2 * QEMU host block devices
4 * Copyright (c) 2003-2008 Fabrice Bellard
6 * This work is licensed under the terms of the GNU GPL, version 2 or
7 * later. See the COPYING file in the top-level directory.
9 * This file incorporates work covered by the following copyright and
12 * Copyright (c) 2003-2008 Fabrice Bellard
14 * Permission is hereby granted, free of charge, to any person obtaining a copy
15 * of this software and associated documentation files (the "Software"), to deal
16 * in the Software without restriction, including without limitation the rights
17 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
18 * copies of the Software, and to permit persons to whom the Software is
19 * furnished to do so, subject to the following conditions:
21 * The above copyright notice and this permission notice shall be included in
22 * all copies or substantial portions of the Software.
24 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
25 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
26 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
27 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
28 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
29 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
33 #include "sysemu/blockdev.h"
34 #include "hw/block/block.h"
35 #include "block/blockjob.h"
36 #include "monitor/monitor.h"
37 #include "qapi/qmp/qerror.h"
38 #include "qemu/option.h"
39 #include "qemu/config-file.h"
40 #include "qapi/qmp/types.h"
41 #include "qapi-visit.h"
42 #include "qapi/qmp-output-visitor.h"
43 #include "sysemu/sysemu.h"
44 #include "block/block_int.h"
45 #include "qmp-commands.h"
47 #include "sysemu/arch_init.h"
49 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
51 static const char *const if_name[IF_COUNT] = {
55 [IF_FLOPPY] = "floppy",
56 [IF_PFLASH] = "pflash",
59 [IF_VIRTIO] = "virtio",
63 static const int if_max_devs[IF_COUNT] = {
65 * Do not change these numbers! They govern how drive option
66 * index maps to unit and bus. That mapping is ABI.
68 * All controllers used to imlement if=T drives need to support
69 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
70 * Otherwise, some index values map to "impossible" bus, unit
73 * For instance, if you change [IF_SCSI] to 255, -drive
74 * if=scsi,index=12 no longer means bus=1,unit=5, but
75 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
76 * the drive can't be set up. Regression.
83 * We automatically delete the drive when a device using it gets
84 * unplugged. Questionable feature, but we can't just drop it.
85 * Device models call blockdev_mark_auto_del() to schedule the
86 * automatic deletion, and generic qdev code calls blockdev_auto_del()
87 * when deletion is actually safe.
89 void blockdev_mark_auto_del(BlockDriverState *bs)
91 DriveInfo *dinfo = drive_get_by_blockdev(bs);
93 if (dinfo && !dinfo->enable_auto_del) {
98 block_job_cancel(bs->job);
105 void blockdev_auto_del(BlockDriverState *bs)
107 DriveInfo *dinfo = drive_get_by_blockdev(bs);
109 if (dinfo && dinfo->auto_del) {
110 drive_put_ref(dinfo);
114 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
116 int max_devs = if_max_devs[type];
117 return max_devs ? index / max_devs : 0;
120 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
122 int max_devs = if_max_devs[type];
123 return max_devs ? index % max_devs : index;
126 QemuOpts *drive_def(const char *optstr)
128 return qemu_opts_parse(qemu_find_opts("drive"), optstr, 0);
131 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
137 opts = drive_def(optstr);
141 if (type != IF_DEFAULT) {
142 qemu_opt_set(opts, "if", if_name[type]);
145 snprintf(buf, sizeof(buf), "%d", index);
146 qemu_opt_set(opts, "index", buf);
149 qemu_opt_set(opts, "file", file);
153 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
157 /* seek interface, bus and unit */
159 QTAILQ_FOREACH(dinfo, &drives, next) {
160 if (dinfo->type == type &&
169 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
171 return drive_get(type,
172 drive_index_to_bus_id(type, index),
173 drive_index_to_unit_id(type, index));
176 int drive_get_max_bus(BlockInterfaceType type)
182 QTAILQ_FOREACH(dinfo, &drives, next) {
183 if(dinfo->type == type &&
184 dinfo->bus > max_bus)
185 max_bus = dinfo->bus;
190 /* Get a block device. This should only be used for single-drive devices
191 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
193 DriveInfo *drive_get_next(BlockInterfaceType type)
195 static int next_block_unit[IF_COUNT];
197 return drive_get(type, 0, next_block_unit[type]++);
200 DriveInfo *drive_get_by_blockdev(BlockDriverState *bs)
204 QTAILQ_FOREACH(dinfo, &drives, next) {
205 if (dinfo->bdrv == bs) {
212 static void bdrv_format_print(void *opaque, const char *name)
214 error_printf(" %s", name);
217 static void drive_uninit(DriveInfo *dinfo)
220 qemu_opts_del(dinfo->opts);
223 bdrv_unref(dinfo->bdrv);
225 QTAILQ_REMOVE(&drives, dinfo, next);
226 g_free(dinfo->serial);
230 void drive_put_ref(DriveInfo *dinfo)
232 assert(dinfo->refcount);
233 if (--dinfo->refcount == 0) {
238 void drive_get_ref(DriveInfo *dinfo)
245 BlockDriverState *bs;
248 static void bdrv_put_ref_bh(void *opaque)
250 BDRVPutRefBH *s = opaque;
253 qemu_bh_delete(s->bh);
258 * Release a BDS reference in a BH
260 * It is not safe to use bdrv_unref() from a callback function when the callers
261 * still need the BlockDriverState. In such cases we schedule a BH to release
264 static void bdrv_put_ref_bh_schedule(BlockDriverState *bs)
268 s = g_new(BDRVPutRefBH, 1);
269 s->bh = qemu_bh_new(bdrv_put_ref_bh, s);
271 qemu_bh_schedule(s->bh);
274 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
276 if (!strcmp(buf, "ignore")) {
277 return BLOCKDEV_ON_ERROR_IGNORE;
278 } else if (!is_read && !strcmp(buf, "enospc")) {
279 return BLOCKDEV_ON_ERROR_ENOSPC;
280 } else if (!strcmp(buf, "stop")) {
281 return BLOCKDEV_ON_ERROR_STOP;
282 } else if (!strcmp(buf, "report")) {
283 return BLOCKDEV_ON_ERROR_REPORT;
285 error_setg(errp, "'%s' invalid %s error action",
286 buf, is_read ? "read" : "write");
291 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
293 if (throttle_conflicting(cfg)) {
294 error_setg(errp, "bps/iops/max total values and read/write values"
295 " cannot be used at the same time");
299 if (!throttle_is_valid(cfg)) {
300 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
307 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
309 /* Takes the ownership of bs_opts */
310 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
311 BlockInterfaceType type,
318 int on_read_error, on_write_error;
327 bool has_driver_specific_opts;
328 BlockDriver *drv = NULL;
330 /* Check common options by copying from bs_opts to opts, all other options
331 * stay in bs_opts for processing by bdrv_open(). */
332 id = qdict_get_try_str(bs_opts, "id");
333 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
334 if (error_is_set(&error)) {
335 error_propagate(errp, error);
339 qemu_opts_absorb_qdict(opts, bs_opts, &error);
340 if (error_is_set(&error)) {
