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.
10 #include "sysemu/blockdev.h"
11 #include "hw/block-common.h"
12 #include "block/blockjob.h"
13 #include "monitor/monitor.h"
14 #include "qapi/qmp/qerror.h"
15 #include "qemu/option.h"
16 #include "qemu/config-file.h"
17 #include "qapi/qmp/types.h"
18 #include "sysemu/sysemu.h"
19 #include "block/block_int.h"
20 #include "qmp-commands.h"
22 #include "sysemu/arch_init.h"
24 static QTAILQ_HEAD(drivelist
, DriveInfo
) drives
= QTAILQ_HEAD_INITIALIZER(drives
);
26 static const char *const if_name
[IF_COUNT
] = {
30 [IF_FLOPPY
] = "floppy",
31 [IF_PFLASH
] = "pflash",
34 [IF_VIRTIO
] = "virtio",
38 static const int if_max_devs
[IF_COUNT
] = {
40 * Do not change these numbers! They govern how drive option
41 * index maps to unit and bus. That mapping is ABI.
43 * All controllers used to imlement if=T drives need to support
44 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
45 * Otherwise, some index values map to "impossible" bus, unit
48 * For instance, if you change [IF_SCSI] to 255, -drive
49 * if=scsi,index=12 no longer means bus=1,unit=5, but
50 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
51 * the drive can't be set up. Regression.
58 * We automatically delete the drive when a device using it gets
59 * unplugged. Questionable feature, but we can't just drop it.
60 * Device models call blockdev_mark_auto_del() to schedule the
61 * automatic deletion, and generic qdev code calls blockdev_auto_del()
62 * when deletion is actually safe.
64 void blockdev_mark_auto_del(BlockDriverState
*bs
)
66 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
69 block_job_cancel(bs
->job
);
76 void blockdev_auto_del(BlockDriverState
*bs
)
78 DriveInfo
*dinfo
= drive_get_by_blockdev(bs
);
80 if (dinfo
&& dinfo
->auto_del
) {
85 static int drive_index_to_bus_id(BlockInterfaceType type
, int index
)
87 int max_devs
= if_max_devs
[type
];
88 return max_devs
? index
/ max_devs
: 0;
91 static int drive_index_to_unit_id(BlockInterfaceType type
, int index
)
93 int max_devs
= if_max_devs
[type
];
94 return max_devs
? index
% max_devs
: index
;
97 QemuOpts
*drive_def(const char *optstr
)
99 return qemu_opts_parse(qemu_find_opts("drive"), optstr
, 0);
102 QemuOpts
*drive_add(BlockInterfaceType type
, int index
, const char *file
,
108 opts
= drive_def(optstr
);
112 if (type
!= IF_DEFAULT
) {
113 qemu_opt_set(opts
, "if", if_name
[type
]);
116 snprintf(buf
, sizeof(buf
), "%d", index
);
117 qemu_opt_set(opts
, "index", buf
);
120 qemu_opt_set(opts
, "file", file
);
124 DriveInfo
*drive_get(BlockInterfaceType type
, int bus
, int unit
)
128 /* seek interface, bus and unit */
130 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
131 if (dinfo
->type
== type
&&
140 DriveInfo
*drive_get_by_index(BlockInterfaceType type
, int index
)
142 return drive_get(type
,
143 drive_index_to_bus_id(type
, index
),
144 drive_index_to_unit_id(type
, index
));
147 int drive_get_max_bus(BlockInterfaceType type
)
153 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
154 if(dinfo
->type
== type
&&
155 dinfo
->bus
> max_bus
)
156 max_bus
= dinfo
->bus
;
161 /* Get a block device. This should only be used for single-drive devices
162 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
164 DriveInfo
*drive_get_next(BlockInterfaceType type
)
166 static int next_block_unit
[IF_COUNT
];
168 return drive_get(type
, 0, next_block_unit
[type
]++);
171 DriveInfo
*drive_get_by_blockdev(BlockDriverState
*bs
)
175 QTAILQ_FOREACH(dinfo
, &drives
, next
) {
176 if (dinfo
->bdrv
== bs
) {
183 static void bdrv_format_print(void *opaque
, const char *name
)
185 error_printf(" %s", name
);
188 static void drive_uninit(DriveInfo
*dinfo
)
190 qemu_opts_del(dinfo
->opts
);
191 bdrv_delete(dinfo
->bdrv
);
193 QTAILQ_REMOVE(&drives
, dinfo
, next
);
197 void drive_put_ref(DriveInfo
*dinfo
)
199 assert(dinfo
->refcount
);
200 if (--dinfo
->refcount
== 0) {
205 void drive_get_ref(DriveInfo
*dinfo
)
215 static void drive_put_ref_bh(void *opaque
)
217 DrivePutRefBH
*s
= opaque
;
219 drive_put_ref(s
->dinfo
);
220 qemu_bh_delete(s
->bh
);
225 * Release a drive reference in a BH
227 * It is not possible to use drive_put_ref() from a callback function when the
