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 "qemu/option.h"
38 #include "qemu/config-file.h"
39 #include "qapi/qmp/types.h"
40 #include "qapi-visit.h"
41 #include "qapi/qmp-output-visitor.h"
42 #include "sysemu/sysemu.h"
43 #include "block/block_int.h"
44 #include "qmp-commands.h"
46 #include "sysemu/arch_init.h"
48 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
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);
92 if (dinfo && !dinfo->enable_auto_del) {
97 block_job_cancel(bs->job);
104 void blockdev_auto_del(BlockDriverState *bs)
106 DriveInfo *dinfo = drive_get_by_blockdev(bs);
108 if (dinfo && dinfo->auto_del) {
113 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
115 int max_devs = if_max_devs[type];
116 return max_devs ? index / max_devs : 0;
119 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
121 int max_devs = if_max_devs[type];
122 return max_devs ? index % max_devs : index;
125 QemuOpts *drive_def(const char *optstr)
127 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
130 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
136 opts = drive_def(optstr);
140 if (type != IF_DEFAULT) {
141 qemu_opt_set(opts, "if", if_name[type]);
144 snprintf(buf, sizeof(buf), "%d", index);
145 qemu_opt_set(opts, "index", buf);
148 qemu_opt_set(opts, "file", file);
152 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
156 /* seek interface, bus and unit */
158 QTAILQ_FOREACH(dinfo, &drives, next) {
159 if (dinfo->type == type &&
168 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
170 return drive_get(type,
171 drive_index_to_bus_id(type, index),
172 drive_index_to_unit_id(type, index));
175 int drive_get_max_bus(BlockInterfaceType type)
181 QTAILQ_FOREACH(dinfo, &drives, next) {
182 if(dinfo->type == type &&
183 dinfo->bus > max_bus)
184 max_bus = dinfo->bus;
189 /* Get a block device. This should only be used for single-drive devices
190 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
192 DriveInfo *drive_get_next(BlockInterfaceType type)
194 static int next_block_unit[IF_COUNT];
196 return drive_get(type, 0, next_block_unit[type]++);
199 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
203 QTAILQ_FOREACH(dinfo, &drives, next) {
204 if (dinfo->bdrv == bs) {
211 static void bdrv_format_print(void *opaque, const char *name)
213 error_printf(" %s", name);
216 void drive_del(DriveInfo *dinfo)
219 qemu_opts_del(dinfo->opts);
222 bdrv_unref(dinfo->bdrv);
224 QTAILQ_REMOVE(&drives, dinfo, next);
225 g_free(dinfo->serial);
231 BlockDriverState *bs;
234 static void bdrv_put_ref_bh(void *opaque)
236 BDRVPutRefBH *s = opaque;
239 qemu_bh_delete(s->bh);
244 * Release a BDS reference in a BH
246 * It is not safe to use bdrv_unref() from a callback function when the callers
247 * still need the BlockDriverState. In such cases we schedule a BH to release
250 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
254 s = g_new(BDRVPutRefBH, 1);
255 s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
257 qemu_bh_schedule(s->bh);
260 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
262 if (!strcmp(buf, "ignore")) {
263 return BLOCKDEV_ON_ERROR_IGNORE;
264 } else if (!is_read && !strcmp(buf, "enospc")) {
265 return BLOCKDEV_ON_ERROR_ENOSPC;
266 } else if (!strcmp(buf, "stop")) {
267 return BLOCKDEV_ON_ERROR_STOP;
268 } else if (!strcmp(buf, "report")) {
269 return BLOCKDEV_ON_ERROR_REPORT;
271 error_setg(errp, "'%s' invalid %s error action",
272 buf, is_read ? "read" : "write");
277 static inline int parse_enum_option(const char *lookup[], const char *buf,
278 int max, int def, Error **errp)
286 for (i = 0; i < max; i++) {
287 if (!strcmp(buf, lookup[i])) {
292 error_setg(errp, "invalid parameter value: %s", buf);
296 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
298 if (throttle_conflicting(cfg)) {
299 error_setg(errp, "bps/iops/max total values and read/write values"
300 " cannot be used at the same time");
304 if (!throttle_is_valid(cfg)) {
305 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
312 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
314 /* Takes the ownership of bs_opts */
315 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
321 int on_read_error, on_write_error;
330 bool has_driver_specific_opts;
331 BlockdevDetectZeroesOptions detect_zeroes;
332 BlockDriver *drv = NULL;
334 /* Check common options by copying from bs_opts to opts, all other options
335 * stay in bs_opts for processing by bdrv_open(). */
336 id = qdict_get_try_str(bs_opts, "id");
337 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
339 error_propagate(errp, error);
343 qemu_opts_absorb_qdict(opts, bs_opts, &error);
345 error_propagate(errp, error);
350 qdict_del(bs_opts, "id");
353 has_driver_specific_opts = !!qdict_size(bs_opts);
355 /* extract parameters */
356 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
357 ro = qemu_opt_get_bool(opts, "read-only", 0);
358 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
360 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
361 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
362 error_setg(errp, "invalid discard option");
367 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
368 bdrv_flags |= BDRV_O_CACHE_WB;
370 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
371 bdrv_flags |= BDRV_O_NOCACHE;
373 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
374 bdrv_flags |= BDRV_O_NO_FLUSH;
377 #ifdef CONFIG_LINUX_AIO
378 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
379 if (!strcmp(buf, "native")) {
380 bdrv_flags |= BDRV_O_NATIVE_AIO;
381 } else if (!strcmp(buf, "threads")) {
382 /* this is the default */
384 error_setg(errp, "invalid aio option");
390 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
391 if (is_help_option(buf)) {
392 error_printf("Supported formats:");
393 bdrv_iterate_format(bdrv_format_print, NULL);
398 drv = bdrv_find_format(buf);
400 error_setg(errp, "'%s' invalid format", buf);
405 /* disk I/O throttling */
406 memset(&cfg, 0, sizeof(cfg));
407 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
408 qemu_opt_get_number(opts, "throttling.bps-total", 0);
409 cfg.buckets[THROTTLE_BPS_READ].avg =
410 qemu_opt_get_number(opts, "throttling.bps-read", 0);
411 cfg.buckets[THROTTLE_BPS_WRITE].avg =
412 qemu_opt_get_number(opts, "throttling.bps-write", 0);
413 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
414 qemu_opt_get_number(opts, "throttling.iops-total", 0);
415 cfg.buckets[THROTTLE_OPS_READ].avg =
416 qemu_opt_get_number(opts, "throttling.iops-read", 0);
417 cfg.buckets[THROTTLE_OPS_WRITE].avg =
418 qemu_opt_get_number(opts, "throttling.iops-write", 0);
420 cfg.buckets[THROTTLE_BPS_TOTAL].max =
421 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
422 cfg.buckets[THROTTLE_BPS_READ].max =
423 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
