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) {
109 drive_put_ref(dinfo);
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 static void drive_uninit(DriveInfo *dinfo)
219 qemu_opts_del(dinfo->opts);
222 bdrv_unref(dinfo->bdrv);
224 QTAILQ_REMOVE(&drives, dinfo, next);
225 g_free(dinfo->serial);
229 void drive_put_ref(DriveInfo *dinfo)
231 assert(dinfo->refcount);
232 if (--dinfo->refcount == 0) {
237 void drive_get_ref(DriveInfo *dinfo)
244 BlockDriverState *bs;
247 static void bdrv_put_ref_bh(void *opaque)
249 BDRVPutRefBH *s = opaque;
252 qemu_bh_delete(s->bh);
257 * Release a BDS reference in a BH
259 * It is not safe to use bdrv_unref() from a callback function when the callers
260 * still need the BlockDriverState. In such cases we schedule a BH to release
263 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
267 s = g_new(BDRVPutRefBH, 1);
268 s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
270 qemu_bh_schedule(s->bh);
273 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
275 if (!strcmp(buf, "ignore")) {
276 return BLOCKDEV_ON_ERROR_IGNORE;
277 } else if (!is_read && !strcmp(buf, "enospc")) {
278 return BLOCKDEV_ON_ERROR_ENOSPC;
279 } else if (!strcmp(buf, "stop")) {
280 return BLOCKDEV_ON_ERROR_STOP;
281 } else if (!strcmp(buf, "report")) {
282 return BLOCKDEV_ON_ERROR_REPORT;
284 error_setg(errp, "'%s' invalid %s error action",
285 buf, is_read ? "read" : "write");
290 static inline int parse_enum_option(const char *lookup[], const char *buf,
291 int max, int def, Error **errp)
299 for (i = 0; i < max; i++) {
300 if (!strcmp(buf, lookup[i])) {
305 error_setg(errp, "invalid parameter value: %s", buf);
309 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
311 if (throttle_conflicting(cfg)) {
312 error_setg(errp, "bps/iops/max total values and read/write values"
313 " cannot be used at the same time");
317 if (!throttle_is_valid(cfg)) {
318 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
325 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
327 /* Takes the ownership of bs_opts */
328 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
335 int on_read_error, on_write_error;
344 bool has_driver_specific_opts;
345 BlockdevDetectZeroesOptions detect_zeroes;
346 BlockDriver *drv = NULL;
348 /* Check common options by copying from bs_opts to opts, all other options
349 * stay in bs_opts for processing by bdrv_open(). */
350 id = qdict_get_try_str(bs_opts, "id");
351 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
353 error_propagate(errp, error);
357 qemu_opts_absorb_qdict(opts, bs_opts, &error);
359 error_propagate(errp, error);
364 qdict_del(bs_opts, "id");
367 has_driver_specific_opts = !!qdict_size(bs_opts);
369 /* extract parameters */
370 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
371 ro = qemu_opt_get_bool(opts, "read-only", 0);
372 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
374 serial = qemu_opt_get(opts, "serial");
376 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
377 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
378 error_setg(errp, "invalid discard option");
383 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
384 bdrv_flags |= BDRV_O_CACHE_WB;
386 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
387 bdrv_flags |= BDRV_O_NOCACHE;
389 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
390 bdrv_flags |= BDRV_O_NO_FLUSH;
393 #ifdef CONFIG_LINUX_AIO
394 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
395 if (!strcmp(buf, "native")) {
396 bdrv_flags |= BDRV_O_NATIVE_AIO;
397 } else if (!strcmp(buf, "threads")) {
398 /* this is the default */
400 error_setg(errp, "invalid aio option");
406 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
407 if (is_help_option(buf)) {
408 error_printf("Supported formats:");
409 bdrv_iterate_format(bdrv_format_print, NULL);
414 drv = bdrv_find_format(buf);
416 error_setg(errp, "'%s' invalid format", buf);
421 /* disk I/O throttling */
422 memset(&cfg, 0, sizeof(cfg));
423 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
424 qemu_opt_get_number(opts, "throttling.bps-total", 0);
425 cfg.buckets[THROTTLE_BPS_READ].avg =
426 qemu_opt_get_number(opts, "throttling.bps-read", 0);
427 cfg.buckets[THROTTLE_BPS_WRITE].avg =
428 qemu_opt_get_number(opts, "throttling.bps-write", 0);
429 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
430 qemu_opt_get_number(opts, "throttling.iops-total", 0);
431 cfg.buckets[THROTTLE_OPS_READ].avg =
432 qemu_opt_get_number(opts, "throttling.iops-read", 0);
433 cfg.buckets[THROTTLE_OPS_WRITE].avg =
434 qemu_opt_get_number(opts, "throttling.iops-write", 0);
436 cfg.buckets[THROTTLE_BPS_TOTAL].max =
437 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
438 cfg.buckets[THROTTLE_BPS_READ].max =
439 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
440 cfg.buckets[THROTTLE_BPS_WRITE].max =
441 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
442 cfg.buckets[THROTTLE_OPS_TOTAL].max =
443 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
444 cfg.buckets[THROTTLE_OPS_READ].max =
445 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
446 cfg.buckets[THROTTLE_OPS_WRITE].max =
447 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
449 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
451 if (!check_throttle_config(&cfg, &error)) {
452 error_propagate(errp, error);
456 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
457 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
458 on_write_error = parse_block_error_action(buf, 0, &error);
460 error_propagate(errp, error);
465 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
466 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
467 on_read_error = parse_block_error_action(buf, 1, &error);
469 error_propagate(errp, error);
475 parse_enum_option(BlockdevDetectZeroesOptions_lookup,
476 qemu_opt_get(opts, "detect-zeroes"),
477 BLOCKDEV_DETECT_ZEROES_OPTIONS_MAX,
478 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
481 error_propagate(errp, error);
485 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP &&
486 !(bdrv_flags & BDRV_O_UNMAP)) {
487 error_setg(errp, "setting detect-zeroes to unmap is not allowed "
488 "without setting discard operation to unmap");
493 dinfo = g_malloc0(sizeof(*dinfo));
494 dinfo->id = g_strdup(qemu_opts_id(opts));
495 dinfo->bdrv = bdrv_new(dinfo->id, &error);
497 error_propagate(errp, error);
500 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
501 dinfo->bdrv->read_only = ro;
502 dinfo->bdrv->detect_zeroes = detect_zeroes;
504 if (serial != NULL) {
505 dinfo->serial = g_strdup(serial);
507 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
509 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
511 /* disk I/O throttling */
512 if (throttle_enabled(&cfg)) {