228 * callers still need the drive. In such cases we schedule a BH to release the
231 static void drive_put_ref_bh_schedule(DriveInfo
*dinfo
)
235 s
= g_new(DrivePutRefBH
, 1);
236 s
->bh
= qemu_bh_new(drive_put_ref_bh
, s
);
238 qemu_bh_schedule(s
->bh
);
241 static int parse_block_error_action(const char *buf
, bool is_read
)
243 if (!strcmp(buf
, "ignore")) {
244 return BLOCKDEV_ON_ERROR_IGNORE
;
245 } else if (!is_read
&& !strcmp(buf
, "enospc")) {
246 return BLOCKDEV_ON_ERROR_ENOSPC
;
247 } else if (!strcmp(buf
, "stop")) {
248 return BLOCKDEV_ON_ERROR_STOP
;
249 } else if (!strcmp(buf
, "report")) {
250 return BLOCKDEV_ON_ERROR_REPORT
;
252 error_report("'%s' invalid %s error action",
253 buf
, is_read
? "read" : "write");
258 static bool do_check_io_limits(BlockIOLimit
*io_limits
, Error
**errp
)
265 bps_flag
= (io_limits
->bps
[BLOCK_IO_LIMIT_TOTAL
] != 0)
266 && ((io_limits
->bps
[BLOCK_IO_LIMIT_READ
] != 0)
267 || (io_limits
->bps
[BLOCK_IO_LIMIT_WRITE
] != 0));
268 iops_flag
= (io_limits
->iops
[BLOCK_IO_LIMIT_TOTAL
] != 0)
269 && ((io_limits
->iops
[BLOCK_IO_LIMIT_READ
] != 0)
270 || (io_limits
->iops
[BLOCK_IO_LIMIT_WRITE
] != 0));
271 if (bps_flag
|| iops_flag
) {
272 error_setg(errp
, "bps(iops) and bps_rd/bps_wr(iops_rd/iops_wr) "
273 "cannot be used at the same time");
277 if (io_limits
->bps
[BLOCK_IO_LIMIT_TOTAL
] < 0 ||
278 io_limits
->bps
[BLOCK_IO_LIMIT_WRITE
] < 0 ||
279 io_limits
->bps
[BLOCK_IO_LIMIT_READ
] < 0 ||
280 io_limits
->iops
[BLOCK_IO_LIMIT_TOTAL
] < 0 ||
281 io_limits
->iops
[BLOCK_IO_LIMIT_WRITE
] < 0 ||
282 io_limits
->iops
[BLOCK_IO_LIMIT_READ
] < 0) {
283 error_setg(errp
, "bps and iops values must be 0 or greater");
290 DriveInfo
*drive_init(QemuOpts
*opts
, BlockInterfaceType block_default_type
)
293 const char *file
= NULL
;
295 const char *mediastr
= "";
296 BlockInterfaceType type
;
297 enum { MEDIA_DISK
, MEDIA_CDROM
} media
;
299 int cyls
, heads
, secs
, translation
;
300 BlockDriver
*drv
= NULL
;
305 int on_read_error
, on_write_error
;
308 BlockIOLimit io_limits
;
314 translation
= BIOS_ATA_TRANSLATION_AUTO
;
317 /* extract parameters */
318 bus_id
= qemu_opt_get_number(opts
, "bus", 0);
319 unit_id
= qemu_opt_get_number(opts
, "unit", -1);
320 index
= qemu_opt_get_number(opts
, "index", -1);
322 cyls
= qemu_opt_get_number(opts
, "cyls", 0);
323 heads
= qemu_opt_get_number(opts
, "heads", 0);
324 secs
= qemu_opt_get_number(opts
, "secs", 0);
326 snapshot
= qemu_opt_get_bool(opts
, "snapshot", 0);
327 ro
= qemu_opt_get_bool(opts
, "readonly", 0);
328 copy_on_read
= qemu_opt_get_bool(opts
, "copy-on-read", false);
330 file
= qemu_opt_get(opts
, "file");
331 serial
= qemu_opt_get(opts
, "serial");
333 if ((buf
= qemu_opt_get(opts
, "if")) != NULL
) {
334 for (type
= 0; type
< IF_COUNT
&& strcmp(buf
, if_name
[type
]); type
++)
336 if (type
== IF_COUNT
) {
337 error_report("unsupported bus type '%s'", buf
);
341 type
= block_default_type
;
344 max_devs
= if_max_devs
[type
];
346 if (cyls
|| heads
|| secs
) {
348 error_report("invalid physical cyls number");
352 error_report("invalid physical heads number");
356 error_report("invalid physical secs number");
361 if ((buf
= qemu_opt_get(opts
, "trans")) != NULL
) {
363 error_report("'%s' trans must be used with cyls, heads and secs",
367 if (!strcmp(buf
, "none"))
368 translation
= BIOS_ATA_TRANSLATION_NONE
;
369 else if (!strcmp(buf
, "lba"))
370 translation
= BIOS_ATA_TRANSLATION_LBA
;
371 else if (!strcmp(buf
, "auto"))
372 translation
= BIOS_ATA_TRANSLATION_AUTO
;
374 error_report("'%s' invalid translation type", buf
);
379 if ((buf
= qemu_opt_get(opts
, "media")) != NULL
) {
380 if (!strcmp(buf
, "disk")) {
382 } else if (!strcmp(buf
, "cdrom")) {
383 if (cyls
|| secs
|| heads
) {
384 error_report("CHS can't be set with media=%s", buf
);
389 error_report("'%s' invalid media", buf
);
394 bdrv_flags
|= BDRV_O_CACHE_WB
;
395 if ((buf
= qemu_opt_get(opts
, "cache")) != NULL
) {
396 if (bdrv_parse_cache_flags(buf
, &bdrv_flags
) != 0) {
397 error_report("invalid cache option");
402 #ifdef CONFIG_LINUX_AIO
403 if ((buf
= qemu_opt_get(opts
, "aio")) != NULL
) {
404 if (!strcmp(buf
, "native")) {
405 bdrv_flags
|= BDRV_O_NATIVE_AIO
;
406 } else if (!strcmp(buf
, "threads")) {
407 /* this is the default */
