2 * QEMU host block devices
4 * Copyright (c) 2003-2008 Fabrice Bellard
6 * This work is licensed under the terms of the GNU GPL, version 2 or
7 * later. See the COPYING file in the top-level directory.
9 * This file incorporates work covered by the following copyright and
12 * Copyright (c) 2003-2008 Fabrice Bellard
14 * Permission is hereby granted, free of charge, to any person obtaining a copy
15 * of this software and associated documentation files (the "Software"), to deal
16 * in the Software without restriction, including without limitation the rights
17 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
18 * copies of the Software, and to permit persons to whom the Software is
19 * furnished to do so, subject to the following conditions:
21 * The above copyright notice and this permission notice shall be included in
22 * all copies or substantial portions of the Software.
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
27 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
33 #include "sysemu/blockdev.h"
34 #include "hw/block/block.h"
35 #include "block/blockjob.h"
36 #include "monitor/monitor.h"
37 #include "qapi/qmp/qerror.h"
38 #include "qemu/option.h"
39 #include "qemu/config-file.h"
40 #include "qapi/qmp/types.h"
41 #include "qapi-visit.h"
42 #include "qapi/qmp-output-visitor.h"
43 #include "sysemu/sysemu.h"
44 #include "block/block_int.h"
45 #include "qmp-commands.h"
47 #include "sysemu/arch_init.h"
49 static QTAILQ_HEAD(drivelist
, DriveInfo
) drives
= QTAILQ_HEAD_INITIALIZER(drives
);
50 extern QemuOptsList qemu_common_drive_opts
;
52 static const char *const if_name
[IF_COUNT
] = {
56 [IF_FLOPPY
] = "floppy",
57 [IF_PFLASH
] = "pflash",
60 [IF_VIRTIO
] = "virtio",
64 static const int if_max_devs
[IF_COUNT
] = {
66 * Do not change these numbers! They govern how drive option
67 * index maps to unit and bus. That mapping is ABI.
69 * All controllers used to imlement if=T drives need to support
70 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
71 * Otherwise, some index values map to "impossible" bus, unit
74 * For instance, if you change [IF_SCSI] to 255, -drive
75 * if=scsi,index=12 no longer means bus=1,unit=5, but
76 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
77 * the drive can't be set up. Regression.
84 * We automatically delete the drive when a device using it gets
85 * unplugged. Questionable feature, but we can't just drop it.
86 * Device models call blockdev_mark_auto_del() to schedule the
87 * automatic deletion, and generic qdev code calls blockdev_auto_del()
88 * when deletion is actually safe.
90 void blockdev_mark_auto_del(BlockDriverState
*bs
)
92 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
94 if (dinfo
&& !dinfo
->enable_auto_del
) {
99 block_job_cancel(bs
->job
);
106 void blockdev_auto_del(BlockDriverState
*bs
)
108 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
110 if (dinfo
&& dinfo
->auto_del
) {
111 drive_put_ref(dinfo
);
115 static int drive_index_to_bus_id(BlockInterfaceType type
, int index
)
117 int max_devs
= if_max_devs
[type
];
118 return max_devs
? index
/ max_devs
: 0;
121 static int drive_index_to_unit_id(BlockInterfaceType type
, int index
)
123 int max_devs
= if_max_devs
[type
];
124 return max_devs
? index
% max_devs
: index
;
127 QemuOpts
*drive_def(const char *optstr
)
129 return qemu_opts_parse(qemu_find_opts("drive"), optstr
, 0);
132 QemuOpts
*drive_add(BlockInterfaceType type
, int index
, const char *file
,
138 opts
= drive_def(optstr
);
142 if (type
!= IF_DEFAULT
) {
143 qemu_opt_set(opts
, "if", if_name
[type
]);
146 snprintf(buf
, sizeof(buf
), "%d", index
);
147 qemu_opt_set(opts
, "index", buf
);
150 qemu_opt_set(opts
, "file", file
);
154 DriveInfo
*drive_get(BlockInterfaceType type
, int bus
, int unit
)
158 /* seek interface, bus and unit */
160 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
161 if (dinfo
->type
== type
&&
170 DriveInfo
*drive_get_by_index(BlockInterfaceType type
, int index
)
172 return drive_get(type
,
173 drive_index_to_bus_id(type
, index
),
174 drive_index_to_unit_id(type
, index
));
177 int drive_get_max_bus(BlockInterfaceType type
)
183 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
184 if(dinfo
->type
== type
&&
185 dinfo
->bus
> max_bus
)
186 max_bus
= dinfo
->bus
;
191 /* Get a block device. This should only be used for single-drive devices
192 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
194 DriveInfo
*drive_get_next(BlockInterfaceType type
)
196 static int next_block_unit
[IF_COUNT
];
198 return drive_get(type
, 0, next_block_unit
[type
]++);
201 DriveInfo
*drive_get_by_blockdev(BlockDriverState
*bs
)
205 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
206 if (dinfo
->bdrv
== bs
) {
213 static void bdrv_format_print(void *opaque
, const char *name
)
215 error_printf(" %s", name
);
218 static void drive_uninit(DriveInfo
*dinfo
)
221 qemu_opts_del(dinfo
->opts
);
224 bdrv_unref(dinfo
->bdrv
);
226 QTAILQ_REMOVE(&drives
, dinfo
, next
);
227 g_free(dinfo
->serial
);
231 void drive_put_ref(DriveInfo
*dinfo
)
233 assert(dinfo
->refcount
);
234 if (--dinfo
->refcount
== 0) {
239 void drive_get_ref(DriveInfo
*dinfo
)
246 BlockDriverState
*bs
;
249 static void bdrv_put_ref_bh(void *opaque
)
251 BDRVPutRefBH
*s
= opaque
;
254 qemu_bh_delete(s
->bh
);
259 * Release a BDS reference in a BH
261 * It is not safe to use bdrv_unref() from a callback function when the callers
262 * still need the BlockDriverState. In such cases we schedule a BH to release
265 static void bdrv_put_ref_bh_schedule(BlockDriverState
*bs
)
269 s
= g_new(BDRVPutRefBH
, 1);
270 s
->bh
= qemu_bh_new(bdrv_put_ref_bh
, s
);
272 qemu_bh_schedule(s
->bh
);
275 static int parse_block_error_action(const char *buf
, bool is_read
, Error
**errp
)
277 if (!strcmp(buf
, "ignore")) {
278 return BLOCKDEV_ON_ERROR_IGNORE
;
279 } else if (!is_read
&& !strcmp(buf
, "enospc")) {
280 return BLOCKDEV_ON_ERROR_ENOSPC
;
281 } else if (!strcmp(buf
, "stop")) {
282 return BLOCKDEV_ON_ERROR_STOP
;
283 } else if (!strcmp(buf
, "report")) {
284 return BLOCKDEV_ON_ERROR_REPORT
;
286 error_setg(errp