513 bdrv_io_limits_enable(dinfo->bdrv);
514 bdrv_set_io_limits(dinfo->bdrv, &cfg);
517 if (!file || !*file) {
518 if (has_driver_specific_opts) {
527 /* always use cache=unsafe with snapshot */
528 bdrv_flags &= ~BDRV_O_CACHE_MASK;
529 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
533 bdrv_flags |= BDRV_O_COPY_ON_READ;
536 if (runstate_check(RUN_STATE_INMIGRATE)) {
537 bdrv_flags |= BDRV_O_INCOMING;
540 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
543 ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
546 error_setg(errp, "could not open disk image %s: %s",
547 file ?: dinfo->id, error_get_pretty(error));
552 if (bdrv_key_required(dinfo->bdrv))
561 bdrv_unref(dinfo->bdrv);
562 QTAILQ_REMOVE(&drives, dinfo, next);
573 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
577 value = qemu_opt_get(opts, from);
579 qemu_opt_set(opts, to, value);
580 qemu_opt_unset(opts, from);
584 QemuOptsList qemu_legacy_drive_opts = {
586 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
590 .type = QEMU_OPT_NUMBER,
591 .help = "bus number",
594 .type = QEMU_OPT_NUMBER,
595 .help = "unit number (i.e. lun for scsi)",
598 .type = QEMU_OPT_NUMBER,
599 .help = "index number",
602 .type = QEMU_OPT_STRING,
603 .help = "media type (disk, cdrom)",
606 .type = QEMU_OPT_STRING,
607 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
610 .type = QEMU_OPT_NUMBER,
611 .help = "number of cylinders (ide disk geometry)",
614 .type = QEMU_OPT_NUMBER,
615 .help = "number of heads (ide disk geometry)",
618 .type = QEMU_OPT_NUMBER,
619 .help = "number of sectors (ide disk geometry)",
622 .type = QEMU_OPT_STRING,
623 .help = "chs translation (auto, lba, none)",
626 .type = QEMU_OPT_BOOL,
627 .help = "(deprecated, ignored)",
630 .type = QEMU_OPT_STRING,
631 .help = "pci address (virtio only)",
634 .type = QEMU_OPT_STRING,
638 /* Options that are passed on, but have special semantics with -drive */
641 .type = QEMU_OPT_BOOL,
642 .help = "open drive file as read-only",
645 .type = QEMU_OPT_STRING,
646 .help = "read error action",
649 .type = QEMU_OPT_STRING,
650 .help = "write error action",
652 .name = "copy-on-read",
653 .type = QEMU_OPT_BOOL,
654 .help = "copy read data from backing file into image file",
657 { /* end of list */ }
661 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
664 DriveInfo *dinfo = NULL;
666 QemuOpts *legacy_opts;
667 DriveMediaType media = MEDIA_DISK;
668 BlockInterfaceType type;
669 int cyls, heads, secs, translation;
670 int max_devs, bus_id, unit_id, index;
672 const char *werror, *rerror;
673 bool read_only = false;
675 const char *filename;
676 Error *local_err = NULL;
678 /* Change legacy command line options into QMP ones */
679 qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
680 qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
681 qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
683 qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
684 qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
685 qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
687 qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
688 qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
689 qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
691 qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
692 qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
693 qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
695 qemu_opt_rename(all_opts,
696 "iops_size", "throttling.iops-size");
698 qemu_opt_rename(all_opts, "readonly", "read-only");
700 value = qemu_opt_get(all_opts, "cache");
704 if (bdrv_parse_cache_flags(value, &flags) != 0) {
705 error_report("invalid cache option");
709 /* Specific options take precedence */
710 if (!qemu_opt_get(all_opts, "cache.writeback")) {
711 qemu_opt_set_bool(all_opts, "cache.writeback",
712 !!(flags & BDRV_O_CACHE_WB));
714 if (!qemu_opt_get(all_opts, "cache.direct")) {
715 qemu_opt_set_bool(all_opts, "cache.direct",
716 !!(flags & BDRV_O_NOCACHE));
718 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
719 qemu_opt_set_bool(all_opts, "cache.no-flush",
720 !!(flags & BDRV_O_NO_FLUSH));
722 qemu_opt_unset(all_opts, "cache");
725 /* Get a QDict for processing the options */
726 bs_opts = qdict_new();
727 qemu_opts_to_qdict(all_opts, bs_opts);
729 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
731 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
733 error_report("%s", error_get_pretty(local_err));
734 error_free(local_err);
738 /* Deprecated option boot=[on|off] */
739 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
740 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
741 "ignored. Future versions will reject this parameter. Please "
742 "update your scripts.\n");
746 value = qemu_opt_get(legacy_opts, "media");
748 if (!strcmp(value, "disk")) {
750 } else if (!strcmp(value, "cdrom")) {
754 error_report("'%s' invalid media", value);
759 /* copy-on-read is disabled with a warning for read-only devices */
760 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
761 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
763 if (read_only && copy_on_read) {
764 error_report("warning: disabling copy-on-read on read-only drive");
765 copy_on_read = false;
768 qdict_put(bs_opts, "read-only",
769 qstring_from_str(read_only ? "on" : "off"));
770 qdict_put(bs_opts, "copy-on-read",
771 qstring_from_str(copy_on_read ? "on" :"off"));
773 /* Controller type */
774 value = qemu_opt_get(legacy_opts, "if");
777 type < IF_COUNT && strcmp(value, if_name[type]);
780 if (type == IF_COUNT) {
781 error_report("unsupported bus type '%s'", value);
785 type = block_default_type;
789 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
790 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
791 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
793 if (cyls || heads || secs) {
795 error_report("invalid physical cyls number");
799 error_report("invalid physical heads number");
803 error_report("invalid physical secs number");
808 translation = BIOS_ATA_TRANSLATION_AUTO;
809 value = qemu_opt_get(legacy_opts, "trans");
812 error_report("'%s' trans must be used with cyls, heads and secs",
816 if (!strcmp(value, "none")) {
817 translation = BIOS_ATA_TRANSLATION_NONE;
818 } else if (!strcmp(value, "lba")) {
819 translation = BIOS_ATA_TRANSLATION_LBA;
820 } else if (!strcmp(value, "large")) {
821 translation = BIOS_ATA_TRANSLATION_LARGE;
822 } else if (!strcmp(value, "rechs")) {
823 translation = BIOS_ATA_TRANSLATION_RECHS;
824 } else if (!strcmp(value, "auto")) {
825 translation = BIOS_ATA_TRANSLATION_AUTO;
827 error_report("'%s' invalid translation type", value);
832 if (media == MEDIA_CDROM) {
833 if (cyls || secs || heads) {
834 error_report("CHS can't be set with media=cdrom");
839 /* Device address specified by bus/unit or index.
