 .help
= "host AIO implementation (threads, native)",
2251 .type
= QEMU_OPT_STRING
,
2252 .help
= "disk format (raw, qcow2, ...)",
2255 .type
= QEMU_OPT_STRING
,
2256 .help
= "disk serial number",
2259 .type
= QEMU_OPT_STRING
,
2260 .help
= "read error action",
2263 .type
= QEMU_OPT_STRING
,
2264 .help
= "write error action",
2267 .type
= QEMU_OPT_STRING
,
2268 .help
= "pci address (virtio only)",
2270 .name
= "read-only",
2271 .type
= QEMU_OPT_BOOL
,
2272 .help
= "open drive file as read-only",
2274 .name
= "throttling.iops-total",
2275 .type
= QEMU_OPT_NUMBER
,
2276 .help
= "limit total I/O operations per second",
2278 .name
= "throttling.iops-read",
2279 .type
= QEMU_OPT_NUMBER
,
2280 .help
= "limit read operations per second",
2282 .name
= "throttling.iops-write",
2283 .type
= QEMU_OPT_NUMBER
,
2284 .help
= "limit write operations per second",
2286 .name
= "throttling.bps-total",
2287 .type
= QEMU_OPT_NUMBER
,
2288 .help
= "limit total bytes per second",
2290 .name
= "throttling.bps-read",
2291 .type
= QEMU_OPT_NUMBER
,
2292 .help
= "limit read bytes per second",
2294 .name
= "throttling.bps-write",
2295 .type
= QEMU_OPT_NUMBER
,
2296 .help
= "limit write bytes per second",
2298 .name
= "throttling.iops-total-max",
2299 .type
= QEMU_OPT_NUMBER
,
2300 .help
= "I/O operations burst",
2302 .name
= "throttling.iops-read-max",
2303 .type
= QEMU_OPT_NUMBER
,
2304 .help
= "I/O operations read burst",
2306 .name
= "throttling.iops-write-max",
2307 .type
= QEMU_OPT_NUMBER
,
2308 .help
= "I/O operations write burst",
2310 .name
= "throttling.bps-total-max",
2311 .type
= QEMU_OPT_NUMBER
,
2312 .help
= "total bytes burst",
2314 .name
= "throttling.bps-read-max",
2315 .type
= QEMU_OPT_NUMBER
,
2316 .help
= "total bytes read burst",
2318 .name
= "throttling.bps-write-max",
2319 .type
= QEMU_OPT_NUMBER
,
2320 .help
= "total bytes write burst",
2322 .name
= "throttling.iops-size",
2323 .type
= QEMU_OPT_NUMBER
,
2324 .help
= "when limiting by iops max size of an I/O in bytes",
2326 .name
= "copy-on-read",
2327 .type
= QEMU_OPT_BOOL
,
2328 .help
= "copy read data from backing file into image file",
2331 .type
= QEMU_OPT_BOOL
,
2332 .help
= "(deprecated, ignored)",
2334 { /* end of list */ }
2338 QemuOptsList qemu_drive_opts
= {
2340 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
2343 * no elements => accept any params
2344 * validation will happen later
2346 { /* end of list */ }