341 error_propagate(errp, error);
346 qdict_del(bs_opts, "id");
349 has_driver_specific_opts = !!qdict_size(bs_opts);
351 /* extract parameters */
352 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
353 ro = qemu_opt_get_bool(opts, "read-only", 0);
354 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
356 serial = qemu_opt_get(opts, "serial");
358 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
359 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
360 error_setg(errp, "invalid discard option");
365 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
366 bdrv_flags |= BDRV_O_CACHE_WB;
368 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
369 bdrv_flags |= BDRV_O_NOCACHE;
371 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
372 bdrv_flags |= BDRV_O_NO_FLUSH;
375 #ifdef CONFIG_LINUX_AIO
376 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
377 if (!strcmp(buf, "native")) {
378 bdrv_flags |= BDRV_O_NATIVE_AIO;
379 } else if (!strcmp(buf, "threads")) {
380 /* this is the default */
382 error_setg(errp, "invalid aio option");
388 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
389 if (is_help_option(buf)) {
390 error_printf("Supported formats:");
391 bdrv_iterate_format(bdrv_format_print, NULL);
396 drv = bdrv_find_format(buf);
398 error_setg(errp, "'%s' invalid format", buf);
403 /* disk I/O throttling */
404 memset(&cfg, 0, sizeof(cfg));
405 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
406 qemu_opt_get_number(opts, "throttling.bps-total", 0);
407 cfg.buckets[THROTTLE_BPS_READ].avg =
408 qemu_opt_get_number(opts, "throttling.bps-read", 0);
409 cfg.buckets[THROTTLE_BPS_WRITE].avg =
410 qemu_opt_get_number(opts, "throttling.bps-write", 0);
411 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
412 qemu_opt_get_number(opts, "throttling.iops-total", 0);
413 cfg.buckets[THROTTLE_OPS_READ].avg =
414 qemu_opt_get_number(opts, "throttling.iops-read", 0);
415 cfg.buckets[THROTTLE_OPS_WRITE].avg =
416 qemu_opt_get_number(opts, "throttling.iops-write", 0);
418 cfg.buckets[THROTTLE_BPS_TOTAL].max =
419 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
420 cfg.buckets[THROTTLE_BPS_READ].max =
421 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
422 cfg.buckets[THROTTLE_BPS_WRITE].max =
423 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
424 cfg.buckets[THROTTLE_OPS_TOTAL].max =
425 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
426 cfg.buckets[THROTTLE_OPS_READ].max =
427 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
428 cfg.buckets[THROTTLE_OPS_WRITE].max =
429 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
431 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
433 if (!check_throttle_config(&cfg, &error)) {
434 error_propagate(errp, error);
438 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
439 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
440 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO && type != IF_NONE) {
441 error_setg(errp, "werror is not supported by this bus type");
445 on_write_error = parse_block_error_action(buf, 0, &error);
446 if (error_is_set(&error)) {
447 error_propagate(errp, error);
452 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
453 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
454 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI && type != IF_NONE) {
455 error_report("rerror is not supported by this bus type");
459 on_read_error = parse_block_error_action(buf, 1, &error);
460 if (error_is_set(&error)) {
461 error_propagate(errp, error);
467 dinfo = g_malloc0(sizeof(*dinfo));
468 dinfo->id = g_strdup(qemu_opts_id(opts));
469 dinfo->bdrv = bdrv_new(dinfo->id);
470 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
471 dinfo->bdrv->read_only = ro;
474 if (serial != NULL) {
475 dinfo->serial = g_strdup(serial);
477 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
479 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
481 /* disk I/O throttling */
482 if (throttle_enabled(&cfg)) {
483 bdrv_io_limits_enable(dinfo->bdrv);
484 bdrv_set_io_limits(dinfo->bdrv, &cfg);
487 if (!file || !*file) {
488 if (has_driver_specific_opts) {
497 /* always use cache=unsafe with snapshot */
498 bdrv_flags &= ~BDRV_O_CACHE_MASK;
499 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
503 bdrv_flags |= BDRV_O_COPY_ON_READ;
506 if (runstate_check(RUN_STATE_INMIGRATE)) {
507 bdrv_flags |= BDRV_O_INCOMING;
510 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
513 ret = bdrv_open(dinfo->bdrv, file, bs_opts, bdrv_flags, drv, &error);
516 error_setg(errp, "could not open disk image %s: %s",
517 file ?: dinfo->id, error_get_pretty(error));
522 if (bdrv_key_required(dinfo->bdrv))
531 bdrv_unref(dinfo->bdrv);
533 QTAILQ_REMOVE(&drives, dinfo, next);
541 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
545 value = qemu_opt_get(opts, from);
547 qemu_opt_set(opts, to, value);
548 qemu_opt_unset(opts, from);
552 QemuOptsList qemu_legacy_drive_opts = {
554 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
558 .type = QEMU_OPT_NUMBER,
559 .help = "bus number",
562 .type = QEMU_OPT_NUMBER,
563 .help = "unit number (i.e. lun for scsi)",
566 .type = QEMU_OPT_NUMBER,
567 .help = "index number",
570 .type = QEMU_OPT_STRING,
571 .help = "media type (disk, cdrom)",
574 .type = QEMU_OPT_STRING,
575 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
578 .type = QEMU_OPT_NUMBER,
579 .help = "number of cylinders (ide disk geometry)",
582 .type = QEMU_OPT_NUMBER,
583 .help = "number of heads (ide disk geometry)",
586 .type = QEMU_OPT_NUMBER,
587 .help = "number of sectors (ide disk geometry)",
590 .type = QEMU_OPT_STRING,
591 .help = "chs translation (auto, lba, none)",
594 .type = QEMU_OPT_BOOL,
595 .help = "(deprecated, ignored)",
598 .type = QEMU_OPT_STRING,
599 .help = "pci address (virtio only)",
602 .type = QEMU_OPT_STRING,
606 /* Options that are passed on, but have special semantics with -drive */
609 .type = QEMU_OPT_BOOL,
610 .help = "open drive file as read-only",
612 .name = "copy-on-read",
613 .type = QEMU_OPT_BOOL,
614 .help = "copy read data from backing file into image file",
617 { /* end of list */ }
621 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
624 DriveInfo *dinfo = NULL;
626 QemuOpts *legacy_opts;
627 DriveMediaType media = MEDIA_DISK;
628 BlockInterfaceType type;
629 int cyls, heads, secs, translation;
630 int max_devs, bus_id, unit_id, index;
632 bool read_only = false;
634 const char *filename;
635 Error *local_err = NULL;
637 /* Change legacy command line options into QMP ones */
638 qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
639 qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
640 qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
642 qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
643 qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
644 qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
646 qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
647 qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
