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
,
317 int on_read_error
, on_write_error
;
326 bool has_driver_specific_opts
;
327 BlockDriver
*drv
= NULL
;
329 /* Check common options by copying from bs_opts to opts, all other options
330 * stay in bs_opts for processing by bdrv_open(). */
331 id
= qdict_get_try_str(bs_opts
, "id");
332 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, &error
);
333 if (error_is_set(&error
)) {
334 error_propagate(errp
, error
);
338 qemu_opts_absorb_qdict(opts
, bs_opts
, &error
);
339 if (error_is_set(&error
)) {
340 error_propagate(errp
, error
);
345 qdict_del(bs_opts
, "id");
348 has_driver_specific_opts
= !!qdict_size(bs_opts
);
350 /* extract parameters */
351 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
352 ro
= qemu_opt_get_bool(opts
, "read-only", 0);
353 copy_on_read
= qemu_opt_get_bool(opts
, "copy-on-read", false);
355 serial
= qemu_opt_get(opts
, "serial");
357 if ((buf
= qemu_opt_get(opts
, "discard")) != NULL
) {
358 if (bdrv_parse_discard_flags(buf
, &bdrv_flags
) != 0) {
359 error_setg(errp
, "invalid discard option");
364 if (qemu_opt_get_bool(opts
, "cache.writeback", true)) {
365 bdrv_flags
|= BDRV_O_CACHE_WB
;
367 if (qemu_opt_get_bool(opts
, "cache.direct", false)) {
368 bdrv_flags
|= BDRV_O_NOCACHE
;
370 if (qemu_opt_get_bool(opts
, "cache.no-flush", false)) {
371 bdrv_flags
|= BDRV_O_NO_FLUSH
;
374 #ifdef CONFIG_LINUX_AIO
375 if ((buf
= qemu_opt_get(opts
, "aio")) != NULL
) {
376 if (!strcmp(buf
, "native")) {
377 bdrv_flags
|= BDRV_O_NATIVE_AIO
;
378 } else if (!strcmp(buf
, "threads")) {
379 /* this is the default */
381 error_setg(errp
, "invalid aio option");
387 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
388 if (is_help_option(buf
)) {
389 error_printf("Supported formats:");
390 bdrv_iterate_format(bdrv_format_print
, NULL
);
395 drv
= bdrv_find_format(buf
);
397 error_setg(errp
, "'%s' invalid format", buf
);
402 /* disk I/O throttling */
403 memset(&cfg
, 0, sizeof(cfg
));
404 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
=
405 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
406 cfg
.buckets
[THROTTLE_BPS_READ
].avg
=
407 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
408 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
=
409 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
410 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
=
411 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
412 cfg
.buckets
[THROTTLE_OPS_READ
].avg
=
413 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
414 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
=
415 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
417 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
=
418 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
419 cfg
.buckets
[THROTTLE_BPS_READ
].max
=
420 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
421 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
=
422 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
423 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
=
424 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
425 cfg
.buckets
[THROTTLE_OPS_READ
].max
=
426 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
427 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
=
428 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
430 cfg
.op_size
= qemu_opt_get_number(opts
, "throttling.iops-size", 0);
432 if (!check_throttle_config(&cfg
, &error
)) {
433 error_propagate(errp
, error
);
437 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
438 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
439 on_write_error
= parse_block_error_action(buf
, 0, &error
);
440 if (error_is_set(&error
)) {
441 error_propagate(errp
, error
);
446 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
447 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
448 on_read_error
= parse_block_error_action(buf
, 1, &error
);
449 if (error_is_set(&error
)) {
450 error_propagate(errp
, error
);
455 if (bdrv_find_node(qemu_opts_id(opts
))) {
456 error_setg(errp
, "device id=%s is conflicting with a node-name",
462 dinfo
= g_malloc0(sizeof(*dinfo
));
463 dinfo
->id
= g_strdup(qemu_opts_id(opts
));
464 dinfo
->bdrv
= bdrv_new(dinfo
->id
);
465 dinfo
->bdrv
->open_flags
= snapshot
? BDRV_O_SNAPSHOT
: 0;
466 dinfo
->bdrv
->read_only
= ro
;
468 if (serial
!= NULL
) {
469 dinfo
->serial
= g_strdup(serial
);
471 QTAILQ_INSERT_TAIL(&drives
, dinfo
, next
);
473 bdrv_set_on_error(dinfo
->bdrv
, on_read_error
, on_write_error
);
475 /* disk I/O throttling */
476 if (throttle_enabled(&cfg
)) {
477 bdrv_io_limits_enable(dinfo
->bdrv
);
478 bdrv_set_io_limits(dinfo
->bdrv
, &cfg
);
481 if (!file
|| !*file
) {
482 if (has_driver_specific_opts
) {
491 /* always use cache=unsafe with snapshot */
492 bdrv_flags
&= ~BDRV_O_CACHE_MASK
;
493 bdrv_flags
|= (BDRV_O_SNAPSHOT
|BDRV_O_CACHE_WB
|BDRV_O_NO_FLUSH
);
497 bdrv_flags
|= BDRV_O_COPY_ON_READ
;
500 if (runstate_check(RUN_STATE_INMIGRATE
)) {
501 bdrv_flags
|= BDRV_O_INCOMING
;
504 bdrv_flags
|= ro
? 0 : BDRV_O_RDWR
;
507 ret
= bdrv_open(dinfo
->bdrv
, file
, bs_opts
, bdrv_flags
, drv
, &error
);
510 error_setg(errp
, "could not open disk image %s: %s",
511 file
?: dinfo
->id
, error_get_pretty(error
));
516 if (bdrv_key_required(dinfo
->bdrv
))
525 bdrv_unref(dinfo
->bdrv
);
527 QTAILQ_REMOVE(&drives
, dinfo
, next
);
535 static void qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
)
539 value
= qemu_opt_get(opts
, from
);
541 qemu_opt_set(opts
, to
, value
);
542 qemu_opt_unset(opts
, from
);
546 QemuOptsList qemu_legacy_drive_opts
= {
548 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
552 .type
= QEMU_OPT_NUMBER
,
553 .help
= "bus number",
556 .type
= QEMU_OPT_NUMBER
,
557 .help
= "unit number (i.e. lun for scsi)",