, "'%s' invalid %s error action",
287 buf
, is_read
? "read" : "write");
292 static bool check_throttle_config(ThrottleConfig
*cfg
, Error
**errp
)
294 if (throttle_conflicting(cfg
)) {
295 error_setg(errp
, "bps/iops/max total values and read/write values"
296 " cannot be used at the same time");
300 if (!throttle_is_valid(cfg
)) {
301 error_setg(errp
, "bps/iops/maxs values must be 0 or greater");
308 typedef enum { MEDIA_DISK
, MEDIA_CDROM
} DriveMediaType
;
310 /* Takes the ownership of bs_opts */
311 static DriveInfo
*blockdev_init(QDict
*bs_opts
,
312 BlockInterfaceType type
,
316 const char *file
= NULL
;
320 int on_read_error
, on_write_error
;
329 bool has_driver_specific_opts
;
330 BlockDriver
*drv
= NULL
;
332 /* Check common options by copying from bs_opts to opts, all other options
333 * stay in bs_opts for processing by bdrv_open(). */
334 id
= qdict_get_try_str(bs_opts
, "id");
335 opts
= qemu_opts_create(&qemu_common_drive_opts
, id
, 1, &error
);
336 if (error_is_set(&error
)) {
337 error_propagate(errp
, error
);
341 qemu_opts_absorb_qdict(opts
, bs_opts
, &error
);
342 if (error_is_set(&error
)) {
343 error_propagate(errp
, error
);
348 qdict_del(bs_opts
, "id");
351 has_driver_specific_opts
= !!qdict_size(bs_opts
);
353 /* extract parameters */
354 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
355 ro
= qemu_opt_get_bool(opts
, "read-only", 0);
356 copy_on_read
= qemu_opt_get_bool(opts
, "copy-on-read", false);
358 file
= qemu_opt_get(opts
, "file");
359 serial
= qemu_opt_get(opts
, "serial");
361 if ((buf
= qemu_opt_get(opts
, "discard")) != NULL
) {
362 if (bdrv_parse_discard_flags(buf
, &bdrv_flags
) != 0) {
363 error_setg(errp
, "invalid discard option");
368 if (qemu_opt_get_bool(opts
, "cache.writeback", true)) {
369 bdrv_flags
|= BDRV_O_CACHE_WB
;
371 if (qemu_opt_get_bool(opts
, "cache.direct", false)) {
372 bdrv_flags
|= BDRV_O_NOCACHE
;
374 if (qemu_opt_get_bool(opts
, "cache.no-flush", false)) {
375 bdrv_flags
|= BDRV_O_NO_FLUSH
;
378 #ifdef CONFIG_LINUX_AIO
379 if ((buf
= qemu_opt_get(opts
, "aio")) != NULL
) {
380 if (!strcmp(buf
, "native")) {
381 bdrv_flags
|= BDRV_O_NATIVE_AIO
;
382 } else if (!strcmp(buf
, "threads")) {
383 /* this is the default */
385 error_setg(errp
, "invalid aio option");
391 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
392 if (is_help_option(buf
)) {
393 error_printf("Supported formats:");
394 bdrv_iterate_format(bdrv_format_print
, NULL
);
399 drv
= bdrv_find_format(buf
);
401 error_setg(errp
, "'%s' invalid format", buf
);
406 /* disk I/O throttling */
407 memset(&cfg
, 0, sizeof(cfg
));
408 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
=
409 qemu_opt_get_number(opts
, "throttling.bps-total", 0);
410 cfg
.buckets
[THROTTLE_BPS_READ
].avg
=
411 qemu_opt_get_number(opts
, "throttling.bps-read", 0);
412 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
=
413 qemu_opt_get_number(opts
, "throttling.bps-write", 0);
414 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
=
415 qemu_opt_get_number(opts
, "throttling.iops-total", 0);
416 cfg
.buckets
[THROTTLE_OPS_READ
].avg
=
417 qemu_opt_get_number(opts
, "throttling.iops-read", 0);
418 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
=
419 qemu_opt_get_number(opts
, "throttling.iops-write", 0);
421 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
=
422 qemu_opt_get_number(opts
, "throttling.bps-total-max", 0);
423 cfg
.buckets
[THROTTLE_BPS_READ
].max
=
424 qemu_opt_get_number(opts
, "throttling.bps-read-max", 0);
425 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
=
426 qemu_opt_get_number(opts
, "throttling.bps-write-max", 0);
427 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
=
428 qemu_opt_get_number(opts
, "throttling.iops-total-max", 0);
429 cfg
.buckets
[THROTTLE_OPS_READ
].max
=
430 qemu_opt_get_number(opts
, "throttling.iops-read-max", 0);
431 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
=
432 qemu_opt_get_number(opts
, "throttling.iops-write-max", 0);
434 cfg
.op_size
= qemu_opt_get_number(opts
, "throttling.iops-size", 0);
436 if (!check_throttle_config(&cfg
, &error
)) {
437 error_propagate(errp
, error
);
441 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
442 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
443 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_NONE
) {
444 error_setg(errp
, "werror is not supported by this bus type");
448 on_write_error
= parse_block_error_action(buf
, 0, &error
);
449 if (error_is_set(&error
)) {
450 error_propagate(errp
, error
);
455 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
456 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
457 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&& type
!= IF_NONE
) {
458 error_report("rerror is not supported by this bus type");
462 on_read_error
= parse_block_error_action(buf
, 1, &error
);
463 if (error_is_set(&error
)) {
464 error_propagate(errp
, error
);
470 dinfo
= g_malloc0(sizeof(*dinfo
));
471 dinfo
->id
= g_strdup(qemu_opts_id(opts
));
472 dinfo
->bdrv
= bdrv_new(dinfo
->id
);
473 dinfo
->bdrv
->open_flags
= snapshot
? BDRV_O_SNAPSHOT
: 0;
474 dinfo
->bdrv
->read_only
= ro
;
477 if (serial
!= NULL
) {
478 dinfo
->serial
= g_strdup(serial
);
480 QTAILQ_INSERT_TAIL(&drives
, dinfo
, next
);
482 bdrv_set_on_error(dinfo
->bdrv
, on_read_error
, on_write_error
);
484 /* disk I/O throttling */
485 if (throttle_enabled(&cfg
)) {
486 bdrv_io_limits_enable(dinfo
->bdrv
);
487 bdrv_set_io_limits(dinfo
->bdrv
, &cfg
);
490 if (!file
|| !*file
) {
491 if (has_driver_specific_opts
) {
500 /* always use cache=unsafe with snapshot */
501 bdrv_flags
&= ~BDRV_O_CACHE_MASK
;
502 bdrv_flags
|= (BDRV_O_SNAPSHOT
|BDRV_O_CACHE_WB
|BDRV_O_NO_FLUSH
);
506 bdrv_flags
|= BDRV_O_COPY_ON_READ
;
509 if (runstate_check(RUN_STATE_INMIGRATE
)) {
510 bdrv_flags
|= BDRV_O_INCOMING
;
513 bdrv_flags
|= ro
? 0 : BDRV_O_RDWR
;
516 ret
= bdrv_open(dinfo
->bdrv
, file
, bs_opts
, bdrv_flags
, drv
, &error
);
519 error_setg(errp
, "could not open disk image %s: %s",
520 file
?: dinfo
->id
, error_get_pretty(error
));
525 if (bdrv_key_required(dinfo
->bdrv