648 qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
650 qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
651 qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
652 qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
654 qemu_opt_rename(all_opts,
655 "iops_size", "throttling.iops-size");
657 qemu_opt_rename(all_opts, "readonly", "read-only");
659 value = qemu_opt_get(all_opts, "cache");
663 if (bdrv_parse_cache_flags(value, &flags) != 0) {
664 error_report("invalid cache option");
668 /* Specific options take precedence */
669 if (!qemu_opt_get(all_opts, "cache.writeback")) {
670 qemu_opt_set_bool(all_opts, "cache.writeback",
671 !!(flags & BDRV_O_CACHE_WB));
673 if (!qemu_opt_get(all_opts, "cache.direct")) {
674 qemu_opt_set_bool(all_opts, "cache.direct",
675 !!(flags & BDRV_O_NOCACHE));
677 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
678 qemu_opt_set_bool(all_opts, "cache.no-flush",
679 !!(flags & BDRV_O_NO_FLUSH));
681 qemu_opt_unset(all_opts, "cache");
684 /* Get a QDict for processing the options */
685 bs_opts = qdict_new();
686 qemu_opts_to_qdict(all_opts, bs_opts);
688 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
690 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
691 if (error_is_set(&local_err)) {
692 qerror_report_err(local_err);
693 error_free(local_err);
697 /* Deprecated option boot=[on|off] */
698 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
699 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
700 "ignored. Future versions will reject this parameter. Please "
701 "update your scripts.\n");
705 value = qemu_opt_get(legacy_opts, "media");
707 if (!strcmp(value, "disk")) {
709 } else if (!strcmp(value, "cdrom")) {
713 error_report("'%s' invalid media", value);
718 /* copy-on-read is disabled with a warning for read-only devices */
719 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
720 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
722 if (read_only && copy_on_read) {
723 error_report("warning: disabling copy-on-read on read-only drive");
724 copy_on_read = false;
727 qdict_put(bs_opts, "read-only",
728 qstring_from_str(read_only ? "on" : "off"));
729 qdict_put(bs_opts, "copy-on-read",
730 qstring_from_str(copy_on_read ? "on" :"off"));
732 /* Controller type */
733 value = qemu_opt_get(legacy_opts, "if");
736 type < IF_COUNT && strcmp(value, if_name[type]);
739 if (type == IF_COUNT) {
740 error_report("unsupported bus type '%s'", value);
744 type = block_default_type;
748 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
749 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
750 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
752 if (cyls || heads || secs) {
754 error_report("invalid physical cyls number");
758 error_report("invalid physical heads number");
762 error_report("invalid physical secs number");
767 translation = BIOS_ATA_TRANSLATION_AUTO;
768 value = qemu_opt_get(legacy_opts, "trans");
771 error_report("'%s' trans must be used with cyls, heads and secs",
775 if (!strcmp(value, "none")) {
776 translation = BIOS_ATA_TRANSLATION_NONE;
777 } else if (!strcmp(value, "lba")) {
778 translation = BIOS_ATA_TRANSLATION_LBA;
779 } else if (!strcmp(value, "large")) {
780 translation = BIOS_ATA_TRANSLATION_LARGE;
781 } else if (!strcmp(value, "rechs")) {
782 translation = BIOS_ATA_TRANSLATION_RECHS;
783 } else if (!strcmp(value, "auto")) {
784 translation = BIOS_ATA_TRANSLATION_AUTO;
786 error_report("'%s' invalid translation type", value);
791 if (media == MEDIA_CDROM) {
792 if (cyls || secs || heads) {
793 error_report("CHS can't be set with media=cdrom");
798 /* Device address specified by bus/unit or index.
799 * If none was specified, try to find the first free one. */
800 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
801 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
802 index = qemu_opt_get_number(legacy_opts, "index", -1);
804 max_devs = if_max_devs[type];
807 if (bus_id != 0 || unit_id != -1) {
808 error_report("index cannot be used with bus and unit");
811 bus_id = drive_index_to_bus_id(type, index);
812 unit_id = drive_index_to_unit_id(type, index);
817 while (drive_get(type, bus_id, unit_id) != NULL) {
819 if (max_devs && unit_id >= max_devs) {
826 if (max_devs && unit_id >= max_devs) {
827 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
831 if (drive_get(type, bus_id, unit_id) != NULL) {
832 error_report("drive with bus=%d, unit=%d (index=%d) exists",
833 bus_id, unit_id, index);
837 /* no id supplied -> create one */
838 if (qemu_opts_id(all_opts) == NULL) {
840 const char *mediastr = "";
841 if (type == IF_IDE || type == IF_SCSI) {
842 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
845 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
848 new_id = g_strdup_printf("%s%s%i", if_name[type],
851 qdict_put(bs_opts, "id", qstring_from_str(new_id));
855 /* Add virtio block device */
856 devaddr = qemu_opt_get(legacy_opts, "addr");
857 if (devaddr && type != IF_VIRTIO) {
858 error_report("addr is not supported by this bus type");
862 if (type == IF_VIRTIO) {
864 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
866 if (arch_type == QEMU_ARCH_S390X) {
867 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
869 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
871 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
873 qemu_opt_set(devopts, "addr", devaddr);
877 filename = qemu_opt_get(legacy_opts, "file");
879 /* Actual block device init: Functionality shared with blockdev-add */
880 dinfo = blockdev_init(filename, bs_opts, type, &local_err);
882 if (error_is_set(&local_err)) {
883 qerror_report_err(local_err);
884 error_free(local_err);
888 assert(!error_is_set(&local_err));
891 /* Set legacy DriveInfo fields */
892 dinfo->enable_auto_del = true;
893 dinfo->opts = all_opts;
896 dinfo->heads = heads;
898 dinfo->trans = translation;
901 dinfo->unit = unit_id;
902 dinfo->devaddr = devaddr;
909 dinfo->media_cd = media == MEDIA_CDROM;
916 qemu_opts_del(legacy_opts);
920 void do_commit(Monitor *mon, const QDict *qdict)
922 const char *device = qdict_get_str(qdict, "device");
923 BlockDriverState *bs;
926 if (!strcmp(device, "all")) {
927 ret = bdrv_commit_all();
929 bs = bdrv_find(device);
931 monitor_printf(mon, "Device '%s' not found\n", device);
934 ret = bdrv_commit(bs);
937 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
942 static void blockdev_do_action(int kind, void *data, Error **errp)
944 TransactionAction action;
945 TransactionActionList list;
949 list.value = &action;
951 qmp_transaction(&list, errp);
954 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
955 bool has_node_name, const char *node_name,
956 const char *snapshot_file,
957 bool has_snapshot_node_name,
958 const char *snapshot_node_name,
959 bool has_format, const char *format,
960 bool has_mode, NewImageMode mode, Error **errp)
962 BlockdevSnapshot snapshot = {
963 .has_device = has_device,
964 .device = (char *) device,
965 .has_node_name = has_node_name,
966 .node_name = (char *) node_name,
967 .snapshot_file = (char *) snapshot_file,
