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
);
51 static const char *const if_name
[IF_COUNT
] = {
55 [IF_FLOPPY
] = "floppy",
56 [IF_PFLASH
] = "pflash",
59 [IF_VIRTIO
] = "virtio",
63 static const int if_max_devs
[IF_COUNT
] = {
65 * Do not change these numbers! They govern how drive option
66 * index maps to unit and bus. That mapping is ABI.
68 * All controllers used to imlement if=T drives need to support
69 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
70 * Otherwise, some index values map to "impossible" bus, unit
73 * For instance, if you change [IF_SCSI] to 255, -drive
74 * if=scsi,index=12 no longer means bus=1,unit=5, but
75 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
76 * the drive can't be set up. Regression.
83 * We automatically delete the drive when a device using it gets
84 * unplugged. Questionable feature, but we can't just drop it.
85 * Device models call blockdev_mark_auto_del() to schedule the
86 * automatic deletion, and generic qdev code calls blockdev_auto_del()
87 * when deletion is actually safe.
89 void blockdev_mark_auto_del(BlockDriverState
*bs
)
91 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
93 if (dinfo
&& !dinfo
->enable_auto_del
) {
98 block_job_cancel(bs
->job
);
105 void blockdev_auto_del(BlockDriverState
*bs
)
107 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
109 if (dinfo
&& dinfo
->auto_del
) {
110 drive_put_ref(dinfo
);
114 static int drive_index_to_bus_id(BlockInterfaceType type
, int index
)
116 int max_devs
= if_max_devs
[type
];
117 return max_devs
? index
/ max_devs
: 0;
120 static int drive_index_to_unit_id(BlockInterfaceType type
, int index
)
122 int max_devs
= if_max_devs
[type
];
123 return max_devs
? index
% max_devs
: index
;
126 QemuOpts
*drive_def(const char *optstr
)
128 return qemu_opts_parse(qemu_find_opts("drive"), optstr
, 0);
131 QemuOpts
*drive_add(BlockInterfaceType type
, int index
, const char *file
,
137 opts
= drive_def(optstr
);
141 if (type
!= IF_DEFAULT
) {
142 qemu_opt_set(opts
, "if", if_name
[type
]);
145 snprintf(buf
, sizeof(buf
), "%d", index
);
146 qemu_opt_set(opts
, "index", buf
);
149 qemu_opt_set(opts
, "file", file
);
153 DriveInfo
*drive_get(BlockInterfaceType type
, int bus
, int unit
)
157 /* seek interface, bus and unit */
159 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
160 if (dinfo
->type
== type
&&
169 DriveInfo
*drive_get_by_index(BlockInterfaceType type
, int index
)
171 return drive_get(type
,
172 drive_index_to_bus_id(type
, index
),
173 drive_index_to_unit_id(type
, index
));
176 int drive_get_max_bus(BlockInterfaceType type
)
182 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
183 if(dinfo
->type
== type
&&
184 dinfo
->bus
> max_bus
)
185 max_bus
= dinfo
->bus
;
190 /* Get a block device. This should only be used for single-drive devices
191 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
193 DriveInfo
*drive_get_next(BlockInterfaceType type
)
195 static int next_block_unit
[IF_COUNT
];
197 return drive_get(type
, 0, next_block_unit
[type
]++);
200 DriveInfo
*drive_get_by_blockdev(BlockDriverState
*bs
)
204 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
205 if (dinfo
->bdrv
== bs
) {
212 static void bdrv_format_print(void *opaque
, const char *name
)
214 error_printf(" %s", name
);
217 static void drive_uninit(DriveInfo
*dinfo
)
220 qemu_opts_del(dinfo
->opts
);
223 bdrv_unref(dinfo
->bdrv
);
225 QTAILQ_REMOVE(&drives
, dinfo
, next
);
226 g_free(dinfo
->serial
);
230 void drive_put_ref(DriveInfo
*dinfo
)
232 assert(dinfo
->refcount
);
233 if (--dinfo
->refcount
== 0) {
238 void drive_get_ref(DriveInfo
*dinfo
)
245 BlockDriverState
*bs
;
248 static void bdrv_put_ref_bh(void *opaque
)
250 BDRVPutRefBH
*s
= opaque
;
253 qemu_bh_delete(s
->bh
);
258 * Release a BDS reference in a BH
260 * It is not safe to use bdrv_unref() from a callback function when the callers
261 * still need the BlockDriverState. In such cases we schedule a BH to release
264 static void bdrv_put_ref_bh_schedule(BlockDriverState
*bs
)
268 s
= g_new(BDRVPutRefBH
, 1);
269 s
->bh
= qemu_bh_new(bdrv_put_ref_bh
, s
);
271 qemu_bh_schedule(s
->bh
);
274 static int parse_block_error_action(const char *buf
, bool is_read
, Error
**errp
)
276 if (!strcmp(buf
, "ignore")) {
277 return BLOCKDEV_ON_ERROR_IGNORE
;
278 } else if (!is_read
&& !strcmp(buf
, "enospc")) {
279 return BLOCKDEV_ON_ERROR_ENOSPC
;
280 } else if (!strcmp(buf
, "stop")) {
281 return BLOCKDEV_ON_ERROR_STOP
;
282 } else if (!strcmp(buf
, "report")) {
283 return BLOCKDEV_ON_ERROR_REPORT
;
285 error_setg(errp
, "'%s' invalid %s error action",
286 buf
, is_read
? "read" : "write");
291 static bool check_throttle_config(ThrottleConfig
*cfg
, Error
**errp
)
293 if (throttle_conflicting(cfg
)) {
294 error_setg(errp
, "bps/iops/max total values and read/write values"
295 " cannot be used at the same time");
299 if (!throttle_is_valid(cfg
)) {
300 error_setg(errp
, "bps/iops/maxs values must be 0 or greater");
307 typedef enum { MEDIA_DISK
, MEDIA_CDROM
} DriveMediaType
;
309 /* Takes the ownership of bs_opts */
310 static DriveInfo
*blockdev_init(QDict
*bs_opts
,
311 BlockInterfaceType type
,
315 const char *file
= NULL
;
319 int on_read_error
, on_write_error
;
328 bool has_driver_specific_opts
;
329 BlockDriver
*drv
= NULL
;
331 /* Check common options by copying from bs_opts to opts, all other options
332 * stay in bs_opts for processing by bdrv_open(). */
333 id
= qdict_get_try_str(bs_opts
, "id");
334 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, &error
);
335 if (error_is_set(&error
)) {
336 error_propagate(errp
, error
);
340 qemu_opts_absorb_qdict(opts
, bs_opts
, &error
);
341 if (error_is_set(&error
)) {
342 error_propagate(errp
, error
);
347 qdict_del(bs_opts
, "id");
350 has_driver_specific_opts
= !!qdict_size(bs_opts
);
352 /* extract parameters */
353 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
354 ro
= qemu_opt_get_bool(opts
, "read-only", 0);
355 copy_on_read
= qemu_opt_get_bool(opts
, "copy-on-read", false);
357 file
= qemu_opt_get(opts
, "file");
358 serial
= qemu_opt_get(opts
, "serial");
360 if ((buf
= qemu_opt_get(opts
, "discard")) != NULL
) {
361 if (bdrv_parse_discard_flags(buf
, &bdrv_flags
) != 0) {
362 error_setg(errp
, "invalid discard option");
367 if (qemu_opt_get_bool(opts
, "cache.writeback", true)) {
368 bdrv_flags
|= BDRV_O_CACHE_WB
;
370 if (qemu_opt_get_bool(opts
, "cache.direct", false)) {
371 bdrv_flags
|= BDRV_O_NOCACHE
;
373 if (qemu_opt_get_bool(opts
, "cache.no-flush", false)) {
374 bdrv_flags
|= BDRV_O_NO_FLUSH
;
377 #ifdef CONFIG_LINUX_AIO
378 if ((buf
= qemu_opt_get(opts
, "aio")) != NULL
) {
379 if (!strcmp(buf
, "native")) {
380 bdrv_flags
|= BDRV_O_NATIVE_AIO
;
381 } else if (!strcmp(buf
, "threads")) {
382 /* this is the default */
384 error_setg(errp
, "invalid aio option");