560 .type
= QEMU_OPT_NUMBER
,
561 .help
= "index number",
564 .type
= QEMU_OPT_STRING
,
565 .help
= "media type (disk, cdrom)",
568 .type
= QEMU_OPT_STRING
,
569 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
572 .type
= QEMU_OPT_NUMBER
,
573 .help
= "number of cylinders (ide disk geometry)",
576 .type
= QEMU_OPT_NUMBER
,
577 .help
= "number of heads (ide disk geometry)",
580 .type
= QEMU_OPT_NUMBER
,
581 .help
= "number of sectors (ide disk geometry)",
584 .type
= QEMU_OPT_STRING
,
585 .help
= "chs translation (auto, lba, none)",
588 .type
= QEMU_OPT_BOOL
,
589 .help
= "(deprecated, ignored)",
592 .type
= QEMU_OPT_STRING
,
593 .help
= "pci address (virtio only)",
596 .type
= QEMU_OPT_STRING
,
600 /* Options that are passed on, but have special semantics with -drive */
603 .type
= QEMU_OPT_BOOL
,
604 .help
= "open drive file as read-only",
607 .type
= QEMU_OPT_STRING
,
608 .help
= "read error action",
611 .type
= QEMU_OPT_STRING
,
612 .help
= "write error action",
614 .name
= "copy-on-read",
615 .type
= QEMU_OPT_BOOL
,
616 .help
= "copy read data from backing file into image file",
619 { /* end of list */ }
623 DriveInfo
*drive_init(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
)
626 DriveInfo
*dinfo
= NULL
;
628 QemuOpts
*legacy_opts
;
629 DriveMediaType media
= MEDIA_DISK
;
630 BlockInterfaceType type
;
631 int cyls
, heads
, secs
, translation
;
632 int max_devs
, bus_id
, unit_id
, index
;
634 const char *werror
, *rerror
;
635 bool read_only
= false;
637 const char *filename
;
638 Error
*local_err
= NULL
;
640 /* Change legacy command line options into QMP ones */
641 qemu_opt_rename(all_opts
, "iops", "throttling.iops-total");
642 qemu_opt_rename(all_opts
, "iops_rd", "throttling.iops-read");
643 qemu_opt_rename(all_opts
, "iops_wr", "throttling.iops-write");
645 qemu_opt_rename(all_opts
, "bps", "throttling.bps-total");
646 qemu_opt_rename(all_opts
, "bps_rd", "throttling.bps-read");
647 qemu_opt_rename(all_opts
, "bps_wr", "throttling.bps-write");
649 qemu_opt_rename(all_opts
, "iops_max", "throttling.iops-total-max");
650 qemu_opt_rename(all_opts
, "iops_rd_max", "throttling.iops-read-max");
651 qemu_opt_rename(all_opts
, "iops_wr_max", "throttling.iops-write-max");
653 qemu_opt_rename(all_opts
, "bps_max", "throttling.bps-total-max");
654 qemu_opt_rename(all_opts
, "bps_rd_max", "throttling.bps-read-max");
655 qemu_opt_rename(all_opts
, "bps_wr_max", "throttling.bps-write-max");
657 qemu_opt_rename(all_opts
,
658 "iops_size", "throttling.iops-size");
660 qemu_opt_rename(all_opts
, "readonly", "read-only");
662 value
= qemu_opt_get(all_opts
, "cache");
666 if (bdrv_parse_cache_flags(value
, &flags
) != 0) {
667 error_report("invalid cache option");
671 /* Specific options take precedence */
672 if (!qemu_opt_get(all_opts
, "cache.writeback")) {
673 qemu_opt_set_bool(all_opts
, "cache.writeback",
674 !!(flags
& BDRV_O_CACHE_WB
));
676 if (!qemu_opt_get(all_opts
, "cache.direct")) {
677 qemu_opt_set_bool(all_opts
, "cache.direct",
678 !!(flags
& BDRV_O_NOCACHE
));
680 if (!qemu_opt_get(all_opts
, "cache.no-flush")) {
681 qemu_opt_set_bool(all_opts
, "cache.no-flush",
682 !!(flags
& BDRV_O_NO_FLUSH
));
684 qemu_opt_unset(all_opts
, "cache");
687 /* Get a QDict for processing the options */
688 bs_opts
= qdict_new();
689 qemu_opts_to_qdict(all_opts
, bs_opts
);
691 legacy_opts
= qemu_opts_create(&qemu_legacy_drive_opts
, NULL
, 0,
693 qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, &local_err
);
694 if (error_is_set(&local_err
)) {
695 qerror_report_err(local_err
);
696 error_free(local_err
);
700 /* Deprecated option boot=[on|off] */
701 if (qemu_opt_get(legacy_opts
, "boot") != NULL
) {
702 fprintf(stderr
, "qemu-kvm: boot=on|off is deprecated and will be "
703 "ignored. Future versions will reject this parameter. Please "
704 "update your scripts.\n");
708 value
= qemu_opt_get(legacy_opts
, "media");
710 if (!strcmp(value
, "disk")) {
712 } else if (!strcmp(value
, "cdrom")) {
716 error_report("'%s' invalid media", value
);
721 /* copy-on-read is disabled with a warning for read-only devices */
722 read_only
|= qemu_opt_get_bool(legacy_opts
, "read-only", false);
723 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
725 if (read_only
&& copy_on_read
) {
726 error_report("warning: disabling copy-on-read on read-only drive");
727 copy_on_read
= false;
730 qdict_put(bs_opts
, "read-only",
731 qstring_from_str(read_only
? "on" : "off"));
732 qdict_put(bs_opts
, "copy-on-read",
733 qstring_from_str(copy_on_read
? "on" :"off"));
735 /* Controller type */
736 value
= qemu_opt_get(legacy_opts
, "if");
739 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
742 if (type
== IF_COUNT
) {
743 error_report("unsupported bus type '%s'", value
);
747 type
= block_default_type
;
751 cyls
= qemu_opt_get_number(legacy_opts
, "cyls", 0);
752 heads
= qemu_opt_get_number(legacy_opts
, "heads", 0);
753 secs
= qemu_opt_get_number(legacy_opts
, "secs", 0);
755 if (cyls
|| heads
|| secs
) {
757 error_report("invalid physical cyls number");
761 error_report("invalid physical heads number");
765 error_report("invalid physical secs number");
770 translation
= BIOS_ATA_TRANSLATION_AUTO
;
771 value
= qemu_opt_get(legacy_opts
, "trans");
774 error_report("'%s' trans must be used with cyls, heads and secs",
778 if (!strcmp(value
, "none")) {
779 translation
= BIOS_ATA_TRANSLATION_NONE
;
780 } else if (!strcmp(value
, "lba")) {
781 translation
= BIOS_ATA_TRANSLATION_LBA
;
782 } else if (!strcmp(value
, "auto")) {
783 translation
= BIOS_ATA_TRANSLATION_AUTO
;
785 error_report("'%s' invalid translation type", value
);
790 if (media
== MEDIA_CDROM
) {
791 if (cyls
|| secs
|| heads
) {
792 error_report("CHS can't be set with media=cdrom");
797 /* Device address specified by bus/unit or index.