))
534 bdrv_unref(dinfo
->bdrv
);
536 QTAILQ_REMOVE(&drives
, dinfo
, next
);
544 static void qemu_opt_rename(QemuOpts
*opts
, const char *from
, const char *to
)
548 value
= qemu_opt_get(opts
, from
);
550 qemu_opt_set(opts
, to
, value
);
551 qemu_opt_unset(opts
, from
);
555 QemuOptsList qemu_legacy_drive_opts
= {
557 .head
= QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts
.head
),
561 .type
= QEMU_OPT_NUMBER
,
562 .help
= "bus number",
565 .type
= QEMU_OPT_NUMBER
,
566 .help
= "unit number (i.e. lun for scsi)",
569 .type
= QEMU_OPT_NUMBER
,
570 .help
= "index number",
573 .type
= QEMU_OPT_STRING
,
574 .help
= "media type (disk, cdrom)",
577 .type
= QEMU_OPT_STRING
,
578 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
581 .type
= QEMU_OPT_NUMBER
,
582 .help
= "number of cylinders (ide disk geometry)",
585 .type
= QEMU_OPT_NUMBER
,
586 .help
= "number of heads (ide disk geometry)",
589 .type
= QEMU_OPT_NUMBER
,
590 .help
= "number of sectors (ide disk geometry)",
593 .type
= QEMU_OPT_STRING
,
594 .help
= "chs translation (auto, lba, none)",
597 .type
= QEMU_OPT_BOOL
,
598 .help
= "(deprecated, ignored)",
601 .type
= QEMU_OPT_STRING
,
602 .help
= "pci address (virtio only)",
605 /* Options that are passed on, but have special semantics with -drive */
608 .type
= QEMU_OPT_BOOL
,
609 .help
= "open drive file as read-only",
611 .name
= "copy-on-read",
612 .type
= QEMU_OPT_BOOL
,
613 .help
= "copy read data from backing file into image file",
616 { /* end of list */ }
620 DriveInfo
*drive_init(QemuOpts
*all_opts
, BlockInterfaceType block_default_type
)
623 DriveInfo
*dinfo
= NULL
;
625 QemuOpts
*legacy_opts
;
626 DriveMediaType media
= MEDIA_DISK
;
627 BlockInterfaceType type
;
628 int cyls
, heads
, secs
, translation
;
629 int max_devs
, bus_id
, unit_id
, index
;
631 bool read_only
= false;
633 Error
*local_err
= NULL
;
635 /* Change legacy command line options into QMP ones */
636 qemu_opt_rename(all_opts
, "iops", "throttling.iops-total");
637 qemu_opt_rename(all_opts
, "iops_rd", "throttling.iops-read");
638 qemu_opt_rename(all_opts
, "iops_wr", "throttling.iops-write");
640 qemu_opt_rename(all_opts
, "bps", "throttling.bps-total");
641 qemu_opt_rename(all_opts
, "bps_rd", "throttling.bps-read");
642 qemu_opt_rename(all_opts
, "bps_wr", "throttling.bps-write");
644 qemu_opt_rename(all_opts
, "iops_max", "throttling.iops-total-max");
645 qemu_opt_rename(all_opts
, "iops_rd_max", "throttling.iops-read-max");
646 qemu_opt_rename(all_opts
, "iops_wr_max", "throttling.iops-write-max");
648 qemu_opt_rename(all_opts
, "bps_max", "throttling.bps-total-max");
649 qemu_opt_rename(all_opts
, "bps_rd_max", "throttling.bps-read-max");
650 qemu_opt_rename(all_opts
, "bps_wr_max", "throttling.bps-write-max");
652 qemu_opt_rename(all_opts
,
653 "iops_size", "throttling.iops-size");
655 qemu_opt_rename(all_opts
, "readonly", "read-only");
657 value
= qemu_opt_get(all_opts
, "cache");
661 if (bdrv_parse_cache_flags(value
, &flags
) != 0) {
662 error_report("invalid cache option");
666 /* Specific options take precedence */
667 if (!qemu_opt_get(all_opts
, "cache.writeback")) {
668 qemu_opt_set_bool(all_opts
, "cache.writeback",
669 !!(flags
& BDRV_O_CACHE_WB
));
671 if (!qemu_opt_get(all_opts
, "cache.direct")) {
672 qemu_opt_set_bool(all_opts
, "cache.direct",
673 !!(flags
& BDRV_O_NOCACHE
));
675 if (!qemu_opt_get(all_opts
, "cache.no-flush")) {
676 qemu_opt_set_bool(all_opts
, "cache.no-flush",
677 !!(flags
& BDRV_O_NO_FLUSH
));
679 qemu_opt_unset(all_opts
, "cache");
682 /* Get a QDict for processing the options */
683 bs_opts
= qdict_new();
684 qemu_opts_to_qdict(all_opts
, bs_opts
);
686 legacy_opts
= qemu_opts_create_nofail(&qemu_legacy_drive_opts
);
687 qemu_opts_absorb_qdict(legacy_opts
, bs_opts
, &local_err
);
688 if (error_is_set(&local_err
)) {
689 qerror_report_err(local_err
);
690 error_free(local_err
);
694 /* Deprecated option boot=[on|off] */
695 if (qemu_opt_get(legacy_opts
, "boot") != NULL
) {
696 fprintf(stderr
, "qemu-kvm: boot=on|off is deprecated and will be "
697 "ignored. Future versions will reject this parameter. Please "
698 "update your scripts.\n");
702 value
= qemu_opt_get(legacy_opts
, "media");
704 if (!strcmp(value
, "disk")) {
706 } else if (!strcmp(value
, "cdrom")) {
710 error_report("'%s' invalid media", value
);
715 /* copy-on-read is disabled with a warning for read-only devices */
716 read_only
|= qemu_opt_get_bool(legacy_opts
, "read-only", false);
717 copy_on_read
= qemu_opt_get_bool(legacy_opts
, "copy-on-read", false);
719 if (read_only
&& copy_on_read
) {
720 error_report("warning: disabling copy-on-read on read-only drive");
721 copy_on_read
= false;
724 qdict_put(bs_opts
, "read-only",
725 qstring_from_str(read_only
? "on" : "off"));
726 qdict_put(bs_opts
, "copy-on-read",
727 qstring_from_str(copy_on_read
? "on" :"off"));
729 /* Controller type */
730 value
= qemu_opt_get(legacy_opts
, "if");
733 type
< IF_COUNT
&& strcmp(value
, if_name
[type
]);
736 if (type
== IF_COUNT
) {
737 error_report("unsupported bus type '%s'", value
);
741 type
= block_default_type
;
745 cyls
= qemu_opt_get_number(legacy_opts
, "cyls", 0);
746 heads
= qemu_opt_get_number(legacy_opts
, "heads", 0);
747 secs
= qemu_opt_get_number(legacy_opts
, "secs", 0);
749 if (cyls
|| heads
|| secs
) {
751 error_report("invalid physical cyls number");
755 error_report("invalid physical heads number");
759 error_report("invalid physical secs number");
764 translation
= BIOS_ATA_TRANSLATION_AUTO
;
765 value
= qemu_opt_get(legacy_opts
, "trans");
768 error_report("'%s' trans must be used with cyls, heads and secs",
772 if (!strcmp(value
, "none")) {
773 translation
= BIOS_ATA_TRANSLATION_NONE
;
774 } else if (!strcmp(value
, "lba")) {
775 translation
= BIOS_ATA_TRANSLATION_LBA
;
776 } else if (!strcmp(value
, "auto")) {
777 translation
= BIOS_ATA_TRANSLATION_AUTO
;
779 error_report("'%s' invalid translation type", value
);
784 if (media
== MEDIA_CDROM
) {
785 if (cyls
|| secs
|| heads
) {
786 error_report("CHS can't be set with media=cdrom");
791 /* Device address specified by bus/unit or index.
