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-common.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 "sysemu/sysemu.h"
42 #include "block/block_int.h"
43 #include "qmp-commands.h"
45 #include "sysemu/arch_init.h"
47 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
48 extern QemuOptsList qemu_common_drive_opts;
50 static const char *const if_name[IF_COUNT] = {
54 [IF_FLOPPY] = "floppy",
55 [IF_PFLASH] = "pflash",
58 [IF_VIRTIO] = "virtio",
62 static const int if_max_devs[IF_COUNT] = {
64 * Do not change these numbers! They govern how drive option
65 * index maps to unit and bus. That mapping is ABI.
67 * All controllers used to imlement if=T drives need to support
68 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
69 * Otherwise, some index values map to "impossible" bus, unit
72 * For instance, if you change [IF_SCSI] to 255, -drive
73 * if=scsi,index=12 no longer means bus=1,unit=5, but
74 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
75 * the drive can't be set up. Regression.
82 * We automatically delete the drive when a device using it gets
83 * unplugged. Questionable feature, but we can't just drop it.
84 * Device models call blockdev_mark_auto_del() to schedule the
85 * automatic deletion, and generic qdev code calls blockdev_auto_del()
86 * when deletion is actually safe.
88 void blockdev_mark_auto_del(BlockDriverState *bs)
90 DriveInfo *dinfo = drive_get_by_blockdev(bs);
93 block_job_cancel(bs->job);
100 void blockdev_auto_del(BlockDriverState *bs)
102 DriveInfo *dinfo = drive_get_by_blockdev(bs);
104 if (dinfo && dinfo->auto_del) {
105 drive_put_ref(dinfo);
109 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
111 int max_devs = if_max_devs[type];
112 return max_devs ? index / max_devs : 0;
115 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
117 int max_devs = if_max_devs[type];
118 return max_devs ? index % max_devs : index;
121 QemuOpts *drive_def(const char *optstr)
123 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
126 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
132 opts = drive_def(optstr);
136 if (type != IF_DEFAULT) {
137 qemu_opt_set(opts, "if", if_name[type]);
140 snprintf(buf, sizeof(buf), "%d", index);
141 qemu_opt_set(opts, "index", buf);
144 qemu_opt_set(opts, "file", file);
148 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
152 /* seek interface, bus and unit */
154 QTAILQ_FOREACH(dinfo, &drives, next) {
155 if (dinfo->type == type &&
164 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
166 return drive_get(type,
167 drive_index_to_bus_id(type, index),
168 drive_index_to_unit_id(type, index));
171 int drive_get_max_bus(BlockInterfaceType type)
177 QTAILQ_FOREACH(dinfo, &drives, next) {
178 if(dinfo->type == type &&
179 dinfo->bus > max_bus)
180 max_bus = dinfo->bus;
185 /* Get a block device. This should only be used for single-drive devices
186 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
188 DriveInfo *drive_get_next(BlockInterfaceType type)
190 static int next_block_unit[IF_COUNT];
192 return drive_get(type, 0, next_block_unit[type]++);
195 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
199 QTAILQ_FOREACH(dinfo, &drives, next) {
200 if (dinfo->bdrv == bs) {
207 static void bdrv_format_print(void *opaque, const char *name)
209 error_printf(" %s", name);
212 static void drive_uninit(DriveInfo *dinfo)
214 qemu_opts_del(dinfo->opts);
215 bdrv_delete(dinfo->bdrv);
217 QTAILQ_REMOVE(&drives, dinfo, next);
218 g_free(dinfo->serial);
222 void drive_put_ref(DriveInfo *dinfo)
224 assert(dinfo->refcount);
225 if (--dinfo->refcount == 0) {
230 void drive_get_ref(DriveInfo *dinfo)
240 static void drive_put_ref_bh(void *opaque)
242 DrivePutRefBH *s = opaque;
244 drive_put_ref(s->dinfo);
245 qemu_bh_delete(s->bh);
250 * Release a drive reference in a BH
252 * It is not possible to use drive_put_ref() from a callback function when the
253 * callers still need the drive. In such cases we schedule a BH to release the
256 static void drive_put_ref_bh_schedule(DriveInfo *dinfo)
260 s = g_new(DrivePutRefBH, 1);
261 s->bh = qemu_bh_new(drive_put_ref_bh, s);
263 qemu_bh_schedule(s->bh);
266 static int parse_block_error_action(const char *buf, bool is_read)