424 cfg.buckets[THROTTLE_BPS_WRITE].max =
425 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
426 cfg.buckets[THROTTLE_OPS_TOTAL].max =
427 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
428 cfg.buckets[THROTTLE_OPS_READ].max =
429 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
430 cfg.buckets[THROTTLE_OPS_WRITE].max =
431 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
433 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
435 if (!check_throttle_config(&cfg, &error)) {
436 error_propagate(errp, error);
440 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
441 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
442 on_write_error = parse_block_error_action(buf, 0, &error);
444 error_propagate(errp, error);
449 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
450 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
451 on_read_error = parse_block_error_action(buf, 1, &error);
453 error_propagate(errp, error);
459 parse_enum_option(BlockdevDetectZeroesOptions_lookup,
460 qemu_opt_get(opts, "detect-zeroes"),
461 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
462 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
465 error_propagate(errp, error);
469 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
470 !(bdrv_flags & BDRV_O_UNMAP)) {
471 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
472 "without setting discard operation to unmap");
477 dinfo = g_malloc0(sizeof(*dinfo));
478 dinfo->id = g_strdup(qemu_opts_id(opts));
479 dinfo->bdrv = bdrv_new(dinfo->id, &error);
481 error_propagate(errp, error);
484 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
485 dinfo->bdrv->read_only = ro;
486 dinfo->bdrv->detect_zeroes = detect_zeroes;
487 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
489 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
491 /* disk I/O throttling */
492 if (throttle_enabled(&cfg)) {
493 bdrv_io_limits_enable(dinfo->bdrv);
494 bdrv_set_io_limits(dinfo->bdrv, &cfg);
497 if (!file || !*file) {
498 if (has_driver_specific_opts) {
507 /* always use cache=unsafe with snapshot */
508 bdrv_flags &= ~BDRV_O_CACHE_MASK;
509 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
513 bdrv_flags |= BDRV_O_COPY_ON_READ;
516 if (runstate_check(RUN_STATE_INMIGRATE)) {
517 bdrv_flags |= BDRV_O_INCOMING;
520 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
523 ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
526 error_setg(errp, "could not open disk image %s: %s",
527 file ?: dinfo->id, error_get_pretty(error));
532 if (bdrv_key_required(dinfo->bdrv))
541 bdrv_unref(dinfo->bdrv);
542 QTAILQ_REMOVE(&drives, dinfo, next);
553 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
557 value = qemu_opt_get(opts, from);
559 qemu_opt_set(opts, to, value);
560 qemu_opt_unset(opts, from);
564 QemuOptsList qemu_legacy_drive_opts = {
566 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
570 .type = QEMU_OPT_NUMBER,
571 .help = "bus number",
574 .type = QEMU_OPT_NUMBER,
575 .help = "unit number (i.e. lun for scsi)",
578 .type = QEMU_OPT_NUMBER,
579 .help = "index number",
582 .type = QEMU_OPT_STRING,
583 .help = "media type (disk, cdrom)",
586 .type = QEMU_OPT_STRING,
587 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
590 .type = QEMU_OPT_NUMBER,
591 .help = "number of cylinders (ide disk geometry)",
594 .type = QEMU_OPT_NUMBER,
595 .help = "number of heads (ide disk geometry)",
598 .type = QEMU_OPT_NUMBER,
599 .help = "number of sectors (ide disk geometry)",
602 .type = QEMU_OPT_STRING,
603 .help = "chs translation (auto, lba, none)",
606 .type = QEMU_OPT_BOOL,
607 .help = "(deprecated, ignored)",
610 .type = QEMU_OPT_STRING,
611 .help = "pci address (virtio only)",
614 .type = QEMU_OPT_STRING,
615 .help = "disk serial number",
618 .type = QEMU_OPT_STRING,
622 /* Options that are passed on, but have special semantics with -drive */
625 .type = QEMU_OPT_BOOL,
626 .help = "open drive file as read-only",
629 .type = QEMU_OPT_STRING,
630 .help = "read error action",
633 .type = QEMU_OPT_STRING,
634 .help = "write error action",
636 .name = "copy-on-read",
637 .type = QEMU_OPT_BOOL,
638 .help = "copy read data from backing file into image file",
641 { /* end of list */ }
645 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type)
648 DriveInfo *dinfo = NULL;
650 QemuOpts *legacy_opts;
651 DriveMediaType media = MEDIA_DISK;
652 BlockInterfaceType type;
653 int cyls, heads, secs, translation;
654 int max_devs, bus_id, unit_id, index;
656 const char *werror, *rerror;
657 bool read_only = false;
660 const char *filename;
661 Error *local_err = NULL;
663 /* Change legacy command line options into QMP ones */
664 qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
665 qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
666 qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
668 qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
669 qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
670 qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
672 qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
673 qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
674 qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
676 qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
677 qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
678 qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
680 qemu_opt_rename(all_opts,
681 "iops_size", "throttling.iops-size");
683 qemu_opt_rename(all_opts, "readonly", "read-only");
685 value = qemu_opt_get(all_opts, "cache");
689 if (bdrv_parse_cache_flags(value, &flags) != 0) {
690 error_report("invalid cache option");
694 /* Specific options take precedence */
695 if (!qemu_opt_get(all_opts, "cache.writeback")) {
696 qemu_opt_set_bool(all_opts, "cache.writeback",
697 !!(flags & BDRV_O_CACHE_WB));
699 if (!qemu_opt_get(all_opts, "cache.direct")) {
700 qemu_opt_set_bool(all_opts, "cache.direct",
701 !!(flags & BDRV_O_NOCACHE));
703 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
704 qemu_opt_set_bool(all_opts, "cache.no-flush",
705 !!(flags & BDRV_O_NO_FLUSH));
707 qemu_opt_unset(all_opts, "cache");
710 /* Get a QDict for processing the options */
711 bs_opts = qdict_new();
712 qemu_opts_to_qdict(all_opts, bs_opts);
714 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
716 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
718 error_report("%s", error_get_pretty(local_err));
719 error_free(local_err);
723 /* Deprecated option boot=[on|off] */
724 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
725 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
726 "ignored. Future versions will reject this parameter. Please "
727 "update your scripts.\n");
731 value = qemu_opt_get(legacy_opts, "media");
733 if (!strcmp(value, "disk")) {
