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(const char *file
, QDict
*bs_opts
,
311 BlockInterfaceType type
,
318 int on_read_error
, on_write_error
;
327 bool has_driver_specific_opts
;
328 BlockDriver
*drv
= NULL
;
330 /* Check common options by copying from bs_opts to opts, all other options
331 * stay in bs_opts for processing by bdrv_open(). */
332 id
= qdict_get_try_str(bs_opts
, "id");
333 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, &error
);
334 if (error_is_set(&error
)) {
335 error_propagate(errp
, error
);
339 qemu_opts_absorb_qdict(opts
, bs_opts
, &error
);
340 if (error_is_set(&error
)) {
341 error_propagate(errp
, error
);
346 qdict_del(bs_opts
, "id");
349 has_driver_specific_opts
= !!qdict_size(bs_opts
);
351 /* extract parameters */
352 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
353 ro
= qemu_opt_get_bool(opts
, "read-only", 0);
354 copy_on_read
= qemu_opt_get_bool(opts
, "copy-on-read", false);
356 serial
= qemu_opt_get(opts
, "serial");
358 if ((buf
= qemu_opt_get(opts
, "discard")) != NULL
) {
359 if (bdrv_parse_discard_flags(buf
, &bdrv_flags
) != 0) {
360 error_setg(errp
, "invalid discard option");
365 if (qemu_opt_get_bool(opts
, "cache.writeback", true)) {
366 bdrv_flags
|= BDRV_O_CACHE_WB
;
368 if (qemu_opt_get_bool(opts
, "cache.direct", false)) {
369 bdrv_flags
|= BDRV_O_NOCACHE
;
371 if (qemu_opt_get_bool(opts
, "cache.no-flush", false)) {
372 bdrv_flags
|= BDRV_O_NO_FLUSH
;
375 #ifdef CONFIG_LINUX_AIO
376 if ((buf
= qemu_opt_get(opts
, "aio")) != NULL
) {
377 if (!strcmp(buf
, "native")) {
378 bdrv_flags
|= BDRV_O_NATIVE_AIO
;
379 } else if (!strcmp(buf
, "threads")) {
380 /* this is the default */
382 error_setg(errp
, "invalid aio option");
388 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
389 if (is_help_option(buf
)) {
390 error_printf("Supported formats:");
391 bdrv_iterate_format(bdrv_format_print
, NULL
);
396 drv
= bdrv_find_format(buf
);
398 error_setg(errp
, "'%s' invalid format", buf
);
403 /* disk I/O throttling */
404 memset(&cfg
, 0, sizeof(cfg
));
405 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
=
406 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
407 cfg
.buckets
[THROTTLE_BPS_READ
].avg
=
408 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
409 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
=
410 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
411 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
=
412 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
413 cfg
.buckets
[THROTTLE_OPS_READ
].avg
=
414 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
415 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
=
416 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
418 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
=
419 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
420 cfg
.buckets
[THROTTLE_BPS_READ
].max
=
421 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
422 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
=
423 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
424 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
=
425 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
426 cfg
.buckets
[THROTTLE_OPS_READ
].max
=
427 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
428 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
=
429 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
431 cfg
.op_size
= qemu_opt_get_number(opts
, "throttling.iops-size", 0);
433 if (!check_throttle_config(&cfg
, &error
)) {
434 error_propagate(errp
, error
);
438 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
439 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
440 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_NONE
) {
441 error_setg(errp
, "werror is not supported by this bus type");
445 on_write_error
= parse_block_error_action(buf
, 0, &error
);
446 if (error_is_set(&error
)) {
447 error_propagate(errp
, error
);
452 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
453 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
454 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&& type
!= IF_NONE
) {
455 error_report("rerror is not supported by this bus type");
459 on_read_error
= parse_block_error_action(buf
, 1, &error
);
460 if (error_is_set(&error
)) {
461 error_propagate(errp
, error
);
467 dinfo
= g_malloc0(sizeof(*dinfo
));
468 dinfo
->id
= g_strdup(qemu_opts_id(opts
));
469 dinfo
->bdrv
= bdrv_new(dinfo
->id
);
470 dinfo
->bdrv
->open_flags
= snapshot
? BDRV_O_SNAPSHOT
: 0;
471 dinfo
->bdrv
->read_only
= ro
;
474 if (serial
!= NULL
) {
475 dinfo
->serial
= g_strdup(serial
);
477 QTAILQ_INSERT_TAIL(&drives
, dinfo
, next
);
479 bdrv_set_on_error(dinfo
->bdrv
, on_read_error
, on_write_error
);
481 /* disk I/O throttling */
482 if (throttle_enabled(&cfg
)) {
483 bdrv_io_limits_enable(dinfo
->bdrv
);
484 bdrv_set_io_limits(dinfo
->bdrv
, &cfg
);
487 if (!file
|| !*file
) {
488 if (has_driver_specific_opts
) {
497 /* always use cache=unsafe with snapshot */
498 bdrv_flags
&= ~BDRV_O_CACHE_MASK
;
499 bdrv_flags
|= (BDRV_O_SNAPSHOT
|BDRV_O_CACHE_WB
|BDRV_O_NO_FLUSH
);
503 bdrv_flags
|= BDRV_O_COPY_ON_READ
;
506 if (runstate_check(RUN_STATE_INMIGRATE
)) {
507 bdrv_flags
|= BDRV_O_INCOMING
;
510 bdrv_flags
|= ro
? 0 : BDRV_O_RDWR
;
513 ret
= bdrv_open(dinfo
->bdrv
, file
, bs_opts
, bdrv_flags
, drv
, &error
);
516 error_setg(errp
, "could not open disk image %s: %s",
517 file
?: dinfo
->id
, error_get_pretty(error
));
522 if (bdrv_key_required(dinfo
->bdrv
))
531 bdrv_unref(dinfo
->bdrv
);
533 QTAILQ_REMOVE(&drives
, dinfo
, next
);
541 static void qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
)
545 value
= qemu_opt_get(opts
, from
);
547 qemu_opt_set(opts
, to
, value
);
548 qemu_opt_unset(opts
, from
);
552 QemuOptsList qemu_legacy_drive_opts