840 * If none was specified, try to find the first free one. */
841 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
842 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
843 index = qemu_opt_get_number(legacy_opts, "index", -1);
845 max_devs = if_max_devs[type];
848 if (bus_id != 0 || unit_id != -1) {
849 error_report("index cannot be used with bus and unit");
852 bus_id = drive_index_to_bus_id(type, index);
853 unit_id = drive_index_to_unit_id(type, index);
858 while (drive_get(type, bus_id, unit_id) != NULL) {
860 if (max_devs && unit_id >= max_devs) {
867 if (max_devs && unit_id >= max_devs) {
868 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
872 if (drive_get(type, bus_id, unit_id) != NULL) {
873 error_report("drive with bus=%d, unit=%d (index=%d) exists",
874 bus_id, unit_id, index);
878 /* no id supplied -> create one */
879 if (qemu_opts_id(all_opts) == NULL) {
881 const char *mediastr = "";
882 if (type == IF_IDE || type == IF_SCSI) {
883 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
886 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
889 new_id = g_strdup_printf("%s%s%i", if_name[type],
892 qdict_put(bs_opts, "id", qstring_from_str(new_id));
896 /* Add virtio block device */
897 devaddr = qemu_opt_get(legacy_opts, "addr");
898 if (devaddr && type != IF_VIRTIO) {
899 error_report("addr is not supported by this bus type");
903 if (type == IF_VIRTIO) {
905 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
907 if (arch_type == QEMU_ARCH_S390X) {
908 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
910 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
912 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
914 qemu_opt_set(devopts, "addr", devaddr);
918 filename = qemu_opt_get(legacy_opts, "file");
920 /* Check werror/rerror compatibility with if=... */
921 werror = qemu_opt_get(legacy_opts, "werror");
922 if (werror != NULL) {
923 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
925 error_report("werror is not supported by this bus type");
928 qdict_put(bs_opts, "werror", qstring_from_str(werror));
931 rerror = qemu_opt_get(legacy_opts, "rerror");
932 if (rerror != NULL) {
933 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
935 error_report("rerror is not supported by this bus type");
938 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
941 /* Actual block device init: Functionality shared with blockdev-add */
942 dinfo = blockdev_init(filename, bs_opts, &local_err);
945 error_report("%s", error_get_pretty(local_err));
946 error_free(local_err);
953 /* Set legacy DriveInfo fields */
954 dinfo->enable_auto_del = true;
955 dinfo->opts = all_opts;
958 dinfo->heads = heads;
960 dinfo->trans = translation;
964 dinfo->unit = unit_id;
965 dinfo->devaddr = devaddr;
972 dinfo->media_cd = media == MEDIA_CDROM;
979 qemu_opts_del(legacy_opts);
983 void do_commit(Monitor *mon, const QDict *qdict)
985 const char *device = qdict_get_str(qdict, "device");
986 BlockDriverState *bs;
989 if (!strcmp(device, "all")) {
990 ret = bdrv_commit_all();
992 bs = bdrv_find(device);
994 monitor_printf(mon, "Device '%s' not found\n", device);
997 ret = bdrv_commit(bs);
1000 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
1005 static void blockdev_do_action(int kind, void *data, Error **errp)
1007 TransactionAction action;
1008 TransactionActionList list;
1012 list.value = &action;
1014 qmp_transaction(&list, errp);
1017 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1018 bool has_node_name, const char *node_name,
1019 const char *snapshot_file,
1020 bool has_snapshot_node_name,
1021 const char *snapshot_node_name,
1022 bool has_format, const char *format,
1023 bool has_mode, NewImageMode mode, Error **errp)
1025 BlockdevSnapshot snapshot = {
1026 .has_device = has_device,
1027 .device = (char *) device,
1028 .has_node_name = has_node_name,
1029 .node_name = (char *) node_name,
1030 .snapshot_file = (char *) snapshot_file,
1031 .has_snapshot_node_name = has_snapshot_node_name,
1032 .snapshot_node_name = (char *) snapshot_node_name,
1033 .has_format = has_format,
1034 .format = (char *) format,
1035 .has_mode = has_mode,
1038 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1042 void qmp_blockdev_snapshot_internal_sync(const char *device,
1046 BlockdevSnapshotInternal snapshot = {
1047 .device = (char *) device,
1048 .name = (char *) name
1051 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1055 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1062 BlockDriverState *bs = bdrv_find(device);
1063 QEMUSnapshotInfo sn;
1064 Error *local_err = NULL;
1065 SnapshotInfo *info = NULL;
1069 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1082 error_setg(errp, "Name or id must be provided");
1086 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1088 error_propagate(errp, local_err);
1093 "Snapshot with id '%s' and name '%s' does not exist on "
1095 STR_OR_NULL(id), STR_OR_NULL(name), device);
1099 bdrv_snapshot_delete(bs, id, name, &local_err);
1101 error_propagate(errp, local_err);
1105 info = g_malloc0(sizeof(SnapshotInfo));
1106 info->id = g_strdup(sn.id_str);
1107 info->name = g_strdup(sn.name);
1108 info->date_nsec = sn.date_nsec;
1109 info->date_sec = sn.date_sec;
1110 info->vm_state_size = sn.vm_state_size;
1111 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1112 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1117 /* New and old BlockDriverState structs for group snapshots */
1119 typedef struct BlkTransactionState BlkTransactionState;
1121 /* Only prepare() may fail. In a single transaction, only one of commit() or
1122 abort() will be called, clean() will always be called if it present. */
1123 typedef struct BdrvActionOps {
1124 /* Size of state struct, in bytes. */
1125 size_t instance_size;
1126 /* Prepare the work, must NOT be NULL. */
1127 void (*prepare)(BlkTransactionState *common, Error **errp);
1128 /* Commit the changes, can be NULL. */
1129 void (*commit)(BlkTransactionState *common);
1130 /* Abort the changes on fail, can be NULL. */
1131 void (*abort)(BlkTransactionState *common);
1132 /* Clean up resource in the end, can be NULL. */
1133 void (*clean)(BlkTransactionState *common);
1137 * This structure must be arranged as first member in child type, assuming
1138 * that compiler will also arrange it to the same address with parent instance.