735 } else if (!strcmp(value, "cdrom")) {
739 error_report("'%s' invalid media", value);
744 /* copy-on-read is disabled with a warning for read-only devices */
745 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
746 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
748 if (read_only && copy_on_read) {
749 error_report("warning: disabling copy-on-read on read-only drive");
750 copy_on_read = false;
753 qdict_put(bs_opts, "read-only",
754 qstring_from_str(read_only ? "on" : "off"));
755 qdict_put(bs_opts, "copy-on-read",
756 qstring_from_str(copy_on_read ? "on" :"off"));
758 /* Controller type */
759 value = qemu_opt_get(legacy_opts, "if");
762 type < IF_COUNT && strcmp(value, if_name[type]);
765 if (type == IF_COUNT) {
766 error_report("unsupported bus type '%s'", value);
770 type = block_default_type;
774 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
775 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
776 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
778 if (cyls || heads || secs) {
780 error_report("invalid physical cyls number");
784 error_report("invalid physical heads number");
788 error_report("invalid physical secs number");
793 translation = BIOS_ATA_TRANSLATION_AUTO;
794 value = qemu_opt_get(legacy_opts, "trans");
797 error_report("'%s' trans must be used with cyls, heads and secs",
801 if (!strcmp(value, "none")) {
802 translation = BIOS_ATA_TRANSLATION_NONE;
803 } else if (!strcmp(value, "lba")) {
804 translation = BIOS_ATA_TRANSLATION_LBA;
805 } else if (!strcmp(value, "large")) {
806 translation = BIOS_ATA_TRANSLATION_LARGE;
807 } else if (!strcmp(value, "rechs")) {
808 translation = BIOS_ATA_TRANSLATION_RECHS;
809 } else if (!strcmp(value, "auto")) {
810 translation = BIOS_ATA_TRANSLATION_AUTO;
812 error_report("'%s' invalid translation type", value);
817 if (media == MEDIA_CDROM) {
818 if (cyls || secs || heads) {
819 error_report("CHS can't be set with media=cdrom");
824 /* Device address specified by bus/unit or index.
825 * If none was specified, try to find the first free one. */
826 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
827 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
828 index = qemu_opt_get_number(legacy_opts, "index", -1);
830 max_devs = if_max_devs[type];
833 if (bus_id != 0 || unit_id != -1) {
834 error_report("index cannot be used with bus and unit");
837 bus_id = drive_index_to_bus_id(type, index);
838 unit_id = drive_index_to_unit_id(type, index);
843 while (drive_get(type, bus_id, unit_id) != NULL) {
845 if (max_devs && unit_id >= max_devs) {
852 if (max_devs && unit_id >= max_devs) {
853 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
857 if (drive_get(type, bus_id, unit_id) != NULL) {
858 error_report("drive with bus=%d, unit=%d (index=%d) exists",
859 bus_id, unit_id, index);
864 serial = qemu_opt_get(legacy_opts, "serial");
866 /* no id supplied -> create one */
867 if (qemu_opts_id(all_opts) == NULL) {
869 const char *mediastr = "";
870 if (type == IF_IDE || type == IF_SCSI) {
871 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
874 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
877 new_id = g_strdup_printf("%s%s%i", if_name[type],
880 qdict_put(bs_opts, "id", qstring_from_str(new_id));
884 /* Add virtio block device */
885 devaddr = qemu_opt_get(legacy_opts, "addr");
886 if (devaddr && type != IF_VIRTIO) {
887 error_report("addr is not supported by this bus type");
891 if (type == IF_VIRTIO) {
893 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
895 if (arch_type == QEMU_ARCH_S390X) {
896 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
898 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
900 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
902 qemu_opt_set(devopts, "addr", devaddr);
906 filename = qemu_opt_get(legacy_opts, "file");
908 /* Check werror/rerror compatibility with if=... */
909 werror = qemu_opt_get(legacy_opts, "werror");
910 if (werror != NULL) {
911 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
913 error_report("werror is not supported by this bus type");
916 qdict_put(bs_opts, "werror", qstring_from_str(werror));
919 rerror = qemu_opt_get(legacy_opts, "rerror");
920 if (rerror != NULL) {
921 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
923 error_report("rerror is not supported by this bus type");
926 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
929 /* Actual block device init: Functionality shared with blockdev-add */
930 dinfo = blockdev_init(filename, bs_opts, &local_err);
934 error_report("%s", error_get_pretty(local_err));
935 error_free(local_err);
942 /* Set legacy DriveInfo fields */
943 dinfo->enable_auto_del = true;
944 dinfo->opts = all_opts;
947 dinfo->heads = heads;
949 dinfo->trans = translation;
953 dinfo->unit = unit_id;
954 dinfo->devaddr = devaddr;
956 dinfo->serial = g_strdup(serial);
963 dinfo->media_cd = media == MEDIA_CDROM;
970 qemu_opts_del(legacy_opts);
975 void do_commit(Monitor *mon, const QDict *qdict)
977 const char *device = qdict_get_str(qdict, "device");
978 BlockDriverState *bs;
981 if (!strcmp(device, "all")) {
982 ret = bdrv_commit_all();
984 bs = bdrv_find(device);
986 monitor_printf(mon, "Device '%s' not found\n", device);
989 ret = bdrv_commit(bs);
992 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
997 static void blockdev_do_action(int kind, void *data, Error **errp)
999 TransactionAction action;
1000 TransactionActionList list;
1004 list.value = &action;
1006 qmp_transaction(&list, errp);
1009 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1010 bool has_node_name, const char *node_name,
1011 const char *snapshot_file,
1012 bool has_snapshot_node_name,
1013 const char *snapshot_node_name,
1014 bool has_format, const char *format,
1015 bool has_mode, NewImageMode mode, Error **errp)
1017 BlockdevSnapshot snapshot = {
1018 .has_device = has_device,
1019 .device = (char *) device,
1020 .has_node_name = has_node_name,
1021 .node_name = (char *) node_name,
1022 .snapshot_file = (char *) snapshot_file,
1023 .has_snapshot_node_name = has_snapshot_node_name,
1024 .snapshot_node_name = (char *) snapshot_node_name,
1025 .has_format = has_format,
1026 .format = (char *) format,
1027 .has_mode = has_mode,
1030 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1034 void qmp_blockdev_snapshot_internal_sync(const char *device,
1038 BlockdevSnapshotInternal snapshot = {
1039 .device = (char *) device,
1040 .name = (char *) name
1043 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1047 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1054 BlockDriverState *bs = bdrv_find(device);
1055 QEMUSnapshotInfo sn;
1056 Error *local_err = NULL;
1057 SnapshotInfo *info = NULL;
1061 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1074 error_setg(errp, "Name or id must be provided");