268 if (!strcmp(buf, "ignore")) {
269 return BLOCKDEV_ON_ERROR_IGNORE;
270 } else if (!is_read && !strcmp(buf, "enospc")) {
271 return BLOCKDEV_ON_ERROR_ENOSPC;
272 } else if (!strcmp(buf, "stop")) {
273 return BLOCKDEV_ON_ERROR_STOP;
274 } else if (!strcmp(buf, "report")) {
275 return BLOCKDEV_ON_ERROR_REPORT;
277 error_report("'%s' invalid %s error action",
278 buf, is_read ? "read" : "write");
283 static bool do_check_io_limits(BlockIOLimit *io_limits, Error **errp)
290 bps_flag = (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] != 0)
291 && ((io_limits->bps[BLOCK_IO_LIMIT_READ] != 0)
292 || (io_limits->bps[BLOCK_IO_LIMIT_WRITE] != 0));
293 iops_flag = (io_limits->iops[BLOCK_IO_LIMIT_TOTAL] != 0)
294 && ((io_limits->iops[BLOCK_IO_LIMIT_READ] != 0)
295 || (io_limits->iops[BLOCK_IO_LIMIT_WRITE] != 0));
296 if (bps_flag || iops_flag) {
297 error_setg(errp, "bps(iops) and bps_rd/bps_wr(iops_rd/iops_wr) "
298 "cannot be used at the same time");
302 if (io_limits->bps[BLOCK_IO_LIMIT_TOTAL] < 0 ||
303 io_limits->bps[BLOCK_IO_LIMIT_WRITE] < 0 ||
304 io_limits->bps[BLOCK_IO_LIMIT_READ] < 0 ||
305 io_limits->iops[BLOCK_IO_LIMIT_TOTAL] < 0 ||
306 io_limits->iops[BLOCK_IO_LIMIT_WRITE] < 0 ||
307 io_limits->iops[BLOCK_IO_LIMIT_READ] < 0) {
308 error_setg(errp, "bps and iops values must be 0 or greater");
315 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
318 const char *file = NULL;
320 const char *mediastr = "";
321 BlockInterfaceType type;
322 enum { MEDIA_DISK, MEDIA_CDROM } media;
324 int cyls, heads, secs, translation;
325 BlockDriver *drv = NULL;
330 int on_read_error, on_write_error;
333 BlockIOLimit io_limits;
342 translation = BIOS_ATA_TRANSLATION_AUTO;
345 /* Check common options by copying from all_opts to opts, all other options
346 * are stored in bs_opts. */
347 id = qemu_opts_id(all_opts);
348 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
349 if (error_is_set(&error)) {
350 qerror_report_err(error);
355 bs_opts = qdict_new();
356 qemu_opts_to_qdict(all_opts, bs_opts);
357 qemu_opts_absorb_qdict(opts, bs_opts, &error);
358 if (error_is_set(&error)) {
359 qerror_report_err(error);
365 qdict_del(bs_opts, "id");
368 /* extract parameters */
369 bus_id = qemu_opt_get_number(opts, "bus", 0);
370 unit_id = qemu_opt_get_number(opts, "unit", -1);
371 index = qemu_opt_get_number(opts, "index", -1);
373 cyls = qemu_opt_get_number(opts, "cyls", 0);
374 heads = qemu_opt_get_number(opts, "heads", 0);
375 secs = qemu_opt_get_number(opts, "secs", 0);
377 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
378 ro = qemu_opt_get_bool(opts, "readonly", 0);
379 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
381 file = qemu_opt_get(opts, "file");
382 serial = qemu_opt_get(opts, "serial");
384 if ((buf = qemu_opt_get(opts, "if")) != NULL) {
385 for (type = 0; type < IF_COUNT && strcmp(buf, if_name[type]); type++)
387 if (type == IF_COUNT) {
388 error_report("unsupported bus type '%s'", buf);
392 type = block_default_type;
395 max_devs = if_max_devs[type];
397 if (cyls || heads || secs) {
399 error_report("invalid physical cyls number");
403 error_report("invalid physical heads number");
407 error_report("invalid physical secs number");
412 if ((buf = qemu_opt_get(opts, "trans")) != NULL) {
414 error_report("'%s' trans must be used with cyls, heads and secs",
418 if (!strcmp(buf, "none"))
419 translation = BIOS_ATA_TRANSLATION_NONE;
420 else if (!strcmp(buf, "lba"))
421 translation = BIOS_ATA_TRANSLATION_LBA;
422 else if (!strcmp(buf, "auto"))
423 translation = BIOS_ATA_TRANSLATION_AUTO;
425 error_report("'%s' invalid translation type", buf);
430 if ((buf = qemu_opt_get(opts, "media")) != NULL) {
431 if (!strcmp(buf, "disk")) {
433 } else if (!strcmp(buf, "cdrom")) {
434 if (cyls || secs || heads) {
435 error_report("CHS can't be set with media=%s", buf);
440 error_report("'%s' invalid media", buf);
445 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
446 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
447 error_report("invalid discard option");
452 bdrv_flags |= BDRV_O_CACHE_WB;
453 if ((buf = qemu_opt_get(opts, "cache")) != NULL) {
454 if (bdrv_parse_cache_flags(buf, &bdrv_flags) != 0) {
455 error_report("invalid cache option");