968 .has_snapshot_node_name = has_snapshot_node_name,
969 .snapshot_node_name = (char *) snapshot_node_name,
970 .has_format = has_format,
971 .format = (char *) format,
972 .has_mode = has_mode,
975 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
979 void qmp_blockdev_snapshot_internal_sync(const char *device,
983 BlockdevSnapshotInternal snapshot = {
984 .device = (char *) device,
985 .name = (char *) name
988 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
992 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
999 BlockDriverState *bs = bdrv_find(device);
1000 QEMUSnapshotInfo sn;
1001 Error *local_err = NULL;
1002 SnapshotInfo *info = NULL;
1006 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1019 error_setg(errp, "Name or id must be provided");
1023 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1024 if (error_is_set(&local_err)) {
1025 error_propagate(errp, local_err);
1030 "Snapshot with id '%s' and name '%s' does not exist on "
1032 STR_OR_NULL(id), STR_OR_NULL(name), device);
1036 bdrv_snapshot_delete(bs, id, name, &local_err);
1037 if (error_is_set(&local_err)) {
1038 error_propagate(errp, local_err);
1042 info = g_malloc0(sizeof(SnapshotInfo));
1043 info->id = g_strdup(sn.id_str);
1044 info->name = g_strdup(sn.name);
1045 info->date_nsec = sn.date_nsec;
1046 info->date_sec = sn.date_sec;
1047 info->vm_state_size = sn.vm_state_size;
1048 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1049 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1054 /* New and old BlockDriverState structs for group snapshots */
1056 typedef struct BlkTransactionState BlkTransactionState;
1058 /* Only prepare() may fail. In a single transaction, only one of commit() or
1059 abort() will be called, clean() will always be called if it present. */
1060 typedef struct BdrvActionOps {
1061 /* Size of state struct, in bytes. */
1062 size_t instance_size;
1063 /* Prepare the work, must NOT be NULL. */
1064 void (*prepare)(BlkTransactionState *common, Error **errp);
1065 /* Commit the changes, can be NULL. */
1066 void (*commit)(BlkTransactionState *common);
1067 /* Abort the changes on fail, can be NULL. */
1068 void (*abort)(BlkTransactionState *common);
1069 /* Clean up resource in the end, can be NULL. */
1070 void (*clean)(BlkTransactionState *common);
1074 * This structure must be arranged as first member in child type, assuming
1075 * that compiler will also arrange it to the same address with parent instance.
1076 * Later it will be used in free().
1078 struct BlkTransactionState {
1079 TransactionAction *action;
1080 const BdrvActionOps *ops;
1081 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1084 /* internal snapshot private data */
1085 typedef struct InternalSnapshotState {
1086 BlkTransactionState common;
1087 BlockDriverState *bs;
1088 QEMUSnapshotInfo sn;
1089 } InternalSnapshotState;
1091 static void internal_snapshot_prepare(BlkTransactionState *common,
1096 BlockDriverState *bs;
1097 QEMUSnapshotInfo old_sn, *sn;
1100 BlockdevSnapshotInternal *internal;
1101 InternalSnapshotState *state;
1104 g_assert(common->action->kind ==
1105 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1106 internal = common->action->blockdev_snapshot_internal_sync;
1107 state = DO_UPCAST(InternalSnapshotState, common, common);
1109 /* 1. parse input */
1110 device = internal->device;
1111 name = internal->name;
1113 /* 2. check for validation */
1114 bs = bdrv_find(device);
1116 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1120 if (!bdrv_is_inserted(bs)) {
1121 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1125 if (bdrv_is_read_only(bs)) {
1126 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1130 if (!bdrv_can_snapshot(bs)) {
1131 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1132 bs->drv->format_name, device, "internal snapshot");
1136 if (!strlen(name)) {
1137 error_setg(errp, "Name is empty");
1141 /* check whether a snapshot with name exist */
1142 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn, errp);
1143 if (error_is_set(errp)) {
1147 "Snapshot with name '%s' already exists on device '%s'",
1152 /* 3. take the snapshot */
1154 pstrcpy(sn->name, sizeof(sn->name), name);
1155 qemu_gettimeofday(&tv);
1156 sn->date_sec = tv.tv_sec;
1157 sn->date_nsec = tv.tv_usec * 1000;
1158 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1160 ret1 = bdrv_snapshot_create(bs, sn);
1162 error_setg_errno(errp, -ret1,
1163 "Failed to create snapshot '%s' on device '%s'",
1168 /* 4. succeed, mark a snapshot is created */
1172 static void internal_snapshot_abort(BlkTransactionState *common)
1174 InternalSnapshotState *state =
1175 DO_UPCAST(InternalSnapshotState, common, common);
1176 BlockDriverState *bs = state->bs;
1177 QEMUSnapshotInfo *sn = &state->sn;
1178 Error *local_error = NULL;
1184 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1185 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1186 "device '%s' in abort: %s",
1189 bdrv_get_device_name(bs),
1190 error_get_pretty(local_error));
1191 error_free(local_error);
1195 /* external snapshot private data */
1196 typedef struct ExternalSnapshotState {
1197 BlkTransactionState common;
1198 BlockDriverState *old_bs;
1199 BlockDriverState *new_bs;
1200 } ExternalSnapshotState;
1202 static void external_snapshot_prepare(BlkTransactionState *common,
1207 QDict *options = NULL;
1208 Error *local_err = NULL;
1209 bool has_device = false;
1211 bool has_node_name = false;
1212 const char *node_name;
1213 bool has_snapshot_node_name = false;
1214 const char *snapshot_node_name;
1215 const char *new_image_file;
1216 const char *format = "qcow2";
1217 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1218 ExternalSnapshotState *state =
1219 DO_UPCAST(ExternalSnapshotState, common, common);
1220 TransactionAction *action = common->action;
1222 /* get parameters */
1223 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1225 has_device = action->blockdev_snapshot_sync->has_device;
1226 device = action->blockdev_snapshot_sync->device;
1227 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1228 node_name = action->blockdev_snapshot_sync->node_name;
1229 has_snapshot_node_name =
1230 action->blockdev_snapshot_sync->has_snapshot_node_name;
1231 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1233 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1234 if (action->blockdev_snapshot_sync->has_format) {
1235 format = action->blockdev_snapshot_sync->format;
1237 if (action->blockdev_snapshot_sync->has_mode) {
1238 mode = action->blockdev_snapshot_sync->mode;
1241 /* start processing */
1242 drv = bdrv_find_format(format);
1244 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1248 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1249 has_node_name ? node_name : NULL,
1251 if (error_is_set(&local_err)) {
1252 error_propagate(errp, local_err);
1256 if (has_node_name && !has_snapshot_node_name) {
1257 error_setg(errp, "New snapshot node name missing");
