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,
317 int on_read_error, on_write_error;
326 bool has_driver_specific_opts;
327 BlockDriver *drv = NULL;
329 /* Check common options by copying from bs_opts to opts, all other options
330 * stay in bs_opts for processing by bdrv_open(). */
331 id = qdict_get_try_str(bs_opts, "id");
332 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
334 error_propagate(errp, error);
338 qemu_opts_absorb_qdict(opts, bs_opts, &error);
340 error_propagate(errp, error);
345 qdict_del(bs_opts, "id");
348 has_driver_specific_opts = !!qdict_size(bs_opts);
350 /* extract parameters */
351 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
352 ro = qemu_opt_get_bool(opts, "read-only", 0);
353 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
355 serial = qemu_opt_get(opts, "serial");
357 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
358 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
359 error_setg(errp, "invalid discard option");
364 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
365 bdrv_flags |= BDRV_O_CACHE_WB;
367 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
368 bdrv_flags |= BDRV_O_NOCACHE;
370 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
371 bdrv_flags |= BDRV_O_NO_FLUSH;
374 #ifdef CONFIG_LINUX_AIO
375 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
376 if (!strcmp(buf, "native")) {
377 bdrv_flags |= BDRV_O_NATIVE_AIO;
378 } else if (!strcmp(buf, "threads")) {
379 /* this is the default */
381 error_setg(errp, "invalid aio option");
387 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
388 if (is_help_option(buf)) {
389 error_printf("Supported formats:");
390 bdrv_iterate_format(bdrv_format_print, NULL);
395 drv = bdrv_find_format(buf);
397 error_setg(errp, "'%s' invalid format", buf);
402 /* disk I/O throttling */
403 memset(&cfg, 0, sizeof(cfg));
404 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
405 qemu_opt_get_number(opts, "throttling.bps-total", 0);
406 cfg.buckets[THROTTLE_BPS_READ].avg =
407 qemu_opt_get_number(opts, "throttling.bps-read", 0);
408 cfg.buckets[THROTTLE_BPS_WRITE].avg =
409 qemu_opt_get_number(opts, "throttling.bps-write", 0);
410 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
411 qemu_opt_get_number(opts, "throttling.iops-total", 0);
412 cfg.buckets[THROTTLE_OPS_READ].avg =
413 qemu_opt_get_number(opts, "throttling.iops-read", 0);
414 cfg.buckets[THROTTLE_OPS_WRITE].avg =
415 qemu_opt_get_number(opts, "throttling.iops-write", 0);
417 cfg.buckets[THROTTLE_BPS_TOTAL].max =
418 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
419 cfg.buckets[THROTTLE_BPS_READ].max =
420 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
421 cfg.buckets[THROTTLE_BPS_WRITE].max =
422 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
423 cfg.buckets[THROTTLE_OPS_TOTAL].max =
424 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
425 cfg.buckets[THROTTLE_OPS_READ].max =
426 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
427 cfg.buckets[THROTTLE_OPS_WRITE].max =
428 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
430 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
432 if (!check_throttle_config(&cfg, &error)) {
433 error_propagate(errp, error);
437 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
438 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
439 on_write_error = parse_block_error_action(buf, 0, &error);
441 error_propagate(errp, error);
446 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
447 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
448 on_read_error = parse_block_error_action(buf, 1, &error);
450 error_propagate(errp, error);
455 if (bdrv_find_node(qemu_opts_id(opts))) {
456 error_setg(errp, "device id=%s is conflicting with a node-name",
462 dinfo = g_malloc0(sizeof(*dinfo));
463 dinfo->id = g_strdup(qemu_opts_id(opts));
464 dinfo->bdrv = bdrv_new(dinfo->id);
465 dinfo->bdrv->open_flags = snapshot ? BDRV_O_SNAPSHOT : 0;
466 dinfo->bdrv->read_only = ro;
468 if (serial != NULL) {
469 dinfo->serial = g_strdup(serial);
471 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
473 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
475 /* disk I/O throttling */
476 if (throttle_enabled(&cfg)) {
477 bdrv_io_limits_enable(dinfo->bdrv);
478 bdrv_set_io_limits(dinfo->bdrv, &cfg);