460 #ifdef CONFIG_LINUX_AIO
461 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
462 if (!strcmp(buf, "native")) {
463 bdrv_flags |= BDRV_O_NATIVE_AIO;
464 } else if (!strcmp(buf, "threads")) {
465 /* this is the default */
467 error_report("invalid aio option");
473 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
474 if (is_help_option(buf)) {
475 error_printf("Supported formats:");
476 bdrv_iterate_format(bdrv_format_print, NULL);
480 drv = bdrv_find_whitelisted_format(buf);
482 error_report("'%s' invalid format", buf);
487 /* disk I/O throttling */
488 io_limits.bps[BLOCK_IO_LIMIT_TOTAL] =
489 qemu_opt_get_number(opts, "bps", 0);
490 io_limits.bps[BLOCK_IO_LIMIT_READ] =
491 qemu_opt_get_number(opts, "bps_rd", 0);
492 io_limits.bps[BLOCK_IO_LIMIT_WRITE] =
493 qemu_opt_get_number(opts, "bps_wr", 0);
494 io_limits.iops[BLOCK_IO_LIMIT_TOTAL] =
495 qemu_opt_get_number(opts, "iops", 0);
496 io_limits.iops[BLOCK_IO_LIMIT_READ] =
497 qemu_opt_get_number(opts, "iops_rd", 0);
498 io_limits.iops[BLOCK_IO_LIMIT_WRITE] =
499 qemu_opt_get_number(opts, "iops_wr", 0);
501 if (!do_check_io_limits(&io_limits, &error)) {
502 error_report("%s", error_get_pretty(error));
507 if (qemu_opt_get(opts, "boot") != NULL) {
508 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
509 "ignored. Future versions will reject this parameter. Please "
510 "update your scripts.\n");
513 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
514 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
515 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
516 error_report("werror is not supported by this bus type");
520 on_write_error = parse_block_error_action(buf, 0);
521 if (on_write_error < 0) {
526 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
527 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
528 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
529 error_report("rerror is not supported by this bus type");
533 on_read_error = parse_block_error_action(buf, 1);
534 if (on_read_error < 0) {
539 if ((devaddr = qemu_opt_get(opts, "addr")) != NULL) {
540 if (type != IF_VIRTIO) {
541 error_report("addr is not supported by this bus type");
546 /* compute bus and unit according index */
549 if (bus_id != 0 || unit_id != -1) {
550 error_report("index cannot be used with bus and unit");
553 bus_id = drive_index_to_bus_id(type, index);
554 unit_id = drive_index_to_unit_id(type, index);
557 /* if user doesn't specify a unit_id,
558 * try to find the first free
563 while (drive_get(type, bus_id, unit_id) != NULL) {
565 if (max_devs && unit_id >= max_devs) {
574 if (max_devs && unit_id >= max_devs) {
575 error_report("unit %d too big (max is %d)",
576 unit_id, max_devs - 1);
581 * catch multiple definitions
584 if (drive_get(type, bus_id, unit_id) != NULL) {
585 error_report("drive with bus=%d, unit=%d (index=%d) exists",
586 bus_id, unit_id, index);
592 dinfo = g_malloc0(sizeof(*dinfo));
593 if ((buf = qemu_opts_id(opts)) != NULL) {
594 dinfo->id = g_strdup(buf);
596 /* no id supplied -> create one */
597 dinfo->id = g_malloc0(32);
598 if (type == IF_IDE || type == IF_SCSI)
599 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
601 snprintf(dinfo->id, 32, "%s%i%s%i",
602 if_name[type], bus_id, mediastr, unit_id);
604 snprintf(dinfo->id, 32, "%s%s%i",
605 if_name[type], mediastr, unit_id);
607 dinfo->bdrv = bdrv_new(dinfo->id);
608 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
609 dinfo->bdrv->read_only = ro;
610 dinfo->devaddr = devaddr;
613 dinfo->unit = unit_id;
615 dinfo->heads = heads;
617 dinfo->trans = translation;
618 dinfo->opts = all_opts;
620 if (serial != NULL) {
621 dinfo->serial = g_strdup(serial);
623 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
625 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
627 /* disk I/O throttling */
628 bdrv_set_io_limits(dinfo->bdrv, &io_limits);
635 dinfo->media_cd = media == MEDIA_CDROM;
644 /* add virtio block device */
646 devopts = qemu_opts_create_nofail(qemu_find_opts("device"));
647 if (arch_type == QEMU_ARCH_S390X) {
648 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
650 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
652 qemu_opt_set(devopts, "drive", dinfo->id);