409 error_report("invalid aio option");
415 if ((buf
= qemu_opt_get(opts
, "format")) != NULL
) {
416 if (is_help_option(buf
)) {
417 error_printf("Supported formats:");
418 bdrv_iterate_format(bdrv_format_print
, NULL
);
422 drv
= bdrv_find_whitelisted_format(buf
);
424 error_report("'%s' invalid format", buf
);
429 /* disk I/O throttling */
430 io_limits
.bps
[BLOCK_IO_LIMIT_TOTAL
] =
431 qemu_opt_get_number(opts
, "bps", 0);
432 io_limits
.bps
[BLOCK_IO_LIMIT_READ
] =
433 qemu_opt_get_number(opts
, "bps_rd", 0);
434 io_limits
.bps
[BLOCK_IO_LIMIT_WRITE
] =
435 qemu_opt_get_number(opts
, "bps_wr", 0);
436 io_limits
.iops
[BLOCK_IO_LIMIT_TOTAL
] =
437 qemu_opt_get_number(opts
, "iops", 0);
438 io_limits
.iops
[BLOCK_IO_LIMIT_READ
] =
439 qemu_opt_get_number(opts
, "iops_rd", 0);
440 io_limits
.iops
[BLOCK_IO_LIMIT_WRITE
] =
441 qemu_opt_get_number(opts
, "iops_wr", 0);
443 if (!do_check_io_limits(&io_limits
, &error
)) {
444 error_report("%s", error_get_pretty(error
));
449 if (qemu_opt_get(opts
, "boot") != NULL
) {
450 fprintf(stderr
, "qemu-kvm: boot=on|off is deprecated and will be "
451 "ignored. Future versions will reject this parameter. Please "
452 "update your scripts.\n");
455 on_write_error
= BLOCKDEV_ON_ERROR_ENOSPC
;
456 if ((buf
= qemu_opt_get(opts
, "werror")) != NULL
) {
457 if (type
!= IF_IDE
&& type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_NONE
) {
458 error_report("werror is not supported by this bus type");
462 on_write_error
= parse_block_error_action(buf
, 0);
463 if (on_write_error
< 0) {
468 on_read_error
= BLOCKDEV_ON_ERROR_REPORT
;
469 if ((buf
= qemu_opt_get(opts
, "rerror")) != NULL
) {
470 if (type
!= IF_IDE
&& type
!= IF_VIRTIO
&& type
!= IF_SCSI
&& type
!= IF_NONE
) {
471 error_report("rerror is not supported by this bus type");
475 on_read_error
= parse_block_error_action(buf
, 1);
476 if (on_read_error
< 0) {
481 if ((devaddr
= qemu_opt_get(opts
, "addr")) != NULL
) {
482 if (type
!= IF_VIRTIO
) {
483 error_report("addr is not supported by this bus type");
488 /* compute bus and unit according index */
491 if (bus_id
!= 0 || unit_id
!= -1) {
492 error_report("index cannot be used with bus and unit");
495 bus_id
= drive_index_to_bus_id(type
, index
);
496 unit_id
= drive_index_to_unit_id(type
, index
);
499 /* if user doesn't specify a unit_id,
500 * try to find the first free
505 while (drive_get(type
, bus_id
, unit_id
) != NULL
) {
507 if (max_devs
&& unit_id
>= max_devs
) {
516 if (max_devs
&& unit_id
>= max_devs
) {
517 error_report("unit %d too big (max is %d)",
518 unit_id
, max_devs
- 1);
523 * catch multiple definitions
526 if (drive_get(type
, bus_id
, unit_id
) != NULL
) {
527 error_report("drive with bus=%d, unit=%d (index=%d) exists",
528 bus_id
, unit_id
, index
);
534 dinfo
= g_malloc0(sizeof(*dinfo
));
535 if ((buf
= qemu_opts_id(opts
)) != NULL
) {
536 dinfo
->id
= g_strdup(buf
);
538 /* no id supplied -> create one */
539 dinfo
->id
= g_malloc0(32);
540 if (type
== IF_IDE
|| type
== IF_SCSI
)
541 mediastr
= (media
== MEDIA_CDROM
) ? "-cd" : "-hd";
543 snprintf(dinfo
->id
, 32, "%s%i%s%i",
544 if_name
[type
], bus_id
, mediastr
, unit_id
);
546 snprintf(dinfo
->id
, 32, "%s%s%i",
547 if_name
[type
], mediastr
, unit_id
);
549 dinfo
->bdrv
= bdrv_new(dinfo
->id
);
550 dinfo
->bdrv
->open_flags
= snapshot
? BDRV_O_SNAPSHOT
: 0;
551 dinfo
->bdrv
->read_only
= ro
;
552 dinfo
->devaddr
= devaddr
;
555 dinfo
->unit
= unit_id
;
557 dinfo
->heads
= heads
;
559 dinfo
->trans
= translation
;
562 dinfo
->serial
= serial
;
563 QTAILQ_INSERT_TAIL(&drives
, dinfo
, next
);
565 bdrv_set_on_error(dinfo
->bdrv
, on_read_error
, on_write_error
);
567 /* disk I/O throttling */
568 bdrv_set_io_limits(dinfo
->bdrv
, &io_limits
);
575 dinfo
->media_cd
= media
== MEDIA_CDROM
;
583 /* add virtio block device */
584 opts
= qemu_opts_create_nofail(qemu_find_opts("device"));
585 if (arch_type
== QEMU_ARCH_S390X
) {
586 qemu_opt_set(opts
, "driver", "virtio-blk-s390");
588 qemu_opt_set(opts
, "driver", "virtio-blk-pci");
590 qemu_opt_set(opts
, "drive", dinfo
->id
);
592 qemu_opt_set(opts
, "addr", devaddr
);
597 if (!file
|| !*file
) {
601 /* always use cache=unsafe with snapshot */
602 bdrv_flags
&= ~BDRV_O_CACHE_MASK
;
603 bdrv_flags
|= (BDRV_O_SNAPSHOT
|BDRV_O_CACHE_WB