1261 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1262 error_setg(errp, "New snapshot node name already existing");
1266 if (!bdrv_is_inserted(state->old_bs)) {
1267 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1271 if (bdrv_in_use(state->old_bs)) {
1272 error_set(errp, QERR_DEVICE_IN_USE, device);
1276 if (!bdrv_is_read_only(state->old_bs)) {
1277 if (bdrv_flush(state->old_bs)) {
1278 error_set(errp, QERR_IO_ERROR);
1283 if (!bdrv_is_first_non_filter(state->old_bs)) {
1284 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1288 flags = state->old_bs->open_flags;
1290 /* create new image w/backing file */
1291 if (mode != NEW_IMAGE_MODE_EXISTING) {
1292 bdrv_img_create(new_image_file, format,
1293 state->old_bs->filename,
1294 state->old_bs->drv->format_name,
1295 NULL, -1, flags, &local_err, false);
1296 if (error_is_set(&local_err)) {
1297 error_propagate(errp, local_err);
1302 if (has_snapshot_node_name) {
1303 options = qdict_new();
1304 qdict_put(options, "node-name",
1305 qstring_from_str(snapshot_node_name));
1308 /* We will manually add the backing_hd field to the bs later */
1309 state->new_bs = bdrv_new("");
1310 /* TODO Inherit bs->options or only take explicit options with an
1311 * extended QMP command? */
1312 ret = bdrv_open(state->new_bs, new_image_file, options,
1313 flags | BDRV_O_NO_BACKING, drv, &local_err);
1315 error_propagate(errp, local_err);
1321 static void external_snapshot_commit(BlkTransactionState *common)
1323 ExternalSnapshotState *state =
1324 DO_UPCAST(ExternalSnapshotState, common, common);
1326 /* This removes our old bs and adds the new bs */
1327 bdrv_append(state->new_bs, state->old_bs);
1328 /* We don't need (or want) to use the transactional
1329 * bdrv_reopen_multiple() across all the entries at once, because we
1330 * don't want to abort all of them if one of them fails the reopen */
1331 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1335 static void external_snapshot_abort(BlkTransactionState *common)
1337 ExternalSnapshotState *state =
1338 DO_UPCAST(ExternalSnapshotState, common, common);
1339 if (state->new_bs) {
1340 bdrv_unref(state->new_bs);
1344 typedef struct DriveBackupState {
1345 BlkTransactionState common;
1346 BlockDriverState *bs;
1350 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1352 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1353 DriveBackup *backup;
1354 Error *local_err = NULL;
1356 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1357 backup = common->action->drive_backup;
1359 qmp_drive_backup(backup->device, backup->target,
1360 backup->has_format, backup->format,
1362 backup->has_mode, backup->mode,
1363 backup->has_speed, backup->speed,
1364 backup->has_on_source_error, backup->on_source_error,
1365 backup->has_on_target_error, backup->on_target_error,
1367 if (error_is_set(&local_err)) {
1368 error_propagate(errp, local_err);
1374 state->bs = bdrv_find(backup->device);
1375 state->job = state->bs->job;
1378 static void drive_backup_abort(BlkTransactionState *common)
1380 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1381 BlockDriverState *bs = state->bs;
1383 /* Only cancel if it's the job we started */
1384 if (bs && bs->job && bs->job == state->job) {
1385 block_job_cancel_sync(bs->job);
1389 static void abort_prepare(BlkTransactionState *common, Error **errp)
1391 error_setg(errp, "Transaction aborted using Abort action");
1394 static void abort_commit(BlkTransactionState *common)
1396 g_assert_not_reached(); /* this action never succeeds */
1399 static const BdrvActionOps actions[] = {
1400 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1401 .instance_size = sizeof(ExternalSnapshotState),
1402 .prepare = external_snapshot_prepare,
1403 .commit = external_snapshot_commit,
1404 .abort = external_snapshot_abort,
1406 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1407 .instance_size = sizeof(DriveBackupState),
1408 .prepare = drive_backup_prepare,
1409 .abort = drive_backup_abort,
1411 [TRANSACTION_ACTION_KIND_ABORT] = {
1412 .instance_size = sizeof(BlkTransactionState),
1413 .prepare = abort_prepare,
1414 .commit = abort_commit,
1416 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1417 .instance_size = sizeof(InternalSnapshotState),
1418 .prepare = internal_snapshot_prepare,
1419 .abort = internal_snapshot_abort,
1424 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1425 * then we do not pivot any of the devices in the group, and abandon the
1428 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1430 TransactionActionList *dev_entry = dev_list;
1431 BlkTransactionState *state, *next;
1432 Error *local_err = NULL;
1434 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1435 QSIMPLEQ_INIT(&snap_bdrv_states);
1437 /* drain all i/o before any snapshots */
1440 /* We don't do anything in this loop that commits us to the snapshot */
1441 while (NULL != dev_entry) {
1442 TransactionAction *dev_info = NULL;
1443 const BdrvActionOps *ops;
1445 dev_info = dev_entry->value;
1446 dev_entry = dev_entry->next;
1448 assert(dev_info->kind < ARRAY_SIZE(actions));
1450 ops = &actions[dev_info->kind];
1451 assert(ops->instance_size > 0);
1453 state = g_malloc0(ops->instance_size);
1455 state->action = dev_info;
1456 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1458 state->ops->prepare(state, &local_err);
1459 if (error_is_set(&local_err)) {
1460 error_propagate(errp, local_err);
1461 goto delete_and_fail;
1465 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1466 if (state->ops->commit) {
1467 state->ops->commit(state);
1476 * failure, and it is all-or-none; abandon each new bs, and keep using
1477 * the original bs for all images
1479 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1480 if (state->ops->abort) {
1481 state->ops->abort(state);
1485 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1486 if (state->ops->clean) {
1487 state->ops->clean(state);
1494 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1496 if (bdrv_in_use(bs)) {
1497 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1500 if (!bdrv_dev_has_removable_media(bs)) {
1501 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1505 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1506 bdrv_dev_eject_request(bs, force);
1508 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1516 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1518 BlockDriverState *bs;
1520 bs = bdrv_find(device);
1522 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1526 eject_device(bs, force, errp);
1529 void qmp_block_passwd(bool has_device, const char *device,
1530 bool has_node_name, const char *node_name,
1531 const char *password, Error **errp)
1533 Error *local_err = NULL;
1534 BlockDriverState *bs;
1537 bs = bdrv_lookup_bs(has_device ? device : NULL,
1538 has_node_name ? node_name : NULL,
1540 if (error_is_set(&local_err)) {
1541 error_propagate(errp, local_err);
1545 err = bdrv_set_key(bs, password);
1546 if (err == -EINVAL) {
1547 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1549 } else if (err < 0) {
1550 error_set(errp, QERR_INVALID_PASSWORD);