654 qemu_opt_set(devopts, "addr", devaddr);
660 if (!file || !*file) {
661 if (qdict_size(bs_opts)) {
668 /* always use cache=unsafe with snapshot */
669 bdrv_flags &= ~BDRV_O_CACHE_MASK;
670 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
674 bdrv_flags |= BDRV_O_COPY_ON_READ;
677 if (runstate_check(RUN_STATE_INMIGRATE)) {
678 bdrv_flags |= BDRV_O_INCOMING;
681 if (media == MEDIA_CDROM) {
682 /* CDROM is fine for any interface, don't check. */
684 } else if (ro == 1) {
685 if (type != IF_SCSI && type != IF_VIRTIO && type != IF_FLOPPY &&
686 type != IF_NONE && type != IF_PFLASH) {
687 error_report("readonly not supported by this bus type");
692 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
694 if (ro && copy_on_read) {
695 error_report("warning: disabling copy_on_read on readonly drive");
698 ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv);
702 if (ret == -EMEDIUMTYPE) {
703 error_report("could not open disk image %s: not in %s format",
704 file ?: dinfo->id, drv->format_name);
706 error_report("could not open disk image %s: %s",
707 file ?: dinfo->id, strerror(-ret));
712 if (bdrv_key_required(dinfo->bdrv))
722 bdrv_delete(dinfo->bdrv);
724 QTAILQ_REMOVE(&drives, dinfo, next);
729 void do_commit(Monitor *mon, const QDict *qdict)
731 const char *device = qdict_get_str(qdict, "device");
732 BlockDriverState *bs;
735 if (!strcmp(device, "all")) {
736 ret = bdrv_commit_all();
738 bs = bdrv_find(device);
740 monitor_printf(mon, "Device '%s' not found\n", device);
743 ret = bdrv_commit(bs);
746 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
751 static void blockdev_do_action(int kind, void *data, Error **errp)
753 BlockdevAction action;
754 BlockdevActionList list;
758 list.value = &action;
760 qmp_transaction(&list, errp);
763 void qmp_blockdev_snapshot_sync(const char *device, const char *snapshot_file,
764 bool has_format, const char *format,
765 bool has_mode, enum NewImageMode mode,
768 BlockdevSnapshot snapshot = {
769 .device = (char *) device,
770 .snapshot_file = (char *) snapshot_file,
771 .has_format = has_format,
772 .format = (char *) format,
773 .has_mode = has_mode,
776 blockdev_do_action(BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC, &snapshot,
781 /* New and old BlockDriverState structs for group snapshots */
782 typedef struct BlkTransactionStates {
783 BlockDriverState *old_bs;
784 BlockDriverState *new_bs;
785 QSIMPLEQ_ENTRY(BlkTransactionStates) entry;
786 } BlkTransactionStates;
789 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
790 * then we do not pivot any of the devices in the group, and abandon the
793 void qmp_transaction(BlockdevActionList *dev_list, Error **errp)
796 BlockdevActionList *dev_entry = dev_list;
797 BlkTransactionStates *states, *next;
798 Error *local_err = NULL;
800 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionStates) snap_bdrv_states;
801 QSIMPLEQ_INIT(&snap_bdrv_states);
803 /* drain all i/o before any snapshots */
806 /* We don't do anything in this loop that commits us to the snapshot */
807 while (NULL != dev_entry) {
808 BlockdevAction *dev_info = NULL;
809 BlockDriver *proto_drv;
812 enum NewImageMode mode;
813 const char *new_image_file;
815 const char *format = "qcow2";
817 dev_info = dev_entry->value;
818 dev_entry = dev_entry->next;
820 states = g_malloc0(sizeof(BlkTransactionStates));
821 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, states, entry);
823 switch (dev_info->kind) {
824 case BLOCKDEV_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
825 device = dev_info->blockdev_snapshot_sync->device;
826 if (!dev_info->blockdev_snapshot_sync->has_mode) {
827 dev_info->blockdev_snapshot_sync->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
829 new_image_file = dev_info->blockdev_snapshot_sync->snapshot_file;
830 if (dev_info->blockdev_snapshot_sync->has_format) {
831 format = dev_info->blockdev_snapshot_sync->format;
833 mode = dev_info->blockdev_snapshot_sync->mode;
839 drv = bdrv_find_format(format);
841 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
842 goto delete_and_fail;
845 states->old_bs = bdrv_find(device);
846 if (!states->old_bs) {
847 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