481 if (!file || !*file) {
482 if (has_driver_specific_opts) {
491 /* always use cache=unsafe with snapshot */
492 bdrv_flags &= ~BDRV_O_CACHE_MASK;
493 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
497 bdrv_flags |= BDRV_O_COPY_ON_READ;
500 if (runstate_check(RUN_STATE_INMIGRATE)) {
501 bdrv_flags |= BDRV_O_INCOMING;
504 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
507 ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
510 error_setg(errp, "could not open disk image %s: %s",
511 file ?: dinfo->id, error_get_pretty(error));
516 if (bdrv_key_required(dinfo->bdrv))
525 bdrv_unref(dinfo->bdrv);
527 QTAILQ_REMOVE(&drives, dinfo, next);
535 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
539 value = qemu_opt_get(opts, from);
541 qemu_opt_set(opts, to, value);
542 qemu_opt_unset(opts, from);
546 QemuOptsList qemu_legacy_drive_opts = {
548 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
552 .type = QEMU_OPT_NUMBER,
553 .help = "bus number",
556 .type = QEMU_OPT_NUMBER,
557 .help = "unit number (i.e. lun for scsi)",
560 .type = QEMU_OPT_NUMBER,
561 .help = "index number",
564 .type = QEMU_OPT_STRING,
565 .help = "media type (disk, cdrom)",
568 .type = QEMU_OPT_STRING,
569 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
572 .type = QEMU_OPT_NUMBER,
573 .help = "number of cylinders (ide disk geometry)",
576 .type = QEMU_OPT_NUMBER,
577 .help = "number of heads (ide disk geometry)",
580 .type = QEMU_OPT_NUMBER,
581 .help = "number of sectors (ide disk geometry)",
584 .type = QEMU_OPT_STRING,
585 .help = "chs translation (auto, lba, none)",
588 .type = QEMU_OPT_BOOL,
589 .help = "(deprecated, ignored)",
592 .type = QEMU_OPT_STRING,
593 .help = "pci address (virtio only)",
596 .type = QEMU_OPT_STRING,
600 /* Options that are passed on, but have special semantics with -drive */
603 .type = QEMU_OPT_BOOL,
604 .help = "open drive file as read-only",
607 .type = QEMU_OPT_STRING,
608 .help = "read error action",
611 .type = QEMU_OPT_STRING,
612 .help = "write error action",
614 .name = "copy-on-read",
615 .type = QEMU_OPT_BOOL,
616 .help = "copy read data from backing file into image file",
619 { /* end of list */ }
623 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
626 DriveInfo *dinfo = NULL;
628 QemuOpts *legacy_opts;
629 DriveMediaType media = MEDIA_DISK;
630 BlockInterfaceType type;
631 int cyls, heads, secs, translation;
632 int max_devs, bus_id, unit_id, index;
634 const char *werror, *rerror;
635 bool read_only = false;
637 const char *filename;
638 Error *local_err = NULL;
640 /* Change legacy command line options into QMP ones */
641 qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
642 qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
643 qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
645 qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
646 qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
647 qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
649 qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
650 qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
651 qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
653 qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
654 qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
655 qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
657 qemu_opt_rename(all_opts,
658 "iops_size", "throttling.iops-size");
660 qemu_opt_rename(all_opts, "readonly", "read-only");
662 value = qemu_opt_get(all_opts, "cache");
666 if (bdrv_parse_cache_flags(value, &flags) != 0) {
667 error_report("invalid cache option");
671 /* Specific options take precedence */
672 if (!qemu_opt_get(all_opts, "cache.writeback")) {
673 qemu_opt_set_bool(all_opts, "cache.writeback",
674 !!(flags & BDRV_O_CACHE_WB));
676 if (!qemu_opt_get(all_opts, "cache.direct")) {
677 qemu_opt_set_bool(all_opts, "cache.direct",
678 !!(flags & BDRV_O_NOCACHE));
680 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
681 qemu_opt_set_bool(all_opts, "cache.no-flush",
682 !!(flags & BDRV_O_NO_FLUSH));
684 qemu_opt_unset(all_opts, "cache");
687 /* Get a QDict for processing the options */
688 bs_opts = qdict_new();
689 qemu_opts_to_qdict(all_opts, bs_opts);
691 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
693 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
695 qerror_report_err(local_err);