1555 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1556 int bdrv_flags, BlockDriver *drv,
1557 const char *password, Error **errp)
1559 Error *local_err = NULL;
1562 ret = bdrv_open(bs, filename, NULL, bdrv_flags, drv, &local_err);
1564 error_propagate(errp, local_err);
1568 if (bdrv_key_required(bs)) {
1570 if (bdrv_set_key(bs, password) < 0) {
1571 error_set(errp, QERR_INVALID_PASSWORD);
1574 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1575 bdrv_get_encrypted_filename(bs));
1577 } else if (password) {
1578 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1582 void qmp_change_blockdev(const char *device, const char *filename,
1583 const char *format, Error **errp)
1585 BlockDriverState *bs;
1586 BlockDriver *drv = NULL;
1590 bs = bdrv_find(device);
1592 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1597 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1599 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1604 eject_device(bs, 0, &err);
1605 if (error_is_set(&err)) {
1606 error_propagate(errp, err);
1610 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1611 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1613 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1616 /* throttling disk I/O limits */
1617 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1624 bool has_bps_rd_max,
1626 bool has_bps_wr_max,
1630 bool has_iops_rd_max,
1631 int64_t iops_rd_max,
1632 bool has_iops_wr_max,
1633 int64_t iops_wr_max,
1635 int64_t iops_size, Error **errp)
1638 BlockDriverState *bs;
1640 bs = bdrv_find(device);
1642 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1646 memset(&cfg, 0, sizeof(cfg));
1647 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1648 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1649 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1651 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1652 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1653 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1656 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1658 if (has_bps_rd_max) {
1659 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1661 if (has_bps_wr_max) {
1662 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1665 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1667 if (has_iops_rd_max) {
1668 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1670 if (has_iops_wr_max) {
1671 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1674 if (has_iops_size) {
1675 cfg.op_size = iops_size;
1678 if (!check_throttle_config(&cfg, errp)) {
1682 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1683 bdrv_io_limits_enable(bs);
1684 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1685 bdrv_io_limits_disable(bs);
1688 if (bs->io_limits_enabled) {
1689 bdrv_set_io_limits(bs, &cfg);
1693 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1695 const char *id = qdict_get_str(qdict, "id");
1696 BlockDriverState *bs;
1700 qerror_report(QERR_DEVICE_NOT_FOUND, id);
1703 if (bdrv_in_use(bs)) {
1704 qerror_report(QERR_DEVICE_IN_USE, id);
1708 /* quiesce block driver; prevent further io */
1713 /* if we have a device attached to this BlockDriverState
1714 * then we need to make the drive anonymous until the device
1715 * can be removed. If this is a drive with no device backing
1716 * then we can just get rid of the block driver state right here.
1718 if (bdrv_get_attached_dev(bs)) {
1721 /* Further I/O must not pause the guest */
1722 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1723 BLOCKDEV_ON_ERROR_REPORT);
1725 drive_uninit(drive_get_by_blockdev(bs));
1731 void qmp_block_resize(bool has_device, const char *device,
1732 bool has_node_name, const char *node_name,
1733 int64_t size, Error **errp)
1735 Error *local_err = NULL;
1736 BlockDriverState *bs;
1739 bs = bdrv_lookup_bs(has_device ? device : NULL,
1740 has_node_name ? node_name : NULL,
1742 if (error_is_set(&local_err)) {
1743 error_propagate(errp, local_err);
1747 if (!bdrv_is_first_non_filter(bs)) {
1748 error_set(errp, QERR_FEATURE_DISABLED, "resize");
1753 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1757 /* complete all in-flight operations before resizing the device */
1760 ret = bdrv_truncate(bs, size);
1765 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1768 error_set(errp, QERR_UNSUPPORTED);
1771 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1774 error_set(errp, QERR_DEVICE_IN_USE, device);
1777 error_setg_errno(errp, -ret, "Could not resize");
1782 static void block_job_cb(void *opaque, int ret)
1784 BlockDriverState *bs = opaque;
1787 trace_block_job_cb(bs, bs->job, ret);
1790 obj = qobject_from_block_job(bs->job);
1792 QDict *dict = qobject_to_qdict(obj);
1793 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1796 if (block_job_is_cancelled(bs->job)) {
1797 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1799 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1801 qobject_decref(obj);
1803 bdrv_put_ref_bh_schedule(bs);
1806 void qmp_block_stream(const char *device, bool has_base,
1807 const char *base, bool has_speed, int64_t speed,
1808 bool has_on_error, BlockdevOnError on_error,
1811 BlockDriverState *bs;
1812 BlockDriverState *base_bs = NULL;
1813 Error *local_err = NULL;
1815 if (!has_on_error) {
1816 on_error = BLOCKDEV_ON_ERROR_REPORT;
1819 bs = bdrv_find(device);
1821 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1826 base_bs = bdrv_find_backing_image(bs, base);
1827 if (base_bs == NULL) {
1828 error_set(errp, QERR_BASE_NOT_FOUND, base);
1833 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1834 on_error, block_job_cb, bs, &local_err);
1835 if (error_is_set(&local_err)) {
1836 error_propagate(errp, local_err);
1840 trace_qmp_block_stream(bs, bs->job);
1843 void qmp_block_commit(const char *device,
1844 bool has_base, const char *base, const char *top,
1845 bool has_speed, int64_t speed,
1848 BlockDriverState *bs;
1849 BlockDriverState *base_bs, *top_bs;
1850 Error *local_err = NULL;
1851 /* This will be part of the QMP command, if/when the
1852 * BlockdevOnError change for blkmirror makes it in
1854 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1856 /* drain all i/o before commits */
1859 bs = bdrv_find(device);
1861 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1865 /* default top_bs is the active layer */
1869 if (strcmp(bs->filename, top) != 0) {
1870 top_bs = bdrv_find_backing_image(bs, top);
1874 if (top_bs == NULL) {
1875 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1879 if (has_base && base) {
1880 base_bs = bdrv_find_backing_image(top_bs, base);
1882 base_bs = bdrv_find_base(top_bs);
1885 if (base_bs == NULL) {
1886 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1891 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1894 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1897 if (local_err != NULL) {
1898 error_propagate(errp, local_err);
1903 void qmp_drive_backup(const char *device, const char *target,
1904 bool has_format, const char *format,