848 goto delete_and_fail;
851 if (!bdrv_is_inserted(states->old_bs)) {
852 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
853 goto delete_and_fail;
856 if (bdrv_in_use(states->old_bs)) {
857 error_set(errp, QERR_DEVICE_IN_USE, device);
858 goto delete_and_fail;
861 if (!bdrv_is_read_only(states->old_bs)) {
862 if (bdrv_flush(states->old_bs)) {
863 error_set(errp, QERR_IO_ERROR);
864 goto delete_and_fail;
868 flags = states->old_bs->open_flags;
870 proto_drv = bdrv_find_protocol(new_image_file);
872 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
873 goto delete_and_fail;
876 /* create new image w/backing file */
877 if (mode != NEW_IMAGE_MODE_EXISTING) {
878 bdrv_img_create(new_image_file, format,
879 states->old_bs->filename,
880 states->old_bs->drv->format_name,
881 NULL, -1, flags, &local_err, false);
882 if (error_is_set(&local_err)) {
883 error_propagate(errp, local_err);
884 goto delete_and_fail;
888 /* We will manually add the backing_hd field to the bs later */
889 states->new_bs = bdrv_new("");
890 /* TODO Inherit bs->options or only take explicit options with an
891 * extended QMP command? */
892 ret = bdrv_open(states->new_bs, new_image_file, NULL,
893 flags | BDRV_O_NO_BACKING, drv);
895 error_set(errp, QERR_OPEN_FILE_FAILED, new_image_file);
896 goto delete_and_fail;
901 /* Now we are going to do the actual pivot. Everything up to this point
902 * is reversible, but we are committed at this point */
903 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
904 /* This removes our old bs from the bdrv_states, and adds the new bs */
905 bdrv_append(states->new_bs, states->old_bs);
906 /* We don't need (or want) to use the transactional
907 * bdrv_reopen_multiple() across all the entries at once, because we
908 * don't want to abort all of them if one of them fails the reopen */
909 bdrv_reopen(states->new_bs, states->new_bs->open_flags & ~BDRV_O_RDWR,
918 * failure, and it is all-or-none; abandon each new bs, and keep using
919 * the original bs for all images
921 QSIMPLEQ_FOREACH(states, &snap_bdrv_states, entry) {
922 if (states->new_bs) {
923 bdrv_delete(states->new_bs);
927 QSIMPLEQ_FOREACH_SAFE(states, &snap_bdrv_states, entry, next) {
933 static void eject_device(BlockDriverState *bs, int force, Error **errp)
935 if (bdrv_in_use(bs)) {
936 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
939 if (!bdrv_dev_has_removable_media(bs)) {
940 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
944 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
945 bdrv_dev_eject_request(bs, force);
947 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
955 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
957 BlockDriverState *bs;
959 bs = bdrv_find(device);
961 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
965 eject_device(bs, force, errp);
968 void qmp_block_passwd(const char *device, const char *password, Error **errp)
970 BlockDriverState *bs;
973 bs = bdrv_find(device);
975 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
979 err = bdrv_set_key(bs, password);
980 if (err == -EINVAL) {
981 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
983 } else if (err < 0) {
984 error_set(errp, QERR_INVALID_PASSWORD);
989 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
990 int bdrv_flags, BlockDriver *drv,
991 const char *password, Error **errp)
993 if (bdrv_open(bs, filename, NULL, bdrv_flags, drv) < 0) {
994 error_set(errp, QERR_OPEN_FILE_FAILED, filename);
998 if (bdrv_key_required(bs)) {
1000 if (bdrv_set_key(bs, password) < 0) {
1001 error_set(errp, QERR_INVALID_PASSWORD);
1004 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1005 bdrv_get_encrypted_filename(bs));
1007 } else if (password) {
1008 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1012 void qmp_change_blockdev(const char *device, const char *filename,
1013 bool has_format, const char *format, Error **errp)
1015 BlockDriverState *bs;
1016 BlockDriver *drv = NULL;
1020 bs = bdrv_find(device);
1022 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1027 drv = bdrv_find_whitelisted_format(format);
1029 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1034 eject_device(bs, 0, &err);
1035 if (error_is_set(&err)) {