= {
554 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
558 .type
= QEMU_OPT_NUMBER
,
559 .help
= "bus number",
562 .type
= QEMU_OPT_NUMBER
,
563 .help
= "unit number (i.e. lun for scsi)",
566 .type
= QEMU_OPT_NUMBER
,
567 .help
= "index number",
570 .type
= QEMU_OPT_STRING
,
571 .help
= "media type (disk, cdrom)",
574 .type
= QEMU_OPT_STRING
,
575 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
578 .type
= QEMU_OPT_NUMBER
,
579 .help
= "number of cylinders (ide disk geometry)",
582 .type
= QEMU_OPT_NUMBER
,
583 .help
= "number of heads (ide disk geometry)",
586 .type
= QEMU_OPT_NUMBER
,
587 .help
= "number of sectors (ide disk geometry)",
590 .type
= QEMU_OPT_STRING
,
591 .help
= "chs translation (auto, lba, none)",
594 .type
= QEMU_OPT_BOOL
,
595 .help
= "(deprecated, ignored)",
598 .type
= QEMU_OPT_STRING
,
599 .help
= "pci address (virtio only)",
602 .type
= QEMU_OPT_STRING
,
606 /* Options that are passed on, but have special semantics with -drive */
609 .type
= QEMU_OPT_BOOL
,
610 .help
= "open drive file as read-only",
612 .name
= "copy-on-read",
613 .type
= QEMU_OPT_BOOL
,
614 .help
= "copy read data from backing file into image file",
617 { /* end of list */ }
621 DriveInfo
*drive_init(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
)
624 DriveInfo
*dinfo
= NULL
;
626 QemuOpts
*legacy_opts
;
627 DriveMediaType media
= MEDIA_DISK
;
628 BlockInterfaceType type
;
629 int cyls
, heads
, secs
, translation
;
630 int max_devs
, bus_id
, unit_id
, index
;
632 bool read_only
= false;
634 const char *filename
;
635 Error
*local_err
= NULL
;
637 /* Change legacy command line options into QMP ones */
638 qemu_opt_rename(all_opts
, "iops", "throttling.iops-total");
639 qemu_opt_rename(all_opts
, "iops_rd", "throttling.iops-read");
640 qemu_opt_rename(all_opts
, "iops_wr", "throttling.iops-write");
642 qemu_opt_rename(all_opts
, "bps", "throttling.bps-total");
643 qemu_opt_rename(all_opts
, "bps_rd", "throttling.bps-read");
644 qemu_opt_rename(all_opts
, "bps_wr", "throttling.bps-write");
646 qemu_opt_rename(all_opts
, "iops_max", "throttling.iops-total-max");
647 qemu_opt_rename(all_opts
, "iops_rd_max", "throttling.iops-read-max");
648 qemu_opt_rename(all_opts
, "iops_wr_max", "throttling.iops-write-max");
650 qemu_opt_rename(all_opts
, "bps_max", "throttling.bps-total-max");
651 qemu_opt_rename(all_opts
, "bps_rd_max", "throttling.bps-read-max");
652 qemu_opt_rename(all_opts
, "bps_wr_max", "throttling.bps-write-max");
654 qemu_opt_rename(all_opts
,
655 "iops_size", "throttling.iops-size");
657 qemu_opt_rename(all_opts
, "readonly", "read-only");
659 value
= qemu_opt_get(all_opts
, "cache");
663 if (bdrv_parse_cache_flags(value
, &flags
) != 0) {
664 error_report("invalid cache option");
668 /* Specific options take precedence */
669 if (!qemu_opt_get(all_opts
, "cache.writeback")) {
670 qemu_opt_set_bool(all_opts
, "cache.writeback",
671 !!(flags
& BDRV_O_CACHE_WB
));
673 if (!qemu_opt_get(all_opts
, "cache.direct")) {
674 qemu_opt_set_bool(all_opts
, "cache.direct",
675 !!(flags
& BDRV_O_NOCACHE
));
677 if (!qemu_opt_get(all_opts
, "cache.no-flush")) {
678 qemu_opt_set_bool(all_opts
, "cache.no-flush",
679 !!(flags
& BDRV_O_NO_FLUSH
));
681 qemu_opt_unset(all_opts
, "cache");
684 /* Get a QDict for processing the options */
685 bs_opts
= qdict_new();
686 qemu_opts_to_qdict(all_opts
, bs_opts
);
688 legacy_opts
= qemu_opts_create(&qemu_legacy_drive_opts
, NULL
, 0,
690 qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, &local_err
);
691 if (error_is_set(&local_err
)) {
692 qerror_report_err(local_err
);
693 error_free(local_err
);
697 /* Deprecated option boot=[on|off] */
698 if (qemu_opt_get(legacy_opts
, "boot") != NULL
) {
699 fprintf(stderr
, "qemu-kvm: boot=on|off is deprecated and will be "
700 "ignored. Future versions will reject this parameter. Please "
701 "update your scripts.\n");
705 value
= qemu_opt_get(legacy_opts
, "media");
707 if (!strcmp(value
, "disk")) {
709 } else if (!strcmp(value
, "cdrom")) {
713 error_report("'%s' invalid media", value
);
718 /* copy-on-read is disabled with a warning for read-only devices */
719 read_only
|= qemu_opt_get_bool(legacy_opts
, "read-only", false);
720 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
722 if (read_only
&& copy_on_read
) {
723 error_report("warning: disabling copy-on-read on read-only drive");
724 copy_on_read
= false;
727 qdict_put(bs_opts
, "read-only",
728 qstring_from_str(read_only
? "on" : "off"));
729 qdict_put(bs_opts
, "copy-on-read",
730 qstring_from_str(copy_on_read
? "on" :"off"));
732 /* Controller type */
733 value
= qemu_opt_get(legacy_opts
, "if");
736 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
739 if (type
== IF_COUNT
) {
740 error_report("unsupported bus type '%s'", value
);
744 type
= block_default_type
;
748 cyls
= qemu_opt_get_number(legacy_opts
, "cyls", 0);
749 heads
= qemu_opt_get_number(legacy_opts
, "heads", 0);
750 secs
= qemu_opt_get_number(legacy_opts
, "secs", 0);
752 if (cyls
|| heads
|| secs
) {
754 error_report("invalid physical cyls number");
758 error_report("invalid physical heads number");
762 error_report("invalid physical secs number");
767 translation
= BIOS_ATA_TRANSLATION_AUTO
;
768 value
= qemu_opt_get(legacy_opts
, "trans");
771 error_report("'%s' trans must be used with cyls, heads and secs",
775 if (!strcmp(value
, "none")) {
776 translation
= BIOS_ATA_TRANSLATION_NONE
;
777 } else if (!strcmp(value
, "lba")) {
778 translation
= BIOS_ATA_TRANSLATION_LBA
;
779 } else if (!strcmp(value
, "auto")) {
780 translation
= BIOS_ATA_TRANSLATION_AUTO
;
782 error_report("'%s' invalid translation type", value
);
787 if (media
== MEDIA_CDROM
) {
788 if (cyls
|| secs
|| heads
) {
789 error_report("CHS can't be set with media=cdrom");
794 /* Device address specified by bus/unit or index.