696 error_free(local_err);
700 /* Deprecated option boot=[on|off] */
701 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
702 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
703 "ignored. Future versions will reject this parameter. Please "
704 "update your scripts.\n");
708 value = qemu_opt_get(legacy_opts, "media");
710 if (!strcmp(value, "disk")) {
712 } else if (!strcmp(value, "cdrom")) {
716 error_report("'%s' invalid media", value);
721 /* copy-on-read is disabled with a warning for read-only devices */
722 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
723 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
725 if (read_only && copy_on_read) {
726 error_report("warning: disabling copy-on-read on read-only drive");
727 copy_on_read = false;
730 qdict_put(bs_opts, "read-only",
731 qstring_from_str(read_only ? "on" : "off"));
732 qdict_put(bs_opts, "copy-on-read",
733 qstring_from_str(copy_on_read ? "on" :"off"));
735 /* Controller type */
736 value = qemu_opt_get(legacy_opts, "if");
739 type < IF_COUNT && strcmp(value, if_name[type]);
742 if (type == IF_COUNT) {
743 error_report("unsupported bus type '%s'", value);
747 type = block_default_type;
751 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
752 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
753 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
755 if (cyls || heads || secs) {
757 error_report("invalid physical cyls number");
761 error_report("invalid physical heads number");
765 error_report("invalid physical secs number");
770 translation = BIOS_ATA_TRANSLATION_AUTO;
771 value = qemu_opt_get(legacy_opts, "trans");
774 error_report("'%s' trans must be used with cyls, heads and secs",
778 if (!strcmp(value, "none")) {
779 translation = BIOS_ATA_TRANSLATION_NONE;
780 } else if (!strcmp(value, "lba")) {
781 translation = BIOS_ATA_TRANSLATION_LBA;
782 } else if (!strcmp(value, "large")) {
783 translation = BIOS_ATA_TRANSLATION_LARGE;
784 } else if (!strcmp(value, "rechs")) {
785 translation = BIOS_ATA_TRANSLATION_RECHS;
786 } else if (!strcmp(value, "auto")) {
787 translation = BIOS_ATA_TRANSLATION_AUTO;
789 error_report("'%s' invalid translation type", value);
794 if (media == MEDIA_CDROM) {
795 if (cyls || secs || heads) {
796 error_report("CHS can't be set with media=cdrom");
801 /* Device address specified by bus/unit or index.
802 * If none was specified, try to find the first free one. */
803 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
804 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
805 index = qemu_opt_get_number(legacy_opts, "index", -1);
807 max_devs = if_max_devs[type];
810 if (bus_id != 0 || unit_id != -1) {
811 error_report("index cannot be used with bus and unit");
814 bus_id = drive_index_to_bus_id(type, index);
815 unit_id = drive_index_to_unit_id(type, index);
820 while (drive_get(type, bus_id, unit_id) != NULL) {
822 if (max_devs && unit_id >= max_devs) {
829 if (max_devs && unit_id >= max_devs) {
830 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
834 if (drive_get(type, bus_id, unit_id) != NULL) {
835 error_report("drive with bus=%d, unit=%d (index=%d) exists",
836 bus_id, unit_id, index);
840 /* no id supplied -> create one */
841 if (qemu_opts_id(all_opts) == NULL) {
843 const char *mediastr = "";
844 if (type == IF_IDE || type == IF_SCSI) {
845 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
848 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
851 new_id = g_strdup_printf("%s%s%i", if_name[type],
854 qdict_put(bs_opts, "id", qstring_from_str(new_id));
858 /* Add virtio block device */
859 devaddr = qemu_opt_get(legacy_opts, "addr");
860 if (devaddr && type != IF_VIRTIO) {
861 error_report("addr is not supported by this bus type");
865 if (type == IF_VIRTIO) {
867 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
869 if (arch_type == QEMU_ARCH_S390X) {
870 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
872 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
874 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
876 qemu_opt_set(devopts, "addr", devaddr);
880 filename = qemu_opt_get(legacy_opts, "file");
882 /* Check werror/rerror compatibility with if=... */
883 werror = qemu_opt_get(legacy_opts, "werror");
884 if (werror != NULL) {
885 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