798 * If none was specified, try to find the first free one. */
799 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
800 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
801 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
803 max_devs
= if_max_devs
[type
];
806 if (bus_id
!= 0 || unit_id
!= -1) {
807 error_report("index cannot be used with bus and unit");
810 bus_id
= drive_index_to_bus_id(type
, index
);
811 unit_id
= drive_index_to_unit_id(type
, index
);
816 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
818 if (max_devs
&& unit_id
>= max_devs
) {
825 if (max_devs
&& unit_id
>= max_devs
) {
826 error_report("unit %d too big (max is %d)", unit_id
, max_devs
- 1);
830 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
831 error_report("drive with bus=%d, unit=%d (index=%d) exists",
832 bus_id
, unit_id
, index
);
836 /* no id supplied -> create one */
837 if (qemu_opts_id(all_opts
) == NULL
) {
839 const char *mediastr
= "";
840 if (type
== IF_IDE
|| type
== IF_SCSI
) {
841 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
844 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
847 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
850 qdict_put(bs_opts
, "id", qstring_from_str(new_id
));
854 /* Add virtio block device */
855 devaddr
= qemu_opt_get(legacy_opts
, "addr");
856 if (devaddr
&& type
!= IF_VIRTIO
) {
857 error_report("addr is not supported by this bus type");
861 if (type
== IF_VIRTIO
) {
863 devopts
= qemu_opts_create(qemu_find_opts("device"), NULL
, 0,
865 if (arch_type
== QEMU_ARCH_S390X
) {
866 qemu_opt_set(devopts
, "driver", "virtio-blk-s390");
868 qemu_opt_set(devopts
, "driver", "virtio-blk-pci");
870 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"));
872 qemu_opt_set(devopts
, "addr", devaddr
);
876 filename
= qemu_opt_get(legacy_opts
, "file");
878 /* Check werror/rerror compatibility with if=... */
879 werror
= qemu_opt_get(legacy_opts
, "werror");
880 if (werror
!= NULL
) {
881 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&&
883 error_report("werror is not supported by this bus type");
886 qdict_put(bs_opts
, "werror", qstring_from_str(werror
));
889 rerror
= qemu_opt_get(legacy_opts
, "rerror");
890 if (rerror
!= NULL
) {
891 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&&
893 error_report("rerror is not supported by this bus type");
896 qdict_put(bs_opts
, "rerror", qstring_from_str(rerror
));
899 /* Actual block device init: Functionality shared with blockdev-add */
900 dinfo
= blockdev_init(filename
, bs_opts
, &local_err
);
902 if (error_is_set(&local_err
)) {
903 qerror_report_err(local_err
);
904 error_free(local_err
);
908 assert(!error_is_set(&local_err
));
911 /* Set legacy DriveInfo fields */
912 dinfo
->enable_auto_del
= true;
913 dinfo
->opts
= all_opts
;
916 dinfo
->heads
= heads
;
918 dinfo
->trans
= translation
;
922 dinfo
->unit
= unit_id
;
923 dinfo
->devaddr
= devaddr
;
930 dinfo
->media_cd
= media
== MEDIA_CDROM
;
937 qemu_opts_del(legacy_opts
);
941 void do_commit(Monitor
*mon
, const QDict
*qdict
)
943 const char *device
= qdict_get_str(qdict
, "device");
944 BlockDriverState
*bs
;
947 if (!strcmp(device
, "all")) {
948 ret
= bdrv_commit_all();
950 bs
= bdrv_find(device
);
952 monitor_printf(mon
, "Device '%s' not found\n", device
);
955 ret
= bdrv_commit(bs
);
958 monitor_printf(mon
, "'commit' error for '%s': %s\n", device
,
963 static void blockdev_do_action(int kind
, void *data
, Error
**errp
)
965 TransactionAction action
;
966 TransactionActionList list
;
970 list
.value
= &action
;
972 qmp_transaction(&list
, errp
);
975 void qmp_blockdev_snapshot_sync(bool has_device
, const char *device
,
976 bool has_node_name
, const char *node_name
,
977 const char *snapshot_file
,
978 bool has_snapshot_node_name
,
979 const char *snapshot_node_name
,
980 bool has_format
, const char *format
,
981 bool has_mode
, NewImageMode mode
, Error
**errp
)
983 BlockdevSnapshot snapshot
= {
984 .has_device
= has_device
,
985 .device
= (char *) device
,
986 .has_node_name
= has_node_name
,
987 .node_name
= (char *) node_name
,
988 .snapshot_file
= (char *) snapshot_file
,
989 .has_snapshot_node_name
= has_snapshot_node_name
,
990 .snapshot_node_name
= (char *) snapshot_node_name
,
991 .has_format
= has_format
,
992 .format
= (char *) format
,
993 .has_mode
= has_mode
,
996 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
1000 void qmp_blockdev_snapshot_internal_sync(const char *device
,
1004 BlockdevSnapshotInternal snapshot
= {
1005 .device
= (char *) device
,
1006 .name
= (char *) name
1009 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
1013 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
1020 BlockDriverState
*bs
= bdrv_find(device
);
1021 QEMUSnapshotInfo sn
;
1022 Error
*local_err
= NULL
;
1023 SnapshotInfo
*info
= NULL
;
1027 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1040 error_setg(errp
, "Name or id must be provided");
1044 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1045 if (error_is_set(&local_err
)) {
1046 error_propagate(errp
, local_err
);
1051 "Snapshot with id '%s' and name '%s' does not exist on "
1053 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1057 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1058 if (error_is_set(&local_err
)) {
1059 error_propagate(errp
, local_err
);
1063 info
= g_malloc0(sizeof(SnapshotInfo
));
1064 info
->id
= g_strdup(sn
.id_str
);
1065 info
->name
= g_strdup(sn
.name
);
1066 info
->date_nsec
= sn
.date_nsec
;
1067 info
->date_sec
= sn
.date_sec
;
1068 info
->vm_state_size
= sn
.vm_state_size
;
1069 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1070 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1075 /* New and old BlockDriverState structs for group snapshots */
1077 typedef struct BlkTransactionState BlkTransactionState
;
1079 /* Only prepare() may fail. In a single transaction, only one of commit() or
1080 abort() will be called, clean() will always be called if it present. */
1081 typedef struct BdrvActionOps
{
1082 /* Size of state struct, in bytes. */
1083 size_t instance_size
;
1084 /* Prepare the work, must NOT be NULL. */
1085 void (*prepare
)(BlkTransactionState
*common
, Error
**errp
);
1086 /* Commit the changes, can be NULL. */
1087 void (*commit
)(BlkTransactionState
*common
);
1088 /* Abort the changes on fail, can be NULL. */
1089 void (*abort
)(BlkTransactionState
*common
);
1090 /* Clean up resource in the end, can be NULL. */
1091 void (*clean
)(BlkTransactionState
*common
);
1095 * This structure must be arranged as first member in child type, assuming
1096 * that compiler will also arrange it to the same address with parent instance.
1097 * Later it will be used in free().