1036 error_propagate(errp, err);
1040 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1041 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1043 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1046 /* throttling disk I/O limits */
1047 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1048 int64_t bps_wr, int64_t iops, int64_t iops_rd,
1049 int64_t iops_wr, Error **errp)
1051 BlockIOLimit io_limits;
1052 BlockDriverState *bs;
1054 bs = bdrv_find(device);
1056 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1060 io_limits.bps[BLOCK_IO_LIMIT_TOTAL] = bps;
1061 io_limits.bps[BLOCK_IO_LIMIT_READ] = bps_rd;
1062 io_limits.bps[BLOCK_IO_LIMIT_WRITE] = bps_wr;
1063 io_limits.iops[BLOCK_IO_LIMIT_TOTAL]= iops;
1064 io_limits.iops[BLOCK_IO_LIMIT_READ] = iops_rd;
1065 io_limits.iops[BLOCK_IO_LIMIT_WRITE]= iops_wr;
1067 if (!do_check_io_limits(&io_limits, errp)) {
1071 bs->io_limits = io_limits;
1073 if (!bs->io_limits_enabled && bdrv_io_limits_enabled(bs)) {
1074 bdrv_io_limits_enable(bs);
1075 } else if (bs->io_limits_enabled && !bdrv_io_limits_enabled(bs)) {
1076 bdrv_io_limits_disable(bs);
1078 if (bs->block_timer) {
1079 qemu_mod_timer(bs->block_timer, qemu_get_clock_ns(vm_clock));
1084 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1086 const char *id = qdict_get_str(qdict, "id");
1087 BlockDriverState *bs;
1091 qerror_report(QERR_DEVICE_NOT_FOUND, id);
1094 if (bdrv_in_use(bs)) {
1095 qerror_report(QERR_DEVICE_IN_USE, id);
1099 /* quiesce block driver; prevent further io */
1104 /* if we have a device attached to this BlockDriverState
1105 * then we need to make the drive anonymous until the device
1106 * can be removed. If this is a drive with no device backing
1107 * then we can just get rid of the block driver state right here.
1109 if (bdrv_get_attached_dev(bs)) {
1112 drive_uninit(drive_get_by_blockdev(bs));
1118 void qmp_block_resize(const char *device, int64_t size, Error **errp)
1120 BlockDriverState *bs;
1122 bs = bdrv_find(device);
1124 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1129 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1133 /* complete all in-flight operations before resizing the device */
1136 switch (bdrv_truncate(bs, size)) {
1140 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1143 error_set(errp, QERR_UNSUPPORTED);
1146 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1149 error_set(errp, QERR_DEVICE_IN_USE, device);
1152 error_set(errp, QERR_UNDEFINED_ERROR);
1157 static void block_job_cb(void *opaque, int ret)
1159 BlockDriverState *bs = opaque;
1162 trace_block_job_cb(bs, bs->job, ret);
1165 obj = qobject_from_block_job(bs->job);
1167 QDict *dict = qobject_to_qdict(obj);
1168 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1171 if (block_job_is_cancelled(bs->job)) {
1172 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1174 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1176 qobject_decref(obj);
1178 drive_put_ref_bh_schedule(drive_get_by_blockdev(bs));
1181 void qmp_block_stream(const char *device, bool has_base,
1182 const char *base, bool has_speed, int64_t speed,
1183 bool has_on_error, BlockdevOnError on_error,
1186 BlockDriverState *bs;
1187 BlockDriverState *base_bs = NULL;
1188 Error *local_err = NULL;
1190 if (!has_on_error) {
1191 on_error = BLOCKDEV_ON_ERROR_REPORT;
1194 bs = bdrv_find(device);
1196 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1201 base_bs = bdrv_find_backing_image(bs, base);
1202 if (base_bs == NULL) {
1203 error_set(errp, QERR_BASE_NOT_FOUND, base);
1208 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1209 on_error, block_job_cb, bs, &local_err);
1210 if (error_is_set(&local_err)) {
1211 error_propagate(errp, local_err);
1215 /* Grab a reference so hotplug does not delete the BlockDriverState from
1218 drive_get_ref(drive_get_by_blockdev(bs));
1220 trace_qmp_block_stream(bs, bs->job);
1223 void qmp_block_commit(const char *device,
1224 bool has_base, const char *base, const char *top,
1225 bool has_speed, int64_t speed,
1228 BlockDriverState *bs;
1229 BlockDriverState *base_bs, *top_bs;
1230 Error *local_err = NULL;
1231 /* This will be part of the QMP command, if/when the
1232 * BlockdevOnError change for blkmirror makes it in