1078 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1080 error_propagate(errp, local_err);
1085 "Snapshot with id '%s' and name '%s' does not exist on "
1087 STR_OR_NULL(id), STR_OR_NULL(name), device);
1091 bdrv_snapshot_delete(bs, id, name, &local_err);
1093 error_propagate(errp, local_err);
1097 info = g_malloc0(sizeof(SnapshotInfo));
1098 info->id = g_strdup(sn.id_str);
1099 info->name = g_strdup(sn.name);
1100 info->date_nsec = sn.date_nsec;
1101 info->date_sec = sn.date_sec;
1102 info->vm_state_size = sn.vm_state_size;
1103 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1104 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1109 /* New and old BlockDriverState structs for group snapshots */
1111 typedef struct BlkTransactionState BlkTransactionState;
1113 /* Only prepare() may fail. In a single transaction, only one of commit() or
1114 abort() will be called, clean() will always be called if it present. */
1115 typedef struct BdrvActionOps {
1116 /* Size of state struct, in bytes. */
1117 size_t instance_size;
1118 /* Prepare the work, must NOT be NULL. */
1119 void (*prepare)(BlkTransactionState *common, Error **errp);
1120 /* Commit the changes, can be NULL. */
1121 void (*commit)(BlkTransactionState *common);
1122 /* Abort the changes on fail, can be NULL. */
1123 void (*abort)(BlkTransactionState *common);
1124 /* Clean up resource in the end, can be NULL. */
1125 void (*clean)(BlkTransactionState *common);
1129 * This structure must be arranged as first member in child type, assuming
1130 * that compiler will also arrange it to the same address with parent instance.
1131 * Later it will be used in free().
1133 struct BlkTransactionState {
1134 TransactionAction *action;
1135 const BdrvActionOps *ops;
1136 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1139 /* internal snapshot private data */
1140 typedef struct InternalSnapshotState {
1141 BlkTransactionState common;
1142 BlockDriverState *bs;
1143 QEMUSnapshotInfo sn;
1144 } InternalSnapshotState;
1146 static void internal_snapshot_prepare(BlkTransactionState *common,
1149 Error *local_err = NULL;
1152 BlockDriverState *bs;
1153 QEMUSnapshotInfo old_sn, *sn;
1156 BlockdevSnapshotInternal *internal;
1157 InternalSnapshotState *state;
1160 g_assert(common->action->kind ==
1161 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1162 internal = common->action->blockdev_snapshot_internal_sync;
1163 state = DO_UPCAST(InternalSnapshotState, common, common);
1165 /* 1. parse input */
1166 device = internal->device;
1167 name = internal->name;
1169 /* 2. check for validation */
1170 bs = bdrv_find(device);
1172 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1176 if (!bdrv_is_inserted(bs)) {
1177 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1181 if (bdrv_is_read_only(bs)) {
1182 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1186 if (!bdrv_can_snapshot(bs)) {
1187 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1188 bs->drv->format_name, device, "internal snapshot");
1192 if (!strlen(name)) {
1193 error_setg(errp, "Name is empty");
1197 /* check whether a snapshot with name exist */
1198 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1201 error_propagate(errp, local_err);
1205 "Snapshot with name '%s' already exists on device '%s'",
1210 /* 3. take the snapshot */
1212 pstrcpy(sn->name, sizeof(sn->name), name);
1213 qemu_gettimeofday(&tv);
1214 sn->date_sec = tv.tv_sec;
1215 sn->date_nsec = tv.tv_usec * 1000;
1216 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1218 ret1 = bdrv_snapshot_create(bs, sn);
1220 error_setg_errno(errp, -ret1,
1221 "Failed to create snapshot '%s' on device '%s'",
1226 /* 4. succeed, mark a snapshot is created */
1230 static void internal_snapshot_abort(BlkTransactionState *common)
1232 InternalSnapshotState *state =
1233 DO_UPCAST(InternalSnapshotState, common, common);
1234 BlockDriverState *bs = state->bs;
1235 QEMUSnapshotInfo *sn = &state->sn;
1236 Error *local_error = NULL;
1242 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1243 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1244 "device '%s' in abort: %s",
1247 bdrv_get_device_name(bs),
1248 error_get_pretty(local_error));
1249 error_free(local_error);
1253 /* external snapshot private data */
1254 typedef struct ExternalSnapshotState {
1255 BlkTransactionState common;
1256 BlockDriverState *old_bs;
1257 BlockDriverState *new_bs;
1258 } ExternalSnapshotState;
1260 static void external_snapshot_prepare(BlkTransactionState *common,
1265 QDict *options = NULL;
1266 Error *local_err = NULL;
1267 bool has_device = false;
1269 bool has_node_name = false;
1270 const char *node_name;
1271 bool has_snapshot_node_name = false;
1272 const char *snapshot_node_name;
1273 const char *new_image_file;
1274 const char *format = "qcow2";
1275 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1276 ExternalSnapshotState *state =
1277 DO_UPCAST(ExternalSnapshotState, common, common);
1278 TransactionAction *action = common->action;
1280 /* get parameters */
1281 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1283 has_device = action->blockdev_snapshot_sync->has_device;
1284 device = action->blockdev_snapshot_sync->device;
1285 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1286 node_name = action->blockdev_snapshot_sync->node_name;
1287 has_snapshot_node_name =
1288 action->blockdev_snapshot_sync->has_snapshot_node_name;
1289 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1291 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1292 if (action->blockdev_snapshot_sync->has_format) {
1293 format = action->blockdev_snapshot_sync->format;
1295 if (action->blockdev_snapshot_sync->has_mode) {
1296 mode = action->blockdev_snapshot_sync->mode;
1299 /* start processing */
1300 drv = bdrv_find_format(format);
1302 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1306 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1307 has_node_name ? node_name : NULL,
1310 error_propagate(errp, local_err);
1314 if (has_node_name && !has_snapshot_node_name) {
1315 error_setg(errp, "New snapshot node name missing");
1319 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1320 error_setg(errp, "New snapshot node name already existing");
1324 if (!bdrv_is_inserted(state->old_bs)) {
1325 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1329 if (bdrv_op_is_blocked(state->old_bs,
1330 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1334 if (!bdrv_is_read_only(state->old_bs)) {
1335 if (bdrv_flush(state->old_bs)) {
1336 error_set(errp, QERR_IO_ERROR);
1341 if (!bdrv_is_first_non_filter(state->old_bs)) {
1342 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1346 flags = state->old_bs->open_flags;
1348 /* create new image w/backing file */
1349 if (mode != NEW_IMAGE_MODE_EXISTING) {