1099 struct BlkTransactionState
{
1100 TransactionAction
*action
;
1101 const BdrvActionOps
*ops
;
1102 QSIMPLEQ_ENTRY(BlkTransactionState
) entry
;
1105 /* internal snapshot private data */
1106 typedef struct InternalSnapshotState
{
1107 BlkTransactionState common
;
1108 BlockDriverState
*bs
;
1109 QEMUSnapshotInfo sn
;
1110 } InternalSnapshotState
;
1112 static void internal_snapshot_prepare(BlkTransactionState
*common
,
1117 BlockDriverState
*bs
;
1118 QEMUSnapshotInfo old_sn
, *sn
;
1121 BlockdevSnapshotInternal
*internal
;
1122 InternalSnapshotState
*state
;
1125 g_assert(common
->action
->kind
==
1126 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1127 internal
= common
->action
->blockdev_snapshot_internal_sync
;
1128 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1130 /* 1. parse input */
1131 device
= internal
->device
;
1132 name
= internal
->name
;
1134 /* 2. check for validation */
1135 bs
= bdrv_find(device
);
1137 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1141 if (!bdrv_is_inserted(bs
)) {
1142 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1146 if (bdrv_is_read_only(bs
)) {
1147 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1151 if (!bdrv_can_snapshot(bs
)) {
1152 error_set(errp
, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
1153 bs
->drv
->format_name
, device
, "internal snapshot");
1157 if (!strlen(name
)) {
1158 error_setg(errp
, "Name is empty");
1162 /* check whether a snapshot with name exist */
1163 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
, errp
);
1164 if (error_is_set(errp
)) {
1168 "Snapshot with name '%s' already exists on device '%s'",
1173 /* 3. take the snapshot */
1175 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1176 qemu_gettimeofday(&tv
);
1177 sn
->date_sec
= tv
.tv_sec
;
1178 sn
->date_nsec
= tv
.tv_usec
* 1000;
1179 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1181 ret1
= bdrv_snapshot_create(bs
, sn
);
1183 error_setg_errno(errp
, -ret1
,
1184 "Failed to create snapshot '%s' on device '%s'",
1189 /* 4. succeed, mark a snapshot is created */
1193 static void internal_snapshot_abort(BlkTransactionState
*common
)
1195 InternalSnapshotState
*state
=
1196 DO_UPCAST(InternalSnapshotState
, common
, common
);
1197 BlockDriverState
*bs
= state
->bs
;
1198 QEMUSnapshotInfo
*sn
= &state
->sn
;
1199 Error
*local_error
= NULL
;
1205 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1206 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1207 "device '%s' in abort: %s",
1210 bdrv_get_device_name(bs
),
1211 error_get_pretty(local_error
));
1212 error_free(local_error
);
1216 /* external snapshot private data */
1217 typedef struct ExternalSnapshotState
{
1218 BlkTransactionState common
;
1219 BlockDriverState
*old_bs
;
1220 BlockDriverState
*new_bs
;
1221 } ExternalSnapshotState
;
1223 static void external_snapshot_prepare(BlkTransactionState
*common
,
1228 QDict
*options
= NULL
;
1229 Error
*local_err
= NULL
;
1230 bool has_device
= false;
1232 bool has_node_name
= false;
1233 const char *node_name
;
1234 bool has_snapshot_node_name
= false;
1235 const char *snapshot_node_name
;
1236 const char *new_image_file
;
1237 const char *format
= "qcow2";
1238 enum NewImageMode mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1239 ExternalSnapshotState
*state
=
1240 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1241 TransactionAction
*action
= common
->action
;
1243 /* get parameters */
1244 g_assert(action
->kind
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
);
1246 has_device
= action
->blockdev_snapshot_sync
->has_device
;
1247 device
= action
->blockdev_snapshot_sync
->device
;
1248 has_node_name
= action
->blockdev_snapshot_sync
->has_node_name
;
1249 node_name
= action
->blockdev_snapshot_sync
->node_name
;
1250 has_snapshot_node_name
=
1251 action
->blockdev_snapshot_sync
->has_snapshot_node_name
;
1252 snapshot_node_name
= action
->blockdev_snapshot_sync
->snapshot_node_name
;
1254 new_image_file
= action
->blockdev_snapshot_sync
->snapshot_file
;
1255 if (action
->blockdev_snapshot_sync
->has_format
) {
1256 format
= action
->blockdev_snapshot_sync
->format
;
1258 if (action
->blockdev_snapshot_sync
->has_mode
) {
1259 mode
= action
->blockdev_snapshot_sync
->mode
;
1262 /* start processing */
1263 drv
= bdrv_find_format(format
);
1265 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1269 state
->old_bs
= bdrv_lookup_bs(has_device
? device
: NULL
,
1270 has_node_name
? node_name
: NULL
,
1272 if (error_is_set(&local_err
)) {
1273 error_propagate(errp
, local_err
);
1277 if (has_node_name
&& !has_snapshot_node_name
) {
1278 error_setg(errp
, "New snapshot node name missing");
1282 if (has_snapshot_node_name
&& bdrv_find_node(snapshot_node_name
)) {
1283 error_setg(errp
, "New snapshot node name already existing");
1287 if (!bdrv_is_inserted(state
->old_bs
)) {
1288 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1292 if (bdrv_in_use(state
->old_bs
)) {
1293 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1297 if (!bdrv_is_read_only(state
->old_bs
)) {
1298 if (bdrv_flush(state
->old_bs
)) {
1299 error_set(errp
, QERR_IO_ERROR
);
1304 if (!bdrv_is_first_non_filter(state
->old_bs
)) {
1305 error_set(errp
, QERR_FEATURE_DISABLED
, "snapshot");
1309 flags
= state
->old_bs
->open_flags
;
1311 /* create new image w/backing file */
1312 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1313 bdrv_img_create(new_image_file
, format
,
1314 state
->old_bs
->filename
,
1315 state
->old_bs
->drv
->format_name
,
1316 NULL
, -1, flags
, &local_err
, false);
1317 if (error_is_set(&local_err
)) {
1318 error_propagate(errp
, local_err
);
1323 if (has_snapshot_node_name
) {
1324 options
= qdict_new();
1325 qdict_put(options
, "node-name",
1326 qstring_from_str(snapshot_node_name
));
1329 /* We will manually add the backing_hd field to the bs later */
1330 state
->new_bs
= bdrv_new("");
1331 /* TODO Inherit bs->options or only take explicit options with an
1332 * extended QMP command? */
1333 ret
= bdrv_open(state
->new_bs
, new_image_file
, options
,
1334 flags
| BDRV_O_NO_BACKING
, drv
, &local_err
);
1336 error_propagate(errp
, local_err
);
1340 static void external_snapshot_commit(BlkTransactionState
*common
)
1342 ExternalSnapshotState
*state
=
1343 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1345 /* This removes our old bs and adds the new bs */
1346 bdrv_append(state
->new_bs
, state
->old_bs
);
1347 /* We don't need (or want) to use the transactional
1348 * bdrv_reopen_multiple() across all the entries at once, because we
1349 * don't want to abort all of them if one of them fails the reopen */
1350 bdrv_reopen(state
->new_bs
, state
->new_bs
->open_flags
& ~BDRV_O_RDWR
,
1354 static void external_snapshot_abort(BlkTransactionState