795 * If none was specified, try to find the first free one. */
796 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
797 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
798 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
800 max_devs
= if_max_devs
[type
];
803 if (bus_id
!= 0 || unit_id
!= -1) {
804 error_report("index cannot be used with bus and unit");
807 bus_id
= drive_index_to_bus_id(type
, index
);
808 unit_id
= drive_index_to_unit_id(type
, index
);
813 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
815 if (max_devs
&& unit_id
>= max_devs
) {
822 if (max_devs
&& unit_id
>= max_devs
) {
823 error_report("unit %d too big (max is %d)", unit_id
, max_devs
- 1);
827 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
828 error_report("drive with bus=%d, unit=%d (index=%d) exists",
829 bus_id
, unit_id
, index
);
833 /* no id supplied -> create one */
834 if (qemu_opts_id(all_opts
) == NULL
) {
836 const char *mediastr
= "";
837 if (type
== IF_IDE
|| type
== IF_SCSI
) {
838 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
841 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
844 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
847 qdict_put(bs_opts
, "id", qstring_from_str(new_id
));
851 /* Add virtio block device */
852 devaddr
= qemu_opt_get(legacy_opts
, "addr");
853 if (devaddr
&& type
!= IF_VIRTIO
) {
854 error_report("addr is not supported by this bus type");
858 if (type
== IF_VIRTIO
) {
860 devopts
= qemu_opts_create(qemu_find_opts("device"), NULL
, 0,
862 if (arch_type
== QEMU_ARCH_S390X
) {
863 qemu_opt_set(devopts
, "driver", "virtio-blk-s390");
865 qemu_opt_set(devopts
, "driver", "virtio-blk-pci");
867 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"));
869 qemu_opt_set(devopts
, "addr", devaddr
);
873 filename
= qemu_opt_get(legacy_opts
, "file");
875 /* Actual block device init: Functionality shared with blockdev-add */
876 dinfo
= blockdev_init(filename
, bs_opts
, type
, &local_err
);
878 if (error_is_set(&local_err
)) {
879 qerror_report_err(local_err
);
880 error_free(local_err
);
884 assert(!error_is_set(&local_err
));
887 /* Set legacy DriveInfo fields */
888 dinfo
->enable_auto_del
= true;
889 dinfo
->opts
= all_opts
;
892 dinfo
->heads
= heads
;
894 dinfo
->trans
= translation
;
897 dinfo
->unit
= unit_id
;
898 dinfo
->devaddr
= devaddr
;
905 dinfo
->media_cd
= media
== MEDIA_CDROM
;
912 qemu_opts_del(legacy_opts
);
916 void do_commit(Monitor
*mon
, const QDict
*qdict
)
918 const char *device
= qdict_get_str(qdict
, "device");
919 BlockDriverState
*bs
;
922 if (!strcmp(device
, "all")) {
923 ret
= bdrv_commit_all();
925 bs
= bdrv_find(device
);
927 monitor_printf(mon
, "Device '%s' not found\n", device
);
930 ret
= bdrv_commit(bs
);
933 monitor_printf(mon
, "'commit' error for '%s': %s\n", device
,
938 static void blockdev_do_action(int kind
, void *data
, Error
**errp
)
940 TransactionAction action
;
941 TransactionActionList list
;
945 list
.value
= &action
;
947 qmp_transaction(&list
, errp
);
950 void qmp_blockdev_snapshot_sync(const char *device
, const char *snapshot_file
,
951 bool has_format
, const char *format
,
952 bool has_mode
, enum NewImageMode mode
,
955 BlockdevSnapshot snapshot
= {
956 .device
= (char *) device
,
957 .snapshot_file
= (char *) snapshot_file
,
958 .has_format
= has_format
,
959 .format
= (char *) format
,
960 .has_mode
= has_mode
,
963 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
967 void qmp_blockdev_snapshot_internal_sync(const char *device
,
971 BlockdevSnapshotInternal snapshot
= {
972 .device
= (char *) device
,
973 .name
= (char *) name
976 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
980 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
987 BlockDriverState
*bs
= bdrv_find(device
);
989 Error
*local_err
= NULL
;
990 SnapshotInfo
*info
= NULL
;
994 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1007 error_setg(errp
, "Name or id must be provided");
1011 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1012 if (error_is_set(&local_err
)) {
1013 error_propagate(errp
, local_err
);
1018 "Snapshot with id '%s' and name '%s' does not exist on "
1020 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1024 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1025 if (error_is_set(&local_err
)) {
1026 error_propagate(errp
, local_err
);
1030 info
= g_malloc0(sizeof(SnapshotInfo
));
1031 info
->id
= g_strdup(sn
.id_str
);
1032 info
->name
= g_strdup(sn
.name
);
1033 info
->date_nsec
= sn
.date_nsec
;
1034 info
->date_sec
= sn
.date_sec
;
1035 info
->vm_state_size
= sn
.vm_state_size
;
1036 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1037 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1042 /* New and old BlockDriverState structs for group snapshots */
1044 typedef struct BlkTransactionState BlkTransactionState
;
1046 /* Only prepare() may fail. In a single transaction, only one of commit() or
1047 abort() will be called, clean() will always be called if it present. */
1048 typedef struct BdrvActionOps
{
1049 /* Size of state struct, in bytes. */
1050 size_t instance_size
;
1051 /* Prepare the work, must NOT be NULL. */
1052 void (*prepare
)(BlkTransactionState
*common
, Error
**errp
);
1053 /* Commit the changes, can be NULL. */
1054 void (*commit
)(BlkTransactionState
*common
);
1055 /* Abort the changes on fail, can be NULL. */
1056 void (*abort
)(BlkTransactionState
*common
);
1057 /* Clean up resource in the end, can be NULL. */
1058 void (*clean
)(BlkTransactionState
*common
);
1062 * This structure must be arranged as first member in child type, assuming
1063 * that compiler will also arrange it to the same address with parent instance.
1064 * Later it will be used in free().