1139 * Later it will be used in free().
1141 struct BlkTransactionState {
1142 TransactionAction *action;
1143 const BdrvActionOps *ops;
1144 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1147 /* internal snapshot private data */
1148 typedef struct InternalSnapshotState {
1149 BlkTransactionState common;
1150 BlockDriverState *bs;
1151 QEMUSnapshotInfo sn;
1152 } InternalSnapshotState;
1154 static void internal_snapshot_prepare(BlkTransactionState *common,
1157 Error *local_err = NULL;
1160 BlockDriverState *bs;
1161 QEMUSnapshotInfo old_sn, *sn;
1164 BlockdevSnapshotInternal *internal;
1165 InternalSnapshotState *state;
1168 g_assert(common->action->kind ==
1169 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1170 internal = common->action->blockdev_snapshot_internal_sync;
1171 state = DO_UPCAST(InternalSnapshotState, common, common);
1173 /* 1. parse input */
1174 device = internal->device;
1175 name = internal->name;
1177 /* 2. check for validation */
1178 bs = bdrv_find(device);
1180 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1184 if (!bdrv_is_inserted(bs)) {
1185 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1189 if (bdrv_is_read_only(bs)) {
1190 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1194 if (!bdrv_can_snapshot(bs)) {
1195 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1196 bs->drv->format_name, device, "internal snapshot");
1200 if (!strlen(name)) {
1201 error_setg(errp, "Name is empty");
1205 /* check whether a snapshot with name exist */
1206 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1209 error_propagate(errp, local_err);
1213 "Snapshot with name '%s' already exists on device '%s'",
1218 /* 3. take the snapshot */
1220 pstrcpy(sn->name, sizeof(sn->name), name);
1221 qemu_gettimeofday(&tv);
1222 sn->date_sec = tv.tv_sec;
1223 sn->date_nsec = tv.tv_usec * 1000;
1224 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1226 ret1 = bdrv_snapshot_create(bs, sn);
1228 error_setg_errno(errp, -ret1,
1229 "Failed to create snapshot '%s' on device '%s'",
1234 /* 4. succeed, mark a snapshot is created */
1238 static void internal_snapshot_abort(BlkTransactionState *common)
1240 InternalSnapshotState *state =
1241 DO_UPCAST(InternalSnapshotState, common, common);
1242 BlockDriverState *bs = state->bs;
1243 QEMUSnapshotInfo *sn = &state->sn;
1244 Error *local_error = NULL;
1250 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1251 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1252 "device '%s' in abort: %s",
1255 bdrv_get_device_name(bs),
1256 error_get_pretty(local_error));
1257 error_free(local_error);
1261 /* external snapshot private data */
1262 typedef struct ExternalSnapshotState {
1263 BlkTransactionState common;
1264 BlockDriverState *old_bs;
1265 BlockDriverState *new_bs;
1266 } ExternalSnapshotState;
1268 static void external_snapshot_prepare(BlkTransactionState *common,
1273 QDict *options = NULL;
1274 Error *local_err = NULL;
1275 bool has_device = false;
1277 bool has_node_name = false;
1278 const char *node_name;
1279 bool has_snapshot_node_name = false;
1280 const char *snapshot_node_name;
1281 const char *new_image_file;
1282 const char *format = "qcow2";
1283 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1284 ExternalSnapshotState *state =
1285 DO_UPCAST(ExternalSnapshotState, common, common);
1286 TransactionAction *action = common->action;
1288 /* get parameters */
1289 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1291 has_device = action->blockdev_snapshot_sync->has_device;
1292 device = action->blockdev_snapshot_sync->device;
1293 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1294 node_name = action->blockdev_snapshot_sync->node_name;
1295 has_snapshot_node_name =
1296 action->blockdev_snapshot_sync->has_snapshot_node_name;
1297 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1299 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1300 if (action->blockdev_snapshot_sync->has_format) {
1301 format = action->blockdev_snapshot_sync->format;
1303 if (action->blockdev_snapshot_sync->has_mode) {
1304 mode = action->blockdev_snapshot_sync->mode;
1307 /* start processing */
1308 drv = bdrv_find_format(format);
1310 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1314 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1315 has_node_name ? node_name : NULL,
1318 error_propagate(errp, local_err);
1322 if (has_node_name && !has_snapshot_node_name) {
1323 error_setg(errp, "New snapshot node name missing");
1327 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1328 error_setg(errp, "New snapshot node name already existing");
1332 if (!bdrv_is_inserted(state->old_bs)) {
1333 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1337 if (bdrv_op_is_blocked(state->old_bs,
1338 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1342 if (!bdrv_is_read_only(state->old_bs)) {
1343 if (bdrv_flush(state->old_bs)) {
1344 error_set(errp, QERR_IO_ERROR);
1349 if (!bdrv_is_first_non_filter(state->old_bs)) {
1350 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1354 flags = state->old_bs->open_flags;
1356 /* create new image w/backing file */
1357 if (mode != NEW_IMAGE_MODE_EXISTING) {
1358 bdrv_img_create(new_image_file, format,
1359 state->old_bs->filename,
1360 state->old_bs->drv->format_name,
1361 NULL, -1, flags, &local_err, false);
1363 error_propagate(errp, local_err);
1368 if (has_snapshot_node_name) {
1369 options = qdict_new();
1370 qdict_put(options, "node-name",
1371 qstring_from_str(snapshot_node_name));
1374 /* TODO Inherit bs->options or only take explicit options with an
1375 * extended QMP command? */
1376 assert(state->new_bs == NULL);
1377 ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1378 flags | BDRV_O_NO_BACKING, drv, &local_err);
1379 /* We will manually add the backing_hd field to the bs later */
1381 error_propagate(errp, local_err);
1385 static void external_snapshot_commit(BlkTransactionState *common)
1387 ExternalSnapshotState *state =
1388 DO_UPCAST(ExternalSnapshotState, common, common);
1390 /* This removes our old bs and adds the new bs */
1391 bdrv_append(state->new_bs, state->old_bs);
1392 /* We don't need (or want) to use the transactional
1393 * bdrv_reopen_multiple() across all the entries at once, because we
1394 * don't want to abort all of them if one of them fails the reopen */