887 error_report("werror is not supported by this bus type");
890 qdict_put(bs_opts, "werror", qstring_from_str(werror));
893 rerror = qemu_opt_get(legacy_opts, "rerror");
894 if (rerror != NULL) {
895 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
897 error_report("rerror is not supported by this bus type");
900 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
903 /* Actual block device init: Functionality shared with blockdev-add */
904 dinfo = blockdev_init(filename, bs_opts, &local_err);
907 qerror_report_err(local_err);
908 error_free(local_err);
915 /* Set legacy DriveInfo fields */
916 dinfo->enable_auto_del = true;
917 dinfo->opts = all_opts;
920 dinfo->heads = heads;
922 dinfo->trans = translation;
926 dinfo->unit = unit_id;
927 dinfo->devaddr = devaddr;
934 dinfo->media_cd = media == MEDIA_CDROM;
941 qemu_opts_del(legacy_opts);
945 void do_commit(Monitor *mon, const QDict *qdict)
947 const char *device = qdict_get_str(qdict, "device");
948 BlockDriverState *bs;
951 if (!strcmp(device, "all")) {
952 ret = bdrv_commit_all();
954 bs = bdrv_find(device);
956 monitor_printf(mon, "Device '%s' not found\n", device);
959 ret = bdrv_commit(bs);
962 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
967 static void blockdev_do_action(int kind, void *data, Error **errp)
969 TransactionAction action;
970 TransactionActionList list;
974 list.value = &action;
976 qmp_transaction(&list, errp);
979 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
980 bool has_node_name, const char *node_name,
981 const char *snapshot_file,
982 bool has_snapshot_node_name,
983 const char *snapshot_node_name,
984 bool has_format, const char *format,
985 bool has_mode, NewImageMode mode, Error **errp)
987 BlockdevSnapshot snapshot = {
988 .has_device = has_device,
989 .device = (char *) device,
990 .has_node_name = has_node_name,
991 .node_name = (char *) node_name,
992 .snapshot_file = (char *) snapshot_file,
993 .has_snapshot_node_name = has_snapshot_node_name,
994 .snapshot_node_name = (char *) snapshot_node_name,
995 .has_format = has_format,
996 .format = (char *) format,
997 .has_mode = has_mode,
1000 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1004 void qmp_blockdev_snapshot_internal_sync(const char *device,
1008 BlockdevSnapshotInternal snapshot = {
1009 .device = (char *) device,
1010 .name = (char *) name
1013 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1017 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1024 BlockDriverState *bs = bdrv_find(device);
1025 QEMUSnapshotInfo sn;
1026 Error *local_err = NULL;
1027 SnapshotInfo *info = NULL;
1031 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1044 error_setg(errp, "Name or id must be provided");
1048 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1050 error_propagate(errp, local_err);
1055 "Snapshot with id '%s' and name '%s' does not exist on "
1057 STR_OR_NULL(id), STR_OR_NULL(name), device);
1061 bdrv_snapshot_delete(bs, id, name, &local_err);
1063 error_propagate(errp, local_err);
1067 info = g_malloc0(sizeof(SnapshotInfo));
1068 info->id = g_strdup(sn.id_str);
1069 info->name = g_strdup(sn.name);
1070 info->date_nsec = sn.date_nsec;
1071 info->date_sec = sn.date_sec;
1072 info->vm_state_size = sn.vm_state_size;
1073 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1074 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1079 /* New and old BlockDriverState structs for group snapshots */
1081 typedef struct BlkTransactionState BlkTransactionState;
1083 /* Only prepare() may fail. In a single transaction, only one of commit() or
1084 abort() will be called, clean() will always be called if it present. */
1085 typedef struct BdrvActionOps {
1086 /* Size of state struct, in bytes. */
1087 size_t instance_size;
1088 /* Prepare the work, must NOT be NULL. */
1089 void (*prepare)(BlkTransactionState *common, Error **errp);
1090 /* Commit the changes, can be NULL. */
1091 void (*commit)(BlkTransactionState *common);
1092 /* Abort the changes on fail, can be NULL. */
1093 void (*abort)(BlkTransactionState *common);
1094 /* Clean up resource in the end, can be NULL. */
1095 void (*clean)(BlkTransactionState *common);
1099 * This structure must be arranged as first member in child type, assuming
1100 * that compiler will also arrange it to the same address with parent instance.