1350 bdrv_img_create(new_image_file, format,
1351 state->old_bs->filename,
1352 state->old_bs->drv->format_name,
1353 NULL, -1, flags, &local_err, false);
1355 error_propagate(errp, local_err);
1360 if (has_snapshot_node_name) {
1361 options = qdict_new();
1362 qdict_put(options, "node-name",
1363 qstring_from_str(snapshot_node_name));
1366 /* TODO Inherit bs->options or only take explicit options with an
1367 * extended QMP command? */
1368 assert(state->new_bs == NULL);
1369 ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1370 flags | BDRV_O_NO_BACKING, drv, &local_err);
1371 /* We will manually add the backing_hd field to the bs later */
1373 error_propagate(errp, local_err);
1377 static void external_snapshot_commit(BlkTransactionState *common)
1379 ExternalSnapshotState *state =
1380 DO_UPCAST(ExternalSnapshotState, common, common);
1382 /* This removes our old bs and adds the new bs */
1383 bdrv_append(state->new_bs, state->old_bs);
1384 /* We don't need (or want) to use the transactional
1385 * bdrv_reopen_multiple() across all the entries at once, because we
1386 * don't want to abort all of them if one of them fails the reopen */
1387 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1391 static void external_snapshot_abort(BlkTransactionState *common)
1393 ExternalSnapshotState *state =
1394 DO_UPCAST(ExternalSnapshotState, common, common);
1395 if (state->new_bs) {
1396 bdrv_unref(state->new_bs);
1400 typedef struct DriveBackupState {
1401 BlkTransactionState common;
1402 BlockDriverState *bs;
1406 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1408 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1409 DriveBackup *backup;
1410 Error *local_err = NULL;
1412 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1413 backup = common->action->drive_backup;
1415 qmp_drive_backup(backup->device, backup->target,
1416 backup->has_format, backup->format,
1418 backup->has_mode, backup->mode,
1419 backup->has_speed, backup->speed,
1420 backup->has_on_source_error, backup->on_source_error,
1421 backup->has_on_target_error, backup->on_target_error,
1424 error_propagate(errp, local_err);
1430 state->bs = bdrv_find(backup->device);
1431 state->job = state->bs->job;
1434 static void drive_backup_abort(BlkTransactionState *common)
1436 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1437 BlockDriverState *bs = state->bs;
1439 /* Only cancel if it's the job we started */
1440 if (bs && bs->job && bs->job == state->job) {
1441 block_job_cancel_sync(bs->job);
1445 static void abort_prepare(BlkTransactionState *common, Error **errp)
1447 error_setg(errp, "Transaction aborted using Abort action");
1450 static void abort_commit(BlkTransactionState *common)
1452 g_assert_not_reached(); /* this action never succeeds */
1455 static const BdrvActionOps actions[] = {
1456 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1457 .instance_size = sizeof(ExternalSnapshotState),
1458 .prepare = external_snapshot_prepare,
1459 .commit = external_snapshot_commit,
1460 .abort = external_snapshot_abort,
1462 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1463 .instance_size = sizeof(DriveBackupState),
1464 .prepare = drive_backup_prepare,
1465 .abort = drive_backup_abort,
1467 [TRANSACTION_ACTION_KIND_ABORT] = {
1468 .instance_size = sizeof(BlkTransactionState),
1469 .prepare = abort_prepare,
1470 .commit = abort_commit,
1472 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1473 .instance_size = sizeof(InternalSnapshotState),
1474 .prepare = internal_snapshot_prepare,
1475 .abort = internal_snapshot_abort,
1480 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1481 * then we do not pivot any of the devices in the group, and abandon the
1484 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1486 TransactionActionList *dev_entry = dev_list;
1487 BlkTransactionState *state, *next;
1488 Error *local_err = NULL;
1490 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1491 QSIMPLEQ_INIT(&snap_bdrv_states);
1493 /* drain all i/o before any snapshots */
1496 /* We don't do anything in this loop that commits us to the snapshot */
1497 while (NULL != dev_entry) {
1498 TransactionAction *dev_info = NULL;
1499 const BdrvActionOps *ops;
1501 dev_info = dev_entry->value;
1502 dev_entry = dev_entry->next;
1504 assert(dev_info->kind < ARRAY_SIZE(actions));
1506 ops = &actions[dev_info->kind];
1507 assert(ops->instance_size > 0);
1509 state = g_malloc0(ops->instance_size);
1511 state->action = dev_info;
1512 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1514 state->ops->prepare(state, &local_err);
1516 error_propagate(errp, local_err);
1517 goto delete_and_fail;
1521 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1522 if (state->ops->commit) {
1523 state->ops->commit(state);
1532 * failure, and it is all-or-none; abandon each new bs, and keep using
1533 * the original bs for all images
1535 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1536 if (state->ops->abort) {
1537 state->ops->abort(state);
1541 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1542 if (state->ops->clean) {
1543 state->ops->clean(state);
1550 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1552 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1555 if (!bdrv_dev_has_removable_media(bs)) {
1556 error_setg(errp, "Device '%s' is not removable",
1557 bdrv_get_device_name(bs));
1561 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1562 bdrv_dev_eject_request(bs, force);
1564 error_setg(errp, "Device '%s' is locked",
1565 bdrv_get_device_name(bs));
1573 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1575 BlockDriverState *bs;
1577 bs = bdrv_find(device);
1579 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1583 eject_device(bs, force, errp);
1586 void qmp_block_passwd(bool has_device, const char *device,
1587 bool has_node_name, const char *node_name,
1588 const char *password, Error **errp)
1590 Error *local_err = NULL;
1591 BlockDriverState *bs;
1594 bs = bdrv_lookup_bs(has_device ? device : NULL,
1595 has_node_name ? node_name : NULL,
1598 error_propagate(errp, local_err);
1602 err = bdrv_set_key(bs, password);
1603 if (err == -EINVAL) {
1604 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1606 } else if (err < 0) {
1607 error_set(errp, QERR_INVALID_PASSWORD);
1612 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1613 int bdrv_flags, BlockDriver *drv,
1614 const char *password, Error **errp)
1616 Error *local_err = NULL;
1619 ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1621 error_propagate(errp, local_err);
1625 if (bdrv_key_required(bs)) {
1627 if (bdrv_set_key(bs, password) < 0) {
1628 error_set(errp, QERR_INVALID_PASSWORD);
1631 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1632 bdrv_get_encrypted_filename(bs));
1634 } else if (password) {