1395 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1399 static void external_snapshot_abort(BlkTransactionState *common)
1401 ExternalSnapshotState *state =
1402 DO_UPCAST(ExternalSnapshotState, common, common);
1403 if (state->new_bs) {
1404 bdrv_unref(state->new_bs);
1408 typedef struct DriveBackupState {
1409 BlkTransactionState common;
1410 BlockDriverState *bs;
1414 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1416 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1417 DriveBackup *backup;
1418 Error *local_err = NULL;
1420 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1421 backup = common->action->drive_backup;
1423 qmp_drive_backup(backup->device, backup->target,
1424 backup->has_format, backup->format,
1426 backup->has_mode, backup->mode,
1427 backup->has_speed, backup->speed,
1428 backup->has_on_source_error, backup->on_source_error,
1429 backup->has_on_target_error, backup->on_target_error,
1432 error_propagate(errp, local_err);
1438 state->bs = bdrv_find(backup->device);
1439 state->job = state->bs->job;
1442 static void drive_backup_abort(BlkTransactionState *common)
1444 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1445 BlockDriverState *bs = state->bs;
1447 /* Only cancel if it's the job we started */
1448 if (bs && bs->job && bs->job == state->job) {
1449 block_job_cancel_sync(bs->job);
1453 static void abort_prepare(BlkTransactionState *common, Error **errp)
1455 error_setg(errp, "Transaction aborted using Abort action");
1458 static void abort_commit(BlkTransactionState *common)
1460 g_assert_not_reached(); /* this action never succeeds */
1463 static const BdrvActionOps actions[] = {
1464 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1465 .instance_size = sizeof(ExternalSnapshotState),
1466 .prepare = external_snapshot_prepare,
1467 .commit = external_snapshot_commit,
1468 .abort = external_snapshot_abort,
1470 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1471 .instance_size = sizeof(DriveBackupState),
1472 .prepare = drive_backup_prepare,
1473 .abort = drive_backup_abort,
1475 [TRANSACTION_ACTION_KIND_ABORT] = {
1476 .instance_size = sizeof(BlkTransactionState),
1477 .prepare = abort_prepare,
1478 .commit = abort_commit,
1480 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1481 .instance_size = sizeof(InternalSnapshotState),
1482 .prepare = internal_snapshot_prepare,
1483 .abort = internal_snapshot_abort,
1488 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1489 * then we do not pivot any of the devices in the group, and abandon the
1492 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1494 TransactionActionList *dev_entry = dev_list;
1495 BlkTransactionState *state, *next;
1496 Error *local_err = NULL;
1498 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1499 QSIMPLEQ_INIT(&snap_bdrv_states);
1501 /* drain all i/o before any snapshots */
1504 /* We don't do anything in this loop that commits us to the snapshot */
1505 while (NULL != dev_entry) {
1506 TransactionAction *dev_info = NULL;
1507 const BdrvActionOps *ops;
1509 dev_info = dev_entry->value;
1510 dev_entry = dev_entry->next;
1512 assert(dev_info->kind < ARRAY_SIZE(actions));
1514 ops = &actions[dev_info->kind];
1515 assert(ops->instance_size > 0);
1517 state = g_malloc0(ops->instance_size);
1519 state->action = dev_info;
1520 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1522 state->ops->prepare(state, &local_err);
1524 error_propagate(errp, local_err);
1525 goto delete_and_fail;
1529 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1530 if (state->ops->commit) {
1531 state->ops->commit(state);
1540 * failure, and it is all-or-none; abandon each new bs, and keep using
1541 * the original bs for all images
1543 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1544 if (state->ops->abort) {
1545 state->ops->abort(state);
1549 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1550 if (state->ops->clean) {
1551 state->ops->clean(state);
1558 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1560 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_EJECT, errp)) {
1563 if (!bdrv_dev_has_removable_media(bs)) {
1564 error_setg(errp, "Device '%s' is not removable",
1565 bdrv_get_device_name(bs));
1569 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1570 bdrv_dev_eject_request(bs, force);
1572 error_setg(errp, "Device '%s' is locked",
1573 bdrv_get_device_name(bs));
1581 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1583 BlockDriverState *bs;
1585 bs = bdrv_find(device);
1587 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1591 eject_device(bs, force, errp);
1594 void qmp_block_passwd(bool has_device, const char *device,
1595 bool has_node_name, const char *node_name,
1596 const char *password, Error **errp)
1598 Error *local_err = NULL;
1599 BlockDriverState *bs;
1602 bs = bdrv_lookup_bs(has_device ? device : NULL,
1603 has_node_name ? node_name : NULL,
1606 error_propagate(errp, local_err);
1610 err = bdrv_set_key(bs, password);
1611 if (err == -EINVAL) {
1612 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1614 } else if (err < 0) {
1615 error_set(errp, QERR_INVALID_PASSWORD);
1620 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1621 int bdrv_flags, BlockDriver *drv,
1622 const char *password, Error **errp)
1624 Error *local_err = NULL;
1627 ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1629 error_propagate(errp, local_err);
1633 if (bdrv_key_required(bs)) {
1635 if (bdrv_set_key(bs, password) < 0) {
1636 error_set(errp, QERR_INVALID_PASSWORD);
1639 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1640 bdrv_get_encrypted_filename(bs));
1642 } else if (password) {
1643 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1647 void qmp_change_blockdev(const char *device, const char *filename,
1648 const char *format, Error **errp)
1650 BlockDriverState *bs;
1651 BlockDriver *drv = NULL;
1655 bs = bdrv_find(device);
1657 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1662 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1664 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1669 eject_device(bs, 0, &err);
1671 error_propagate(errp, err);
1675 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1676 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1678 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1681 /* throttling disk I/O limits */