1066 struct BlkTransactionState
{
1067 TransactionAction
*action
;
1068 const BdrvActionOps
*ops
;
1069 QSIMPLEQ_ENTRY(BlkTransactionState
) entry
;
1072 /* internal snapshot private data */
1073 typedef struct InternalSnapshotState
{
1074 BlkTransactionState common
;
1075 BlockDriverState
*bs
;
1076 QEMUSnapshotInfo sn
;
1077 } InternalSnapshotState
;
1079 static void internal_snapshot_prepare(BlkTransactionState
*common
,
1084 BlockDriverState
*bs
;
1085 QEMUSnapshotInfo old_sn
, *sn
;
1088 BlockdevSnapshotInternal
*internal
;
1089 InternalSnapshotState
*state
;
1092 g_assert(common
->action
->kind
==
1093 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1094 internal
= common
->action
->blockdev_snapshot_internal_sync
;
1095 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1097 /* 1. parse input */
1098 device
= internal
->device
;
1099 name
= internal
->name
;
1101 /* 2. check for validation */
1102 bs
= bdrv_find(device
);
1104 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1108 if (!bdrv_is_inserted(bs
)) {
1109 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1113 if (bdrv_is_read_only(bs
)) {
1114 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1118 if (!bdrv_can_snapshot(bs
)) {
1119 error_set(errp
, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
1120 bs
->drv
->format_name
, device
, "internal snapshot");
1124 if (!strlen(name
)) {
1125 error_setg(errp
, "Name is empty");
1129 /* check whether a snapshot with name exist */
1130 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
, errp
);
1131 if (error_is_set(errp
)) {
1135 "Snapshot with name '%s' already exists on device '%s'",
1140 /* 3. take the snapshot */
1142 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1143 qemu_gettimeofday(&tv
);
1144 sn
->date_sec
= tv
.tv_sec
;
1145 sn
->date_nsec
= tv
.tv_usec
* 1000;
1146 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1148 ret1
= bdrv_snapshot_create(bs
, sn
);
1150 error_setg_errno(errp
, -ret1
,
1151 "Failed to create snapshot '%s' on device '%s'",
1156 /* 4. succeed, mark a snapshot is created */
1160 static void internal_snapshot_abort(BlkTransactionState
*common
)
1162 InternalSnapshotState
*state
=
1163 DO_UPCAST(InternalSnapshotState
, common
, common
);
1164 BlockDriverState
*bs
= state
->bs
;
1165 QEMUSnapshotInfo
*sn
= &state
->sn
;
1166 Error
*local_error
= NULL
;
1172 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1173 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1174 "device '%s' in abort: %s",
1177 bdrv_get_device_name(bs
),
1178 error_get_pretty(local_error
));
1179 error_free(local_error
);
1183 /* external snapshot private data */
1184 typedef struct ExternalSnapshotState
{
1185 BlkTransactionState common
;
1186 BlockDriverState
*old_bs
;
1187 BlockDriverState
*new_bs
;
1188 } ExternalSnapshotState
;
1190 static void external_snapshot_prepare(BlkTransactionState
*common
,
1195 Error
*local_err
= NULL
;
1197 const char *new_image_file
;
1198 const char *format
= "qcow2";
1199 enum NewImageMode mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1200 ExternalSnapshotState
*state
=
1201 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1202 TransactionAction
*action
= common
->action
;
1204 /* get parameters */
1205 g_assert(action
->kind
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
);
1207 device
= action
->blockdev_snapshot_sync
->device
;
1208 new_image_file
= action
->blockdev_snapshot_sync
->snapshot_file
;
1209 if (action
->blockdev_snapshot_sync
->has_format
) {
1210 format
= action
->blockdev_snapshot_sync
->format
;
1212 if (action
->blockdev_snapshot_sync
->has_mode
) {
1213 mode
= action
->blockdev_snapshot_sync
->mode
;
1216 /* start processing */
1217 drv
= bdrv_find_format(format
);
1219 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1223 state
->old_bs
= bdrv_find(device
);
1224 if (!state
->old_bs
) {
1225 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1229 if (!bdrv_is_inserted(state
->old_bs
)) {
1230 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1234 if (bdrv_in_use(state
->old_bs
)) {
1235 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1239 if (!bdrv_is_read_only(state
->old_bs
)) {
1240 if (bdrv_flush(state
->old_bs
)) {
1241 error_set(errp
, QERR_IO_ERROR
);
1246 if (bdrv_check_ext_snapshot(state
->old_bs
) != EXT_SNAPSHOT_ALLOWED
) {
1247 error_set(errp
, QERR_FEATURE_DISABLED
, "snapshot");
1251 flags
= state
->old_bs
->open_flags
;
1253 /* create new image w/backing file */
1254 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1255 bdrv_img_create(new_image_file
, format
,
1256 state
->old_bs
->filename
,
1257 state
->old_bs
->drv
->format_name
,
1258 NULL
, -1, flags
, &local_err
, false);
1259 if (error_is_set(&local_err
)) {
1260 error_propagate(errp
, local_err
);
1265 /* We will manually add the backing_hd field to the bs later */
1266 state
->new_bs
= bdrv_new("");
1267 /* TODO Inherit bs->options or only take explicit options with an
1268 * extended QMP command? */
1269 ret
= bdrv_open(state
->new_bs
, new_image_file
, NULL
,
1270 flags
| BDRV_O_NO_BACKING
, drv
, &local_err
);
1272 error_propagate(errp
, local_err
);
1276 static void external_snapshot_commit(BlkTransactionState
*common
)
1278 ExternalSnapshotState
*state
=
1279 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1281 /* This removes our old bs and adds the new bs */
1282 bdrv_append(state
->new_bs
, state
->old_bs
);
1283 /* We don't need (or want) to use the transactional
1284 * bdrv_reopen_multiple() across all the entries at once, because we
1285 * don't want to abort all of them if one of them fails the reopen */
1286 bdrv_reopen(state
->new_bs
, state
->new_bs
->open_flags
& ~BDRV_O_RDWR
,
1290 static void external_snapshot_abort(BlkTransactionState
*common
)
1292 ExternalSnapshotState
*state
=
1293 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1294 if (state
->new_bs
) {
1295 bdrv_unref(state
->new_bs
);
1299 typedef struct DriveBackupState
{
1300 BlkTransactionState common
;
1301 BlockDriverState
*bs
;
1305 static void drive_backup_prepare(BlkTransactionState
*common
, Error
**errp
)
1307 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1308 DriveBackup
*backup
;
1309 Error
*local_err
= NULL
;
1311 assert(common
->action
->kind
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1312 backup
= common
->action
->drive_backup
;