390 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
391 if (is_help_option(buf
)) {
392 error_printf("Supported formats:");
393 bdrv_iterate_format(bdrv_format_print
, NULL
);
398 drv
= bdrv_find_format(buf
);
400 error_setg(errp
, "'%s' invalid format", buf
);
405 /* disk I/O throttling */
406 memset(&cfg
, 0, sizeof(cfg
));
407 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
=
408 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
409 cfg
.buckets
[THROTTLE_BPS_READ
].avg
=
410 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
411 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
=
412 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
413 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
=
414 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
415 cfg
.buckets
[THROTTLE_OPS_READ
].avg
=
416 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
417 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
=
418 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
420 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
=
421 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
422 cfg
.buckets
[THROTTLE_BPS_READ
].max
=
423 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
424 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
=
425 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
426 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
=
427 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
428 cfg
.buckets
[THROTTLE_OPS_READ
].max
=
429 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
430 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
=
431 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
433 cfg
.op_size
= qemu_opt_get_number(opts
, "throttling.iops-size", 0);
435 if (!check_throttle_config(&cfg
, &error
)) {
436 error_propagate(errp
, error
);
440 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
441 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
442 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_NONE
) {
443 error_setg(errp
, "werror is not supported by this bus type");
447 on_write_error
= parse_block_error_action(buf
, 0, &error
);
448 if (error_is_set(&error
)) {
449 error_propagate(errp
, error
);
454 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
455 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
456 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&& type
!= IF_NONE
) {
457 error_report("rerror is not supported by this bus type");
461 on_read_error
= parse_block_error_action(buf
, 1, &error
);
462 if (error_is_set(&error
)) {
463 error_propagate(errp
, error
);
469 dinfo
= g_malloc0(sizeof(*dinfo
));
470 dinfo
->id
= g_strdup(qemu_opts_id(opts
));
471 dinfo
->bdrv
= bdrv_new(dinfo
->id
);
472 dinfo
->bdrv
->open_flags
= snapshot
? BDRV_O_SNAPSHOT
: 0;
473 dinfo
->bdrv
->read_only
= ro
;
476 if (serial
!= NULL
) {
477 dinfo
->serial
= g_strdup(serial
);
479 QTAILQ_INSERT_TAIL(&drives
, dinfo
, next
);
481 bdrv_set_on_error(dinfo
->bdrv
, on_read_error
, on_write_error
);
483 /* disk I/O throttling */
484 if (throttle_enabled(&cfg
)) {
485 bdrv_io_limits_enable(dinfo
->bdrv
);
486 bdrv_set_io_limits(dinfo
->bdrv
, &cfg
);
489 if (!file
|| !*file
) {
490 if (has_driver_specific_opts
) {
499 /* always use cache=unsafe with snapshot */
500 bdrv_flags
&= ~BDRV_O_CACHE_MASK
;
501 bdrv_flags
|= (BDRV_O_SNAPSHOT
|BDRV_O_CACHE_WB
|BDRV_O_NO_FLUSH
);
505 bdrv_flags
|= BDRV_O_COPY_ON_READ
;
508 if (runstate_check(RUN_STATE_INMIGRATE
)) {
509 bdrv_flags
|= BDRV_O_INCOMING
;
512 bdrv_flags
|= ro
? 0 : BDRV_O_RDWR
;
515 ret
= bdrv_open(dinfo
->bdrv
, file
, bs_opts
, bdrv_flags
, drv
, &error
);
518 error_setg(errp
, "could not open disk image %s: %s",
519 file
?: dinfo
->id
, error_get_pretty(error
));
524 if (bdrv_key_required(dinfo
->bdrv
))
533 bdrv_unref(dinfo
->bdrv
);
535 QTAILQ_REMOVE(&drives
, dinfo
, next
);
543 static void qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
)
547 value
= qemu_opt_get(opts
, from
);
549 qemu_opt_set(opts
, to
, value
);
550 qemu_opt_unset(opts
, from
);
554 QemuOptsList qemu_legacy_drive_opts
= {
556 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
560 .type
= QEMU_OPT_NUMBER
,
561 .help
= "bus number",
564 .type
= QEMU_OPT_NUMBER
,
565 .help
= "unit number (i.e. lun for scsi)",
568 .type
= QEMU_OPT_NUMBER
,
569 .help
= "index number",
572 .type
= QEMU_OPT_STRING
,
573 .help
= "media type (disk, cdrom)",
576 .type
= QEMU_OPT_STRING
,
577 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
580 .type
= QEMU_OPT_NUMBER
,
581 .help
= "number of cylinders (ide disk geometry)",
584 .type
= QEMU_OPT_NUMBER
,
585 .help
= "number of heads (ide disk geometry)",
588 .type
= QEMU_OPT_NUMBER
,
589 .help
= "number of sectors (ide disk geometry)",
592 .type
= QEMU_OPT_STRING
,
593 .help
= "chs translation (auto, lba, none)",
596 .type
= QEMU_OPT_BOOL
,
597 .help
= "(deprecated, ignored)",
600 .type
= QEMU_OPT_STRING
,
601 .help
= "pci address (virtio only)",
604 /* Options that are passed on, but have special semantics with -drive */
607 .type
= QEMU_OPT_BOOL
,
608 .help
= "open drive file as read-only",
610 .name
= "copy-on-read",
611 .type
= QEMU_OPT_BOOL
,
612 .help
= "copy read data from backing file into image file",
615 { /* end of list */ }
619 DriveInfo
*drive_init(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
)
622 DriveInfo
*dinfo
= NULL
;
624 QemuOpts
*legacy_opts
;
625 DriveMediaType media
= MEDIA_DISK
;
626 BlockInterfaceType type
;
627 int cyls
, heads
, secs
, translation
;
628 int max_devs
, bus_id
, unit_id
, index
;
630 bool read_only
= false;
632 Error
*local_err
= NULL
;
634 /* Change legacy command line options into QMP ones */
635 qemu_opt_rename(all_opts
, "iops", "throttling.iops-total");
636 qemu_opt_rename(all_opts
, "iops_rd", "throttling.iops-read");
637 qemu_opt_rename(all_opts
, "iops_wr", "throttling.iops-write");
639 qemu_opt_rename(all_opts
, "bps", "throttling.bps-total");
640 qemu_opt_rename(all_opts
, "bps_rd", "throttling.bps-read");
641 qemu_opt_rename(all_opts
, "bps_wr", "throttling.bps-write");
643 qemu_opt_rename(all_opts
, "iops_max", "throttling.iops-total-max");
644 qemu_opt_rename(all_opts
, "iops_rd_max", "throttling.iops-read-max");
645 qemu_opt_rename(all_opts
, "iops_wr_max", "throttling.iops-write-max");
647 qemu_opt_rename(all_opts
, "bps_max", "throttling.bps-total-max");
648 qemu_opt_rename(all_opts
, "bps_rd_max", "throttling.bps-read-max");
649 qemu_opt_rename(all_opts
, "bps_wr_max", "throttling.bps-write-max");
651 qemu_opt_rename(all_opts
,
652 "iops_size", "throttling.iops-size");
654 qemu_opt_rename(all_opts
, "readonly", "read-only");
656 value
= qemu_opt_get(all_opts
, "cache");
660 if (bdrv_parse_cache_flags(value
, &flags
) != 0) {
661 error_report("invalid cache option");
665 /* Specific options take precedence */