1682 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1689 bool has_bps_rd_max,
1691 bool has_bps_wr_max,
1695 bool has_iops_rd_max,
1696 int64_t iops_rd_max,
1697 bool has_iops_wr_max,
1698 int64_t iops_wr_max,
1700 int64_t iops_size, Error **errp)
1703 BlockDriverState *bs;
1705 bs = bdrv_find(device);
1707 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1711 memset(&cfg, 0, sizeof(cfg));
1712 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1713 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1714 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1716 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1717 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1718 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1721 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1723 if (has_bps_rd_max) {
1724 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1726 if (has_bps_wr_max) {
1727 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1730 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1732 if (has_iops_rd_max) {
1733 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1735 if (has_iops_wr_max) {
1736 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1739 if (has_iops_size) {
1740 cfg.op_size = iops_size;
1743 if (!check_throttle_config(&cfg, errp)) {
1747 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1748 bdrv_io_limits_enable(bs);
1749 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1750 bdrv_io_limits_disable(bs);
1753 if (bs->io_limits_enabled) {
1754 bdrv_set_io_limits(bs, &cfg);
1758 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1760 const char *id = qdict_get_str(qdict, "id");
1761 BlockDriverState *bs;
1762 Error *local_err = NULL;
1766 error_report("Device '%s' not found", id);
1769 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, &local_err)) {
1770 error_report("%s", error_get_pretty(local_err));
1771 error_free(local_err);
1775 /* quiesce block driver; prevent further io */
1780 /* if we have a device attached to this BlockDriverState
1781 * then we need to make the drive anonymous until the device
1782 * can be removed. If this is a drive with no device backing
1783 * then we can just get rid of the block driver state right here.
1785 if (bdrv_get_attached_dev(bs)) {
1788 /* Further I/O must not pause the guest */
1789 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1790 BLOCKDEV_ON_ERROR_REPORT);
1792 drive_uninit(drive_get_by_blockdev(bs));
1798 void qmp_block_resize(bool has_device, const char *device,
1799 bool has_node_name, const char *node_name,
1800 int64_t size, Error **errp)
1802 Error *local_err = NULL;
1803 BlockDriverState *bs;
1806 bs = bdrv_lookup_bs(has_device ? device : NULL,
1807 has_node_name ? node_name : NULL,
1810 error_propagate(errp, local_err);
1814 if (!bdrv_is_first_non_filter(bs)) {
1815 error_set(errp, QERR_FEATURE_DISABLED, "resize");
1820 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1824 /* complete all in-flight operations before resizing the device */
1827 ret = bdrv_truncate(bs, size);
1832 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1835 error_set(errp, QERR_UNSUPPORTED);
1838 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1841 error_set(errp, QERR_DEVICE_IN_USE, device);
1844 error_setg_errno(errp, -ret, "Could not resize");
1849 static void block_job_cb(void *opaque, int ret)
1851 BlockDriverState *bs = opaque;
1854 trace_block_job_cb(bs, bs->job, ret);
1857 obj = qobject_from_block_job(bs->job);
1859 QDict *dict = qobject_to_qdict(obj);
1860 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1863 if (block_job_is_cancelled(bs->job)) {
1864 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1866 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1868 qobject_decref(obj);
1870 bdrv_put_ref_bh_schedule(bs);
1873 void qmp_block_stream(const char *device, bool has_base,
1874 const char *base, bool has_speed, int64_t speed,
1875 bool has_on_error, BlockdevOnError on_error,
1878 BlockDriverState *bs;
1879 BlockDriverState *base_bs = NULL;
1880 Error *local_err = NULL;
1882 if (!has_on_error) {
1883 on_error = BLOCKDEV_ON_ERROR_REPORT;
1886 bs = bdrv_find(device);
1888 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1892 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
1897 base_bs = bdrv_find_backing_image(bs, base);
1898 if (base_bs == NULL) {
1899 error_set(errp, QERR_BASE_NOT_FOUND, base);
1904 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1905 on_error, block_job_cb, bs, &local_err);
1907 error_propagate(errp, local_err);
1911 trace_qmp_block_stream(bs, bs->job);
1914 void qmp_block_commit(const char *device,
1915 bool has_base, const char *base, const char *top,
1916 bool has_speed, int64_t speed,
1919 BlockDriverState *bs;
1920 BlockDriverState *base_bs, *top_bs;
1921 Error *local_err = NULL;
1922 /* This will be part of the QMP command, if/when the
1923 * BlockdevOnError change for blkmirror makes it in
1925 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1931 /* drain all i/o before commits */
1934 bs = bdrv_find(device);
1936 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1940 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT, errp)) {
1944 /* default top_bs is the active layer */
1948 if (strcmp(bs->filename, top) != 0) {
1949 top_bs = bdrv_find_backing_image(bs, top);
1953 if (top_bs == NULL) {
1954 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1958 if (has_base && base) {
1959 base_bs = bdrv_find_backing_image(top_bs, base);
1961 base_bs = bdrv_find_base(top_bs);
1964 if (base_bs == NULL) {
1965 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1970 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1973 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1976 if (local_err != NULL) {
1977 error_propagate(errp, local_err);
1982 void qmp_drive_backup(const char *device, const char *target,
1983 bool has_format, const char *format,
1984 enum MirrorSyncMode sync,
1985 bool has_mode, enum NewImageMode mode,
1986 bool has_speed, int64_t speed,
1987 bool has_on_source_error, BlockdevOnError on_source_error,
1988 bool has_on_target_error, BlockdevOnError on_target_error,
1991 BlockDriverState *bs;
1992 BlockDriverState *target_bs;
1993 BlockDriverState *source = NULL;
1994 BlockDriver *drv = NULL;
1995 Error *local_err = NULL;
2003 if (!has_on_source_error) {