1314 qmp_drive_backup(backup
->device
, backup
->target
,
1315 backup
->has_format
, backup
->format
,
1317 backup
->has_mode
, backup
->mode
,
1318 backup
->has_speed
, backup
->speed
,
1319 backup
->has_on_source_error
, backup
->on_source_error
,
1320 backup
->has_on_target_error
, backup
->on_target_error
,
1322 if (error_is_set(&local_err
)) {
1323 error_propagate(errp
, local_err
);
1329 state
->bs
= bdrv_find(backup
->device
);
1330 state
->job
= state
->bs
->job
;
1333 static void drive_backup_abort(BlkTransactionState
*common
)
1335 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1336 BlockDriverState
*bs
= state
->bs
;
1338 /* Only cancel if it's the job we started */
1339 if (bs
&& bs
->job
&& bs
->job
== state
->job
) {
1340 block_job_cancel_sync(bs
->job
);
1344 static void abort_prepare(BlkTransactionState
*common
, Error
**errp
)
1346 error_setg(errp
, "Transaction aborted using Abort action");
1349 static void abort_commit(BlkTransactionState
*common
)
1351 g_assert_not_reached(); /* this action never succeeds */
1354 static const BdrvActionOps actions
[] = {
1355 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
1356 .instance_size
= sizeof(ExternalSnapshotState
),
1357 .prepare
= external_snapshot_prepare
,
1358 .commit
= external_snapshot_commit
,
1359 .abort
= external_snapshot_abort
,
1361 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
1362 .instance_size
= sizeof(DriveBackupState
),
1363 .prepare
= drive_backup_prepare
,
1364 .abort
= drive_backup_abort
,
1366 [TRANSACTION_ACTION_KIND_ABORT
] = {
1367 .instance_size
= sizeof(BlkTransactionState
),
1368 .prepare
= abort_prepare
,
1369 .commit
= abort_commit
,
1371 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
1372 .instance_size
= sizeof(InternalSnapshotState
),
1373 .prepare
= internal_snapshot_prepare
,
1374 .abort
= internal_snapshot_abort
,
1379 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1380 * then we do not pivot any of the devices in the group, and abandon the
1383 void qmp_transaction(TransactionActionList
*dev_list
, Error
**errp
)
1385 TransactionActionList
*dev_entry
= dev_list
;
1386 BlkTransactionState
*state
, *next
;
1387 Error
*local_err
= NULL
;
1389 QSIMPLEQ_HEAD(snap_bdrv_states
, BlkTransactionState
) snap_bdrv_states
;
1390 QSIMPLEQ_INIT(&snap_bdrv_states
);
1392 /* drain all i/o before any snapshots */
1395 /* We don't do anything in this loop that commits us to the snapshot */
1396 while (NULL
!= dev_entry
) {
1397 TransactionAction
*dev_info
= NULL
;
1398 const BdrvActionOps
*ops
;
1400 dev_info
= dev_entry
->value
;
1401 dev_entry
= dev_entry
->next
;
1403 assert(dev_info
->kind
< ARRAY_SIZE(actions
));
1405 ops
= &actions
[dev_info
->kind
];
1406 assert(ops
->instance_size
> 0);
1408 state
= g_malloc0(ops
->instance_size
);
1410 state
->action
= dev_info
;
1411 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
1413 state
->ops
->prepare(state
, &local_err
);
1414 if (error_is_set(&local_err
)) {
1415 error_propagate(errp
, local_err
);
1416 goto delete_and_fail
;
1420 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1421 if (state
->ops
->commit
) {
1422 state
->ops
->commit(state
);
1431 * failure, and it is all-or-none; abandon each new bs, and keep using
1432 * the original bs for all images
1434 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1435 if (state
->ops
->abort
) {
1436 state
->ops
->abort(state
);
1440 QSIMPLEQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
1441 if (state
->ops
->clean
) {
1442 state
->ops
->clean(state
);
1449 static void eject_device(BlockDriverState
*bs
, int force
, Error
**errp
)
1451 if (bdrv_in_use(bs
)) {
1452 error_set(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
1455 if (!bdrv_dev_has_removable_media(bs
)) {
1456 error_set(errp
, QERR_DEVICE_NOT_REMOVABLE
, bdrv_get_device_name(bs
));
1460 if (bdrv_dev_is_medium_locked(bs
) && !bdrv_dev_is_tray_open(bs
)) {
1461 bdrv_dev_eject_request(bs
, force
);
1463 error_set(errp
, QERR_DEVICE_LOCKED
, bdrv_get_device_name(bs
));
1471 void qmp_eject(const char *device
, bool has_force
, bool force
, Error
**errp
)
1473 BlockDriverState
*bs
;
1475 bs
= bdrv_find(device
);
1477 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1481 eject_device(bs
, force
, errp
);
1484 void qmp_block_passwd(const char *device
, const char *password
, Error
**errp
)
1486 BlockDriverState
*bs
;
1489 bs
= bdrv_find(device
);
1491 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1495 err
= bdrv_set_key(bs
, password
);
1496 if (err
== -EINVAL
) {
1497 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1499 } else if (err
< 0) {
1500 error_set(errp
, QERR_INVALID_PASSWORD
);
1505 static void qmp_bdrv_open_encrypted(BlockDriverState
*bs
, const char *filename
,
1506 int bdrv_flags
, BlockDriver
*drv
,
1507 const char *password
, Error
**errp
)
1509 Error
*local_err
= NULL
;
1512 ret
= bdrv_open(bs
, filename
, NULL
, bdrv_flags
, drv
, &local_err
);
1514 error_propagate(errp
, local_err
);
1518 if (bdrv_key_required(bs
)) {
1520 if (bdrv_set_key(bs
, password
) < 0) {
1521 error_set(errp
, QERR_INVALID_PASSWORD
);
1524 error_set(errp
, QERR_DEVICE_ENCRYPTED
, bdrv_get_device_name(bs
),
1525 bdrv_get_encrypted_filename(bs
));
1527 } else if (password
) {
1528 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1532 void qmp_change_blockdev(const char *device
, const char *filename
,
1533 const char *format
, Error
**errp
)
1535 BlockDriverState
*bs
;
1536 BlockDriver
*drv
= NULL
;
1540 bs
= bdrv_find(device
);
1542 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1547 drv
= bdrv_find_whitelisted_format(format
, bs
->read_only
);
1549 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1554 eject_device(bs
, 0, &err
);
1555 if (error_is_set(&err
)) {
1556 error_propagate(errp
, err
);
1560 bdrv_flags
= bdrv_is_read_only(bs
) ? 0 : BDRV_O_RDWR
;
1561 bdrv_flags
|= bdrv_is_snapshot(bs
) ? BDRV_O_SNAPSHOT
: 0;
1563 qmp_bdrv_open_encrypted(bs
, filename
, bdrv_flags
, drv
, NULL
, errp
);
1566 /* throttling disk I/O limits */
1567 void qmp_block_set_io_throttle(const char *device
, int64_t bps
, int64_t bps_rd
,
1574 bool has_bps_rd_max
,
1576 bool has_bps_wr_max
,
1580 bool has_iops_rd_max
,
1581 int64_t iops_rd_max
,
1582 bool has_iops_wr_max
,
1583 int64_t iops_wr_max
,
1585 int64_t iops_size
, Error
**errp
)
1588 BlockDriverState
*bs
;
1590 bs
= bdrv_find(device