*common
)
1356 ExternalSnapshotState
*state
=
1357 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1358 if (state
->new_bs
) {
1359 bdrv_unref(state
->new_bs
);
1363 typedef struct DriveBackupState
{
1364 BlkTransactionState common
;
1365 BlockDriverState
*bs
;
1369 static void drive_backup_prepare(BlkTransactionState
*common
, Error
**errp
)
1371 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1372 DriveBackup
*backup
;
1373 Error
*local_err
= NULL
;
1375 assert(common
->action
->kind
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1376 backup
= common
->action
->drive_backup
;
1378 qmp_drive_backup(backup
->device
, backup
->target
,
1379 backup
->has_format
, backup
->format
,
1381 backup
->has_mode
, backup
->mode
,
1382 backup
->has_speed
, backup
->speed
,
1383 backup
->has_on_source_error
, backup
->on_source_error
,
1384 backup
->has_on_target_error
, backup
->on_target_error
,
1386 if (error_is_set(&local_err
)) {
1387 error_propagate(errp
, local_err
);
1393 state
->bs
= bdrv_find(backup
->device
);
1394 state
->job
= state
->bs
->job
;
1397 static void drive_backup_abort(BlkTransactionState
*common
)
1399 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1400 BlockDriverState
*bs
= state
->bs
;
1402 /* Only cancel if it's the job we started */
1403 if (bs
&& bs
->job
&& bs
->job
== state
->job
) {
1404 block_job_cancel_sync(bs
->job
);
1408 static void abort_prepare(BlkTransactionState
*common
, Error
**errp
)
1410 error_setg(errp
, "Transaction aborted using Abort action");
1413 static void abort_commit(BlkTransactionState
*common
)
1415 g_assert_not_reached(); /* this action never succeeds */
1418 static const BdrvActionOps actions
[] = {
1419 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
1420 .instance_size
= sizeof(ExternalSnapshotState
),
1421 .prepare
= external_snapshot_prepare
,
1422 .commit
= external_snapshot_commit
,
1423 .abort
= external_snapshot_abort
,
1425 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
1426 .instance_size
= sizeof(DriveBackupState
),
1427 .prepare
= drive_backup_prepare
,
1428 .abort
= drive_backup_abort
,
1430 [TRANSACTION_ACTION_KIND_ABORT
] = {
1431 .instance_size
= sizeof(BlkTransactionState
),
1432 .prepare
= abort_prepare
,
1433 .commit
= abort_commit
,
1435 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
1436 .instance_size
= sizeof(InternalSnapshotState
),
1437 .prepare
= internal_snapshot_prepare
,
1438 .abort
= internal_snapshot_abort
,
1443 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1444 * then we do not pivot any of the devices in the group, and abandon the
1447 void qmp_transaction(TransactionActionList
*dev_list
, Error
**errp
)
1449 TransactionActionList
*dev_entry
= dev_list
;
1450 BlkTransactionState
*state
, *next
;
1451 Error
*local_err
= NULL
;
1453 QSIMPLEQ_HEAD(snap_bdrv_states
, BlkTransactionState
) snap_bdrv_states
;
1454 QSIMPLEQ_INIT(&snap_bdrv_states
);
1456 /* drain all i/o before any snapshots */
1459 /* We don't do anything in this loop that commits us to the snapshot */
1460 while (NULL
!= dev_entry
) {
1461 TransactionAction
*dev_info
= NULL
;
1462 const BdrvActionOps
*ops
;
1464 dev_info
= dev_entry
->value
;
1465 dev_entry
= dev_entry
->next
;
1467 assert(dev_info
->kind
< ARRAY_SIZE(actions
));
1469 ops
= &actions
[dev_info
->kind
];
1470 assert(ops
->instance_size
> 0);
1472 state
= g_malloc0(ops
->instance_size
);
1474 state
->action
= dev_info
;
1475 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
1477 state
->ops
->prepare(state
, &local_err
);
1478 if (error_is_set(&local_err
)) {
1479 error_propagate(errp
, local_err
);
1480 goto delete_and_fail
;
1484 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1485 if (state
->ops
->commit
) {
1486 state
->ops
->commit(state
);
1495 * failure, and it is all-or-none; abandon each new bs, and keep using
1496 * the original bs for all images
1498 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1499 if (state
->ops
->abort
) {
1500 state
->ops
->abort(state
);
1504 QSIMPLEQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
1505 if (state
->ops
->clean
) {
1506 state
->ops
->clean(state
);
1513 static void eject_device(BlockDriverState
*bs
, int force
, Error
**errp
)
1515 if (bdrv_in_use(bs
)) {
1516 error_set(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
1519 if (!bdrv_dev_has_removable_media(bs
)) {
1520 error_set(errp
, QERR_DEVICE_NOT_REMOVABLE
, bdrv_get_device_name(bs
));
1524 if (bdrv_dev_is_medium_locked(bs
) && !bdrv_dev_is_tray_open(bs
)) {
1525 bdrv_dev_eject_request(bs
, force
);
1527 error_set(errp
, QERR_DEVICE_LOCKED
, bdrv_get_device_name(bs
));
1535 void qmp_eject(const char *device
, bool has_force
, bool force
, Error
**errp
)
1537 BlockDriverState
*bs
;
1539 bs
= bdrv_find(device
);
1541 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1545 eject_device(bs
, force
, errp
);
1548 void qmp_block_passwd(bool has_device
, const char *device
,
1549 bool has_node_name
, const char *node_name
,
1550 const char *password
, Error
**errp
)
1552 Error
*local_err
= NULL
;
1553 BlockDriverState
*bs
;
1556 bs
= bdrv_lookup_bs(has_device
? device
: NULL
,
1557 has_node_name
? node_name
: NULL
,
1559 if (error_is_set(&local_err
)) {
1560 error_propagate(errp
, local_err
);
1564 err
= bdrv_set_key(bs
, password
);
1565 if (err
== -EINVAL
) {
1566 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1568 } else if (err
< 0) {
1569 error_set(errp
, QERR_INVALID_PASSWORD
);
1574 static void qmp_bdrv_open_encrypted(BlockDriverState
*bs
, const char *filename
,
1575 int bdrv_flags
, BlockDriver
*drv
,
1576 const char *password
, Error
**errp
)
1578 Error
*local_err
= NULL
;
1581 ret
= bdrv_open(bs
, filename
, NULL
, bdrv_flags
, drv
, &local_err
);
1583 error_propagate(errp
, local_err
);
1587 if (bdrv_key_required(bs
)) {
1589 if (bdrv_set_key(bs
, password
) < 0) {
1590 error_set(errp
, QERR_INVALID_PASSWORD
);
1593 error_set(errp
, QERR_DEVICE_ENCRYPTED
, bdrv_get_device_name(bs
),
1594 bdrv_get_encrypted_filename(bs
));
1596 } else if (password
) {
1597 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1601 void qmp_change_blockdev(const char *device
, const char *filename
,
1602 const char *format
, Error
**errp
)
1604 BlockDriverState
*bs
;
1605 BlockDriver
*drv
= NULL
;
1609 bs
= bdrv_find(device
);
1611 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1616 drv
= bdrv_find_whitelisted_format(format
, bs
->read_only
);
1618 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1623 eject_device(bs
, 0, &err
);
1624 if (error_is_set(&err
)) {
1625 error_propagate(errp
, err
);
1629 bdrv_flags
= bdrv_is_read_only(bs
) ? 0 : BDRV_O_RDWR
;
1630 bdrv_flags
|= bdrv_is_snapshot(bs
) ? BDRV_O_SNAPSHOT
: 0;
1632 qmp_bdrv_open_encrypted(bs
, filename
, bdrv_flags