792 * If none was specified, try to find the first free one. */
793 bus_id
= qemu_opt_get_number(legacy_opts
, "bus", 0);
794 unit_id
= qemu_opt_get_number(legacy_opts
, "unit", -1);
795 index
= qemu_opt_get_number(legacy_opts
, "index", -1);
797 max_devs
= if_max_devs
[type
];
800 if (bus_id
!= 0 || unit_id
!= -1) {
801 error_report("index cannot be used with bus and unit");
804 bus_id
= drive_index_to_bus_id(type
, index
);
805 unit_id
= drive_index_to_unit_id(type
, index
);
810 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
812 if (max_devs
&& unit_id
>= max_devs
) {
819 if (max_devs
&& unit_id
>= max_devs
) {
820 error_report("unit %d too big (max is %d)", unit_id
, max_devs
- 1);
824 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
825 error_report("drive with bus=%d, unit=%d (index=%d) exists",
826 bus_id
, unit_id
, index
);
830 /* no id supplied -> create one */
831 if (qemu_opts_id(all_opts
) == NULL
) {
833 const char *mediastr
= "";
834 if (type
== IF_IDE
|| type
== IF_SCSI
) {
835 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
838 new_id
= g_strdup_printf("%s%i%s%i", if_name
[type
], bus_id
,
841 new_id
= g_strdup_printf("%s%s%i", if_name
[type
],
844 qdict_put(bs_opts
, "id", qstring_from_str(new_id
));
848 /* Add virtio block device */
849 devaddr
= qemu_opt_get(legacy_opts
, "addr");
850 if (devaddr
&& type
!= IF_VIRTIO
) {
851 error_report("addr is not supported by this bus type");
855 if (type
== IF_VIRTIO
) {
857 devopts
= qemu_opts_create_nofail(qemu_find_opts("device"));
858 if (arch_type
== QEMU_ARCH_S390X
) {
859 qemu_opt_set(devopts
, "driver", "virtio-blk-s390");
861 qemu_opt_set(devopts
, "driver", "virtio-blk-pci");
863 qemu_opt_set(devopts
, "drive", qdict_get_str(bs_opts
, "id"));
865 qemu_opt_set(devopts
, "addr", devaddr
);
869 /* Actual block device init: Functionality shared with blockdev-add */
870 dinfo
= blockdev_init(bs_opts
, type
, &local_err
);
872 if (error_is_set(&local_err
)) {
873 qerror_report_err(local_err
);
874 error_free(local_err
);
878 assert(!error_is_set(&local_err
));
881 /* Set legacy DriveInfo fields */
882 dinfo
->enable_auto_del
= true;
883 dinfo
->opts
= all_opts
;
886 dinfo
->heads
= heads
;
888 dinfo
->trans
= translation
;
891 dinfo
->unit
= unit_id
;
892 dinfo
->devaddr
= devaddr
;
899 dinfo
->media_cd
= media
== MEDIA_CDROM
;
906 qemu_opts_del(legacy_opts
);
910 void do_commit(Monitor
*mon
, const QDict
*qdict
)
912 const char *device
= qdict_get_str(qdict
, "device");
913 BlockDriverState
*bs
;
916 if (!strcmp(device
, "all")) {
917 ret
= bdrv_commit_all();
919 bs
= bdrv_find(device
);
921 monitor_printf(mon
, "Device '%s' not found\n", device
);
924 ret
= bdrv_commit(bs
);
927 monitor_printf(mon
, "'commit' error for '%s': %s\n", device
,
932 static void blockdev_do_action(int kind
, void *data
, Error
**errp
)
934 TransactionAction action
;
935 TransactionActionList list
;
939 list
.value
= &action
;
941 qmp_transaction(&list
, errp
);
944 void qmp_blockdev_snapshot_sync(const char *device
, const char *snapshot_file
,
945 bool has_format
, const char *format
,
946 bool has_mode
, enum NewImageMode mode
,
949 BlockdevSnapshot snapshot
= {
950 .device
= (char *) device
,
951 .snapshot_file
= (char *) snapshot_file
,
952 .has_format
= has_format
,
953 .format
= (char *) format
,
954 .has_mode
= has_mode
,
957 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
,
961 void qmp_blockdev_snapshot_internal_sync(const char *device
,
965 BlockdevSnapshotInternal snapshot
= {
966 .device
= (char *) device
,
967 .name
= (char *) name
970 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
,
974 SnapshotInfo
*qmp_blockdev_snapshot_delete_internal_sync(const char *device
,
981 BlockDriverState
*bs
= bdrv_find(device
);
983 Error
*local_err
= NULL
;
984 SnapshotInfo
*info
= NULL
;
988 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1001 error_setg(errp
, "Name or id must be provided");
1005 ret
= bdrv_snapshot_find_by_id_and_name(bs
, id
, name
, &sn
, &local_err
);
1006 if (error_is_set(&local_err
)) {
1007 error_propagate(errp
, local_err
);
1012 "Snapshot with id '%s' and name '%s' does not exist on "
1014 STR_OR_NULL(id
), STR_OR_NULL(name
), device
);
1018 bdrv_snapshot_delete(bs
, id
, name
, &local_err
);
1019 if (error_is_set(&local_err
)) {
1020 error_propagate(errp
, local_err
);
1024 info
= g_malloc0(sizeof(SnapshotInfo
));
1025 info
->id
= g_strdup(sn
.id_str
);
1026 info
->name
= g_strdup(sn
.name
);
1027 info
->date_nsec
= sn
.date_nsec
;
1028 info
->date_sec
= sn
.date_sec
;
1029 info
->vm_state_size
= sn
.vm_state_size
;
1030 info
->vm_clock_nsec
= sn
.vm_clock_nsec
% 1000000000;
1031 info
->vm_clock_sec
= sn
.vm_clock_nsec
/ 1000000000;
1036 /* New and old BlockDriverState structs for group snapshots */
1038 typedef struct BlkTransactionState BlkTransactionState
;
1040 /* Only prepare() may fail. In a single transaction, only one of commit() or
1041 abort() will be called, clean() will always be called if it present. */
1042 typedef struct BdrvActionOps
{
1043 /* Size of state struct, in bytes. */
1044 size_t instance_size
;
1045 /* Prepare the work, must NOT be NULL. */
1046 void (*prepare
)(BlkTransactionState
*common
, Error
**errp
);
1047 /* Commit the changes, can be NULL. */
1048 void (*commit
)(BlkTransactionState
*common
);
1049 /* Abort the changes on fail, can be NULL. */
1050 void (*abort
)(BlkTransactionState
*common
);
1051 /* Clean up resource in the end, can be NULL. */
1052 void (*clean
)(BlkTransactionState
*common
);
1056 * This structure must be arranged as first member in child type, assuming
1057 * that compiler will also arrange it to the same address with parent instance.
1058 * Later it will be used in free().
1060 struct BlkTransactionState
{
1061 TransactionAction
*action
;
1062 const BdrvActionOps
*ops
;
1063 QSIMPLEQ_ENTRY(BlkTransactionState
) entry
;
1066 /* internal snapshot private data */
1067 typedef struct InternalSnapshotState
{
1068 BlkTransactionState common
;
1069 BlockDriverState
*bs
;
1070 QEMUSnapshotInfo sn
;
1071 } InternalSnapshotState
;
1073 static void internal_snapshot_prepare(BlkTransactionState
*common
,
1078 BlockDriverState
*bs
;
1079 QEMUSnapshotInfo old_sn
, *sn
;
1082 BlockdevSnapshotInternal
*internal
;
1083 InternalSnapshotState
*state
;
1086 g_assert(common
->action
->kind
==
1087 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
);
1088 internal
= common
->action
->blockdev_snapshot_internal_sync
;
1089 state
= DO_UPCAST(InternalSnapshotState
, common
, common
);
1091 /* 1. parse input */
1092 device
= internal
->device
;
1093 name
= internal
->name
;
1095 /* 2. check for validation */
1096 bs
= bdrv_find(device
);
1098 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1102 if (!bdrv_is_inserted(bs
)) {
1103 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1107 if (bdrv_is_read_only(bs