2004 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2006 if (!has_on_target_error) {
2007 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2010 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2013 bs = bdrv_find(device);
2015 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2019 if (!bdrv_is_inserted(bs)) {
2020 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2025 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2028 drv = bdrv_find_format(format);
2030 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2035 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
2039 flags = bs->open_flags | BDRV_O_RDWR;
2041 /* See if we have a backing HD we can use to create our new image
2043 if (sync == MIRROR_SYNC_MODE_TOP) {
2044 source = bs->backing_hd;
2046 sync = MIRROR_SYNC_MODE_FULL;
2049 if (sync == MIRROR_SYNC_MODE_NONE) {
2053 size = bdrv_getlength(bs);
2055 error_setg_errno(errp, -size, "bdrv_getlength failed");
2059 if (mode != NEW_IMAGE_MODE_EXISTING) {
2060 assert(format && drv);
2062 bdrv_img_create(target, format, source->filename,
2063 source->drv->format_name, NULL,
2064 size, flags, &local_err, false);
2066 bdrv_img_create(target, format, NULL, NULL, NULL,
2067 size, flags, &local_err, false);
2072 error_propagate(errp, local_err);
2077 ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2079 error_propagate(errp, local_err);
2083 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2084 block_job_cb, bs, &local_err);
2085 if (local_err != NULL) {
2086 bdrv_unref(target_bs);
2087 error_propagate(errp, local_err);
2092 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2094 return bdrv_named_nodes_list();
2097 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2099 void qmp_drive_mirror(const char *device, const char *target,
2100 bool has_format, const char *format,
2101 enum MirrorSyncMode sync,
2102 bool has_mode, enum NewImageMode mode,
2103 bool has_speed, int64_t speed,
2104 bool has_granularity, uint32_t granularity,
2105 bool has_buf_size, int64_t buf_size,
2106 bool has_on_source_error, BlockdevOnError on_source_error,
2107 bool has_on_target_error, BlockdevOnError on_target_error,
2110 BlockDriverState *bs;
2111 BlockDriverState *source, *target_bs;
2112 BlockDriver *drv = NULL;
2113 Error *local_err = NULL;
2121 if (!has_on_source_error) {
2122 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2124 if (!has_on_target_error) {
2125 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2128 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2130 if (!has_granularity) {
2133 if (!has_buf_size) {
2134 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2137 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2138 error_set(errp, QERR_INVALID_PARAMETER, device);
2141 if (granularity & (granularity - 1)) {
2142 error_set(errp, QERR_INVALID_PARAMETER, device);
2146 bs = bdrv_find(device);
2148 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2152 if (!bdrv_is_inserted(bs)) {
2153 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2158 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2161 drv = bdrv_find_format(format);
2163 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2168 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR, errp)) {
2172 flags = bs->open_flags | BDRV_O_RDWR;
2173 source = bs->backing_hd;
2174 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2175 sync = MIRROR_SYNC_MODE_FULL;
2177 if (sync == MIRROR_SYNC_MODE_NONE) {
2181 size = bdrv_getlength(bs);
2183 error_setg_errno(errp, -size, "bdrv_getlength failed");
2187 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2188 && mode != NEW_IMAGE_MODE_EXISTING)
2190 /* create new image w/o backing file */
2191 assert(format && drv);
2192 bdrv_img_create(target, format,
2193 NULL, NULL, NULL, size, flags, &local_err, false);
2196 case NEW_IMAGE_MODE_EXISTING:
2198 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2199 /* create new image with backing file */
2200 bdrv_img_create(target, format,
2202 source->drv->format_name,
2203 NULL, size, flags, &local_err, false);
2211 error_propagate(errp, local_err);
2215 /* Mirroring takes care of copy-on-write using the source's backing
2219 ret = bdrv_open(&target_bs, target, NULL, NULL, flags | BDRV_O_NO_BACKING,
2222 error_propagate(errp, local_err);
2226 mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2227 on_source_error, on_target_error,
2228 block_job_cb, bs, &local_err);
2229 if (local_err != NULL) {
2230 bdrv_unref(target_bs);
2231 error_propagate(errp, local_err);
2236 static BlockJob *find_block_job(const char *device)
2238 BlockDriverState *bs;
2240 bs = bdrv_find(device);
2241 if (!bs || !bs->job) {
2247 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2249 BlockJob *job = find_block_job(device);
2252 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2256 block_job_set_speed(job, speed, errp);
2259 void qmp_block_job_cancel(const char *device,
2260 bool has_force, bool force, Error **errp)
2262 BlockJob *job = find_block_job(device);
2269 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2272 if (job->paused && !force) {
2273 error_setg(errp, "The block job for device '%s' is currently paused",
2278 trace_qmp_block_job_cancel(job);
2279 block_job_cancel(job);
2282 void qmp_block_job_pause(const char *device, Error **errp)
2284 BlockJob *job = find_block_job(device);
2287 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2291 trace_qmp_block_job_pause(job);
2292 block_job_pause(job);
2295 void qmp_block_job_resume(const char *device, Error **errp)
2297 BlockJob *job = find_block_job(device);
2300 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2304 trace_qmp_block_job_resume(job);
2305 block_job_resume(job);
2308 void qmp_block_job_complete(const char *device, Error **errp)
2310 BlockJob *job = find_block_job(device);
2313 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2317 trace_qmp_block_job_complete(job);
2318 block_job_complete(job, errp);
2321 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2323 QmpOutputVisitor *ov = qmp_output_visitor_new();
2327 Error *local_err = NULL;
2329 /* Require an ID in the top level */
2330 if (!options->has_id) {
2331 error_setg(errp, "Block device needs an ID");
2335 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2336 * cache.direct=false instead of silently switching to aio=threads, except
2337 * if called from drive_init.