);
1592 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1596 memset(&cfg
, 0, sizeof(cfg
));
1597 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
= bps
;
1598 cfg
.buckets
[THROTTLE_BPS_READ
].avg
= bps_rd
;
1599 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
= bps_wr
;
1601 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
= iops
;
1602 cfg
.buckets
[THROTTLE_OPS_READ
].avg
= iops_rd
;
1603 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
= iops_wr
;
1606 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
= bps_max
;
1608 if (has_bps_rd_max
) {
1609 cfg
.buckets
[THROTTLE_BPS_READ
].max
= bps_rd_max
;
1611 if (has_bps_wr_max
) {
1612 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
= bps_wr_max
;
1615 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
= iops_max
;
1617 if (has_iops_rd_max
) {
1618 cfg
.buckets
[THROTTLE_OPS_READ
].max
= iops_rd_max
;
1620 if (has_iops_wr_max
) {
1621 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
= iops_wr_max
;
1624 if (has_iops_size
) {
1625 cfg
.op_size
= iops_size
;
1628 if (!check_throttle_config(&cfg
, errp
)) {
1632 if (!bs
->io_limits_enabled
&& throttle_enabled(&cfg
)) {
1633 bdrv_io_limits_enable(bs
);
1634 } else if (bs
->io_limits_enabled
&& !throttle_enabled(&cfg
)) {
1635 bdrv_io_limits_disable(bs
);
1638 if (bs
->io_limits_enabled
) {
1639 bdrv_set_io_limits(bs
, &cfg
);
1643 int do_drive_del(Monitor
*mon
, const QDict
*qdict
, QObject
**ret_data
)
1645 const char *id
= qdict_get_str(qdict
, "id");
1646 BlockDriverState
*bs
;
1650 qerror_report(QERR_DEVICE_NOT_FOUND
, id
);
1653 if (bdrv_in_use(bs
)) {
1654 qerror_report(QERR_DEVICE_IN_USE
, id
);
1658 /* quiesce block driver; prevent further io */
1663 /* if we have a device attached to this BlockDriverState
1664 * then we need to make the drive anonymous until the device
1665 * can be removed. If this is a drive with no device backing
1666 * then we can just get rid of the block driver state right here.
1668 if (bdrv_get_attached_dev(bs
)) {
1671 /* Further I/O must not pause the guest */
1672 bdrv_set_on_error(bs
, BLOCKDEV_ON_ERROR_REPORT
,
1673 BLOCKDEV_ON_ERROR_REPORT
);
1675 drive_uninit(drive_get_by_blockdev(bs
));
1681 void qmp_block_resize(const char *device
, int64_t size
, Error
**errp
)
1683 BlockDriverState
*bs
;
1686 bs
= bdrv_find(device
);
1688 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1693 error_set(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
1697 /* complete all in-flight operations before resizing the device */
1700 ret
= bdrv_truncate(bs
, size
);
1705 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1708 error_set(errp
, QERR_UNSUPPORTED
);
1711 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1714 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1717 error_setg_errno(errp
, -ret
, "Could not resize");
1722 static void block_job_cb(void *opaque
, int ret
)
1724 BlockDriverState
*bs
= opaque
;
1727 trace_block_job_cb(bs
, bs
->job
, ret
);
1730 obj
= qobject_from_block_job(bs
->job
);
1732 QDict
*dict
= qobject_to_qdict(obj
);
1733 qdict_put(dict
, "error", qstring_from_str(strerror(-ret
)));
1736 if (block_job_is_cancelled(bs
->job
)) {
1737 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED
, obj
);
1739 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED
, obj
);
1741 qobject_decref(obj
);
1743 bdrv_put_ref_bh_schedule(bs
);
1746 void qmp_block_stream(const char *device
, bool has_base
,
1747 const char *base
, bool has_speed
, int64_t speed
,
1748 bool has_on_error
, BlockdevOnError on_error
,
1751 BlockDriverState
*bs
;
1752 BlockDriverState
*base_bs
= NULL
;
1753 Error
*local_err
= NULL
;
1755 if (!has_on_error
) {
1756 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1759 bs
= bdrv_find(device
);
1761 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1766 base_bs
= bdrv_find_backing_image(bs
, base
);
1767 if (base_bs
== NULL
) {
1768 error_set(errp
, QERR_BASE_NOT_FOUND
, base
);
1773 stream_start(bs
, base_bs
, base
, has_speed
? speed
: 0,
1774 on_error
, block_job_cb
, bs
, &local_err
);
1775 if (error_is_set(&local_err
)) {
1776 error_propagate(errp
, local_err
);
1780 trace_qmp_block_stream(bs
, bs
->job
);
1783 void qmp_block_commit(const char *device
,
1784 bool has_base
, const char *base
, const char *top
,
1785 bool has_speed
, int64_t speed
,
1788 BlockDriverState
*bs
;
1789 BlockDriverState
*base_bs
, *top_bs
;
1790 Error
*local_err
= NULL
;
1791 /* This will be part of the QMP command, if/when the
1792 * BlockdevOnError change for blkmirror makes it in
1794 BlockdevOnError on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1796 /* drain all i/o before commits */
1799 bs
= bdrv_find(device
);
1801 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1805 /* default top_bs is the active layer */
1809 if (strcmp(bs
->filename
, top
) != 0) {
1810 top_bs
= bdrv_find_backing_image(bs
, top
);
1814 if (top_bs
== NULL
) {
1815 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
1819 if (has_base
&& base
) {
1820 base_bs
= bdrv_find_backing_image(top_bs
, base
);
1822 base_bs
= bdrv_find_base(top_bs
);
1825 if (base_bs
== NULL
) {
1826 error_set(errp
, QERR_BASE_NOT_FOUND
, base
? base
: "NULL");
1831 commit_active_start(bs
, base_bs
, speed
, on_error
, block_job_cb
,
1834 commit_start(bs
, base_bs
, top_bs
, speed
, on_error
, block_job_cb
, bs
,
1837 if (local_err
!= NULL
) {
1838 error_propagate(errp
, local_err
);
1843 void qmp_drive_backup(const char *device
, const char *target
,
1844 bool has_format
, const char *format
,
1845 enum MirrorSyncMode sync
,
1846 bool has_mode
, enum NewImageMode mode
,
1847 bool has_speed
, int64_t speed
,
1848 bool has_on_source_error
, BlockdevOnError on_source_error
,
1849 bool has_on_target_error
, BlockdevOnError on_target_error
,
1852 BlockDriverState
*bs
;
1853 BlockDriverState
*target_bs
;
1854 BlockDriverState
*source
= NULL
;
1855 BlockDriver
*drv
= NULL
;
1856 Error
*local_err
= NULL
;
1864 if (!has_on_source_error
) {
1865 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1867 if (!has_on_target_error
) {
1868 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1871 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1874 bs
= bdrv_find(device
);
1876 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1880 if (!bdrv_is_inserted(bs
)) {