|BDRV_O_NO_FLUSH
);
607 bdrv_flags
|= BDRV_O_COPY_ON_READ
;
610 if (runstate_check(RUN_STATE_INMIGRATE
)) {
611 bdrv_flags
|= BDRV_O_INCOMING
;
614 if (media
== MEDIA_CDROM
) {
615 /* CDROM is fine for any interface, don't check. */
617 } else if (ro
== 1) {
618 if (type
!= IF_SCSI
&& type
!= IF_VIRTIO
&& type
!= IF_FLOPPY
&&
619 type
!= IF_NONE
&& type
!= IF_PFLASH
) {
620 error_report("readonly not supported by this bus type");
625 bdrv_flags
|= ro
? 0 : BDRV_O_RDWR
;
627 if (ro
&& copy_on_read
) {
628 error_report("warning: disabling copy_on_read on readonly drive");
631 ret
= bdrv_open(dinfo
->bdrv
, file
, bdrv_flags
, drv
);
633 if (ret
== -EMEDIUMTYPE
) {
634 error_report("could not open disk image %s: not in %s format",
635 file
, drv
->format_name
);
637 error_report("could not open disk image %s: %s",
638 file
, strerror(-ret
));
643 if (bdrv_key_required(dinfo
->bdrv
))
648 bdrv_delete(dinfo
->bdrv
);
650 QTAILQ_REMOVE(&drives
, dinfo
, next
);
655 void do_commit(Monitor
*mon
, const QDict
*qdict
)
657 const char *device
= qdict_get_str(qdict
, "device");
658 BlockDriverState
*bs
;
661 if (!strcmp(device
, "all")) {
662 ret
= bdrv_commit_all();
664 bs
= bdrv_find(device
);
666 monitor_printf(mon
, "Device '%s' not found\n", device
);
669 ret
= bdrv_commit(bs
);
672 monitor_printf(mon
, "'commit' error for '%s': %s\n", device
,
677 static void blockdev_do_action(int kind
, void *data
, Error
**errp
)
679 BlockdevAction action
;
680 BlockdevActionList list
;
684 list
.value
= &action
;
686 qmp_transaction(&list
, errp
);
689 void qmp_blockdev_snapshot_sync(const char *device
, const char *snapshot_file
,
690 bool has_format
, const char *format
,
691 bool has_mode
, enum NewImageMode mode
,
694 BlockdevSnapshot snapshot
= {
695 .device
= (char *) device
,
696 .snapshot_file
= (char *) snapshot_file
,
697 .has_format
= has_format
,
698 .format
= (char *) format
,
699 .has_mode
= has_mode
,
702 blockdev_do_action(BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
, &snapshot
,
707 /* New and old BlockDriverState structs for group snapshots */
708 typedef struct BlkTransactionStates
{
709 BlockDriverState
*old_bs
;
710 BlockDriverState
*new_bs
;
711 QSIMPLEQ_ENTRY(BlkTransactionStates
) entry
;
712 } BlkTransactionStates
;
715 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
716 * then we do not pivot any of the devices in the group, and abandon the
719 void qmp_transaction(BlockdevActionList
*dev_list
, Error
**errp
)
722 BlockdevActionList
*dev_entry
= dev_list
;
723 BlkTransactionStates
*states
, *next
;
724 Error
*local_err
= NULL
;
726 QSIMPLEQ_HEAD(snap_bdrv_states
, BlkTransactionStates
) snap_bdrv_states
;
727 QSIMPLEQ_INIT(&snap_bdrv_states
);
729 /* drain all i/o before any snapshots */
732 /* We don't do anything in this loop that commits us to the snapshot */
733 while (NULL
!= dev_entry
) {
734 BlockdevAction
*dev_info
= NULL
;
735 BlockDriver
*proto_drv
;
738 enum NewImageMode mode
;
739 const char *new_image_file
;
741 const char *format
= "qcow2";
743 dev_info
= dev_entry
->value
;
744 dev_entry
= dev_entry
->next
;
746 states
= g_malloc0(sizeof(BlkTransactionStates
));
747 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states
, states
, entry
);
749 switch (dev_info
->kind
) {
750 case BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC
:
751 device
= dev_info
->blockdev_snapshot_sync
->device
;
752 if (!dev_info
->blockdev_snapshot_sync
->has_mode
) {
753 dev_info
->blockdev_snapshot_sync
->mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
755 new_image_file
= dev_info
->blockdev_snapshot_sync
->snapshot_file
;
756 if (dev_info
->blockdev_snapshot_sync
->has_format
) {
757 format
= dev_info
->blockdev_snapshot_sync
->format
;
759 mode
= dev_info
->blockdev_snapshot_sync
->mode
;
765 drv
= bdrv_find_format(format
);
767 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
768 goto delete_and_fail
;
771 states
->old_bs
= bdrv_find(device
);
772 if (!states
->old_bs
) {
773 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
774 goto delete_and_fail
;
777 if (!bdrv_is_inserted(states
->old_bs
)) {
778 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
779 goto delete_and_fail
;
782 if (bdrv_in_use(states
->old_bs
)) {