, drv
, NULL
, errp
);
1635 /* throttling disk I/O limits */
1636 void qmp_block_set_io_throttle(const char *device
, int64_t bps
, int64_t bps_rd
,
1643 bool has_bps_rd_max
,
1645 bool has_bps_wr_max
,
1649 bool has_iops_rd_max
,
1650 int64_t iops_rd_max
,
1651 bool has_iops_wr_max
,
1652 int64_t iops_wr_max
,
1654 int64_t iops_size
, Error
**errp
)
1657 BlockDriverState
*bs
;
1659 bs
= bdrv_find(device
);
1661 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1665 memset(&cfg
, 0, sizeof(cfg
));
1666 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
= bps
;
1667 cfg
.buckets
[THROTTLE_BPS_READ
].avg
= bps_rd
;
1668 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
= bps_wr
;
1670 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
= iops
;
1671 cfg
.buckets
[THROTTLE_OPS_READ
].avg
= iops_rd
;
1672 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
= iops_wr
;
1675 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
= bps_max
;
1677 if (has_bps_rd_max
) {
1678 cfg
.buckets
[THROTTLE_BPS_READ
].max
= bps_rd_max
;
1680 if (has_bps_wr_max
) {
1681 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
= bps_wr_max
;
1684 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
= iops_max
;
1686 if (has_iops_rd_max
) {
1687 cfg
.buckets
[THROTTLE_OPS_READ
].max
= iops_rd_max
;
1689 if (has_iops_wr_max
) {
1690 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
= iops_wr_max
;
1693 if (has_iops_size
) {
1694 cfg
.op_size
= iops_size
;
1697 if (!check_throttle_config(&cfg
, errp
)) {
1701 if (!bs
->io_limits_enabled
&& throttle_enabled(&cfg
)) {
1702 bdrv_io_limits_enable(bs
);
1703 } else if (bs
->io_limits_enabled
&& !throttle_enabled(&cfg
)) {
1704 bdrv_io_limits_disable(bs
);
1707 if (bs
->io_limits_enabled
) {
1708 bdrv_set_io_limits(bs
, &cfg
);
1712 int do_drive_del(Monitor
*mon
, const QDict
*qdict
, QObject
**ret_data
)
1714 const char *id
= qdict_get_str(qdict
, "id");
1715 BlockDriverState
*bs
;
1719 qerror_report(QERR_DEVICE_NOT_FOUND
, id
);
1722 if (bdrv_in_use(bs
)) {
1723 qerror_report(QERR_DEVICE_IN_USE
, id
);
1727 /* quiesce block driver; prevent further io */
1732 /* if we have a device attached to this BlockDriverState
1733 * then we need to make the drive anonymous until the device
1734 * can be removed. If this is a drive with no device backing
1735 * then we can just get rid of the block driver state right here.
1737 if (bdrv_get_attached_dev(bs
)) {
1740 /* Further I/O must not pause the guest */
1741 bdrv_set_on_error(bs
, BLOCKDEV_ON_ERROR_REPORT
,
1742 BLOCKDEV_ON_ERROR_REPORT
);
1744 drive_uninit(drive_get_by_blockdev(bs
));
1750 void qmp_block_resize(bool has_device
, const char *device
,
1751 bool has_node_name
, const char *node_name
,
1752 int64_t size
, Error
**errp
)
1754 Error
*local_err
= NULL
;
1755 BlockDriverState
*bs
;
1758 bs
= bdrv_lookup_bs(has_device
? device
: NULL
,
1759 has_node_name
? node_name
: NULL
,
1761 if (error_is_set(&local_err
)) {
1762 error_propagate(errp
, local_err
);
1766 if (!bdrv_is_first_non_filter(bs
)) {
1767 error_set(errp
, QERR_FEATURE_DISABLED
, "resize");
1772 error_set(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
1776 /* complete all in-flight operations before resizing the device */
1779 ret
= bdrv_truncate(bs
, size
);
1784 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1787 error_set(errp
, QERR_UNSUPPORTED
);
1790 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1793 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1796 error_setg_errno(errp
, -ret
, "Could not resize");
1801 static void block_job_cb(void *opaque
, int ret
)
1803 BlockDriverState
*bs
= opaque
;
1806 trace_block_job_cb(bs
, bs
->job
, ret
);
1809 obj
= qobject_from_block_job(bs
->job
);
1811 QDict
*dict
= qobject_to_qdict(obj
);
1812 qdict_put(dict
, "error", qstring_from_str(strerror(-ret
)));
1815 if (block_job_is_cancelled(bs
->job
)) {
1816 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED
, obj
);
1818 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED
, obj
);
1820 qobject_decref(obj
);
1822 bdrv_put_ref_bh_schedule(bs
);
1825 void qmp_block_stream(const char *device
, bool has_base
,
1826 const char *base
, bool has_speed
, int64_t speed
,
1827 bool has_on_error
, BlockdevOnError on_error
,
1830 BlockDriverState
*bs
;
1831 BlockDriverState
*base_bs
= NULL
;
1832 Error
*local_err
= NULL
;
1834 if (!has_on_error
) {
1835 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1838 bs
= bdrv_find(device
);
1840 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1845 base_bs
= bdrv_find_backing_image(bs
, base
);
1846 if (base_bs
== NULL
) {
1847 error_set(errp
, QERR_BASE_NOT_FOUND
, base
);
1852 stream_start(bs
, base_bs
, base
, has_speed
? speed
: 0,
1853 on_error
, block_job_cb
, bs
, &local_err
);
1854 if (error_is_set(&local_err
)) {
1855 error_propagate(errp
, local_err
);
1859 trace_qmp_block_stream(bs
, bs
->job
);
1862 void qmp_block_commit(const char *device
,
1863 bool has_base
, const char *base
, const char *top
,
1864 bool has_speed
, int64_t speed
,
1867 BlockDriverState
*bs
;
1868 BlockDriverState
*base_bs
, *top_bs
;
1869 Error
*local_err
= NULL
;
1870 /* This will be part of the QMP command, if/when the
1871 * BlockdevOnError change for blkmirror makes it in
1873 BlockdevOnError on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1875 /* drain all i/o before commits */
1878 bs
= bdrv_find(device
);
1880 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1884 /* default top_bs is the active layer */
1888 if (strcmp(bs
->filename
, top
) != 0) {
1889 top_bs
= bdrv_find_backing_image(bs
, top
);
1893 if (top_bs
== NULL
) {
1894 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
1898 if (has_base
&& base
) {
1899 base_bs
= bdrv_find_backing_image(top_bs
, base
);
1901 base_bs
= bdrv_find_base(top_bs
);
1904 if (base_bs
== NULL
) {
1905 error_set(errp
, QERR_BASE_NOT_FOUND
, base
? base
: "NULL");
1910 commit_active_start(bs
, base_bs
, speed
, on_error
, block_job_cb
,
1913 commit_start(bs
, base_bs
, top_bs
, speed
, on_error
, block_job_cb
, bs
,
1916 if (local_err
!= NULL
) {
1917 error_propagate(errp
, local_err
);
1922 void qmp_drive_backup(const char *device
, const char *target
,
1923 bool has_format
, const char *format
,
1924 enum MirrorSyncMode sync
,
1925 bool has_mode
, enum NewImageMode mode
,
1926 bool has_speed
, int64_t speed
,
1927 bool has_on_source_error
, BlockdevOnError on_source_error
,
1928 bool has_on_target_error
, BlockdevOnError on_target_error
,
1931 BlockDriverState
*bs
;
1932 BlockDriverState
*target_bs
;
1933 BlockDriverState
*source
= NULL
;
1934 BlockDriver
*drv
= NULL
;
1935 Error
*local_err
= NULL
;
1943 if (!has_on_source_error
) {
1944 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1946 if (!has_on_target_error