1101 * Later it will be used in free().
1103 struct BlkTransactionState {
1104 TransactionAction *action;
1105 const BdrvActionOps *ops;
1106 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1109 /* internal snapshot private data */
1110 typedef struct InternalSnapshotState {
1111 BlkTransactionState common;
1112 BlockDriverState *bs;
1113 QEMUSnapshotInfo sn;
1114 } InternalSnapshotState;
1116 static void internal_snapshot_prepare(BlkTransactionState *common,
1121 BlockDriverState *bs;
1122 QEMUSnapshotInfo old_sn, *sn;
1125 BlockdevSnapshotInternal *internal;
1126 InternalSnapshotState *state;
1129 g_assert(common->action->kind ==
1130 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1131 internal = common->action->blockdev_snapshot_internal_sync;
1132 state = DO_UPCAST(InternalSnapshotState, common, common);
1134 /* 1. parse input */
1135 device = internal->device;
1136 name = internal->name;
1138 /* 2. check for validation */
1139 bs = bdrv_find(device);
1141 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1145 if (!bdrv_is_inserted(bs)) {
1146 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1150 if (bdrv_is_read_only(bs)) {
1151 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1155 if (!bdrv_can_snapshot(bs)) {
1156 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1157 bs->drv->format_name, device, "internal snapshot");
1161 if (!strlen(name)) {
1162 error_setg(errp, "Name is empty");
1166 /* check whether a snapshot with name exist */
1167 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn, errp);
1168 if (error_is_set(errp)) {
1172 "Snapshot with name '%s' already exists on device '%s'",
1177 /* 3. take the snapshot */
1179 pstrcpy(sn->name, sizeof(sn->name), name);
1180 qemu_gettimeofday(&tv);
1181 sn->date_sec = tv.tv_sec;
1182 sn->date_nsec = tv.tv_usec * 1000;
1183 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1185 ret1 = bdrv_snapshot_create(bs, sn);
1187 error_setg_errno(errp, -ret1,
1188 "Failed to create snapshot '%s' on device '%s'",
1193 /* 4. succeed, mark a snapshot is created */
1197 static void internal_snapshot_abort(BlkTransactionState *common)
1199 InternalSnapshotState *state =
1200 DO_UPCAST(InternalSnapshotState, common, common);
1201 BlockDriverState *bs = state->bs;
1202 QEMUSnapshotInfo *sn = &state->sn;
1203 Error *local_error = NULL;
1209 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1210 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1211 "device '%s' in abort: %s",
1214 bdrv_get_device_name(bs),
1215 error_get_pretty(local_error));
1216 error_free(local_error);
1220 /* external snapshot private data */
1221 typedef struct ExternalSnapshotState {
1222 BlkTransactionState common;
1223 BlockDriverState *old_bs;
1224 BlockDriverState *new_bs;
1225 } ExternalSnapshotState;
1227 static void external_snapshot_prepare(BlkTransactionState *common,
1232 QDict *options = NULL;
1233 Error *local_err = NULL;
1234 bool has_device = false;
1236 bool has_node_name = false;
1237 const char *node_name;
1238 bool has_snapshot_node_name = false;
1239 const char *snapshot_node_name;
1240 const char *new_image_file;
1241 const char *format = "qcow2";
1242 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1243 ExternalSnapshotState *state =
1244 DO_UPCAST(ExternalSnapshotState, common, common);
1245 TransactionAction *action = common->action;
1247 /* get parameters */
1248 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1250 has_device = action->blockdev_snapshot_sync->has_device;
1251 device = action->blockdev_snapshot_sync->device;
1252 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1253 node_name = action->blockdev_snapshot_sync->node_name;
1254 has_snapshot_node_name =
1255 action->blockdev_snapshot_sync->has_snapshot_node_name;
1256 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1258 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1259 if (action->blockdev_snapshot_sync->has_format) {
1260 format = action->blockdev_snapshot_sync->format;
1262 if (action->blockdev_snapshot_sync->has_mode) {
1263 mode = action->blockdev_snapshot_sync->mode;
1266 /* start processing */
1267 drv = bdrv_find_format(format);
1269 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1273 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1274 has_node_name ? node_name : NULL,
1277 error_propagate(errp, local_err);
1281 if (has_node_name && !has_snapshot_node_name) {
1282 error_setg(errp, "New snapshot node name missing");