666 if (!qemu_opt_get(all_opts
, "cache.writeback")) {
667 qemu_opt_set_bool(all_opts
, "cache.writeback",
668 !!(flags
& BDRV_O_CACHE_WB
));
670 if (!qemu_opt_get(all_opts
, "cache.direct")) {
671 qemu_opt_set_bool(all_opts
, "cache.direct",
672 !!(flags
& BDRV_O_NOCACHE
));
674 if (!qemu_opt_get(all_opts
, "cache.no-flush")) {
675 qemu_opt_set_bool(all_opts
, "cache.no-flush",
676 !!(flags
& BDRV_O_NO_FLUSH
));
678 qemu_opt_unset(all_opts
, "cache");
681 /* Get a QDict for processing the options */
682 bs_opts
= qdict_new();
683 qemu_opts_to_qdict(all_opts
, bs_opts
);
685 legacy_opts
= qemu_opts_create(&qemu_legacy_drive_opts
, NULL
, 0,
687 qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, &local_err
);
688 if (error_is_set(&local_err
)) {
689 qerror_report_err(local_err
);
690 error_free(local_err
);
694 /* Deprecated option boot=[on|off] */
695 if (qemu_opt_get(legacy_opts
, "boot") != NULL
) {
696 fprintf(stderr
, "qemu-kvm: boot=on|off is deprecated and will be "
697 "ignored. Future versions will reject this parameter. Please "
698 "update your scripts.\n");
702 value
= qemu_opt_get(legacy_opts
, "media");
704 if (!strcmp(value
, "disk")) {
706 } else if (!strcmp(value
, "cdrom")) {
710 error_report("'%s' invalid media", value
);
715 /* copy-on-read is disabled with a warning for read-only devices */
716 read_only
|= qemu_opt_get_bool(legacy_opts
, "read-only", false);
717 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
719 if (read_only
&& copy_on_read
) {
720 error_report("warning: disabling copy-on-read on read-only drive");
721 copy_on_read
= false;
724 qdict_put(bs_opts
, "read-only",
725 qstring_from_str(read_only
? "on" : "off"));
726 qdict_put(bs_opts
, "copy-on-read",
727 qstring_from_str(copy_on_read
? "on" :"off"));
729 /* Controller type */
730 value
= qemu_opt_get(legacy_opts
, "if");
733 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
736 if (type
== IF_COUNT
) {
737 error_report("unsupported bus type '%s'", value
);
741 type
= block_default_type
;
745 cyls
= qemu_opt_get_number(legacy_opts
, "cyls", 0);
746 heads
= qemu_opt_get_number(legacy_opts
, "heads", 0);
747 secs
= qemu_opt_get_number(legacy_opts
, "secs", 0);
749 if (cyls
|| heads
|| secs
) {
751 error_report("invalid physical cyls number");
755 error_report("invalid physical heads number");
759 error_report("invalid physical secs number");
764 translation
= BIOS_ATA_TRANSLATION_AUTO
;
765 value
= qemu_opt_get(legacy_opts
, "trans");
768 error_report("'%s' trans must be used with cyls, heads and secs",
772 if (!strcmp(value
, "none")) {
773 translation
= BIOS_ATA_TRANSLATION_NONE
;
774 } else if (!strcmp(value
, "lba")) {
775 translation
= BIOS_ATA_TRANSLATION_LBA
;
776 } else if (!strcmp(value
, "auto")) {
777 translation
= BIOS_ATA_TRANSLATION_AUTO
;
779 error_report("'%s' invalid translation type", value
);
784 if (media
== MEDIA_CDROM
) {
785 if (cyls
|| secs
|| heads
) {
786 error_report("CHS can't be set with media=cdrom");
791 /* Device address specified by bus/unit or index.
792 * If none was specified, try to find the first free one. */
793 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
794 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
795 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
797 max_devs
= if_max_devs
[type
];
800 if (bus_id
!= 0 || unit_id
!= -1) {
801 error_report("index cannot be used with bus and unit");
804 bus_id
= drive_index_to_bus_id(type
, index
);
805 unit_id
= drive_index_to_unit_id(type
, index
);
810 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
812 if (max_devs
&& unit_id
>= max_devs
) {
819 if (max_devs
&& unit_id
>= max_devs
) {
820 error_report("unit %d too big (max is %d)", unit_id
, max_devs
- 1);
824 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
825 error_report("drive with bus=%d, unit=%d (index=%d) exists",
826 bus_id
, unit_id
, index
);
830 /* no id supplied -> create one */
831 if (qemu_opts_id(all_opts
) == NULL
) {
833 const char *mediastr
= "";
834 if (type
== IF_IDE
|| type
== IF_SCSI
) {
835 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
838 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
841 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
844 qdict_put(bs_opts
, "id", qstring_from_str(new_id
));
848 /* Add virtio block device */
849 devaddr
= qemu_opt_get(legacy_opts
, "addr");
850 if (devaddr
&& type
!= IF_VIRTIO
) {
851 error_report("addr is not supported by this bus type");
855 if (type
== IF_VIRTIO
) {
857 devopts
= qemu_opts_create(qemu_find_opts("device"), NULL
, 0,
859 if (arch_type
== QEMU_ARCH_S390X
) {
860 qemu_opt_set(devopts
, "driver", "virtio-blk-s390");
862 qemu_opt_set(devopts
, "driver", "virtio-blk-pci");
864 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"));
866 qemu_opt_set(devopts
, "addr", devaddr
);
870 /* Actual block device init: Functionality shared with blockdev-add */
871 dinfo
= blockdev_init(bs_opts
, type
, &local_err
);
873 if (error_is_set(&local_err
)) {
874 qerror_report_err(local_err
);
875 error_free(local_err
);
879 assert(!error_is_set(&local_err
));
882 /* Set legacy DriveInfo fields */
883 dinfo
->enable_auto_del
= true;
884 dinfo
->opts
= all_opts
;
887 dinfo
->heads
= heads
;
889 dinfo
->trans
= translation
;
892 dinfo
->unit
= unit_id
;
893 dinfo
->devaddr
= devaddr
;
900 dinfo
->media_cd
= media
== MEDIA_CDROM
;
907 qemu_opts_del(legacy_opts
);
911 void do_commit(Monitor
*mon
, const QDict
*qdict
)
913 const char *device
= qdict_get_str(qdict
, "device");
914 BlockDriverState
*bs
;
917 if (!strcmp(device
, "all")) {
918 ret
= bdrv_commit_all();
920 bs
= bdrv_find(device
);
922 monitor_printf(mon
, "Device '%s' not found\n", device
);
925 ret
= bdrv_commit(bs
);
928 monitor_printf(mon
, "'commit' error for '%s': %s\n", device
,
933 static void blockdev_do_action(int kind
, void *data
, Error
**errp
)
935 TransactionAction action
;
936 TransactionActionList list
;
940 list
.value
= &action
;
942 qmp_transaction(&list
, errp
);
945 void qmp_blockdev_snapshot_sync(const char *device
, const char *snapshot_file
,
946 bool has_format
, const char *format
,
947 bool has_mode
, enum NewImageMode mode
,
950 BlockdevSnapshot snapshot
= {
951 .device
= (char *) device
,
952 .snapshot_file
= (char *) snapshot_file
,
953 .has_format
= has_format
,
954 .format
= (char *) format
,
955 .has_mode
= has_mode
,
958 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
962 void qmp_blockdev_snapshot_internal_sync(const char *device
,
966 BlockdevSnapshotInternal snapshot
= {
967 .device
= (char *) device
,
968 .name
= (char *) name
971 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
975 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
982 BlockDriverState
*bs
= bdrv_find(device
);
984 Error
*local_err
= NULL
;
985 SnapshotInfo
*info
= NULL
;