1881 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1886 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1889 drv
= bdrv_find_format(format
);
1891 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1896 if (bdrv_in_use(bs
)) {
1897 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1901 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1903 /* See if we have a backing HD we can use to create our new image
1905 if (sync
== MIRROR_SYNC_MODE_TOP
) {
1906 source
= bs
->backing_hd
;
1908 sync
= MIRROR_SYNC_MODE_FULL
;
1911 if (sync
== MIRROR_SYNC_MODE_NONE
) {
1915 size
= bdrv_getlength(bs
);
1917 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1921 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1922 assert(format
&& drv
);
1924 bdrv_img_create(target
, format
, source
->filename
,
1925 source
->drv
->format_name
, NULL
,
1926 size
, flags
, &local_err
, false);
1928 bdrv_img_create(target
, format
, NULL
, NULL
, NULL
,
1929 size
, flags
, &local_err
, false);
1933 if (error_is_set(&local_err
)) {
1934 error_propagate(errp
, local_err
);
1938 target_bs
= bdrv_new("");
1939 ret
= bdrv_open(target_bs
, target
, NULL
, flags
, drv
, &local_err
);
1941 bdrv_unref(target_bs
);
1942 error_propagate(errp
, local_err
);
1946 backup_start(bs
, target_bs
, speed
, sync
, on_source_error
, on_target_error
,
1947 block_job_cb
, bs
, &local_err
);
1948 if (local_err
!= NULL
) {
1949 bdrv_unref(target_bs
);
1950 error_propagate(errp
, local_err
);
1955 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1957 void qmp_drive_mirror(const char *device
, const char *target
,
1958 bool has_format
, const char *format
,
1959 enum MirrorSyncMode sync
,
1960 bool has_mode
, enum NewImageMode mode
,
1961 bool has_speed
, int64_t speed
,
1962 bool has_granularity
, uint32_t granularity
,
1963 bool has_buf_size
, int64_t buf_size
,
1964 bool has_on_source_error
, BlockdevOnError on_source_error
,
1965 bool has_on_target_error
, BlockdevOnError on_target_error
,
1968 BlockDriverState
*bs
;
1969 BlockDriverState
*source
, *target_bs
;
1970 BlockDriver
*drv
= NULL
;
1971 Error
*local_err
= NULL
;
1979 if (!has_on_source_error
) {
1980 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1982 if (!has_on_target_error
) {
1983 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1986 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1988 if (!has_granularity
) {
1991 if (!has_buf_size
) {
1992 buf_size
= DEFAULT_MIRROR_BUF_SIZE
;
1995 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
1996 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1999 if (granularity
& (granularity
- 1)) {
2000 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
2004 bs
= bdrv_find(device
);
2006 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
2010 if (!bdrv_is_inserted(bs
)) {
2011 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
2016 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
2019 drv
= bdrv_find_format(format
);
2021 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
2026 if (bdrv_in_use(bs
)) {
2027 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
2031 flags
= bs
->open_flags
| BDRV_O_RDWR
;
2032 source
= bs
->backing_hd
;
2033 if (!source
&& sync
== MIRROR_SYNC_MODE_TOP
) {
2034 sync
= MIRROR_SYNC_MODE_FULL
;
2036 if (sync
== MIRROR_SYNC_MODE_NONE
) {
2040 size
= bdrv_getlength(bs
);
2042 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
2046 if ((sync
== MIRROR_SYNC_MODE_FULL
|| !source
)
2047 && mode
!= NEW_IMAGE_MODE_EXISTING
)
2049 /* create new image w/o backing file */
2050 assert(format
&& drv
);
2051 bdrv_img_create(target
, format
,
2052 NULL
, NULL
, NULL
, size
, flags
, &local_err
, false);
2055 case NEW_IMAGE_MODE_EXISTING
:
2057 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
2058 /* create new image with backing file */
2059 bdrv_img_create(target
, format
,
2061 source
->drv
->format_name
,
2062 NULL
, size
, flags
, &local_err
, false);
2069 if (error_is_set(&local_err
)) {
2070 error_propagate(errp
, local_err
);
2074 /* Mirroring takes care of copy-on-write using the source's backing
2077 target_bs
= bdrv_new("");
2078 ret
= bdrv_open(target_bs
, target
, NULL
, flags
| BDRV_O_NO_BACKING
, drv
,
2081 bdrv_unref(target_bs
);
2082 error_propagate(errp
, local_err
);
2086 mirror_start(bs
, target_bs
, speed
, granularity
, buf_size
, sync
,
2087 on_source_error
, on_target_error
,
2088 block_job_cb
, bs
, &local_err
);
2089 if (local_err
!= NULL
) {
2090 bdrv_unref(target_bs
);
2091 error_propagate(errp
, local_err
);
2096 static BlockJob
*find_block_job(const char *device
)
2098 BlockDriverState
*bs
;
2100 bs
= bdrv_find(device
);
2101 if (!bs
|| !bs
->job
) {
2107 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
2109 BlockJob
*job
= find_block_job(device
);
2112 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2116 block_job_set_speed(job
, speed
, errp
);
2119 void qmp_block_job_cancel(const char *device
,
2120 bool has_force
, bool force
, Error
**errp
)
2122 BlockJob
*job
= find_block_job(device
);
2129 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2132 if (job
->paused
&& !force
) {
2133 error_set(errp
, QERR_BLOCK_JOB_PAUSED
, device
);
2137 trace_qmp_block_job_cancel(job
);
2138 block_job_cancel(job
);
2141 void qmp_block_job_pause(const char *device
, Error
**errp
)
2143 BlockJob
*job
= find_block_job(device
);
2146 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2150 trace_qmp_block_job_pause(job
);
2151 block_job_pause(job
);
2154 void qmp_block_job_resume(const char *device
, Error
**errp
)
2156 BlockJob
*job
= find_block_job(device
);
2159 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2163 trace_qmp_block_job_resume(job
);
2164 block_job_resume(job
);
2167 void qmp_block_job_complete(const char *device
, Error
**errp
)
2169 BlockJob
*job
= find_block_job(device
);
2172 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2176 trace_qmp_block_job_complete(job
);
2177 block_job_complete(job
, errp
);
2180 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
2182 QmpOutputVisitor
*ov
= qmp_output_visitor_new();
2185 Error
*local_err
= NULL
;
2187 /* Require an ID in the top level */
2188 if (!options
->has_id
) {
2189 error_setg(errp
, "Block device needs an ID");
2193 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2194 * cache.direct=false instead of silently switching to aio=threads, except
2195 * if called from drive_init.