1635 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1639 void qmp_change_blockdev(const char *device, const char *filename,
1640 const char *format, Error **errp)
1642 BlockDriverState *bs;
1643 BlockDriver *drv = NULL;
1647 bs = bdrv_find(device);
1649 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1654 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1656 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1661 eject_device(bs, 0, &err);
1663 error_propagate(errp, err);
1667 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1668 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1670 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1673 /* throttling disk I/O limits */
1674 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1681 bool has_bps_rd_max,
1683 bool has_bps_wr_max,
1687 bool has_iops_rd_max,
1688 int64_t iops_rd_max,
1689 bool has_iops_wr_max,
1690 int64_t iops_wr_max,
1692 int64_t iops_size, Error **errp)
1695 BlockDriverState *bs;
1696 AioContext *aio_context;
1698 bs = bdrv_find(device);
1700 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1704 memset(&cfg, 0, sizeof(cfg));
1705 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1706 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1707 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1709 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1710 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1711 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1714 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1716 if (has_bps_rd_max) {
1717 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1719 if (has_bps_wr_max) {
1720 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1723 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1725 if (has_iops_rd_max) {
1726 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1728 if (has_iops_wr_max) {
1729 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1732 if (has_iops_size) {
1733 cfg.op_size = iops_size;
1736 if (!check_throttle_config(&cfg, errp)) {
1740 aio_context = bdrv_get_aio_context(bs);
1741 aio_context_acquire(aio_context);
1743 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1744 bdrv_io_limits_enable(bs);
1745 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1746 bdrv_io_limits_disable(bs);
1749 if (bs->io_limits_enabled) {
1750 bdrv_set_io_limits(bs, &cfg);
1753 aio_context_release(aio_context);
1756 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1758 const char *id = qdict_get_str(qdict, "id");
1759 BlockDriverState *bs;
1760 Error *local_err = NULL;
1764 error_report("Device '%s' not found", id);
1767 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1768 error_report("%s", error_get_pretty(local_err));
1769 error_free(local_err);
1773 /* quiesce block driver; prevent further io */
1778 /* if we have a device attached to this BlockDriverState
1779 * then we need to make the drive anonymous until the device
1780 * can be removed. If this is a drive with no device backing
1781 * then we can just get rid of the block driver state right here.
1783 if (bdrv_get_attached_dev(bs)) {
1786 /* Further I/O must not pause the guest */
1787 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1788 BLOCKDEV_ON_ERROR_REPORT);
1790 drive_del(drive_get_by_blockdev(bs));
1796 void qmp_block_resize(bool has_device, const char *device,
1797 bool has_node_name, const char *node_name,
1798 int64_t size, Error **errp)
1800 Error *local_err = NULL;
1801 BlockDriverState *bs;
1804 bs = bdrv_lookup_bs(has_device ? device : NULL,
1805 has_node_name ? node_name : NULL,
1808 error_propagate(errp, local_err);
1812 if (!bdrv_is_first_non_filter(bs)) {
1813 error_set(errp, QERR_FEATURE_DISABLED, "resize");
1818 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1822 /* complete all in-flight operations before resizing the device */
1825 ret = bdrv_truncate(bs, size);
1830 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1833 error_set(errp, QERR_UNSUPPORTED);
1836 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1839 error_set(errp, QERR_DEVICE_IN_USE, device);
1842 error_setg_errno(errp, -ret, "Could not resize");
1847 static void block_job_cb(void *opaque, int ret)
1849 BlockDriverState *bs = opaque;
1852 trace_block_job_cb(bs, bs->job, ret);
1855 obj = qobject_from_block_job(bs->job);
1857 QDict *dict = qobject_to_qdict(obj);
1858 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1861 if (block_job_is_cancelled(bs->job)) {
1862 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1864 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1866 qobject_decref(obj);
1868 bdrv_put_ref_bh_schedule(bs);
1871 void qmp_block_stream(const char *device, bool has_base,
1872 const char *base, bool has_speed, int64_t speed,
1873 bool has_on_error, BlockdevOnError on_error,
1876 BlockDriverState *bs;
1877 BlockDriverState *base_bs = NULL;
1878 Error *local_err = NULL;
1880 if (!has_on_error) {
1881 on_error = BLOCKDEV_ON_ERROR_REPORT;
1884 bs = bdrv_find(device);
1886 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1890 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
1895 base_bs = bdrv_find_backing_image(bs, base);
1896 if (base_bs == NULL) {
1897 error_set(errp, QERR_BASE_NOT_FOUND, base);
1902 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1903 on_error, block_job_cb, bs, &local_err);
1905 error_propagate(errp, local_err);
1909 trace_qmp_block_stream(bs, bs->job);
1912 void qmp_block_commit(const char *device,
1913 bool has_base, const char *base, const char *top,
1914 bool has_speed, int64_t speed,
1917 BlockDriverState *bs;
1918 BlockDriverState *base_bs, *top_bs;
1919 Error *local_err = NULL;
1920 /* This will be part of the QMP command, if/when the
1921 * BlockdevOnError change for blkmirror makes it in
1923 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1929 /* drain all i/o before commits */
1932 bs = bdrv_find(device);
1934 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1938 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, errp)) {
1942 /* default top_bs is the active layer */
1946 if (strcmp(bs->filename, top) != 0) {
1947 top_bs = bdrv_find_backing_image(bs, top);
1951 if (top_bs == NULL) {
1952 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1956 if (has_base && base) {
1957 base_bs = bdrv_find_backing_image(top_bs, base);
1959 base_bs = bdrv_find_base(top_bs);
1962 if (base_bs == NULL) {
1963 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1968 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1971 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1974 if (local_err != NULL) {
1975 error_propagate(errp, local_err);
1980 void qmp_drive_backup(const char *device, const char *target,
1981 bool has_format, const char *format,
1982 enum MirrorSyncMode sync,