783 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
784 goto delete_and_fail
;
787 if (!bdrv_is_read_only(states
->old_bs
)) {
788 if (bdrv_flush(states
->old_bs
)) {
789 error_set(errp
, QERR_IO_ERROR
);
790 goto delete_and_fail
;
794 flags
= states
->old_bs
->open_flags
;
796 proto_drv
= bdrv_find_protocol(new_image_file
);
798 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
799 goto delete_and_fail
;
802 /* create new image w/backing file */
803 if (mode
!= NEW_IMAGE_MODE_EXISTING
) {
804 bdrv_img_create(new_image_file
, format
,
805 states
->old_bs
->filename
,
806 states
->old_bs
->drv
->format_name
,
807 NULL
, -1, flags
, &local_err
, false);
808 if (error_is_set(&local_err
)) {
809 error_propagate(errp
, local_err
);
810 goto delete_and_fail
;
814 /* We will manually add the backing_hd field to the bs later */
815 states
->new_bs
= bdrv_new("");
816 ret
= bdrv_open(states
->new_bs
, new_image_file
,
817 flags
| BDRV_O_NO_BACKING
, drv
);
819 error_set(errp
, QERR_OPEN_FILE_FAILED
, new_image_file
);
820 goto delete_and_fail
;
825 /* Now we are going to do the actual pivot. Everything up to this point
826 * is reversible, but we are committed at this point */
827 QSIMPLEQ_FOREACH(states
, &snap_bdrv_states
, entry
) {
828 /* This removes our old bs from the bdrv_states, and adds the new bs */
829 bdrv_append(states
->new_bs
, states
->old_bs
);
830 /* We don't need (or want) to use the transactional
831 * bdrv_reopen_multiple() across all the entries at once, because we
832 * don't want to abort all of them if one of them fails the reopen */
833 bdrv_reopen(states
->new_bs
, states
->new_bs
->open_flags
& ~BDRV_O_RDWR
,
842 * failure, and it is all-or-none; abandon each new bs, and keep using
843 * the original bs for all images
845 QSIMPLEQ_FOREACH(states
, &snap_bdrv_states
, entry
) {
846 if (states
->new_bs
) {
847 bdrv_delete(states
->new_bs
);
851 QSIMPLEQ_FOREACH_SAFE(states
, &snap_bdrv_states
, entry
, next
) {
857 static void eject_device(BlockDriverState
*bs
, int force
, Error
**errp
)
859 if (bdrv_in_use(bs
)) {
860 error_set(errp
, QERR_DEVICE_IN_USE
, bdrv_get_device_name(bs
));
863 if (!bdrv_dev_has_removable_media(bs
)) {
864 error_set(errp
, QERR_DEVICE_NOT_REMOVABLE
, bdrv_get_device_name(bs
));
868 if (bdrv_dev_is_medium_locked(bs
) && !bdrv_dev_is_tray_open(bs
)) {
869 bdrv_dev_eject_request(bs
, force
);
871 error_set(errp
, QERR_DEVICE_LOCKED
, bdrv_get_device_name(bs
));
879 void qmp_eject(const char *device
, bool has_force
, bool force
, Error
**errp
)
881 BlockDriverState
*bs
;
883 bs
= bdrv_find(device
);
885 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
889 eject_device(bs
, force
, errp
);
892 void qmp_block_passwd(const char *device
, const char *password
, Error
**errp
)
894 BlockDriverState
*bs
;
897 bs
= bdrv_find(device
);
899 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
903 err
= bdrv_set_key(bs
, password
);
904 if (err
== -EINVAL
) {
905 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
907 } else if (err
< 0) {
908 error_set(errp
, QERR_INVALID_PASSWORD
);
913 static void qmp_bdrv_open_encrypted(BlockDriverState
*bs
, const char *filename
,
914 int bdrv_flags
, BlockDriver
*drv
,
915 const char *password
, Error
**errp
)
917 if (bdrv_open(bs
, filename
, bdrv_flags
, drv
) < 0) {
918 error_set(errp
, QERR_OPEN_FILE_FAILED
, filename
);
922 if (bdrv_key_required(bs
)) {
924 if (bdrv_set_key(bs
, password
) < 0) {
925 error_set(errp
, QERR_INVALID_PASSWORD
);
928 error_set(errp
, QERR_DEVICE_ENCRYPTED
, bdrv_get_device_name(bs
),
929 bdrv_get_encrypted_filename(bs
));
931 } else if (password
) {
932 error_set(errp
, QERR_DEVICE_NOT_ENCRYPTED
, bdrv_get_device_name(bs
));
936 void qmp_change_blockdev(const char *device
, const char *filename
,
937 bool has_format
, const char *format
, Error
**errp
)
939 BlockDriverState
*bs
;
940 BlockDriver
*drv
= NULL
;
944 bs
= bdrv_find(device
);
946 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
951 drv
= bdrv_find_whitelisted_format(format
);
953 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
958 eject_device(bs
, 0, &err
);
959 if (error_is_set(&err
)) {
960 error_propagate(errp
, err
);
964 bdrv_flags
= bdrv_is_read_only(bs
) ? 0 : BDRV_O_RDWR
;
965 bdrv_flags