1905 enum MirrorSyncMode sync,
1906 bool has_mode, enum NewImageMode mode,
1907 bool has_speed, int64_t speed,
1908 bool has_on_source_error, BlockdevOnError on_source_error,
1909 bool has_on_target_error, BlockdevOnError on_target_error,
1912 BlockDriverState *bs;
1913 BlockDriverState *target_bs;
1914 BlockDriverState *source = NULL;
1915 BlockDriver *drv = NULL;
1916 Error *local_err = NULL;
1924 if (!has_on_source_error) {
1925 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1927 if (!has_on_target_error) {
1928 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1931 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1934 bs = bdrv_find(device);
1936 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1940 if (!bdrv_is_inserted(bs)) {
1941 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1946 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1949 drv = bdrv_find_format(format);
1951 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1956 if (bdrv_in_use(bs)) {
1957 error_set(errp, QERR_DEVICE_IN_USE, device);
1961 flags = bs->open_flags | BDRV_O_RDWR;
1963 /* See if we have a backing HD we can use to create our new image
1965 if (sync == MIRROR_SYNC_MODE_TOP) {
1966 source = bs->backing_hd;
1968 sync = MIRROR_SYNC_MODE_FULL;
1971 if (sync == MIRROR_SYNC_MODE_NONE) {
1975 size = bdrv_getlength(bs);
1977 error_setg_errno(errp, -size, "bdrv_getlength failed");
1981 if (mode != NEW_IMAGE_MODE_EXISTING) {
1982 assert(format && drv);
1984 bdrv_img_create(target, format, source->filename,
1985 source->drv->format_name, NULL,
1986 size, flags, &local_err, false);
1988 bdrv_img_create(target, format, NULL, NULL, NULL,
1989 size, flags, &local_err, false);
1993 if (error_is_set(&local_err)) {
1994 error_propagate(errp, local_err);
1998 target_bs = bdrv_new("");
1999 ret = bdrv_open(target_bs, target, NULL, flags, drv, &local_err);
2001 bdrv_unref(target_bs);
2002 error_propagate(errp, local_err);
2006 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2007 block_job_cb, bs, &local_err);
2008 if (local_err != NULL) {
2009 bdrv_unref(target_bs);
2010 error_propagate(errp, local_err);
2015 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2017 return bdrv_named_nodes_list();
2020 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2022 void qmp_drive_mirror(const char *device, const char *target,
2023 bool has_format, const char *format,
2024 enum MirrorSyncMode sync,
2025 bool has_mode, enum NewImageMode mode,
2026 bool has_speed, int64_t speed,
2027 bool has_granularity, uint32_t granularity,
2028 bool has_buf_size, int64_t buf_size,
2029 bool has_on_source_error, BlockdevOnError on_source_error,
2030 bool has_on_target_error, BlockdevOnError on_target_error,
2033 BlockDriverState *bs;
2034 BlockDriverState *source, *target_bs;
2035 BlockDriver *drv = NULL;
2036 Error *local_err = NULL;
2044 if (!has_on_source_error) {
2045 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2047 if (!has_on_target_error) {
2048 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2051 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2053 if (!has_granularity) {
2056 if (!has_buf_size) {
2057 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2060 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2061 error_set(errp, QERR_INVALID_PARAMETER, device);
2064 if (granularity & (granularity - 1)) {
2065 error_set(errp, QERR_INVALID_PARAMETER, device);
2069 bs = bdrv_find(device);
2071 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2075 if (!bdrv_is_inserted(bs)) {
2076 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2081 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2084 drv = bdrv_find_format(format);
2086 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2091 if (bdrv_in_use(bs)) {
2092 error_set(errp, QERR_DEVICE_IN_USE, device);
2096 flags = bs->open_flags | BDRV_O_RDWR;
2097 source = bs->backing_hd;
2098 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2099 sync = MIRROR_SYNC_MODE_FULL;
2101 if (sync == MIRROR_SYNC_MODE_NONE) {
2105 size = bdrv_getlength(bs);
2107 error_setg_errno(errp, -size, "bdrv_getlength failed");
2111 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2112 && mode != NEW_IMAGE_MODE_EXISTING)
2114 /* create new image w/o backing file */
2115 assert(format && drv);
2116 bdrv_img_create(target, format,
2117 NULL, NULL, NULL, size, flags, &local_err, false);
2120 case NEW_IMAGE_MODE_EXISTING:
2122 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2123 /* create new image with backing file */
2124 bdrv_img_create(target, format,
2126 source->drv->format_name,
2127 NULL, size, flags, &local_err, false);
2134 if (error_is_set(&local_err)) {
2135 error_propagate(errp, local_err);
2139 /* Mirroring takes care of copy-on-write using the source's backing
2142 target_bs = bdrv_new("");
2143 ret = bdrv_open(target_bs, target, NULL, flags | BDRV_O_NO_BACKING, drv,
2146 bdrv_unref(target_bs);
2147 error_propagate(errp, local_err);
2151 mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2152 on_source_error, on_target_error,
2153 block_job_cb, bs, &local_err);
2154 if (local_err != NULL) {
2155 bdrv_unref(target_bs);
2156 error_propagate(errp, local_err);
2161 static BlockJob *find_block_job(const char *device)
2163 BlockDriverState *bs;
2165 bs = bdrv_find(device);
2166 if (!bs || !bs->job) {
2172 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2174 BlockJob *job = find_block_job(device);
2177 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2181 block_job_set_speed(job, speed, errp);
2184 void qmp_block_job_cancel(const char *device,
2185 bool has_force, bool force, Error **errp)
2187 BlockJob *job = find_block_job(device);
2194 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2197 if (job->paused && !force) {
2198 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
2202 trace_qmp_block_job_cancel(job);
2203 block_job_cancel(job);
2206 void qmp_block_job_pause(const char *device, Error **errp)
2208 BlockJob *job = find_block_job(device);
2211 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2215 trace_qmp_block_job_pause(job);
2216 block_job_pause(job);
2219 void qmp_block_job_resume(const char *device, Error **errp)
2221 BlockJob *job = find_block_job(device);
2224 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2228 trace_qmp_block_job_resume(job);
2229 block_job_resume(job);
2232 void qmp_block_job_complete(const char *device, Error **errp)
2234 BlockJob *job = find_block_job(device);
2237 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2241 trace_qmp_block_job_complete(job);
2242 block_job_complete(job, errp);
2245 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2247 QmpOutputVisitor *ov = qmp_output_visitor_new();
2250 Error *local_err = NULL;
2252 /* Require an ID in the top level */
2253 if (!options->has_id) {
2254 error_setg(errp, "Block device needs an ID");
2258 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2259 * cache.direct=false instead of silently switching to aio=threads, except
2260 * if called from drive_init.