989 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1002 error_setg(errp
, "Name or id must be provided");
1006 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1007 if (error_is_set(&local_err
)) {
1008 error_propagate(errp
, local_err
);
1013 "Snapshot with id '%s' and name '%s' does not exist on "
1015 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1019 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1020 if (error_is_set(&local_err
)) {
1021 error_propagate(errp
, local_err
);
1025 info
= g_malloc0(sizeof(SnapshotInfo
));
1026 info
->id
= g_strdup(sn
.id_str
);
1027 info
->name
= g_strdup(sn
.name
);
1028 info
->date_nsec
= sn
.date_nsec
;
1029 info
->date_sec
= sn
.date_sec
;
1030 info
->vm_state_size
= sn
.vm_state_size
;
1031 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1032 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1037 /* New and old BlockDriverState structs for group snapshots */
1039 typedef struct BlkTransactionState BlkTransactionState
;
1041 /* Only prepare() may fail. In a single transaction, only one of commit() or
1042 abort() will be called, clean() will always be called if it present. */
1043 typedef struct BdrvActionOps
{
1044 /* Size of state struct, in bytes. */
1045 size_t instance_size
;
1046 /* Prepare the work, must NOT be NULL. */
1047 void (*prepare
)(BlkTransactionState
*common
, Error
**errp
);
1048 /* Commit the changes, can be NULL. */
1049 void (*commit
)(BlkTransactionState
*common
);
1050 /* Abort the changes on fail, can be NULL. */
1051 void (*abort
)(BlkTransactionState
*common
);
1052 /* Clean up resource in the end, can be NULL. */
1053 void (*clean
)(BlkTransactionState
*common
);
1057 * This structure must be arranged as first member in child type, assuming
1058 * that compiler will also arrange it to the same address with parent instance.
1059 * Later it will be used in free().
1061 struct BlkTransactionState
{
1062 TransactionAction
*action
;
1063 const BdrvActionOps
*ops
;
1064 QSIMPLEQ_ENTRY(BlkTransactionState
) entry
;
1067 /* internal snapshot private data */
1068 typedef struct InternalSnapshotState
{
1069 BlkTransactionState common
;
1070 BlockDriverState
*bs
;
1071 QEMUSnapshotInfo sn
;
1072 } InternalSnapshotState
;
1074 static void internal_snapshot_prepare(BlkTransactionState
*common
,
1079 BlockDriverState
*bs
;
1080 QEMUSnapshotInfo old_sn
, *sn
;
1083 BlockdevSnapshotInternal
*internal
;
1084 InternalSnapshotState
*state
;
1087 g_assert(common
->action
->kind
==
1088 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1089 internal
= common
->action
->blockdev_snapshot_internal_sync
;
1090 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1092 /* 1. parse input */
1093 device
= internal
->device
;
1094 name
= internal
->name
;
1096 /* 2. check for validation */
1097 bs
= bdrv_find(device
);
1099 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1103 if (!bdrv_is_inserted(bs
)) {
1104 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1108 if (bdrv_is_read_only(bs
)) {
1109 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1113 if (!bdrv_can_snapshot(bs
)) {
1114 error_set(errp
, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
1115 bs
->drv
->format_name
, device
, "internal snapshot");
1119 if (!strlen(name
)) {
1120 error_setg(errp
, "Name is empty");
1124 /* check whether a snapshot with name exist */
1125 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
, errp
);
1126 if (error_is_set(errp
)) {
1130 "Snapshot with name '%s' already exists on device '%s'",
1135 /* 3. take the snapshot */
1137 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1138 qemu_gettimeofday(&tv
);
1139 sn
->date_sec
= tv
.tv_sec
;
1140 sn
->date_nsec
= tv
.tv_usec
* 1000;
1141 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1143 ret1
= bdrv_snapshot_create(bs
, sn
);
1145 error_setg_errno(errp
, -ret1
,
1146 "Failed to create snapshot '%s' on device '%s'",
1151 /* 4. succeed, mark a snapshot is created */
1155 static void internal_snapshot_abort(BlkTransactionState
*common
)
1157 InternalSnapshotState
*state
=
1158 DO_UPCAST(InternalSnapshotState
, common
, common
);
1159 BlockDriverState
*bs
= state
->bs
;
1160 QEMUSnapshotInfo
*sn
= &state
->sn
;
1161 Error
*local_error
= NULL
;
1167 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1168 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1169 "device '%s' in abort: %s",
1172 bdrv_get_device_name(bs
),
1173 error_get_pretty(local_error
));
1174 error_free(local_error
);
1178 /* external snapshot private data */
1179 typedef struct ExternalSnapshotState
{
1180 BlkTransactionState common
;
1181 BlockDriverState
*old_bs
;
1182 BlockDriverState
*new_bs
;
1183 } ExternalSnapshotState
;
1185 static void external_snapshot_prepare(BlkTransactionState
*common
,
1190 Error
*local_err
= NULL
;
1192 const char *new_image_file
;
1193 const char *format
= "qcow2";
1194 enum NewImageMode mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1195 ExternalSnapshotState
*state
=
1196 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1197 TransactionAction
*action
= common
->action
;
1199 /* get parameters */
1200 g_assert(action
->kind
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
);
1202 device
= action
->blockdev_snapshot_sync
->device
;
1203 new_image_file
= action
->blockdev_snapshot_sync
->snapshot_file
;
1204 if (action
->blockdev_snapshot_sync
->has_format
) {
1205 format
= action
->blockdev_snapshot_sync
->format
;
1207 if (action
->blockdev_snapshot_sync
->has_mode
) {
1208 mode
= action
->blockdev_snapshot_sync
->mode
;
1211 /* start processing */
1212 drv
= bdrv_find_format(format
);
1214 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1218 state
->old_bs
= bdrv_find(device
);
1219 if (!state
->old_bs
) {
1220 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1224 if (!bdrv_is_inserted(state
->old_bs
)) {
1225 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1229 if (bdrv_in_use(state
->old_bs
)) {
1230 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1234 if (!bdrv_is_read_only(state
->old_bs
)) {
1235 if (bdrv_flush(state
->old_bs
)) {
1236 error_set(errp
, QERR_IO_ERROR
);
1241 if (bdrv_check_ext_snapshot(state
->old_bs
) != EXT_SNAPSHOT_ALLOWED
) {
1242 error_set(errp
, QERR_FEATURE_DISABLED
, "snapshot");
1246 flags
= state
->old_bs
->open_flags
;
1248 /* create new image w/backing file */
1249 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1250 bdrv_img_create(new_image_file
, format
,
1251 state
->old_bs
->filename
,
1252 state
->old_bs
->drv
->format_name
,
1253 NULL
, -1, flags
, &local_err
, false);
1254 if (error_is_set(&local_err
)) {
1255 error_propagate(errp
, local_err
);
1260 /* We will manually add the backing_hd field to the bs later */
1261 state
->new_bs
= bdrv_new("");
1262 /* TODO Inherit bs->options or only take explicit options with an
1263 * extended QMP command? */