) {
1947 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1950 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1953 bs
= bdrv_find(device
);
1955 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1959 if (!bdrv_is_inserted(bs
)) {
1960 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1965 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1968 drv
= bdrv_find_format(format
);
1970 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1975 if (bdrv_in_use(bs
)) {
1976 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1980 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1982 /* See if we have a backing HD we can use to create our new image
1984 if (sync
== MIRROR_SYNC_MODE_TOP
) {
1985 source
= bs
->backing_hd
;
1987 sync
= MIRROR_SYNC_MODE_FULL
;
1990 if (sync
== MIRROR_SYNC_MODE_NONE
) {
1994 size
= bdrv_getlength(bs
);
1996 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
2000 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
2001 assert(format
&& drv
);
2003 bdrv_img_create(target
, format
, source
->filename
,
2004 source
->drv
->format_name
, NULL
,
2005 size
, flags
, &local_err
, false);
2007 bdrv_img_create(target
, format
, NULL
, NULL
, NULL
,
2008 size
, flags
, &local_err
, false);
2012 if (error_is_set(&local_err
)) {
2013 error_propagate(errp
, local_err
);
2017 target_bs
= bdrv_new("");
2018 ret
= bdrv_open(target_bs
, target
, NULL
, flags
, drv
, &local_err
);
2020 bdrv_unref(target_bs
);
2021 error_propagate(errp
, local_err
);
2025 backup_start(bs
, target_bs
, speed
, sync
, on_source_error
, on_target_error
,
2026 block_job_cb
, bs
, &local_err
);
2027 if (local_err
!= NULL
) {
2028 bdrv_unref(target_bs
);
2029 error_propagate(errp
, local_err
);
2034 BlockDeviceInfoList
*qmp_query_named_block_nodes(Error
**errp
)
2036 return bdrv_named_nodes_list();
2039 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2041 void qmp_drive_mirror(const char *device
, const char *target
,
2042 bool has_format
, const char *format
,
2043 enum MirrorSyncMode sync
,
2044 bool has_mode
, enum NewImageMode mode
,
2045 bool has_speed
, int64_t speed
,
2046 bool has_granularity
, uint32_t granularity
,
2047 bool has_buf_size
, int64_t buf_size
,
2048 bool has_on_source_error
, BlockdevOnError on_source_error
,
2049 bool has_on_target_error
, BlockdevOnError on_target_error
,
2052 BlockDriverState
*bs
;
2053 BlockDriverState
*source
, *target_bs
;
2054 BlockDriver
*drv
= NULL
;
2055 Error
*local_err
= NULL
;
2063 if (!has_on_source_error
) {
2064 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
2066 if (!has_on_target_error
) {
2067 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
2070 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
2072 if (!has_granularity
) {
2075 if (!has_buf_size
) {
2076 buf_size
= DEFAULT_MIRROR_BUF_SIZE
;
2079 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
2080 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
2083 if (granularity
& (granularity
- 1)) {
2084 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
2088 bs
= bdrv_find(device
);
2090 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
2094 if (!bdrv_is_inserted(bs
)) {
2095 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
2100 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
2103 drv
= bdrv_find_format(format
);
2105 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
2110 if (bdrv_in_use(bs
)) {
2111 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
2115 flags
= bs
->open_flags
| BDRV_O_RDWR
;
2116 source
= bs
->backing_hd
;
2117 if (!source
&& sync
== MIRROR_SYNC_MODE_TOP
) {
2118 sync
= MIRROR_SYNC_MODE_FULL
;
2120 if (sync
== MIRROR_SYNC_MODE_NONE
) {
2124 size
= bdrv_getlength(bs
);
2126 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
2130 if ((sync
== MIRROR_SYNC_MODE_FULL
|| !source
)
2131 && mode
!= NEW_IMAGE_MODE_EXISTING
)
2133 /* create new image w/o backing file */
2134 assert(format
&& drv
);
2135 bdrv_img_create(target
, format
,
2136 NULL
, NULL
, NULL
, size
, flags
, &local_err
, false);
2139 case NEW_IMAGE_MODE_EXISTING
:
2141 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
2142 /* create new image with backing file */
2143 bdrv_img_create(target
, format
,
2145 source
->drv
->format_name
,
2146 NULL
, size
, flags
, &local_err
, false);
2153 if (error_is_set(&local_err
)) {
2154 error_propagate(errp
, local_err
);
2158 /* Mirroring takes care of copy-on-write using the source's backing
2161 target_bs
= bdrv_new("");
2162 ret
= bdrv_open(target_bs
, target
, NULL
, flags
| BDRV_O_NO_BACKING
, drv
,
2165 bdrv_unref(target_bs
);
2166 error_propagate(errp
, local_err
);
2170 mirror_start(bs
, target_bs
, speed
, granularity
, buf_size
, sync
,
2171 on_source_error
, on_target_error
,
2172 block_job_cb
, bs
, &local_err
);
2173 if (local_err
!= NULL
) {
2174 bdrv_unref(target_bs
);
2175 error_propagate(errp
, local_err
);
2180 static BlockJob
*find_block_job(const char *device
)
2182 BlockDriverState
*bs
;
2184 bs
= bdrv_find(device
);
2185 if (!bs
|| !bs
->job
) {
2191 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
2193 BlockJob
*job
= find_block_job(device
);
2196 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2200 block_job_set_speed(job
, speed
, errp
);
2203 void qmp_block_job_cancel(const char *device
,
2204 bool has_force
, bool force
, Error
**errp
)
2206 BlockJob
*job
= find_block_job(device
);
2213 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2216 if (job
->paused
&& !force
) {
2217 error_set(errp
, QERR_BLOCK_JOB_PAUSED
, device
);
2221 trace_qmp_block_job_cancel(job
);
2222 block_job_cancel(job
);
2225 void qmp_block_job_pause(const char *device
, Error
**errp
)
2227 BlockJob
*job
= find_block_job(device
);
2230 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2234 trace_qmp_block_job_pause(job
);
2235 block_job_pause(job
);
2238 void qmp_block_job_resume(const char *device
, Error
**errp
)
2240 BlockJob
*job
= find_block_job(device
);
2243 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2247 trace_qmp_block_job_resume(job
);
2248 block_job_resume(job
);
2251 void qmp_block_job_complete(const char *device
, Error
**errp
)
2253 BlockJob
*job
= find_block_job(device
);
2256 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2260 trace_qmp_block_job_complete(job
);
2261 block_job_complete(job
, errp
);
2264 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
2266 QmpOutputVisitor
*ov
= qmp_output_visitor_new();
2269 Error
*local_err
= NULL
;
2271 /* Require an ID in the top level */
2272 if (!options
->has_id
) {
2273 error_setg(errp
, "Block device needs an ID");
2277 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2278 * cache.direct=false instead of silently switching to aio=threads, except
2279 * if called from drive_init.