)) {
1108 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1112 if (!bdrv_can_snapshot(bs
)) {
1113 error_set(errp
, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED
,
1114 bs
->drv
->format_name
, device
, "internal snapshot");
1118 if (!strlen(name
)) {
1119 error_setg(errp
, "Name is empty");
1123 /* check whether a snapshot with name exist */
1124 ret
= bdrv_snapshot_find_by_id_and_name(bs
, NULL
, name
, &old_sn
, errp
);
1125 if (error_is_set(errp
)) {
1129 "Snapshot with name '%s' already exists on device '%s'",
1134 /* 3. take the snapshot */
1136 pstrcpy(sn
->name
, sizeof(sn
->name
), name
);
1137 qemu_gettimeofday(&tv
);
1138 sn
->date_sec
= tv
.tv_sec
;
1139 sn
->date_nsec
= tv
.tv_usec
* 1000;
1140 sn
->vm_clock_nsec
= qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL
);
1142 ret1
= bdrv_snapshot_create(bs
, sn
);
1144 error_setg_errno(errp
, -ret1
,
1145 "Failed to create snapshot '%s' on device '%s'",
1150 /* 4. succeed, mark a snapshot is created */
1154 static void internal_snapshot_abort(BlkTransactionState
*common
)
1156 InternalSnapshotState
*state
=
1157 DO_UPCAST(InternalSnapshotState
, common
, common
);
1158 BlockDriverState
*bs
= state
->bs
;
1159 QEMUSnapshotInfo
*sn
= &state
->sn
;
1160 Error
*local_error
= NULL
;
1166 if (bdrv_snapshot_delete(bs
, sn
->id_str
, sn
->name
, &local_error
) < 0) {
1167 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1168 "device '%s' in abort: %s",
1171 bdrv_get_device_name(bs
),
1172 error_get_pretty(local_error
));
1173 error_free(local_error
);
1177 /* external snapshot private data */
1178 typedef struct ExternalSnapshotState
{
1179 BlkTransactionState common
;
1180 BlockDriverState
*old_bs
;
1181 BlockDriverState
*new_bs
;
1182 } ExternalSnapshotState
;
1184 static void external_snapshot_prepare(BlkTransactionState
*common
,
1189 Error
*local_err
= NULL
;
1191 const char *new_image_file
;
1192 const char *format
= "qcow2";
1193 enum NewImageMode mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1194 ExternalSnapshotState
*state
=
1195 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1196 TransactionAction
*action
= common
->action
;
1198 /* get parameters */
1199 g_assert(action
->kind
== TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
);
1201 device
= action
->blockdev_snapshot_sync
->device
;
1202 new_image_file
= action
->blockdev_snapshot_sync
->snapshot_file
;
1203 if (action
->blockdev_snapshot_sync
->has_format
) {
1204 format
= action
->blockdev_snapshot_sync
->format
;
1206 if (action
->blockdev_snapshot_sync
->has_mode
) {
1207 mode
= action
->blockdev_snapshot_sync
->mode
;
1210 /* start processing */
1211 drv
= bdrv_find_format(format
);
1213 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1217 state
->old_bs
= bdrv_find(device
);
1218 if (!state
->old_bs
) {
1219 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1223 if (!bdrv_is_inserted(state
->old_bs
)) {
1224 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1228 if (bdrv_in_use(state
->old_bs
)) {
1229 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1233 if (!bdrv_is_read_only(state
->old_bs
)) {
1234 if (bdrv_flush(state
->old_bs
)) {
1235 error_set(errp
, QERR_IO_ERROR
);
1240 if (bdrv_check_ext_snapshot(state
->old_bs
) != EXT_SNAPSHOT_ALLOWED
) {
1241 error_set(errp
, QERR_FEATURE_DISABLED
, "snapshot");
1245 flags
= state
->old_bs
->open_flags
;
1247 /* create new image w/backing file */
1248 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1249 bdrv_img_create(new_image_file
, format
,
1250 state
->old_bs
->filename
,
1251 state
->old_bs
->drv
->format_name
,
1252 NULL
, -1, flags
, &local_err
, false);
1253 if (error_is_set(&local_err
)) {
1254 error_propagate(errp
, local_err
);
1259 /* We will manually add the backing_hd field to the bs later */
1260 state
->new_bs
= bdrv_new("");
1261 /* TODO Inherit bs->options or only take explicit options with an
1262 * extended QMP command? */
1263 ret
= bdrv_open(state
->new_bs
, new_image_file
, NULL
,
1264 flags
| BDRV_O_NO_BACKING
, drv
, &local_err
);
1266 error_propagate(errp
, local_err
);
1270 static void external_snapshot_commit(BlkTransactionState
*common
)
1272 ExternalSnapshotState
*state
=
1273 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1275 /* This removes our old bs and adds the new bs */
1276 bdrv_append(state
->new_bs
, state
->old_bs
);
1277 /* We don't need (or want) to use the transactional
1278 * bdrv_reopen_multiple() across all the entries at once, because we
1279 * don't want to abort all of them if one of them fails the reopen */
1280 bdrv_reopen(state
->new_bs
, state
->new_bs
->open_flags
& ~BDRV_O_RDWR
,
1284 static void external_snapshot_abort(BlkTransactionState
*common
)
1286 ExternalSnapshotState
*state
=
1287 DO_UPCAST(ExternalSnapshotState
, common
, common
);
1288 if (state
->new_bs
) {
1289 bdrv_unref(state
->new_bs
);
1293 typedef struct DriveBackupState
{
1294 BlkTransactionState common
;
1295 BlockDriverState
*bs
;
1299 static void drive_backup_prepare(BlkTransactionState
*common
, Error
**errp
)
1301 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1302 DriveBackup
*backup
;
1303 Error
*local_err
= NULL
;
1305 assert(common
->action
->kind
== TRANSACTION_ACTION_KIND_DRIVE_BACKUP
);
1306 backup
= common
->action
->drive_backup
;
1308 qmp_drive_backup(backup
->device
, backup
->target
,
1309 backup
->has_format
, backup
->format
,
1311 backup
->has_mode
, backup
->mode
,
1312 backup
->has_speed
, backup
->speed
,
1313 backup
->has_on_source_error
, backup
->on_source_error
,
1314 backup
->has_on_target_error
, backup
->on_target_error
,
1316 if (error_is_set(&local_err
)) {
1317 error_propagate(errp
, local_err
);
1323 state
->bs
= bdrv_find(backup
->device
);
1324 state
->job
= state
->bs
->job
;
1327 static void drive_backup_abort(BlkTransactionState
*common
)
1329 DriveBackupState
*state
= DO_UPCAST(DriveBackupState
, common
, common
);
1330 BlockDriverState
*bs
= state
->bs
;
1332 /* Only cancel if it's the job we started */
1333 if (bs
&& bs
->job
&& bs
->job
== state
->job
) {
1334 block_job_cancel_sync(bs
->job
);
1338 static void abort_prepare(BlkTransactionState
*common
, Error
**errp
)
1340 error_setg(errp
, "Transaction aborted using Abort action");
1343 static void abort_commit(BlkTransactionState
*common
)
1345 g_assert_not_reached(); /* this action never succeeds */
1348 static const BdrvActionOps actions
[] = {
1349 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
] = {
1350 .instance_size
= sizeof(ExternalSnapshotState
),
1351 .prepare
= external_snapshot_prepare
,
1352 .commit
= external_snapshot_commit
,
1353 .abort
= external_snapshot_abort
,
1355 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP
] = {
1356 .instance_size
= sizeof(DriveBackupState
),
1357 .prepare
= drive_backup_prepare