1983 bool has_mode, enum NewImageMode mode,
1984 bool has_speed, int64_t speed,
1985 bool has_on_source_error, BlockdevOnError on_source_error,
1986 bool has_on_target_error, BlockdevOnError on_target_error,
1989 BlockDriverState *bs;
1990 BlockDriverState *target_bs;
1991 BlockDriverState *source = NULL;
1992 BlockDriver *drv = NULL;
1993 Error *local_err = NULL;
2001 if (!has_on_source_error) {
2002 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2004 if (!has_on_target_error) {
2005 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2008 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2011 bs = bdrv_find(device);
2013 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2017 if (!bdrv_is_inserted(bs)) {
2018 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2023 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2026 drv = bdrv_find_format(format);
2028 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2033 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2037 flags = bs->open_flags | BDRV_O_RDWR;
2039 /* See if we have a backing HD we can use to create our new image
2041 if (sync == MIRROR_SYNC_MODE_TOP) {
2042 source = bs->backing_hd;
2044 sync = MIRROR_SYNC_MODE_FULL;
2047 if (sync == MIRROR_SYNC_MODE_NONE) {
2051 size = bdrv_getlength(bs);
2053 error_setg_errno(errp, -size, "bdrv_getlength failed");
2057 if (mode != NEW_IMAGE_MODE_EXISTING) {
2058 assert(format && drv);
2060 bdrv_img_create(target, format, source->filename,
2061 source->drv->format_name, NULL,
2062 size, flags, &local_err, false);
2064 bdrv_img_create(target, format, NULL, NULL, NULL,
2065 size, flags, &local_err, false);
2070 error_propagate(errp, local_err);
2075 ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2077 error_propagate(errp, local_err);
2081 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2082 block_job_cb, bs, &local_err);
2083 if (local_err != NULL) {
2084 bdrv_unref(target_bs);
2085 error_propagate(errp, local_err);
2090 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2092 return bdrv_named_nodes_list();
2095 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2097 void qmp_drive_mirror(const char *device, const char *target,
2098 bool has_format, const char *format,
2099 enum MirrorSyncMode sync,
2100 bool has_mode, enum NewImageMode mode,
2101 bool has_speed, int64_t speed,
2102 bool has_granularity, uint32_t granularity,
2103 bool has_buf_size, int64_t buf_size,
2104 bool has_on_source_error, BlockdevOnError on_source_error,
2105 bool has_on_target_error, BlockdevOnError on_target_error,
2108 BlockDriverState *bs;
2109 BlockDriverState *source, *target_bs;
2110 BlockDriver *drv = NULL;
2111 Error *local_err = NULL;
2119 if (!has_on_source_error) {
2120 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2122 if (!has_on_target_error) {
2123 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2126 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2128 if (!has_granularity) {
2131 if (!has_buf_size) {
2132 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2135 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2136 error_set(errp, QERR_INVALID_PARAMETER, device);
2139 if (granularity & (granularity - 1)) {
2140 error_set(errp, QERR_INVALID_PARAMETER, device);
2144 bs = bdrv_find(device);
2146 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2150 if (!bdrv_is_inserted(bs)) {
2151 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2156 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2159 drv = bdrv_find_format(format);
2161 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2166 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2170 flags = bs->open_flags | BDRV_O_RDWR;
2171 source = bs->backing_hd;
2172 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2173 sync = MIRROR_SYNC_MODE_FULL;
2175 if (sync == MIRROR_SYNC_MODE_NONE) {
2179 size = bdrv_getlength(bs);
2181 error_setg_errno(errp, -size, "bdrv_getlength failed");
2185 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2186 && mode != NEW_IMAGE_MODE_EXISTING)
2188 /* create new image w/o backing file */
2189 assert(format && drv);
2190 bdrv_img_create(target, format,
2191 NULL, NULL, NULL, size, flags, &local_err, false);
2194 case NEW_IMAGE_MODE_EXISTING:
2196 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2197 /* create new image with backing file */
2198 bdrv_img_create(target, format,
2200 source->drv->format_name,
2201 NULL, size, flags, &local_err, false);
2209 error_propagate(errp, local_err);
2213 /* Mirroring takes care of copy-on-write using the source's backing
2217 ret = bdrv_open(&target_bs, target, NULL, NULL, flags | BDRV_O_NO_BACKING,
2220 error_propagate(errp, local_err);
2224 mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2225 on_source_error, on_target_error,
2226 block_job_cb, bs, &local_err);
2227 if (local_err != NULL) {
2228 bdrv_unref(target_bs);
2229 error_propagate(errp, local_err);
2234 static BlockJob *find_block_job(const char *device)
2236 BlockDriverState *bs;
2238 bs = bdrv_find(device);
2239 if (!bs || !bs->job) {
2245 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2247 BlockJob *job = find_block_job(device);
2250 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2254 block_job_set_speed(job, speed, errp);
2257 void qmp_block_job_cancel(const char *device,
2258 bool has_force, bool force, Error **errp)
2260 BlockJob *job = find_block_job(device);
2267 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2270 if (job->paused && !force) {
2271 error_setg(errp, "The block job for device '%s' is currently paused",
2276 trace_qmp_block_job_cancel(job);
2277 block_job_cancel(job);
2280 void qmp_block_job_pause(const char *device, Error **errp)
2282 BlockJob *job = find_block_job(device);
2285 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2289 trace_qmp_block_job_pause(job);
2290 block_job_pause(job);
2293 void qmp_block_job_resume(const char *device, Error **errp)
2295 BlockJob *job = find_block_job(device);
2298 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2302 trace_qmp_block_job_resume(job);
2303 block_job_resume(job);
2306 void qmp_block_job_complete(const char *device, Error **errp)
2308 BlockJob *job = find_block_job(device);
2311 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2315 trace_qmp_block_job_complete(job);
2316 block_job_complete(job, errp);
2319 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2321 QmpOutputVisitor *ov = qmp_output_visitor_new();
2325 Error *local_err = NULL;
2327 /* Require an ID in the top level */
2328 if (!options->has_id) {
2329 error_setg(errp, "Block device needs an ID");
2333 /* TODO Sort it out in raw-posix and drive_new(): Reject aio=native with
2334 * cache.direct=false instead of silently switching to aio=threads, except
2335 * when called from drive_new().