|= bdrv_is_snapshot(bs
) ? BDRV_O_SNAPSHOT
: 0;
967 qmp_bdrv_open_encrypted(bs
, filename
, bdrv_flags
, drv
, NULL
, errp
);
970 /* throttling disk I/O limits */
971 void qmp_block_set_io_throttle(const char *device
, int64_t bps
, int64_t bps_rd
,
972 int64_t bps_wr
, int64_t iops
, int64_t iops_rd
,
973 int64_t iops_wr
, Error
**errp
)
975 BlockIOLimit io_limits
;
976 BlockDriverState
*bs
;
978 bs
= bdrv_find(device
);
980 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
984 io_limits
.bps
[BLOCK_IO_LIMIT_TOTAL
] = bps
;
985 io_limits
.bps
[BLOCK_IO_LIMIT_READ
] = bps_rd
;
986 io_limits
.bps
[BLOCK_IO_LIMIT_WRITE
] = bps_wr
;
987 io_limits
.iops
[BLOCK_IO_LIMIT_TOTAL
]= iops
;
988 io_limits
.iops
[BLOCK_IO_LIMIT_READ
] = iops_rd
;
989 io_limits
.iops
[BLOCK_IO_LIMIT_WRITE
]= iops_wr
;
991 if (!do_check_io_limits(&io_limits
, errp
)) {
995 bs
->io_limits
= io_limits
;
996 bs
->slice_time
= BLOCK_IO_SLICE_TIME
;
998 if (!bs
->io_limits_enabled
&& bdrv_io_limits_enabled(bs
)) {
999 bdrv_io_limits_enable(bs
);
1000 } else if (bs
->io_limits_enabled
&& !bdrv_io_limits_enabled(bs
)) {
1001 bdrv_io_limits_disable(bs
);
1003 if (bs
->block_timer
) {
1004 qemu_mod_timer(bs
->block_timer
, qemu_get_clock_ns(vm_clock
));
1009 int do_drive_del(Monitor
*mon
, const QDict
*qdict
, QObject
**ret_data
)
1011 const char *id
= qdict_get_str(qdict
, "id");
1012 BlockDriverState
*bs
;
1016 qerror_report(QERR_DEVICE_NOT_FOUND
, id
);
1019 if (bdrv_in_use(bs
)) {
1020 qerror_report(QERR_DEVICE_IN_USE
, id
);
1024 /* quiesce block driver; prevent further io */
1029 /* if we have a device attached to this BlockDriverState
1030 * then we need to make the drive anonymous until the device
1031 * can be removed. If this is a drive with no device backing
1032 * then we can just get rid of the block driver state right here.
1034 if (bdrv_get_attached_dev(bs
)) {
1037 drive_uninit(drive_get_by_blockdev(bs
));
1043 void qmp_block_resize(const char *device
, int64_t size
, Error
**errp
)
1045 BlockDriverState
*bs
;
1047 bs
= bdrv_find(device
);
1049 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1054 error_set(errp
, QERR_INVALID_PARAMETER_VALUE
, "size", "a >0 size");
1058 switch (bdrv_truncate(bs
, size
)) {
1062 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1065 error_set(errp
, QERR_UNSUPPORTED
);
1068 error_set(errp
, QERR_DEVICE_IS_READ_ONLY
, device
);
1071 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1074 error_set(errp
, QERR_UNDEFINED_ERROR
);
1079 static void block_job_cb(void *opaque
, int ret
)
1081 BlockDriverState
*bs
= opaque
;
1084 trace_block_job_cb(bs
, bs
->job
, ret
);
1087 obj
= qobject_from_block_job(bs
->job
);
1089 QDict
*dict
= qobject_to_qdict(obj
);
1090 qdict_put(dict
, "error", qstring_from_str(strerror(-ret
)));
1093 if (block_job_is_cancelled(bs
->job
)) {
1094 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED
, obj
);
1096 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED
, obj
);
1098 qobject_decref(obj
);
1100 drive_put_ref_bh_schedule(drive_get_by_blockdev(bs
));
1103 void qmp_block_stream(const char *device
, bool has_base
,
1104 const char *base
, bool has_speed
, int64_t speed
,
1105 bool has_on_error
, BlockdevOnError on_error
,
1108 BlockDriverState
*bs
;
1109 BlockDriverState
*base_bs
= NULL
;
1110 Error
*local_err
= NULL
;
1112 if (!has_on_error
) {
1113 on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1116 bs
= bdrv_find(device
);
1118 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1123 base_bs
= bdrv_find_backing_image(bs
, base
);
1124 if (base_bs
== NULL
) {
1125 error_set(errp
, QERR_BASE_NOT_FOUND
, base
);
1130 stream_start(bs
, base_bs
, base
, has_speed
? speed
: 0,
1131 on_error
, block_job_cb
, bs
, &local_err
);
1132 if (error_is_set(&local_err
)) {
1133 error_propagate(errp
, local_err
);
1137 /* Grab a reference so hotplug does not delete the BlockDriverState from
1140 drive_get_ref(drive_get_by_blockdev(bs
));
1142 trace_qmp_block_stream(bs
, bs
->job
);
1145 void qmp_block_commit(const char *device
,
1146 bool has_base
, const char *base
, const char *top
,
1147 bool has_speed
, int64_t speed
,
1150 BlockDriverState
*bs
;
1151 BlockDriverState
*base_bs
, *top_bs
;
1152 Error
*local_err
= NULL
;