,
1358 .abort
= drive_backup_abort
,
1360 [TRANSACTION_ACTION_KIND_ABORT
] = {
1361 .instance_size
= sizeof(BlkTransactionState
),
1362 .prepare
= abort_prepare
,
1363 .commit
= abort_commit
,
1365 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC
] = {
1366 .instance_size
= sizeof(InternalSnapshotState
),
1367 .prepare
= internal_snapshot_prepare
,
1368 .abort
= internal_snapshot_abort
,
1373 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1374 * then we do not pivot any of the devices in the group, and abandon the
1377 void qmp_transaction(TransactionActionList
*dev_list
, Error
**errp
)
1379 TransactionActionList
*dev_entry
= dev_list
;
1380 BlkTransactionState
*state
, *next
;
1381 Error
*local_err
= NULL
;
1383 QSIMPLEQ_HEAD(snap_bdrv_states
, BlkTransactionState
) snap_bdrv_states
;
1384 QSIMPLEQ_INIT(&snap_bdrv_states
);
1386 /* drain all i/o before any snapshots */
1389 /* We don't do anything in this loop that commits us to the snapshot */
1390 while (NULL
!= dev_entry
) {
1391 TransactionAction
*dev_info
= NULL
;
1392 const BdrvActionOps
*ops
;
1394 dev_info
= dev_entry
->value
;
1395 dev_entry
= dev_entry
->next
;
1397 assert(dev_info
->kind
< ARRAY_SIZE(actions
));
1399 ops
= &actions
[dev_info
->kind
];
1400 assert(ops
->instance_size
> 0);
1402 state
= g_malloc0(ops
->instance_size
);
1404 state
->action
= dev_info
;
1405 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states
, state
, entry
);
1407 state
->ops
->prepare(state
, &local_err
);
1408 if (error_is_set(&local_err
)) {
1409 error_propagate(errp
, local_err
);
1410 goto delete_and_fail
;
1414 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1415 if (state
->ops
->commit
) {
1416 state
->ops
->commit(state
);
1425 * failure, and it is all-or-none; abandon each new bs, and keep using
1426 * the original bs for all images
1428 QSIMPLEQ_FOREACH(state
, &snap_bdrv_states
, entry
) {
1429 if (state
->ops
->abort
) {
1430 state
->ops
->abort(state
);
1434 QSIMPLEQ_FOREACH_SAFE(state
, &snap_bdrv_states
, entry
, next
) {
1435 if (state
->ops
->clean
) {
1436 state
->ops
->clean(state
);
1443 static void eject_device(BlockDriverState
*bs
, int force
, Error
**errp
)
1445 if (bdrv_in_use(bs
)) {
1446 error_set(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
1449 if (!bdrv_dev_has_removable_media(bs
)) {
1450 error_set(errp
, QERR_DEVICE_NOT_REMOVABLE
, bdrv_get_device_name(bs
));
1454 if (bdrv_dev_is_medium_locked(bs
) && !bdrv_dev_is_tray_open(bs
)) {
1455 bdrv_dev_eject_request(bs
, force
);
1457 error_set(errp
, QERR_DEVICE_LOCKED
, bdrv_get_device_name(bs
));
1465 void qmp_eject(const char *device
, bool has_force
, bool force
, Error
**errp
)
1467 BlockDriverState
*bs
;
1469 bs
= bdrv_find(device
);
1471 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1475 eject_device(bs
, force
, errp
);
1478 void qmp_block_passwd(const char *device
, const char *password
, Error
**errp
)
1480 BlockDriverState
*bs
;
1483 bs
= bdrv_find(device
);
1485 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1489 err
= bdrv_set_key(bs
, password
);
1490 if (err
== -EINVAL
) {
1491 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1493 } else if (err
< 0) {
1494 error_set(errp
, QERR_INVALID_PASSWORD
);
1499 static void qmp_bdrv_open_encrypted(BlockDriverState
*bs
, const char *filename
,
1500 int bdrv_flags
, BlockDriver
*drv
,
1501 const char *password
, Error
**errp
)
1503 Error
*local_err
= NULL
;
1506 ret
= bdrv_open(bs
, filename
, NULL
, bdrv_flags
, drv
, &local_err
);
1508 error_propagate(errp
, local_err
);
1512 if (bdrv_key_required(bs
)) {
1514 if (bdrv_set_key(bs
, password
) < 0) {
1515 error_set(errp
, QERR_INVALID_PASSWORD
);
1518 error_set(errp
, QERR_DEVICE_ENCRYPTED
, bdrv_get_device_name(bs
),
1519 bdrv_get_encrypted_filename(bs
));
1521 } else if (password
) {
1522 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
1526 void qmp_change_blockdev(const char *device
, const char *filename
,
1527 bool has_format
, const char *format
, Error
**errp
)
1529 BlockDriverState
*bs
;
1530 BlockDriver
*drv
= NULL
;
1534 bs
= bdrv_find(device
);
1536 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1541 drv
= bdrv_find_whitelisted_format(format
, bs
->read_only
);
1543 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1548 eject_device(bs
, 0, &err
);
1549 if (error_is_set(&err
)) {
1550 error_propagate(errp
, err
);
1554 bdrv_flags
= bdrv_is_read_only(bs
) ? 0 : BDRV_O_RDWR
;
1555 bdrv_flags
|= bdrv_is_snapshot(bs
) ? BDRV_O_SNAPSHOT
: 0;
1557 qmp_bdrv_open_encrypted(bs
, filename
, bdrv_flags
, drv
, NULL
, errp
);
1560 /* throttling disk I/O limits */
1561 void qmp_block_set_io_throttle(const char *device
, int64_t bps
, int64_t bps_rd
,
1568 bool has_bps_rd_max
,
1570 bool has_bps_wr_max
,
1574 bool has_iops_rd_max
,
1575 int64_t iops_rd_max
,
1576 bool has_iops_wr_max
,
1577 int64_t iops_wr_max
,
1579 int64_t iops_size
, Error
**errp
)
1582 BlockDriverState
*bs
;
1584 bs
= bdrv_find(device
);
1586 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1590 memset(&cfg
, 0, sizeof(cfg
));
1591 cfg
.buckets
[THROTTLE_BPS_TOTAL
].avg
= bps
;
1592 cfg
.buckets
[THROTTLE_BPS_READ
].avg
= bps_rd
;
1593 cfg
.buckets
[THROTTLE_BPS_WRITE
].avg
= bps_wr
;
1595 cfg
.buckets
[THROTTLE_OPS_TOTAL
].avg
= iops
;
1596 cfg
.buckets
[THROTTLE_OPS_READ
].avg
= iops_rd
;
1597 cfg
.buckets
[THROTTLE_OPS_WRITE
].avg
= iops_wr
;
1600 cfg
.buckets
[THROTTLE_BPS_TOTAL
].max
= bps_max
;
1602 if (has_bps_rd_max
) {
1603 cfg
.buckets
[THROTTLE_BPS_READ
].max
= bps_rd_max
;
1605 if (has_bps_wr_max
) {
1606 cfg
.buckets
[THROTTLE_BPS_WRITE
].max
= bps_wr_max
;
1609 cfg
.buckets
[THROTTLE_OPS_TOTAL
].max
= iops_max
;
1611 if (has_iops_rd_max
) {
1612 cfg
.buckets
[THROTTLE_OPS_READ
].max
= iops_rd_max
;
1614 if (has_iops_wr_max
) {
1615 cfg
.buckets
[THROTTLE_OPS_WRITE
].max
= iops_wr_max
;
1618 if (has_iops_size
) {
1619 cfg
.op_size
= iops_size
;
1622 if (!check_throttle_config(&cfg
, errp
)) {
1626 if (!bs
->io_limits_enabled
&& throttle_enabled(&cfg
)) {
1627 bdrv_io_limits_enable(bs
);
1628 } else if (bs
->io_limits_enabled
&& !throttle_enabled(&cfg
)) {
1629 bdrv_io_limits_disable(bs
);
1632 if (bs
->io_limits_enabled
) {
1633 bdrv_set_io_limits(bs
, &cfg
);
1637 int do_drive_del(Monitor
*mon
, const QDict
*qdict
, QObject
**ret_data
)
1639 const char *id
= qdict_get_str(qdict
, "id");
1640 BlockDriverState
*bs
;
1644 qerror_report(QERR_DEVICE_NOT_FOUND
, id
);
1647 if (bdrv_in_use(bs
)) {
1648 qerror_report(QERR_DEVICE_IN_USE
, id
);
1652 /* quiesce block driver; prevent further io */
1657 /* if we have a device attached to this BlockDriverState
1658 * then we need to make the drive anonymous until the device
1659 * can be removed. If this is a drive with no device backing
1660 * then we can just get rid of the block driver state right here.