2197 * For now, simply forbidding the combination for all drivers will do. */
2198 if (options
->has_aio
&& options
->aio
== BLOCKDEV_AIO_OPTIONS_NATIVE
) {
2199 bool direct
= options
->cache
->has_direct
&& options
->cache
->direct
;
2200 if (!options
->has_cache
&& !direct
) {
2201 error_setg(errp
, "aio=native requires cache.direct=true");
2206 visit_type_BlockdevOptions(qmp_output_get_visitor(ov
),
2207 &options
, NULL
, &local_err
);
2208 if (error_is_set(&local_err
)) {
2209 error_propagate(errp
, local_err
);
2213 obj
= qmp_output_get_qobject(ov
);
2214 qdict
= qobject_to_qdict(obj
);
2216 qdict_flatten(qdict
);
2218 blockdev_init(NULL
, qdict
, IF_NONE
, &local_err
);
2219 if (error_is_set(&local_err
)) {
2220 error_propagate(errp
, local_err
);
2225 qmp_output_visitor_cleanup(ov
);
2228 static void do_qmp_query_block_jobs_one(void *opaque
, BlockDriverState
*bs
)
2230 BlockJobInfoList
**prev
= opaque
;
2231 BlockJob
*job
= bs
->job
;
2234 BlockJobInfoList
*elem
= g_new0(BlockJobInfoList
, 1);
2235 elem
->value
= block_job_query(bs
->job
);
2236 (*prev
)->next
= elem
;
2241 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
2243 /* Dummy is a fake list element for holding the head pointer */
2244 BlockJobInfoList dummy
= {};
2245 BlockJobInfoList
*prev
= &dummy
;
2246 bdrv_iterate(do_qmp_query_block_jobs_one
, &prev
);
2250 QemuOptsList qemu_common_drive_opts
= {
2252 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
2256 .type
= QEMU_OPT_BOOL
,
2257 .help
= "enable/disable snapshot mode",
2260 .type
= QEMU_OPT_STRING
,
2261 .help
= "discard operation (ignore/off, unmap/on)",
2263 .name
= "cache.writeback",
2264 .type
= QEMU_OPT_BOOL
,
2265 .help
= "enables writeback mode for any caches",
2267 .name
= "cache.direct",
2268 .type
= QEMU_OPT_BOOL
,
2269 .help
= "enables use of O_DIRECT (bypass the host page cache)",
2271 .name
= "cache.no-flush",
2272 .type
= QEMU_OPT_BOOL
,
2273 .help
= "ignore any flush requests for the device",
2276 .type
= QEMU_OPT_STRING
,
2277 .help
= "host AIO implementation (threads, native)",
2280 .type
= QEMU_OPT_STRING
,
2281 .help
= "disk format (raw, qcow2, ...)",
2284 .type
= QEMU_OPT_STRING
,
2285 .help
= "disk serial number",
2288 .type
= QEMU_OPT_STRING
,
2289 .help
= "read error action",
2292 .type
= QEMU_OPT_STRING
,
2293 .help
= "write error action",
2295 .name
= "read-only",
2296 .type
= QEMU_OPT_BOOL
,
2297 .help
= "open drive file as read-only",
2299 .name
= "throttling.iops-total",
2300 .type
= QEMU_OPT_NUMBER
,
2301 .help
= "limit total I/O operations per second",
2303 .name
= "throttling.iops-read",
2304 .type
= QEMU_OPT_NUMBER
,
2305 .help
= "limit read operations per second",
2307 .name
= "throttling.iops-write",
2308 .type
= QEMU_OPT_NUMBER
,
2309 .help
= "limit write operations per second",
2311 .name
= "throttling.bps-total",
2312 .type
= QEMU_OPT_NUMBER
,
2313 .help
= "limit total bytes per second",
2315 .name
= "throttling.bps-read",
2316 .type
= QEMU_OPT_NUMBER
,
2317 .help
= "limit read bytes per second",
2319 .name
= "throttling.bps-write",
2320 .type
= QEMU_OPT_NUMBER
,
2321 .help
= "limit write bytes per second",
2323 .name
= "throttling.iops-total-max",
2324 .type
= QEMU_OPT_NUMBER
,
2325 .help
= "I/O operations burst",
2327 .name
= "throttling.iops-read-max",
2328 .type
= QEMU_OPT_NUMBER
,
2329 .help
= "I/O operations read burst",
2331 .name
= "throttling.iops-write-max",
2332 .type
= QEMU_OPT_NUMBER
,
2333 .help
= "I/O operations write burst",
2335 .name
= "throttling.bps-total-max",
2336 .type
= QEMU_OPT_NUMBER
,
2337 .help
= "total bytes burst",
2339 .name
= "throttling.bps-read-max",
2340 .type
= QEMU_OPT_NUMBER
,
2341 .help
= "total bytes read burst",
2343 .name
= "throttling.bps-write-max",
2344 .type
= QEMU_OPT_NUMBER
,
2345 .help
= "total bytes write burst",
2347 .name
= "throttling.iops-size",
2348 .type
= QEMU_OPT_NUMBER
,
2349 .help
= "when limiting by iops max size of an I/O in bytes",
2351 .name
= "copy-on-read",
2352 .type
= QEMU_OPT_BOOL
,
2353 .help
= "copy read data from backing file into image file",
2355 { /* end of list */ }
2359 QemuOptsList qemu_drive_opts
= {
2361 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
2364 * no elements => accept any params
2365 * validation will happen later
2367 { /* end of list */ }