1234 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1236 /* drain all i/o before commits */
1239 bs = bdrv_find(device);
1241 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1245 /* default top_bs is the active layer */
1249 if (strcmp(bs->filename, top) != 0) {
1250 top_bs = bdrv_find_backing_image(bs, top);
1254 if (top_bs == NULL) {
1255 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1259 if (has_base && base) {
1260 base_bs = bdrv_find_backing_image(top_bs, base);
1262 base_bs = bdrv_find_base(top_bs);
1265 if (base_bs == NULL) {
1266 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1270 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1272 if (local_err != NULL) {
1273 error_propagate(errp, local_err);
1276 /* Grab a reference so hotplug does not delete the BlockDriverState from
1279 drive_get_ref(drive_get_by_blockdev(bs));
1282 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
1284 void qmp_drive_mirror(const char *device, const char *target,
1285 bool has_format, const char *format,
1286 enum MirrorSyncMode sync,
1287 bool has_mode, enum NewImageMode mode,
1288 bool has_speed, int64_t speed,
1289 bool has_granularity, uint32_t granularity,
1290 bool has_buf_size, int64_t buf_size,
1291 bool has_on_source_error, BlockdevOnError on_source_error,
1292 bool has_on_target_error, BlockdevOnError on_target_error,
1295 BlockDriverState *bs;
1296 BlockDriverState *source, *target_bs;
1297 BlockDriver *proto_drv;
1298 BlockDriver *drv = NULL;
1299 Error *local_err = NULL;
1307 if (!has_on_source_error) {
1308 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1310 if (!has_on_target_error) {
1311 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1314 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1316 if (!has_granularity) {
1319 if (!has_buf_size) {
1320 buf_size = DEFAULT_MIRROR_BUF_SIZE;
1323 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
1324 error_set(errp, QERR_INVALID_PARAMETER, device);
1327 if (granularity & (granularity - 1)) {
1328 error_set(errp, QERR_INVALID_PARAMETER, device);
1332 bs = bdrv_find(device);
1334 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1338 if (!bdrv_is_inserted(bs)) {
1339 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1344 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1347 drv = bdrv_find_format(format);
1349 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1354 if (bdrv_in_use(bs)) {
1355 error_set(errp, QERR_DEVICE_IN_USE, device);
1359 flags = bs->open_flags | BDRV_O_RDWR;
1360 source = bs->backing_hd;
1361 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
1362 sync = MIRROR_SYNC_MODE_FULL;
1365 proto_drv = bdrv_find_protocol(target);
1367 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1371 bdrv_get_geometry(bs, &size);
1373 if (sync == MIRROR_SYNC_MODE_FULL && mode != NEW_IMAGE_MODE_EXISTING) {
1374 /* create new image w/o backing file */
1375 assert(format && drv);
1376 bdrv_img_create(target, format,
1377 NULL, NULL, NULL, size, flags, &local_err, false);
1380 case NEW_IMAGE_MODE_EXISTING:
1383 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
1384 /* create new image with backing file */
1385 bdrv_img_create(target, format,
1387 source->drv->format_name,
1388 NULL, size, flags, &local_err, false);
1395 if (error_is_set(&local_err)) {
1396 error_propagate(errp, local_err);
1400 /* Mirroring takes care of copy-on-write using the source's backing
1403 target_bs = bdrv_new("");
1404 ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv);
1407 bdrv_delete(target_bs);
1408 error_set(errp, QERR_OPEN_FILE_FAILED, target);
1412 mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
1413 on_source_error, on_target_error,
1414 block_job_cb, bs, &local_err);
1415 if (local_err != NULL) {
1416 bdrv_delete(target_bs);
1417 error_propagate(errp, local_err);
1421 /* Grab a reference so hotplug does not delete the BlockDriverState from
1424 drive_get_ref(drive_get_by_blockdev(bs));
1427 static BlockJob *find_block_job(const char *device)
1429 BlockDriverState *bs;
1431 bs = bdrv_find(device);
1432 if (!bs || !bs->job) {
1438 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
1440 BlockJob *job = find_block_job(device);