2281 * For now, simply forbidding the combination for all drivers will do. */
2282 if (options
->has_aio
&& options
->aio
== BLOCKDEV_AIO_OPTIONS_NATIVE
) {
2283 bool direct
= options
->cache
->has_direct
&& options
->cache
->direct
;
2284 if (!options
->has_cache
&& !direct
) {
2285 error_setg(errp
, "aio=native requires cache.direct=true");
2290 visit_type_BlockdevOptions(qmp_output_get_visitor(ov
),
2291 &options
, NULL
, &local_err
);
2292 if (error_is_set(&local_err
)) {
2293 error_propagate(errp
, local_err
);
2297 obj
= qmp_output_get_qobject(ov
);
2298 qdict
= qobject_to_qdict(obj
);
2300 qdict_flatten(qdict
);
2302 blockdev_init(NULL
, qdict
, &local_err
);
2303 if (error_is_set(&local_err
)) {
2304 error_propagate(errp
, local_err
);
2309 qmp_output_visitor_cleanup(ov
);
2312 static void do_qmp_query_block_jobs_one(void *opaque
, BlockDriverState
*bs
)
2314 BlockJobInfoList
**prev
= opaque
;
2315 BlockJob
*job
= bs
->job
;
2318 BlockJobInfoList
*elem
= g_new0(BlockJobInfoList
, 1);
2319 elem
->value
= block_job_query(bs
->job
);
2320 (*prev
)->next
= elem
;
2325 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
2327 /* Dummy is a fake list element for holding the head pointer */
2328 BlockJobInfoList dummy
= {};
2329 BlockJobInfoList
*prev
= &dummy
;
2330 bdrv_iterate(do_qmp_query_block_jobs_one
, &prev
);
2334 QemuOptsList qemu_common_drive_opts
= {
2336 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
2340 .type
= QEMU_OPT_BOOL
,
2341 .help
= "enable/disable snapshot mode",
2344 .type
= QEMU_OPT_STRING
,
2345 .help
= "discard operation (ignore/off, unmap/on)",
2347 .name
= "cache.writeback",
2348 .type
= QEMU_OPT_BOOL
,
2349 .help
= "enables writeback mode for any caches",
2351 .name
= "cache.direct",
2352 .type
= QEMU_OPT_BOOL
,
2353 .help
= "enables use of O_DIRECT (bypass the host page cache)",
2355 .name
= "cache.no-flush",
2356 .type
= QEMU_OPT_BOOL
,
2357 .help
= "ignore any flush requests for the device",
2360 .type
= QEMU_OPT_STRING
,
2361 .help
= "host AIO implementation (threads, native)",
2364 .type
= QEMU_OPT_STRING
,
2365 .help
= "disk format (raw, qcow2, ...)",
2368 .type
= QEMU_OPT_STRING
,
2369 .help
= "disk serial number",
2372 .type
= QEMU_OPT_STRING
,
2373 .help
= "read error action",
2376 .type
= QEMU_OPT_STRING
,
2377 .help
= "write error action",
2379 .name
= "read-only",
2380 .type
= QEMU_OPT_BOOL
,
2381 .help
= "open drive file as read-only",
2383 .name
= "throttling.iops-total",
2384 .type
= QEMU_OPT_NUMBER
,
2385 .help
= "limit total I/O operations per second",
2387 .name
= "throttling.iops-read",
2388 .type
= QEMU_OPT_NUMBER
,
2389 .help
= "limit read operations per second",
2391 .name
= "throttling.iops-write",
2392 .type
= QEMU_OPT_NUMBER
,
2393 .help
= "limit write operations per second",
2395 .name
= "throttling.bps-total",
2396 .type
= QEMU_OPT_NUMBER
,
2397 .help
= "limit total bytes per second",
2399 .name
= "throttling.bps-read",
2400 .type
= QEMU_OPT_NUMBER
,
2401 .help
= "limit read bytes per second",
2403 .name
= "throttling.bps-write",
2404 .type
= QEMU_OPT_NUMBER
,
2405 .help
= "limit write bytes per second",
2407 .name
= "throttling.iops-total-max",
2408 .type
= QEMU_OPT_NUMBER
,
2409 .help
= "I/O operations burst",
2411 .name
= "throttling.iops-read-max",
2412 .type
= QEMU_OPT_NUMBER
,
2413 .help
= "I/O operations read burst",
2415 .name
= "throttling.iops-write-max",
2416 .type
= QEMU_OPT_NUMBER
,
2417 .help
= "I/O operations write burst",
2419 .name
= "throttling.bps-total-max",
2420 .type
= QEMU_OPT_NUMBER
,
2421 .help
= "total bytes burst",
2423 .name
= "throttling.bps-read-max",
2424 .type
= QEMU_OPT_NUMBER
,
2425 .help
= "total bytes read burst",
2427 .name
= "throttling.bps-write-max",
2428 .type
= QEMU_OPT_NUMBER
,
2429 .help
= "total bytes write burst",
2431 .name
= "throttling.iops-size",
2432 .type
= QEMU_OPT_NUMBER
,
2433 .help
= "when limiting by iops max size of an I/O in bytes",
2435 .name
= "copy-on-read",
2436 .type
= QEMU_OPT_BOOL
,
2437 .help
= "copy read data from backing file into image file",
2439 { /* end of list */ }
2443 QemuOptsList qemu_drive_opts
= {
2445 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
2448 * no elements => accept any params
2449 * validation will happen later
2451 { /* end of list */ }