1286 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1287 error_setg(errp, "New snapshot node name already existing");
1291 if (!bdrv_is_inserted(state->old_bs)) {
1292 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1296 if (bdrv_in_use(state->old_bs)) {
1297 error_set(errp, QERR_DEVICE_IN_USE, device);
1301 if (!bdrv_is_read_only(state->old_bs)) {
1302 if (bdrv_flush(state->old_bs)) {
1303 error_set(errp, QERR_IO_ERROR);
1308 if (!bdrv_is_first_non_filter(state->old_bs)) {
1309 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1313 flags = state->old_bs->open_flags;
1315 /* create new image w/backing file */
1316 if (mode != NEW_IMAGE_MODE_EXISTING) {
1317 bdrv_img_create(new_image_file, format,
1318 state->old_bs->filename,
1319 state->old_bs->drv->format_name,
1320 NULL, -1, flags, &local_err, false);
1322 error_propagate(errp, local_err);
1327 if (has_snapshot_node_name) {
1328 options = qdict_new();
1329 qdict_put(options, "node-name",
1330 qstring_from_str(snapshot_node_name));
1333 /* TODO Inherit bs->options or only take explicit options with an
1334 * extended QMP command? */
1335 assert(state->new_bs == NULL);
1336 ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1337 flags | BDRV_O_NO_BACKING, drv, &local_err);
1338 /* We will manually add the backing_hd field to the bs later */
1340 error_propagate(errp, local_err);
1344 static void external_snapshot_commit(BlkTransactionState *common)
1346 ExternalSnapshotState *state =
1347 DO_UPCAST(ExternalSnapshotState, common, common);
1349 /* This removes our old bs and adds the new bs */
1350 bdrv_append(state->new_bs, state->old_bs);
1351 /* We don't need (or want) to use the transactional
1352 * bdrv_reopen_multiple() across all the entries at once, because we
1353 * don't want to abort all of them if one of them fails the reopen */
1354 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1358 static void external_snapshot_abort(BlkTransactionState *common)
1360 ExternalSnapshotState *state =
1361 DO_UPCAST(ExternalSnapshotState, common, common);
1362 if (state->new_bs) {
1363 bdrv_unref(state->new_bs);
1367 typedef struct DriveBackupState {
1368 BlkTransactionState common;
1369 BlockDriverState *bs;
1373 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1375 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1376 DriveBackup *backup;
1377 Error *local_err = NULL;
1379 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1380 backup = common->action->drive_backup;
1382 qmp_drive_backup(backup->device, backup->target,
1383 backup->has_format, backup->format,
1385 backup->has_mode, backup->mode,
1386 backup->has_speed, backup->speed,
1387 backup->has_on_source_error, backup->on_source_error,
1388 backup->has_on_target_error, backup->on_target_error,
1391 error_propagate(errp, local_err);
1397 state->bs = bdrv_find(backup->device);
1398 state->job = state->bs->job;
1401 static void drive_backup_abort(BlkTransactionState *common)
1403 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1404 BlockDriverState *bs = state->bs;
1406 /* Only cancel if it's the job we started */
1407 if (bs && bs->job && bs->job == state->job) {
1408 block_job_cancel_sync(bs->job);
1412 static void abort_prepare(BlkTransactionState *common, Error **errp)
1414 error_setg(errp, "Transaction aborted using Abort action");
1417 static void abort_commit(BlkTransactionState *common)
1419 g_assert_not_reached(); /* this action never succeeds */
1422 static const BdrvActionOps actions[] = {
1423 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1424 .instance_size = sizeof(ExternalSnapshotState),
1425 .prepare = external_snapshot_prepare,
1426 .commit = external_snapshot_commit,
1427 .abort = external_snapshot_abort,
1429 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1430 .instance_size = sizeof(DriveBackupState),
1431 .prepare = drive_backup_prepare,
1432 .abort = drive_backup_abort,
1434 [TRANSACTION_ACTION_KIND_ABORT] = {
1435 .instance_size = sizeof(BlkTransactionState),
1436 .prepare = abort_prepare,
1437 .commit = abort_commit,
1439 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1440 .instance_size = sizeof(InternalSnapshotState),
1441 .prepare = internal_snapshot_prepare,
1442 .abort = internal_snapshot_abort,
1447 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1448 * then we do not pivot any of the devices in the group, and abandon the
1451 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1453 TransactionActionList *dev_entry = dev_list;
1454 BlkTransactionState *state, *next;