1153 /* This will be part of the QMP command, if/when the
1154 * BlockdevOnError change for blkmirror makes it in
1156 BlockdevOnError on_error
= BLOCKDEV_ON_ERROR_REPORT
;
1158 /* drain all i/o before commits */
1161 bs
= bdrv_find(device
);
1163 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1167 /* default top_bs is the active layer */
1171 if (strcmp(bs
->filename
, top
) != 0) {
1172 top_bs
= bdrv_find_backing_image(bs
, top
);
1176 if (top_bs
== NULL
) {
1177 error_setg(errp
, "Top image file %s not found", top
? top
: "NULL");
1181 if (has_base
&& base
) {
1182 base_bs
= bdrv_find_backing_image(top_bs
, base
);
1184 base_bs
= bdrv_find_base(top_bs
);
1187 if (base_bs
== NULL
) {
1188 error_set(errp
, QERR_BASE_NOT_FOUND
, base
? base
: "NULL");
1192 commit_start(bs
, base_bs
, top_bs
, speed
, on_error
, block_job_cb
, bs
,
1194 if (local_err
!= NULL
) {
1195 error_propagate(errp
, local_err
);
1198 /* Grab a reference so hotplug does not delete the BlockDriverState from
1201 drive_get_ref(drive_get_by_blockdev(bs
));
1204 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1206 void qmp_drive_mirror(const char *device
, const char *target
,
1207 bool has_format
, const char *format
,
1208 enum MirrorSyncMode sync
,
1209 bool has_mode
, enum NewImageMode mode
,
1210 bool has_speed
, int64_t speed
,
1211 bool has_granularity
, uint32_t granularity
,
1212 bool has_buf_size
, int64_t buf_size
,
1213 bool has_on_source_error
, BlockdevOnError on_source_error
,
1214 bool has_on_target_error
, BlockdevOnError on_target_error
,
1217 BlockDriverState
*bs
;
1218 BlockDriverState
*source
, *target_bs
;
1219 BlockDriver
*proto_drv
;
1220 BlockDriver
*drv
= NULL
;
1221 Error
*local_err
= NULL
;
1229 if (!has_on_source_error
) {
1230 on_source_error
= BLOCKDEV_ON_ERROR_REPORT
;
1232 if (!has_on_target_error
) {
1233 on_target_error
= BLOCKDEV_ON_ERROR_REPORT
;
1236 mode
= NEW_IMAGE_MODE_ABSOLUTE_PATHS
;
1238 if (!has_granularity
) {
1241 if (!has_buf_size
) {
1242 buf_size
= DEFAULT_MIRROR_BUF_SIZE
;
1245 if (granularity
!= 0 && (granularity
< 512 || granularity
> 1048576 * 64)) {
1246 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1249 if (granularity
& (granularity
- 1)) {
1250 error_set(errp
, QERR_INVALID_PARAMETER
, device
);
1254 bs
= bdrv_find(device
);
1256 error_set(errp
, QERR_DEVICE_NOT_FOUND
, device
);
1260 if (!bdrv_is_inserted(bs
)) {
1261 error_set(errp
, QERR_DEVICE_HAS_NO_MEDIUM
, device
);
1266 format
= mode
== NEW_IMAGE_MODE_EXISTING
? NULL
: bs
->drv
->format_name
;
1269 drv
= bdrv_find_format(format
);
1271 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1276 if (bdrv_in_use(bs
)) {
1277 error_set(errp
, QERR_DEVICE_IN_USE
, device
);
1281 flags
= bs
->open_flags
| BDRV_O_RDWR
;
1282 source
= bs
->backing_hd
;
1283 if (!source
&& sync
== MIRROR_SYNC_MODE_TOP
) {
1284 sync
= MIRROR_SYNC_MODE_FULL
;
1287 proto_drv
= bdrv_find_protocol(target
);
1289 error_set(errp
, QERR_INVALID_BLOCK_FORMAT
, format
);
1293 bdrv_get_geometry(bs
, &size
);
1295 if (sync
== MIRROR_SYNC_MODE_FULL
&& mode
!= NEW_IMAGE_MODE_EXISTING
) {
1296 /* create new image w/o backing file */
1297 assert(format
&& drv
);
1298 bdrv_img_create(target
, format
,
1299 NULL
, NULL
, NULL
, size
, flags
, &local_err
, false);
1302 case NEW_IMAGE_MODE_EXISTING
:
1305 case NEW_IMAGE_MODE_ABSOLUTE_PATHS
:
1306 /* create new image with backing file */
1307 bdrv_img_create(target
, format
,
1309 source
->drv
->format_name
,
1310 NULL
, size
, flags
, &local_err
, false);
1317 if (error_is_set(&local_err
)) {
1318 error_propagate(errp
, local_err
);
1322 /* Mirroring takes care of copy-on-write using the source's backing
1325 target_bs
= bdrv_new("");
1326 ret
= bdrv_open(target_bs
, target
, flags
| BDRV_O_NO_BACKING
, drv
);
1329 bdrv_delete(target_bs
);
1330 error_set(errp
, QERR_OPEN_FILE_FAILED
, target
);
1334 mirror_start(bs
, target_bs
, speed
, granularity
, buf_size
, sync
,
1335 on_source_error
, on_target_error
,
1336 block_job_cb
, bs
, &local_err
);
1337 if (local_err
!= NULL
) {
1338 bdrv_delete(target_bs
);
1339 error_propagate(errp
, local_err
);
1343 /* Grab a reference so hotplug does not delete the BlockDriverState from
1346 drive_get_ref(drive_get_by_blockdev(bs
));
1349 static BlockJob