1264 ret
= bdrv_open(state
->new_bs
, new_image_file
, NULL
,
1265 flags
| BDRV_O_NO_BACKING
, drv
, &local_err
);
1267 error_propagate(errp
, local_err
);
1271 static void external_snapshot_commit(BlkTransactionState
*common
)
1273 ExternalSnapshotState
*state
=
1274 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1276 /* This removes our old bs and adds the new bs */
1277 bdrv_append(state
->new_bs
, state
->old_bs
);
1278 /* We don't need (or want) to use the transactional
1279 * bdrv_reopen_multiple() across all the entries at once, because we
1280 * don't want to abort all of them if one of them fails the reopen */
1281 bdrv_reopen(state
->new_bs
, state
->new_bs
->open_flags
& ~BDRV_O_RDWR
,
1285 static void external_snapshot_abort(BlkTransactionState
*common
)
1287 ExternalSnapshotState
*state
=
1288 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1289 if (state
->new_bs
) {
1290 bdrv_unref(state
->new_bs
);
1294 typedef struct DriveBackupState
{
1295 BlkTransactionState common
;
1296 BlockDriverState
*bs
;
1300 static void drive_backup_prepare(BlkTransactionState
*common
, Error
**errp
)
1302 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1303 DriveBackup
*backup
;
1304 Error
*local_err
= NULL
;
1306 assert(common
->action
->kind
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1307 backup
= common
->action
->drive_backup
;
1309 qmp_drive_backup(backup
->device
, backup
->target
,
1310 backup
->has_format
, backup
->format
,
1312 backup
->has_mode
, backup
->mode
,
1313 backup
->has_speed
, backup
->speed
,
1314 backup
->has_on_source_error
, backup
->on_source_error
,
1315 backup
->has_on_target_error
, backup
->on_target_error
,
1317 if (error_is_set(&local_err
)) {
1318 error_propagate(errp
, local_err
);
1324 state
->bs
= bdrv_find(backup
->device
);
1325 state
->job
= state
->bs
->job
;
1328 static void drive_backup_abort(BlkTransactionState
*common
)
1330 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1331 BlockDriverState
*bs
= state
->bs
;
1333 /* Only cancel if it's the job we started */
1334 if (bs
&& bs
->job
&& bs
->job
== state
->job
) {
1335 block_job_cancel_sync(bs
->job
);
1339 static void abort_prepare(BlkTransactionState
*common
, Error
**errp
)
1341 error_setg(errp
, "Transaction aborted using Abort action");
1344 static void abort_commit(BlkTransactionState
*common
)
1346 g_assert_not_reached(); /* this action never succeeds */
1349 static const BdrvActionOps actions
[] = {
1350 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
1351 .instance_size
= sizeof(ExternalSnapshotState
),
1352 .prepare
= external_snapshot_prepare
,
1353 .commit
= external_snapshot_commit
,
1354 .abort
= external_snapshot_abort
,
1356 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
1357 .instance_size
= sizeof(DriveBackupState
),
1358 .prepare
= drive_backup_prepare
,
1359 .abort
= drive_backup_abort
,
1361 [TRANSACTION_ACTION_KIND_ABORT
] = {
1362 .instance_size
= sizeof(BlkTransactionState
),
1363 .prepare
= abort_prepare
,
1364 .commit
= abort_commit
,
1366 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
1367 .instance_size
= sizeof(InternalSnapshotState
),
1368 .prepare
= internal_snapshot_prepare
,
1369 .abort
= internal_snapshot_abort
,
1374 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1375 * then we do not pivot any of the devices in the group, and abandon the
1378 void qmp_transaction(TransactionActionList
*dev_list
, Error
**errp
)
1380 TransactionActionList
*dev_entry
= dev_list
;
1381 BlkTransactionState
*state
, *next
;
1382 Error
*local_err
= NULL
;
1384 QSIMPLEQ_HEAD(snap_bdrv_states
, BlkTransactionState
) snap_bdrv_states
;
1385 QSIMPLEQ_INIT(&snap_bdrv_states
);
1387 /* drain all i/o before any snapshots */
1390 /* We don't do anything in this loop that commits us to the snapshot */
1391 while (NULL
!= dev_entry
) {
1392 TransactionAction
*dev_info
= NULL
;
1393 const BdrvActionOps
*ops
;
1395 dev_info
= dev_entry
->value
;
1396 dev_entry
= dev_entry
->next
;
1398 assert(dev_info
->kind
< ARRAY_SIZE(actions
));
1400 ops
= &actions
[dev_info
->kind
];
1401 assert(ops
->instance_size
> 0);
1403 state
= g_malloc0(ops
->instance_size
);
1405 state
->action
= dev_info
;
1406 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
1408 state
->ops
->prepare(state
, &local_err
);
1409 if (error_is_set(&local_err
)) {
1410 error_propagate(errp
, local_err
);
1411 goto delete_and_fail
;
1415 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1416 if (state
->ops
->commit
) {
1417 state
->ops
->commit(state
);
1426 * failure, and it is all-or-none; abandon each new bs, and keep using
1427 * the original bs for all images
1429 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1430 if (state
->ops
->abort
) {
1431 state
->ops
->abort(state
);
1435 QSIMPLEQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
1436 if (state
->ops
->clean
) {
1437 state
->ops
->clean(state
);
1444 static void eject_device(BlockDriverState
*bs
, int force
, Error
**errp
)
1446 if (bdrv_in_use(bs
)) {
1447 error_set(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
1450 if (!bdrv_dev_has_removable_media(bs
)) {
1451 error_set(errp
, QERR_DEVICE_NOT_REMOVABLE
, bdrv_get_device_name(bs
));
1455 if (bdrv_dev_is_medium_locked(bs
) && !bdrv_dev_is_tray_open(bs
)) {
1456 bdrv_dev_eject_request(bs
, force
);
1458 error_set(errp
, QERR_DEVICE_LOCKED
, bdrv_get_device_name(bs
));
1466 void qmp_eject(const char *device
, bool has_force
, bool force
, Error
**errp
)
1468 BlockDriverState
*bs
;
1470 bs
= bdrv_find(device
);
1472 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1476 eject_device(bs
, force
, errp
);
1479 void qmp_block_passwd(const char *device
, const char *password
, Error
**errp
)
1481 BlockDriverState
*bs
;
1484 bs
= bdrv_find(device
);
1486 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1490 err
= bdrv_set_key(bs
, password
);
1491 if (err
== -EINVAL
) {
1492 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1494 } else if (err
< 0) {
1495 error_set(errp
, QERR_INVALID_PASSWORD
);
1500 static void qmp_bdrv_open_encrypted(BlockDriverState
*bs
, const char *filename
,
1501 int bdrv_flags
, BlockDriver
*drv
,
1502 const char *password
, Error
**errp
)
1504 Error
*local_err
= NULL
;
1507 ret
= bdrv_open(bs
, filename
, NULL
, bdrv_flags
, drv
, &local_err
);
1509 error_propagate(errp
, local_err
);
1513 if (bdrv_key_required(bs
)) {
1515 if (bdrv_set_key(bs
, password
) < 0) {
1516 error_set(errp
, QERR_INVALID_PASSWORD
);
1519 error_set(errp
, QERR_DEVICE_ENCRYPTED
, bdrv_get_device_name(bs
),
1520 bdrv_get_encrypted_filename(bs
));
1522 } else if (password
) {
1523 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1527 void qmp_change_blockdev(const char *device
, const char *filename
,
1528 const char *format
, Error
**errp
)
1530 BlockDriverState
*bs