1443 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1447 block_job_set_speed(job, speed, errp);
1450 void qmp_block_job_cancel(const char *device,
1451 bool has_force, bool force, Error **errp)
1453 BlockJob *job = find_block_job(device);
1460 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1463 if (job->paused && !force) {
1464 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
1468 trace_qmp_block_job_cancel(job);
1469 block_job_cancel(job);
1472 void qmp_block_job_pause(const char *device, Error **errp)
1474 BlockJob *job = find_block_job(device);
1477 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1481 trace_qmp_block_job_pause(job);
1482 block_job_pause(job);
1485 void qmp_block_job_resume(const char *device, Error **errp)
1487 BlockJob *job = find_block_job(device);
1490 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1494 trace_qmp_block_job_resume(job);
1495 block_job_resume(job);
1498 void qmp_block_job_complete(const char *device, Error **errp)
1500 BlockJob *job = find_block_job(device);
1503 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
1507 trace_qmp_block_job_complete(job);
1508 block_job_complete(job, errp);
1511 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
1513 BlockJobInfoList **prev = opaque;
1514 BlockJob *job = bs->job;
1517 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
1518 elem->value = block_job_query(bs->job);
1519 (*prev)->next = elem;
1524 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
1526 /* Dummy is a fake list element for holding the head pointer */
1527 BlockJobInfoList dummy = {};
1528 BlockJobInfoList *prev = &dummy;
1529 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
1533 QemuOptsList qemu_common_drive_opts = {
1535 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
1539 .type = QEMU_OPT_NUMBER,
1540 .help = "bus number",
1543 .type = QEMU_OPT_NUMBER,
1544 .help = "unit number (i.e. lun for scsi)",
1547 .type = QEMU_OPT_STRING,
1548 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
1551 .type = QEMU_OPT_NUMBER,
1552 .help = "index number",
1555 .type = QEMU_OPT_NUMBER,
1556 .help = "number of cylinders (ide disk geometry)",
1559 .type = QEMU_OPT_NUMBER,
1560 .help = "number of heads (ide disk geometry)",
1563 .type = QEMU_OPT_NUMBER,
1564 .help = "number of sectors (ide disk geometry)",
1567 .type = QEMU_OPT_STRING,
1568 .help = "chs translation (auto, lba. none)",
1571 .type = QEMU_OPT_STRING,
1572 .help = "media type (disk, cdrom)",
1575 .type = QEMU_OPT_BOOL,
1576 .help = "enable/disable snapshot mode",
1579 .type = QEMU_OPT_STRING,
1580 .help = "disk image",
1583 .type = QEMU_OPT_STRING,
1584 .help = "discard operation (ignore/off, unmap/on)",
1587 .type = QEMU_OPT_STRING,
1588 .help = "host cache usage (none, writeback, writethrough, "
1589 "directsync, unsafe)",
1592 .type = QEMU_OPT_STRING,
1593 .help = "host AIO implementation (threads, native)",
1596 .type = QEMU_OPT_STRING,
1597 .help = "disk format (raw, qcow2, ...)",
1600 .type = QEMU_OPT_STRING,
1601 .help = "disk serial number",
1604 .type = QEMU_OPT_STRING,
1605 .help = "read error action",
1608 .type = QEMU_OPT_STRING,
1609 .help = "write error action",
1612 .type = QEMU_OPT_STRING,
1613 .help = "pci address (virtio only)",
1616 .type = QEMU_OPT_BOOL,
1617 .help = "open drive file as read-only",
1620 .type = QEMU_OPT_NUMBER,
1621 .help = "limit total I/O operations per second",
1624 .type = QEMU_OPT_NUMBER,
1625 .help = "limit read operations per second",
1628 .type = QEMU_OPT_NUMBER,
1629 .help = "limit write operations per second",
1632 .type = QEMU_OPT_NUMBER,
1633 .help = "limit total bytes per second",
1636 .type = QEMU_OPT_NUMBER,
1637 .help = "limit read bytes per second",
1640 .type = QEMU_OPT_NUMBER,
1641 .help = "limit write bytes per second",
1643 .name = "copy-on-read",
1644 .type = QEMU_OPT_BOOL,
1645 .help = "copy read data from backing file into image file",
1648 .type = QEMU_OPT_BOOL,
1649 .help = "(deprecated, ignored)",
1651 { /* end of list */ }
1655 QemuOptsList qemu_drive_opts = {
1657 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
1660 * no elements => accept any params
1661 * validation will happen later
1663 { /* end of list */ }