1662 if (bdrv_get_attached_dev(bs
)) {
1665 /* Further I/O must not pause the guest */
1666 bdrv_set_on_error(bs
, BLOCKDEV_ON_ERROR_REPORT
,
1667 BLOCKDEV_ON_ERROR_REPORT
);
1669 drive_uninit(drive_get_by_blockdev(bs
));
1675 void qmp_block_resize(const char *device
, int64_t size
, Error
**errp
)
1677 BlockDriverState
*bs
;
1680 bs
= bdrv_find(device
);
1682 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1687 error_set(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
1691 /* complete all in-flight operations before resizing the device */
1694 ret
= bdrv_truncate(bs
, size
);
1699 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1702 error_set(errp
, QERR_UNSUPPORTED
);
1705 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1708 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1711 error_setg_errno(errp
, -ret
, "Could not resize");
1716 static void block_job_cb(void *opaque
, int ret
)
1718 BlockDriverState
*bs
= opaque
;
1721 trace_block_job_cb(bs
, bs
->job
, ret
);
1724 obj
= qobject_from_block_job(bs
->job
);
1726 QDict
*dict
= qobject_to_qdict(obj
);
1727 qdict_put(dict
, "error", qstring_from_str(strerror(-ret
)));
1730 if (block_job_is_cancelled(bs
->job
)) {
1731 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED
, obj
);
1733 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED
, obj
);
1735 qobject_decref(obj
);
1737 bdrv_put_ref_bh_schedule(bs
);
1740 void qmp_block_stream(const char *device
, bool has_base
,
1741 const char *base
, bool has_speed
, int64_t speed
,
1742 bool has_on_error
, BlockdevOnError on_error
,
1745 BlockDriverState
*bs
;
1746 BlockDriverState
*base_bs
= NULL
;
1747 Error
*local_err
= NULL
;
1749 if (!has_on_error
) {
1750 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1753 bs
= bdrv_find(device
);
1755 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1760 base_bs
= bdrv_find_backing_image(bs
, base
);
1761 if (base_bs
== NULL
) {
1762 error_set(errp
, QERR_BASE_NOT_FOUND
, base
);
1767 stream_start(bs
, base_bs
, base
, has_speed
? speed
: 0,
1768 on_error
, block_job_cb
, bs
, &local_err
);
1769 if (error_is_set(&local_err
)) {
1770 error_propagate(errp
, local_err
);
1774 trace_qmp_block_stream(bs
, bs
->job
);
1777 void qmp_block_commit(const char *device
,
1778 bool has_base
, const char *base
, const char *top
,
1779 bool has_speed
, int64_t speed
,
1782 BlockDriverState
*bs
;
1783 BlockDriverState
*base_bs
, *top_bs
;
1784 Error
*local_err
= NULL
;
1785 /* This will be part of the QMP command, if/when the
1786 * BlockdevOnError change for blkmirror makes it in
1788 BlockdevOnError on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1790 /* drain all i/o before commits */
1793 bs
= bdrv_find(device
);
1795 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1799 /* default top_bs is the active layer */
1803 if (strcmp(bs
->filename
, top
) != 0) {
1804 top_bs
= bdrv_find_backing_image(bs
, top
);
1808 if (top_bs
== NULL
) {
1809 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
1813 if (has_base
&& base
) {
1814 base_bs
= bdrv_find_backing_image(top_bs
, base
);
1816 base_bs
= bdrv_find_base(top_bs
);
1819 if (base_bs
== NULL
) {
1820 error_set(errp
, QERR_BASE_NOT_FOUND
, base
? base
: "NULL");
1824 commit_start(bs
, base_bs
, top_bs
, speed
, on_error
, block_job_cb
, bs
,
1826 if (local_err
!= NULL
) {
1827 error_propagate(errp
, local_err
);
1832 void qmp_drive_backup(const char *device
, const char *target
,
1833 bool has_format
, const char *format
,
1834 enum MirrorSyncMode sync
,
1835 bool has_mode
, enum NewImageMode mode
,
1836 bool has_speed
, int64_t speed
,
1837 bool has_on_source_error
, BlockdevOnError on_source_error
,
1838 bool has_on_target_error
, BlockdevOnError on_target_error
,
1841 BlockDriverState
*bs
;
1842 BlockDriverState
*target_bs
;
1843 BlockDriverState
*source
= NULL
;
1844 BlockDriver
*drv
= NULL
;
1845 Error
*local_err
= NULL
;
1853 if (!has_on_source_error
) {
1854 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1856 if (!has_on_target_error
) {
1857 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1860 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1863 bs
= bdrv_find(device
);
1865 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1869 if (!bdrv_is_inserted(bs
)) {
1870 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1875 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1878 drv
= bdrv_find_format(format
);
1880 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1885 if (bdrv_in_use(bs
)) {
1886 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1890 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1892 /* See if we have a backing HD we can use to create our new image
1894 if (sync
== MIRROR_SYNC_MODE_TOP
) {
1895 source
= bs
->backing_hd
;
1897 sync
= MIRROR_SYNC_MODE_FULL
;
1900 if (sync
== MIRROR_SYNC_MODE_NONE
) {
1904 size
= bdrv_getlength(bs
);
1906 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
1910 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
1911 assert(format
&& drv
);
1913 bdrv_img_create(target
, format
, source
->filename
,
1914 source
->drv
->format_name
, NULL
,
1915 size
, flags
, &local_err
, false);
1917 bdrv_img_create(target
, format
, NULL
, NULL
, NULL
,
1918 size
, flags
, &local_err
, false);
1922 if (error_is_set(&local_err
)) {
1923 error_propagate(errp
, local_err
);
1927 target_bs
= bdrv_new("");
1928 ret
= bdrv_open(target_bs
, target
, NULL
, flags
, drv
, &local_err
);
1930 bdrv_unref(target_bs
);
1931 error_propagate(errp
, local_err
);
1935 backup_start(bs
, target_bs
, speed
, sync
, on_source_error
, on_target_error
,
1936 block_job_cb
, bs
, &local_err
);
1937 if (local_err
!= NULL
) {
1938 bdrv_unref(target_bs
);
1939 error_propagate(errp
, local_err
);
1944 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1946 void qmp_drive_mirror(const char *device
, const char *target
,
1947 bool has_format
, const char *format
,
1948 enum MirrorSyncMode sync
,
1949 bool has_mode
, enum NewImageMode mode
,
1950 bool has_speed
, int64_t speed
,
1951 bool has_granularity
, uint32_t granularity
,
1952 bool has_buf_size
, int64_t buf_size
,
1953 bool has_on_source_error
, BlockdevOnError on_source_error
,
1954 bool has_on_target_error
, BlockdevOnError on_target_error
,
1957 BlockDriverState
*bs
;
1958 BlockDriverState
*source
, *target_bs
;
1959 BlockDriver
*drv
= NULL
;
1960 Error
*local_err
= NULL
;
1968 if (!has_on_source_error
) {
1969 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1971 if (!has_on_target_error
) {
1972 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1975 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1977 if (!has_granularity
) {
1980 if (!has_buf_size
) {
1981 buf_size
= DEFAULT_MIRROR_BUF_SIZE
;
1984 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
1985 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1988 if (granularity
& (granularity
- 1)) {
1989 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1993 bs
= bdrv_find(device
);
1995 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1999 if (!bdrv_is_inserted(bs
)) {
2000 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
2005 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
2008 drv
= bdrv_find_format(format
);
2010 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
2015 if (bdrv_in_use(bs
)) {
2016 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
2020 flags
= bs
->open_flags
| BDRV_O_RDWR
;
2021 source
= bs
->backing_hd
;
2022 if (!source
&& sync
== MIRROR_SYNC_MODE_TOP
) {
2023 sync
= MIRROR_SYNC_MODE_FULL
;
2026 size
= bdrv_getlength(bs
);
2028 error_setg_errno(errp
, -size
, "bdrv_getlength failed");
2032 if (sync
== MIRROR_SYNC_MODE_FULL
&& mode
!= NEW_IMAGE_MODE_EXISTING
) {
2033 /* create new image w/o backing file */
2034 assert(format
&& drv
);
2035 bdrv_img_create(target
, format
,
2036 NULL
, NULL
, NULL
, size
, flags
, &local_err
, false);
2039 case NEW_IMAGE_MODE_EXISTING
:
2041 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
2042 /* create new image with backing file */
2043 bdrv_img_create(target
, format
,
2045 source
->drv
->format_name
,
2046 NULL
, size
, flags
, &local_err
, false);
2053 if (error_is_set(&local_err
)) {
2054 error_propagate(errp
, local_err
);
2058 /* Mirroring takes care of copy-on-write using the source's backing
2061 target_bs
= bdrv_new("");
2062 ret
= bdrv_open(target_bs
, target
, NULL
, flags
| BDRV_O_NO_BACKING
, drv
,
2065 bdrv_unref(target_bs
);
2066 error_propagate(errp
, local_err
);
2070 mirror_start(bs
, target_bs
, speed
, granularity
, buf_size
, sync
,
2071 on_source_error
, on_target_error
,
2072 block_job_cb
, bs
, &local_err
);
2073 if (local_err
!= NULL
) {
2074 bdrv_unref(target_bs
);
2075 error_propagate(errp
, local_err
);
2080 static BlockJob
*find_block_job(const char *device
)
2082 BlockDriverState
*bs
;
2084 bs
= bdrv_find(device
);
2085 if (!bs
|| !bs
->job
) {
2091 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
2093 BlockJob
*job
= find_block_job(device
);
2096 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2100 block_job_set_speed(job
, speed
, errp
);
2103 void qmp_block_job_cancel(const char *device
,
2104 bool has_force
, bool force
, Error
**errp
)
2106 BlockJob
*job
= find_block_job(device
);
2113 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2116 if (job
->paused
&& !force
) {
2117 error_set(errp
, QERR_BLOCK_JOB_PAUSED
, device
);
2121 trace_qmp_block_job_cancel(job
);
2122 block_job_cancel(job
);
2125 void qmp_block_job_pause(const char *device
, Error
**errp
)
2127 BlockJob
*job
= find_block_job(device
);
2130 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2134 trace_qmp_block_job_pause(job
);
2135 block_job_pause(job
);
2138 void qmp_block_job_resume(const char *device
, Error
**errp
)
2140 BlockJob
*job
= find_block_job(device
);
2143 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2147 trace_qmp_block_job_resume(job
);
2148 block_job_resume(job
);
2151 void qmp_block_job_complete(const char *device
, Error
**errp
)
2153 BlockJob
*job
= find_block_job(device
);
2156 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
2160 trace_qmp_block_job_complete(job
);
2161 block_job_complete(job
, errp
);
2164 void qmp_blockdev_add(BlockdevOptions
*options
, Error
**errp
)
2166 QmpOutputVisitor
*ov
= qmp_output_visitor_new();
2169 Error
*local_err
= NULL
;
2171 /* Require an ID in the top level */
2172 if (!options
->has_id
) {
2173 error_setg(errp
, "Block device needs an ID");
2177 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2178 * cache.direct=false instead of silently switching to aio=threads, except
2179 * if called from drive_init.