2337 * For now, simply forbidding the combination for all drivers will do. */
2338 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2339 bool direct = options->has_cache &&
2340 options->cache->has_direct &&
2341 options->cache->direct;
2343 error_setg(errp, "aio=native requires cache.direct=true");
2348 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2349 &options, NULL, &local_err);
2351 error_propagate(errp, local_err);
2355 obj = qmp_output_get_qobject(ov);
2356 qdict = qobject_to_qdict(obj);
2358 qdict_flatten(qdict);
2360 dinfo = blockdev_init(NULL, qdict, &local_err);
2362 error_propagate(errp, local_err);
2366 if (bdrv_key_required(dinfo->bdrv)) {
2368 error_setg(errp, "blockdev-add doesn't support encrypted devices");
2373 qmp_output_visitor_cleanup(ov);
2376 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2378 BlockJobInfoList **prev = opaque;
2379 BlockJob *job = bs->job;
2382 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2383 elem->value = block_job_query(bs->job);
2384 (*prev)->next = elem;
2389 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2391 /* Dummy is a fake list element for holding the head pointer */
2392 BlockJobInfoList dummy = {};
2393 BlockJobInfoList *prev = &dummy;
2394 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2398 QemuOptsList qemu_common_drive_opts = {
2400 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2404 .type = QEMU_OPT_BOOL,
2405 .help = "enable/disable snapshot mode",
2408 .type = QEMU_OPT_STRING,
2409 .help = "discard operation (ignore/off, unmap/on)",
2411 .name = "cache.writeback",
2412 .type = QEMU_OPT_BOOL,
2413 .help = "enables writeback mode for any caches",
2415 .name = "cache.direct",
2416 .type = QEMU_OPT_BOOL,
2417 .help = "enables use of O_DIRECT (bypass the host page cache)",
2419 .name = "cache.no-flush",
2420 .type = QEMU_OPT_BOOL,
2421 .help = "ignore any flush requests for the device",
2424 .type = QEMU_OPT_STRING,
2425 .help = "host AIO implementation (threads, native)",
2428 .type = QEMU_OPT_STRING,
2429 .help = "disk format (raw, qcow2, ...)",
2432 .type = QEMU_OPT_STRING,
2433 .help = "read error action",
2436 .type = QEMU_OPT_STRING,
2437 .help = "write error action",
2439 .name = "read-only",
2440 .type = QEMU_OPT_BOOL,
2441 .help = "open drive file as read-only",
2443 .name = "throttling.iops-total",
2444 .type = QEMU_OPT_NUMBER,
2445 .help = "limit total I/O operations per second",
2447 .name = "throttling.iops-read",
2448 .type = QEMU_OPT_NUMBER,
2449 .help = "limit read operations per second",
2451 .name = "throttling.iops-write",
2452 .type = QEMU_OPT_NUMBER,
2453 .help = "limit write operations per second",
2455 .name = "throttling.bps-total",
2456 .type = QEMU_OPT_NUMBER,
2457 .help = "limit total bytes per second",
2459 .name = "throttling.bps-read",
2460 .type = QEMU_OPT_NUMBER,
2461 .help = "limit read bytes per second",
2463 .name = "throttling.bps-write",
2464 .type = QEMU_OPT_NUMBER,
2465 .help = "limit write bytes per second",
2467 .name = "throttling.iops-total-max",
2468 .type = QEMU_OPT_NUMBER,
2469 .help = "I/O operations burst",
2471 .name = "throttling.iops-read-max",
2472 .type = QEMU_OPT_NUMBER,
2473 .help = "I/O operations read burst",
2475 .name = "throttling.iops-write-max",
2476 .type = QEMU_OPT_NUMBER,
2477 .help = "I/O operations write burst",
2479 .name = "throttling.bps-total-max",
2480 .type = QEMU_OPT_NUMBER,
2481 .help = "total bytes burst",
2483 .name = "throttling.bps-read-max",
2484 .type = QEMU_OPT_NUMBER,
2485 .help = "total bytes read burst",
2487 .name = "throttling.bps-write-max",
2488 .type = QEMU_OPT_NUMBER,
2489 .help = "total bytes write burst",
2491 .name = "throttling.iops-size",
2492 .type = QEMU_OPT_NUMBER,
2493 .help = "when limiting by iops max size of an I/O in bytes",
2495 .name = "copy-on-read",
2496 .type = QEMU_OPT_BOOL,
2497 .help = "copy read data from backing file into image file",
2499 .name = "detect-zeroes",
2500 .type = QEMU_OPT_STRING,
2501 .help = "try to optimize zero writes (off, on, unmap)",
2503 { /* end of list */ }
2507 QemuOptsList qemu_drive_opts = {
2509 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2512 * no elements => accept any params
2513 * validation will happen later
2515 { /* end of list */ }