2262 * For now, simply forbidding the combination for all drivers will do. */
2263 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2264 bool direct = options->cache->has_direct && options->cache->direct;
2265 if (!options->has_cache && !direct) {
2266 error_setg(errp, "aio=native requires cache.direct=true");
2271 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2272 &options, NULL, &local_err);
2273 if (error_is_set(&local_err)) {
2274 error_propagate(errp, local_err);
2278 obj = qmp_output_get_qobject(ov);
2279 qdict = qobject_to_qdict(obj);
2281 qdict_flatten(qdict);
2283 blockdev_init(NULL, qdict, IF_NONE, &local_err);
2284 if (error_is_set(&local_err)) {
2285 error_propagate(errp, local_err);
2290 qmp_output_visitor_cleanup(ov);
2293 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2295 BlockJobInfoList **prev = opaque;
2296 BlockJob *job = bs->job;
2299 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2300 elem->value = block_job_query(bs->job);
2301 (*prev)->next = elem;
2306 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2308 /* Dummy is a fake list element for holding the head pointer */
2309 BlockJobInfoList dummy = {};
2310 BlockJobInfoList *prev = &dummy;
2311 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2315 QemuOptsList qemu_common_drive_opts = {
2317 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2321 .type = QEMU_OPT_BOOL,
2322 .help = "enable/disable snapshot mode",
2325 .type = QEMU_OPT_STRING,
2326 .help = "discard operation (ignore/off, unmap/on)",
2328 .name = "cache.writeback",
2329 .type = QEMU_OPT_BOOL,
2330 .help = "enables writeback mode for any caches",
2332 .name = "cache.direct",
2333 .type = QEMU_OPT_BOOL,
2334 .help = "enables use of O_DIRECT (bypass the host page cache)",
2336 .name = "cache.no-flush",
2337 .type = QEMU_OPT_BOOL,
2338 .help = "ignore any flush requests for the device",
2341 .type = QEMU_OPT_STRING,
2342 .help = "host AIO implementation (threads, native)",
2345 .type = QEMU_OPT_STRING,
2346 .help = "disk format (raw, qcow2, ...)",
2349 .type = QEMU_OPT_STRING,
2350 .help = "disk serial number",
2353 .type = QEMU_OPT_STRING,
2354 .help = "read error action",
2357 .type = QEMU_OPT_STRING,
2358 .help = "write error action",
2360 .name = "read-only",
2361 .type = QEMU_OPT_BOOL,
2362 .help = "open drive file as read-only",
2364 .name = "throttling.iops-total",
2365 .type = QEMU_OPT_NUMBER,
2366 .help = "limit total I/O operations per second",
2368 .name = "throttling.iops-read",
2369 .type = QEMU_OPT_NUMBER,
2370 .help = "limit read operations per second",
2372 .name = "throttling.iops-write",
2373 .type = QEMU_OPT_NUMBER,
2374 .help = "limit write operations per second",
2376 .name = "throttling.bps-total",
2377 .type = QEMU_OPT_NUMBER,
2378 .help = "limit total bytes per second",
2380 .name = "throttling.bps-read",
2381 .type = QEMU_OPT_NUMBER,
2382 .help = "limit read bytes per second",
2384 .name = "throttling.bps-write",
2385 .type = QEMU_OPT_NUMBER,
2386 .help = "limit write bytes per second",
2388 .name = "throttling.iops-total-max",
2389 .type = QEMU_OPT_NUMBER,
2390 .help = "I/O operations burst",
2392 .name = "throttling.iops-read-max",
2393 .type = QEMU_OPT_NUMBER,
2394 .help = "I/O operations read burst",
2396 .name = "throttling.iops-write-max",
2397 .type = QEMU_OPT_NUMBER,
2398 .help = "I/O operations write burst",
2400 .name = "throttling.bps-total-max",
2401 .type = QEMU_OPT_NUMBER,
2402 .help = "total bytes burst",
2404 .name = "throttling.bps-read-max",
2405 .type = QEMU_OPT_NUMBER,
2406 .help = "total bytes read burst",
2408 .name = "throttling.bps-write-max",
2409 .type = QEMU_OPT_NUMBER,
2410 .help = "total bytes write burst",
2412 .name = "throttling.iops-size",
2413 .type = QEMU_OPT_NUMBER,
2414 .help = "when limiting by iops max size of an I/O in bytes",
2416 .name = "copy-on-read",
2417 .type = QEMU_OPT_BOOL,
2418 .help = "copy read data from backing file into image file",
2420 { /* end of list */ }
2424 QemuOptsList qemu_drive_opts = {
2426 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2429 * no elements => accept any params
2430 * validation will happen later
2432 { /* end of list */ }