2339 * For now, simply forbidding the combination for all drivers will do. */
2340 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2341 bool direct = options->has_cache &&
2342 options->cache->has_direct &&
2343 options->cache->direct;
2345 error_setg(errp, "aio=native requires cache.direct=true");
2350 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2351 &options, NULL, &local_err);
2353 error_propagate(errp, local_err);
2357 obj = qmp_output_get_qobject(ov);
2358 qdict = qobject_to_qdict(obj);
2360 qdict_flatten(qdict);
2362 dinfo = blockdev_init(NULL, qdict, &local_err);
2364 error_propagate(errp, local_err);
2368 if (bdrv_key_required(dinfo->bdrv)) {
2369 drive_uninit(dinfo);
2370 error_setg(errp, "blockdev-add doesn't support encrypted devices");
2375 qmp_output_visitor_cleanup(ov);
2378 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2380 BlockJobInfoList **prev = opaque;
2381 BlockJob *job = bs->job;
2384 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2385 elem->value = block_job_query(bs->job);
2386 (*prev)->next = elem;
2391 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2393 /* Dummy is a fake list element for holding the head pointer */
2394 BlockJobInfoList dummy = {};
2395 BlockJobInfoList *prev = &dummy;
2396 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2400 QemuOptsList qemu_common_drive_opts = {
2402 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2406 .type = QEMU_OPT_BOOL,
2407 .help = "enable/disable snapshot mode",
2410 .type = QEMU_OPT_STRING,
2411 .help = "discard operation (ignore/off, unmap/on)",
2413 .name = "cache.writeback",
2414 .type = QEMU_OPT_BOOL,
2415 .help = "enables writeback mode for any caches",
2417 .name = "cache.direct",
2418 .type = QEMU_OPT_BOOL,
2419 .help = "enables use of O_DIRECT (bypass the host page cache)",
2421 .name = "cache.no-flush",
2422 .type = QEMU_OPT_BOOL,
2423 .help = "ignore any flush requests for the device",
2426 .type = QEMU_OPT_STRING,
2427 .help = "host AIO implementation (threads, native)",
2430 .type = QEMU_OPT_STRING,
2431 .help = "disk format (raw, qcow2, ...)",
2434 .type = QEMU_OPT_STRING,
2435 .help = "disk serial number",
2438 .type = QEMU_OPT_STRING,
2439 .help = "read error action",
2442 .type = QEMU_OPT_STRING,
2443 .help = "write error action",
2445 .name = "read-only",
2446 .type = QEMU_OPT_BOOL,
2447 .help = "open drive file as read-only",
2449 .name = "throttling.iops-total",
2450 .type = QEMU_OPT_NUMBER,
2451 .help = "limit total I/O operations per second",
2453 .name = "throttling.iops-read",
2454 .type = QEMU_OPT_NUMBER,
2455 .help = "limit read operations per second",
2457 .name = "throttling.iops-write",
2458 .type = QEMU_OPT_NUMBER,
2459 .help = "limit write operations per second",
2461 .name = "throttling.bps-total",
2462 .type = QEMU_OPT_NUMBER,
2463 .help = "limit total bytes per second",
2465 .name = "throttling.bps-read",
2466 .type = QEMU_OPT_NUMBER,
2467 .help = "limit read bytes per second",
2469 .name = "throttling.bps-write",
2470 .type = QEMU_OPT_NUMBER,
2471 .help = "limit write bytes per second",
2473 .name = "throttling.iops-total-max",
2474 .type = QEMU_OPT_NUMBER,
2475 .help = "I/O operations burst",
2477 .name = "throttling.iops-read-max",
2478 .type = QEMU_OPT_NUMBER,
2479 .help = "I/O operations read burst",
2481 .name = "throttling.iops-write-max",
2482 .type = QEMU_OPT_NUMBER,
2483 .help = "I/O operations write burst",
2485 .name = "throttling.bps-total-max",
2486 .type = QEMU_OPT_NUMBER,
2487 .help = "total bytes burst",
2489 .name = "throttling.bps-read-max",
2490 .type = QEMU_OPT_NUMBER,
2491 .help = "total bytes read burst",
2493 .name = "throttling.bps-write-max",
2494 .type = QEMU_OPT_NUMBER,
2495 .help = "total bytes write burst",
2497 .name = "throttling.iops-size",
2498 .type = QEMU_OPT_NUMBER,
2499 .help = "when limiting by iops max size of an I/O in bytes",
2501 .name = "copy-on-read",
2502 .type = QEMU_OPT_BOOL,
2503 .help = "copy read data from backing file into image file",
2505 .name = "detect-zeroes",
2506 .type = QEMU_OPT_STRING,
2507 .help = "try to optimize zero writes (off, on, unmap)",
2509 { /* end of list */ }
2513 QemuOptsList qemu_drive_opts = {
2515 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2518 * no elements => accept any params
2519 * validation will happen later
2521 { /* end of list */ }