*find_block_job(const char *device
)
1351 BlockDriverState
*bs
;
1353 bs
= bdrv_find(device
);
1354 if (!bs
|| !bs
->job
) {
1360 void qmp_block_job_set_speed(const char *device
, int64_t speed
, Error
**errp
)
1362 BlockJob
*job
= find_block_job(device
);
1365 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
1369 block_job_set_speed(job
, speed
, errp
);
1372 void qmp_block_job_cancel(const char *device
,
1373 bool has_force
, bool force
, Error
**errp
)
1375 BlockJob
*job
= find_block_job(device
);
1382 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
1385 if (job
->paused
&& !force
) {
1386 error_set(errp
, QERR_BLOCK_JOB_PAUSED
, device
);
1390 trace_qmp_block_job_cancel(job
);
1391 block_job_cancel(job
);
1394 void qmp_block_job_pause(const char *device
, Error
**errp
)
1396 BlockJob
*job
= find_block_job(device
);
1399 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
1403 trace_qmp_block_job_pause(job
);
1404 block_job_pause(job
);
1407 void qmp_block_job_resume(const char *device
, Error
**errp
)
1409 BlockJob
*job
= find_block_job(device
);
1412 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
1416 trace_qmp_block_job_resume(job
);
1417 block_job_resume(job
);
1420 void qmp_block_job_complete(const char *device
, Error
**errp
)
1422 BlockJob
*job
= find_block_job(device
);
1425 error_set(errp
, QERR_BLOCK_JOB_NOT_ACTIVE
, device
);
1429 trace_qmp_block_job_complete(job
);
1430 block_job_complete(job
, errp
);
1433 static void do_qmp_query_block_jobs_one(void *opaque
, BlockDriverState
*bs
)
1435 BlockJobInfoList
**prev
= opaque
;
1436 BlockJob
*job
= bs
->job
;
1439 BlockJobInfoList
*elem
= g_new0(BlockJobInfoList
, 1);
1440 elem
->value
= block_job_query(bs
->job
);
1441 (*prev
)->next
= elem
;
1446 BlockJobInfoList
*qmp_query_block_jobs(Error
**errp
)
1448 /* Dummy is a fake list element for holding the head pointer */
1449 BlockJobInfoList dummy
= {};
1450 BlockJobInfoList
*prev
= &dummy
;
1451 bdrv_iterate(do_qmp_query_block_jobs_one
, &prev
);
1455 QemuOptsList qemu_drive_opts
= {
1457 .head
= QTAILQ_HEAD_INITIALIZER(qemu_drive_opts
.head
),
1461 .type
= QEMU_OPT_NUMBER
,
1462 .help
= "bus number",
1465 .type
= QEMU_OPT_NUMBER
,
1466 .help
= "unit number (i.e. lun for scsi)",
1469 .type
= QEMU_OPT_STRING
,
1470 .help
= "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1473 .type
= QEMU_OPT_NUMBER
,
1474 .help
= "index number",
1477 .type
= QEMU_OPT_NUMBER
,
1478 .help
= "number of cylinders (ide disk geometry)",
1481 .type
= QEMU_OPT_NUMBER
,
1482 .help
= "number of heads (ide disk geometry)",
1485 .type
= QEMU_OPT_NUMBER
,
1486 .help
= "number of sectors (ide disk geometry)",
1489 .type
= QEMU_OPT_STRING
,
1490 .help
= "chs translation (auto, lba. none)",
1493 .type
= QEMU_OPT_STRING
,
1494 .help
= "media type (disk, cdrom)",
1497 .type
= QEMU_OPT_BOOL
,
1498 .help
= "enable/disable snapshot mode",
1501 .type
= QEMU_OPT_STRING
,
1502 .help
= "disk image",
1505 .type
= QEMU_OPT_STRING
,
1506 .help
= "host cache usage (none, writeback, writethrough, "
1507 "directsync, unsafe)",
1510 .type
= QEMU_OPT_STRING
,
1511 .help
= "host AIO implementation (threads, native)",
1514 .type
= QEMU_OPT_STRING
,
1515 .help
= "disk format (raw, qcow2, ...)",
1518 .type
= QEMU_OPT_STRING
,
1519 .help
= "disk serial number",
1522 .type
= QEMU_OPT_STRING
,
1523 .help
= "read error action",
1526 .type
= QEMU_OPT_STRING
,
1527 .help
= "write error action",
1530 .type
= QEMU_OPT_STRING
,
1531 .help
= "pci address (virtio only)",
1534 .type
= QEMU_OPT_BOOL
,
1535 .help
= "open drive file as read-only",
1538 .type
= QEMU_OPT_NUMBER
,
1539 .help
= "limit total I/O operations per second",
1542 .type
= QEMU_OPT_NUMBER
,
1543 .help
= "limit read operations per second",
1546 .type
= QEMU_OPT_NUMBER
,
1547 .help
= "limit write operations per second",
1550 .type
= QEMU_OPT_NUMBER
,
1551 .help
= "limit total bytes per second",
1554 .type
= QEMU_OPT_NUMBER
,
1555 .help
= "limit read bytes per second",
1558 .type
= QEMU_OPT_NUMBER
,
1559 .help
= "limit write bytes per second",
1561 .name
= "copy-on-read",
1562 .type
= QEMU_OPT_BOOL
,
1563 .help
= "copy read data from backing file into image file",
1566 .type
= QEMU_OPT_BOOL
,
1567 .help
= "(deprecated, ignored)",
1569 { /* end of list */ }