;
1531 BlockDriver
*drv
= NULL
;
1535 bs
= bdrv_find(device
);
1537 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1542 drv
= bdrv_find_whitelisted_format(format
, bs
->read_only
);
1544 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1549 eject_device(bs
, 0, &err
);
1550 if (error_is_set(&err
)) {
1551 error_propagate(errp
, err
);
1555 bdrv_flags
= bdrv_is_read_only(bs
) ? 0 : BDRV_O_RDWR
;
1556 bdrv_flags
|= bdrv_is_snapshot(bs
) ? BDRV_O_SNAPSHOT
: 0;
1558 qmp_bdrv_open_encrypted(bs
, filename
, bdrv_flags
, drv
, NULL
, errp
);
1561 /* throttling disk I/O limits */
1562 void qmp_block_set_io_throttle(const char *device
, int64_t bps
, int64_t bps_rd
,
1569 bool has_bps_rd_max
,
1571 bool has_bps_wr_max
,
1575 bool has_iops_rd_max
,
1576 int64_t iops_rd_max
,
1577 bool has_iops_wr_max
,
1578 int64_t iops_wr_max
,
1580 int64_t iops_size
, Error
**errp
)
1583 BlockDriverState
*bs
;
1585 bs
= bdrv_find(device
);
1587 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1591 memset(&cfg
, 0, sizeof(cfg
));
1592 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
= bps
;
1593 cfg
.buckets
[THROTTLE_BPS_READ
].avg
= bps_rd
;
1594 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
= bps_wr
;
1596 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
= iops
;
1597 cfg
.buckets
[THROTTLE_OPS_READ
].avg
= iops_rd
;
1598 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
= iops_wr
;
1601 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
= bps_max
;
1603 if (has_bps_rd_max
) {
1604 cfg
.buckets
[THROTTLE_BPS_READ
].max
= bps_rd_max
;
1606 if (has_bps_wr_max
) {
1607 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
= bps_wr_max
;
1610 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
= iops_max
;
1612 if (has_iops_rd_max
) {
1613 cfg
.buckets
[THROTTLE_OPS_READ
].max
= iops_rd_max
;
1615 if (has_iops_wr_max
) {
1616 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
= iops_wr_max
;
1619 if (has_iops_size
) {
1620 cfg
.op_size
= iops_size
;
1623 if (!check_throttle_config(&cfg
, errp
)) {
1627 if (!bs
->io_limits_enabled
&& throttle_enabled(&cfg
)) {
1628 bdrv_io_limits_enable(bs
);
1629 } else if (bs
->io_limits_enabled
&& !throttle_enabled(&cfg
)) {
1630 bdrv_io_limits_disable(bs
);
1633 if (bs
->io_limits_enabled
) {
1634 bdrv_set_io_limits(bs
, &cfg
);
1638 int do_drive_del(Monitor
*mon
, const QDict
*qdict
, QObject
**ret_data
)
1640 const char *id
= qdict_get_str(qdict
, "id");
1641 BlockDriverState
*bs
;
1645 qerror_report(QERR_DEVICE_NOT_FOUND
, id
);
1648 if (bdrv_in_use(bs
)) {
1649 qerror_report(QERR_DEVICE_IN_USE
, id
);
1653 /* quiesce block driver; prevent further io */
1658 /* if we have a device attached to this BlockDriverState
1659 * then we need to make the drive anonymous until the device
1660 * can be removed. If this is a drive with no device backing
1661 * then we can just get rid of the block driver state right here.
1663 if (bdrv_get_attached_dev(bs
)) {
1666 /* Further I/O must not pause the guest */
1667 bdrv_set_on_error(bs
, BLOCKDEV_ON_ERROR_REPORT
,
1668 BLOCKDEV_ON_ERROR_REPORT
);
1670 drive_uninit(drive_get_by_blockdev(bs
));
1676 void qmp_block_resize(const char *device
, int64_t size
, Error
**errp
)
1678 BlockDriverState
*bs
;
1681 bs
= bdrv_find(device
);
1683 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1688 error_set(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
1692 /* complete all in-flight operations before resizing the device */
1695 ret
= bdrv_truncate(bs
, size
);
1700 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1703 error_set(errp
, QERR_UNSUPPORTED
);
1706 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1709 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1712 error_setg_errno(errp
, -ret
, "Could not resize");
1717 static void block_job_cb(void *opaque
, int ret
)
1719 BlockDriverState
*bs
= opaque
;
1722 trace_block_job_cb(bs
, bs
->job
, ret
);
1725 obj
= qobject_from_block_job(bs
->job
);
1727 QDict
*dict
= qobject_to_qdict(obj
);
1728 qdict_put(dict
, "error", qstring_from_str(strerror(-ret
)));
1731 if (block_job_is_cancelled(bs
->job
)) {
1732 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED
, obj
);
1734 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED
, obj
);
1736 qobject_decref(obj
);
1738 bdrv_put_ref_bh_schedule(bs
);
1741 void qmp_block_stream(const char *device
, bool has_base
,
1742 const char *base
, bool has_speed
, int64_t speed
,
1743 bool has_on_error
, BlockdevOnError on_error
,
1746 BlockDriverState
*bs
;
1747 BlockDriverState
*base_bs
= NULL
;
1748 Error
*local_err
= NULL
;
1750 if (!has_on_error
) {
1751 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1754 bs
= bdrv_find(device
);
1756 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1761 base_bs
= bdrv_find_backing_image(bs
, base
);
1762 if (base_bs
== NULL
) {
1763 error_set(errp
, QERR_BASE_NOT_FOUND
, base
);
1768 stream_start(bs
, base_bs
, base
, has_speed
? speed
: 0,
1769 on_error
, block_job_cb
, bs
, &local_err
);
1770 if (error_is_set(&local_err
)) {
1771 error_propagate(errp
, local_err
);
1775 trace_qmp_block_stream(bs
, bs
->job
);
1778 void qmp_block_commit(const char *device
,
1779 bool has_base
, const char *base
, const char *top
,
1780 bool has_speed
, int64_t speed
,
1783 BlockDriverState
*bs
;
1784 BlockDriverState
*base_bs
, *top_bs
;
1785 Error
*local_err
= NULL
;
1786 /* This will be part of the QMP command, if/when the
1787 * BlockdevOnError change for blkmirror makes it in
1789 BlockdevOnError on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1791 /* drain all i/o before commits */
1794 bs
= bdrv_find(device
);
1796 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1800 /* default top_bs is the active layer */
1804 if (strcmp(bs
->filename
, top
) != 0) {
1805 top_bs
= bdrv_find_backing_image(bs
, top
);
1809 if (top_bs
== NULL
) {
1810 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
1814 if (has_base
&& base
) {
1815 base_bs
= bdrv_find_backing_image(top_bs
, base
);
1817 base_bs
= bdrv_find_base(top_bs
);
1820 if (base_bs
== NULL
) {
1821 error_set(errp
, QERR_BASE_NOT_FOUND
, base
? base
: "NULL");
1826 commit_active_start(bs
, base_bs
, speed
, on_error
, block_job_cb
,
1829 commit_start(bs
, base_bs
, top_bs
, speed
, on_error
, block_job_cb
, bs
,
1832 if (local_err
!= NULL
) {
1833 error_propagate(errp
, local_err
);
1838 void qmp_drive_backup(const char *device
, const char *target
,
1839 bool has_format
, const char *format
,
1840 enum MirrorSyncMode sync
,
1841 bool has_mode
, enum NewImageMode mode
,
1842 bool has_speed
, int64_t speed
,
1843 bool has_on_source_error
, BlockdevOnError on_source_error
,
1844 bool has_on_target_error
, BlockdevOnError on_target_error
,
1847 BlockDriverState
*bs
;
1848 BlockDriverState
*target_bs
;
1849 BlockDriverState
*source
= NULL
;