2181 * For now, simply forbidding the combination for all drivers will do. */
2182 if (options
->has_aio
&& options
->aio
== BLOCKDEV_AIO_OPTIONS_NATIVE
) {
2183 bool direct
= options
->cache
->has_direct
&& options
->cache
->direct
;
2184 if (!options
->has_cache
&& !direct
) {
2185 error_setg(errp
, "aio=native requires cache.direct=true");
2190 visit_type_BlockdevOptions(qmp_output_get_visitor(ov
),
2191 &options
, NULL
, &local_err
);
2192 if (error_is_set(&local_err
)) {
2193 error_propagate(errp
, local_err
);
2197 obj
= qmp_output_get_qobject(ov
);
2198 qdict
= qobject_to_qdict(obj
);
2200 qdict_flatten(qdict
);
2202 blockdev_init(qdict
, IF_NONE
, &local_err
);
2203 if (error_is_set(&local_err
)) {
2204 error_propagate(errp
, local_err
);
2209 qmp_output_visitor_cleanup(ov
);
2212 static void do_qmp_query_block_jobs_one(void *opaque
, BlockDriverState
*bs
)
2214 BlockJobInfoList
**prev
= opaque
;
2215 BlockJob
*job
= bs
->job
;
2218 BlockJobInfoList
*elem
= g_new0(BlockJobInfoList
, 1);
2219 elem
->value
= block_job_query(bs
->job
);
2220 (*prev
)->next
= elem
;
2225 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
2227 /* Dummy is a fake list element for holding the head pointer */
2228 BlockJobInfoList dummy
= {};
2229 BlockJobInfoList
*prev
= &dummy
;
2230 bdrv_iterate(do_qmp_query_block_jobs_one
, &prev
);
2234 QemuOptsList qemu_common_drive_opts
= {
2236 .head
= QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts
.head
),
2240 .type
= QEMU_OPT_BOOL
,
2241 .help
= "enable/disable snapshot mode",
2244 .type
= QEMU_OPT_STRING
,
2245 .help
= "disk image",
2248 .type
= QEMU_OPT_STRING
,
2249 .help
= "discard operation (ignore/off, unmap/on)",
2251 .name
= "cache.writeback",
2252 .type
= QEMU_OPT_BOOL
,
2253 .help
= "enables writeback mode for any caches",
2255 .name
= "cache.direct",
2256 .type
= QEMU_OPT_BOOL
,
2257 .help
= "enables use of O_DIRECT (bypass the host page cache)",
2259 .name
= "cache.no-flush",
2260 .type
= QEMU_OPT_BOOL
,
2261 .help
= "ignore any flush requests for the device",
2264 .type
= QEMU_OPT_STRING
,
2265 .help
= "host AIO implementation (threads, native)",
2268 .type
= QEMU_OPT_STRING
,
2269 .help
= "disk format (raw, qcow2, ...)",
2272 .type
= QEMU_OPT_STRING
,
2273 .help
= "disk serial number",
2276 .type
= QEMU_OPT_STRING
,
2277 .help
= "read error action",
2280 .type
= QEMU_OPT_STRING
,
2281 .help
= "write error action",
2283 .name
= "read-only",
2284 .type
= QEMU_OPT_BOOL
,
2285 .help
= "open drive file as read-only",
2287 .name
= "throttling.iops-total",
2288 .type
= QEMU_OPT_NUMBER
,
2289 .help
= "limit total I/O operations per second",
2291 .name
= "throttling.iops-read",
2292 .type
= QEMU_OPT_NUMBER
,
2293 .help
= "limit read operations per second",
2295 .name
= "throttling.iops-write",
2296 .type
= QEMU_OPT_NUMBER
,
2297 .help
= "limit write operations per second",
2299 .name
= "throttling.bps-total",
2300 .type
= QEMU_OPT_NUMBER
,
2301 .help
= "limit total bytes per second",
2303 .name
= "throttling.bps-read",
2304 .type
= QEMU_OPT_NUMBER
,
2305 .help
= "limit read bytes per second",
2307 .name
= "throttling.bps-write",
2308 .type
= QEMU_OPT_NUMBER
,
2309 .help
= "limit write bytes per second",
2311 .name
= "throttling.iops-total-max",
2312 .type
= QEMU_OPT_NUMBER
,
2313 .help
= "I/O operations burst",
2315 .name
= "throttling.iops-read-max",
2316 .type
= QEMU_OPT_NUMBER
,
2317 .help
= "I/O operations read burst",
2319 .name
= "throttling.iops-write-max",
2320 .type
= QEMU_OPT_NUMBER
,
2321 .help
= "I/O operations write burst",
2323 .name
= "throttling.bps-total-max",
2324 .type
= QEMU_OPT_NUMBER
,
2325 .help
= "total bytes burst",
2327 .name
= "throttling.bps-read-max",
2328 .type
= QEMU_OPT_NUMBER
,
2329 .help
= "total bytes read burst",
2331 .name
= "throttling.bps-write-max",
2332 .type
= QEMU_OPT_NUMBER
,
2333 .help
= "total bytes write burst",
2335 .name
= "throttling.iops-size",
2336 .type
= QEMU_OPT_NUMBER
,
2337 .help
= "when limiting by iops max size of an I/O in bytes",
2339 .name
= "copy-on-read",
2340 .type
= QEMU_OPT_BOOL
,
2341 .help
= "copy read data from backing file into image file",
2343 { /* end of list */ }
2347 QemuOptsList qemu_drive_opts
= {
2349 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
2352 * no elements => accept any params
2353 * validation will happen later
2355 { /* end of list */ }