1850 BlockDriver
*drv
= NULL
;
1851 Error
*local_err
= NULL
;
1859 if (!has_on_source_error
) {
1860 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1862 if (!has_on_target_error
) {
1863 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1866 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1869 bs
= bdrv_find(device
);
1871 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1875 if (!bdrv_is_inserted(bs
)) {
1876 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1881 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1884 drv
= bdrv_find_format(format
);
1886 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1891 if (bdrv_in_use(bs
)) {
1892 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1896 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1898 /* See if we have a backing HD we can use to create our new image
1900 if (sync
== MIRROR_SYNC_MODE_TOP
) {
1901 source
= bs
->backing_hd
;
1903 sync
= MIRROR_SYNC_MODE_FULL
;
1906 if (sync
== MIRROR_SYNC_MODE_NONE
) {
1910 size
= bdrv_getlength(bs
);
1912 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1916 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1917 assert(format
&& drv
);
1919 bdrv_img_create(target
, format
, source
->filename
,
1920 source
->drv
->format_name
, NULL
,
1921 size
, flags
, &local_err
, false);
1923 bdrv_img_create(target
, format
, NULL
, NULL
, NULL
,
1924 size
, flags
, &local_err
, false);
1928 if (error_is_set(&local_err
)) {
1929 error_propagate(errp
, local_err
);
1933 target_bs
= bdrv_new("");
1934 ret
= bdrv_open(target_bs
, target
, NULL
, flags
, drv
, &local_err
);
1936 bdrv_unref(target_bs
);
1937 error_propagate(errp
, local_err
);
1941 backup_start(bs
, target_bs
, speed
, sync
, on_source_error
, on_target_error
,
1942 block_job_cb
, bs
, &local_err
);
1943 if (local_err
!= NULL
) {
1944 bdrv_unref(target_bs
);
1945 error_propagate(errp
, local_err
);
1950 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1952 void qmp_drive_mirror(const char *device
, const char *target
,
1953 bool has_format
, const char *format
,
1954 enum MirrorSyncMode sync
,
1955 bool has_mode
, enum NewImageMode mode
,
1956 bool has_speed
, int64_t speed
,
1957 bool has_granularity
, uint32_t granularity
,
1958 bool has_buf_size
, int64_t buf_size
,
1959 bool has_on_source_error
, BlockdevOnError on_source_error
,
1960 bool has_on_target_error
, BlockdevOnError on_target_error
,
1963 BlockDriverState
*bs
;
1964 BlockDriverState
*source
, *target_bs
;
1965 BlockDriver
*drv
= NULL
;
1966 Error
*local_err
= NULL
;
1974 if (!has_on_source_error
) {
1975 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1977 if (!has_on_target_error
) {
1978 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1981 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1983 if (!has_granularity
) {
1986 if (!has_buf_size
) {
1987 buf_size
= DEFAULT_MIRROR_BUF_SIZE
;
1990 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
1991 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1994 if (granularity
& (granularity
- 1)) {
1995 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1999 bs
= bdrv_find(device
);
2001 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
2005 if (!bdrv_is_inserted(bs
)) {
2006 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
2011 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
2014 drv
= bdrv_find_format(format
);
2016 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
2021 if (bdrv_in_use(bs
)) {
2022 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
2026 flags
= bs
->open_flags
| BDRV_O_RDWR
;
2027 source
= bs
->backing_hd
;
2028 if (!source
&& sync
== MIRROR_SYNC_MODE_TOP
) {
2029 sync
= MIRROR_SYNC_MODE_FULL
;
2031 if (sync
== MIRROR_SYNC_MODE_NONE
) {
2035 size
= bdrv_getlength(bs
);
2037 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
2041 if ((sync
== MIRROR_SYNC_MODE_FULL
|| !source
)
2042 && mode
!= NEW_IMAGE_MODE_EXISTING
)
2044 /* create new image w/o backing file */
2045 assert(format
&& drv
);
2046 bdrv_img_create(target
, format
,
2047 NULL
, NULL
, NULL
, size
, flags
, &local_err
, false);
2050 case NEW_IMAGE_MODE_EXISTING
:
2052 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
2053 /* create new image with backing file */
2054 bdrv_img_create(target
, format
,
2056 source
->drv
->format_name
,
2057 NULL
, size
, flags
, &local_err
, false);
2064 if (error_is_set(&local_err
)) {
2065 error_propagate(errp
, local_err
);
2069 /* Mirroring takes care of copy-on-write using the source's backing
2072 target_bs
= bdrv_new("");
2073 ret
= bdrv_open(target_bs
, target
, NULL
, flags
| BDRV_O_NO_BACKING
, drv
,
2076 bdrv_unref(target_bs
);
2077 error_propagate(errp
, local_err
);
2081 mirror_start(bs
, target_bs
, speed
, granularity
, buf_size
, sync
,
2082 on_source_error
, on_target_error
,
2083 block_job_cb
, bs
, &local_err
);
2084 if (local_err
!= NULL
) {
2085 bdrv_unref(target_bs
);
2086 error_propagate(errp
, local_err
);
2091 static BlockJob
*find_block_job(const char *device
)
2093 BlockDriverState
*bs
;
2095 bs
= bdrv_find(device
);
2096 if (!bs
|| !bs
->job
) {
2102 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
2104 BlockJob
*job
= find_block_job(device
);
2107 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2111 block_job_set_speed(job
, speed
, errp
);
2114 void qmp_block_job_cancel(const char *device
,
2115 bool has_force
, bool force
, Error
**errp
)
2117 BlockJob
*job
= find_block_job(device
);
2124 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2127 if (job
->paused
&& !force
) {
2128 error_set(errp
, QERR_BLOCK_JOB_PAUSED
, device
);
2132 trace_qmp_block_job_cancel(job
);
2133 block_job_cancel(job
);
2136 void qmp_block_job_pause(const char *device
, Error
**errp
)
2138 BlockJob
*job
= find_block_job(device
);
2141 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2145 trace_qmp_block_job_pause(job
);
2146 block_job_pause(job
);
2149 void qmp_block_job_resume(const char *device
, Error
**errp
)
2151 BlockJob
*job
= find_block_job(device
);
2154 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2158 trace_qmp_block_job_resume(job
);
2159 block_job_resume(job
);
2162 void qmp_block_job_complete(const char *device
, Error
**errp
)
2164 BlockJob
*job
= find_block_job(device
);
2167 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2171 trace_qmp_block_job_complete(job
);
2172 block_job_complete(job
, errp
);
2175 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
2177 QmpOutputVisitor
*ov
= qmp_output_visitor_new();
2180 Error
*local_err
= NULL
;
2182 /* Require an ID in the top level */
2183 if (!options
->has_id
) {
2184 error_setg(errp
, "Block device needs an ID");
2188 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2189 * cache.direct=false instead of silently switching to aio=threads, except
2190 * if called from drive_init.
2192 * For now, simply forbidding the combination for all drivers will do. */
2193 if (options
->has_aio
&& options
->aio
== BLOCKDEV_AIO_OPTIONS_NATIVE
) {
2194 bool direct
= options
->cache
->has_direct
&& options
->cache
->direct
;
2195 if (!options
->has_cache
&& !direct
) {
2196 error_setg(errp
, "aio=native requires cache.direct=true");
2201 visit_type_BlockdevOptions(qmp_output_get_visitor(ov
),
2202 &options
, NULL
, &local_err
);
2203 if (error_is_set(&local_err
)) {
2204 error_propagate(errp
, local_err
);
2208 obj
= qmp_output_get_qobject(ov
);
2209 qdict
= qobject_to_qdict(obj
);
2211 qdict_flatten(qdict
);
2213 blockdev_init(qdict
, IF_NONE
, &local_err
);
2214 if (error_is_set(&local_err
)) {
2215 error_propagate(errp
, local_err
);
2220 qmp_output_visitor_cleanup(ov
);
2223 static void do_qmp_query_block_jobs_one(void *opaque
, BlockDriverState
*bs
)
2225 BlockJobInfoList
**prev
= opaque
;
2226 BlockJob
*job
= bs
->job
;
2229 BlockJobInfoList
*elem
= g_new0(BlockJobInfoList
, 1);
2230 elem
->value
= block_job_query(bs
->job
);
2231 (*prev
)->next
= elem
;
2236 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
2238 /* Dummy is a fake list element for holding the head pointer */
2239 BlockJobInfoList dummy
= {};
2240 BlockJobInfoList
*prev
= &dummy
;
2241 bdrv_iterate(do_qmp_query_block_jobs_one
, &prev
);
2245 QemuOptsList qemu_common_drive_opts
= {
2247 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
2251 .type
= QEMU_OPT_BOOL
,
2252 .help
= "enable/disable snapshot mode",
2255 .type
= QEMU_OPT_STRING
,
2256 .help
= "disk image",
2259 .type
= QEMU_OPT_STRING
,
2260 .help
= "discard operation (ignore/off, unmap/on)",
2262 .name
= "cache.writeback",
2263 .type
= QEMU_OPT_BOOL
,
2264 .help
= "enables writeback mode for any caches",
2266 .name
= "cache.direct",
2267 .type
= QEMU_OPT_BOOL
,
2268 .help
= "enables use of O_DIRECT (bypass the host page cache)",
2270 .name
= "cache.no-flush",
2271 .type
= QEMU_OPT_BOOL
,
2272 .help
= "ignore any flush requests for the device",
2275 .type
= QEMU_OPT_STRING
,
2276 .help
= "host AIO implementation (threads, native)",
2279 .type
= QEMU_OPT_STRING
,
2280 .help
= "disk format (raw, qcow2, ...)",
2283 .type
= QEMU_OPT_STRING
,
2284 .help
= "disk serial number",
2287 .type
= QEMU_OPT_STRING
,
2288 .help
= "read error action",
2291 .type
= QEMU_OPT_STRING
,
2292 .help
= "write error action",
2294 .name
= "read-only",
2295 .type
= QEMU_OPT_BOOL
,
2296 .help
= "open drive file as read-only",
2298 .name
= "throttling.iops-total",
2299 .type
= QEMU_OPT_NUMBER
,
2300 .help
= "limit total I/O operations per second",
2302 .name
= "throttling.iops-read",
2303 .type
= QEMU_OPT_NUMBER
,
2304 .help
= "limit read operations per second",
2306 .name
= "throttling.iops-write",
2307 .type
= QEMU_OPT_NUMBER
,
2308 .help
= "limit write operations per second",
2310 .name
= "throttling.bps-total",
2311 .type
= QEMU_OPT_NUMBER
,
2312 .help
= "limit total bytes per second",
2314 .name
= "throttling.bps-read",
2315 .type
= QEMU_OPT_NUMBER
,
2316 .help
= "limit read bytes per second",
2318 .name
= "throttling.bps-write",
2319 .type
= QEMU_OPT_NUMBER
,
2320 .help
= "limit write bytes per second",
2322 .name
= "throttling.iops-total-max",
2323 .type
= QEMU_OPT_NUMBER
,
2324 .help
= "I/O operations burst",
2326 .name
= "throttling.iops-read-max",
2327 .type
= QEMU_OPT_NUMBER
,
2328 .help
= "I/O operations read burst",
2330 .name
= "throttling.iops-write-max",
2331 .type
= QEMU_OPT_NUMBER
,
2332 .help
= "I/O operations write burst",
2334 .name
= "throttling.bps-total-max",
2335 .type
= QEMU_OPT_NUMBER
,
2336 .help
= "total bytes burst",
2338 .name
= "throttling.bps-read-max",
2339 .type
= QEMU_OPT_NUMBER
,
2340 .help
= "total bytes read burst",
2342 .name
= "throttling.bps-write-max",
2343 .type
= QEMU_OPT_NUMBER
,
2344 .help
= "total bytes write burst",
2346 .name
= "throttling.iops-size",
2347 .type
= QEMU_OPT_NUMBER
,
2348 .help
= "when limiting by iops max size of an I/O in bytes",
2350 .name
= "copy-on-read",
2351 .type
= QEMU_OPT_BOOL
,
2352 .help
= "copy read data from backing file into image file",
2354 { /* end of list */ }
2358 QemuOptsList qemu_drive_opts
= {
2360 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
2363 * no elements => accept any params
2364 * validation will happen later
2366 { /* end of list */ }