1455 Error *local_err = NULL;
1457 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1458 QSIMPLEQ_INIT(&snap_bdrv_states);
1460 /* drain all i/o before any snapshots */
1463 /* We don't do anything in this loop that commits us to the snapshot */
1464 while (NULL != dev_entry) {
1465 TransactionAction *dev_info = NULL;
1466 const BdrvActionOps *ops;
1468 dev_info = dev_entry->value;
1469 dev_entry = dev_entry->next;
1471 assert(dev_info->kind < ARRAY_SIZE(actions));
1473 ops = &actions[dev_info->kind];
1474 assert(ops->instance_size > 0);
1476 state = g_malloc0(ops->instance_size);
1478 state->action = dev_info;
1479 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1481 state->ops->prepare(state, &local_err);
1483 error_propagate(errp, local_err);
1484 goto delete_and_fail;
1488 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1489 if (state->ops->commit) {
1490 state->ops->commit(state);
1499 * failure, and it is all-or-none; abandon each new bs, and keep using
1500 * the original bs for all images
1502 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1503 if (state->ops->abort) {
1504 state->ops->abort(state);
1508 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1509 if (state->ops->clean) {
1510 state->ops->clean(state);
1517 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1519 if (bdrv_in_use(bs)) {
1520 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1523 if (!bdrv_dev_has_removable_media(bs)) {
1524 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1528 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1529 bdrv_dev_eject_request(bs, force);
1531 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1539 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1541 BlockDriverState *bs;
1543 bs = bdrv_find(device);
1545 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1549 eject_device(bs, force, errp);
1552 void qmp_block_passwd(bool has_device, const char *device,
1553 bool has_node_name, const char *node_name,
1554 const char *password, Error **errp)
1556 Error *local_err = NULL;
1557 BlockDriverState *bs;
1560 bs = bdrv_lookup_bs(has_device ? device : NULL,
1561 has_node_name ? node_name : NULL,
1564 error_propagate(errp, local_err);
1568 err = bdrv_set_key(bs, password);
1569 if (err == -EINVAL) {
1570 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1572 } else if (err < 0) {
1573 error_set(errp, QERR_INVALID_PASSWORD);
1578 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1579 int bdrv_flags, BlockDriver *drv,
1580 const char *password, Error **errp)
1582 Error *local_err = NULL;
1585 ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1587 error_propagate(errp, local_err);
1591 if (bdrv_key_required(bs)) {
1593 if (bdrv_set_key(bs, password) < 0) {
1594 error_set(errp, QERR_INVALID_PASSWORD);
1597 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1598 bdrv_get_encrypted_filename(bs));
1600 } else if (password) {
1601 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1605 void qmp_change_blockdev(const char *device, const char *filename,
1606 const char *format, Error **errp)
1608 BlockDriverState *bs;
1609 BlockDriver *drv = NULL;
1613 bs = bdrv_find(device);
1615 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1620 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1622 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1627 eject_device(bs, 0, &err);
1629 error_propagate(errp, err);
1633 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1634 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1636 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1639 /* throttling disk I/O limits */
1640 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1647 bool has_bps_rd_max,
1649 bool has_bps_wr_max,
1653 bool has_iops_rd_max,
1654 int64_t iops_rd_max,
1655 bool has_iops_wr_max,
1656 int64_t iops_wr_max,
1658 int64_t iops_size, Error **errp)
1661 BlockDriverState *bs;
1663 bs = bdrv_find(device);
1665 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1669 memset(&cfg, 0, sizeof(cfg));
1670 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1671 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1672 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1674 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1675 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1676 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1679 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1681 if (has_bps_rd_max) {
1682 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1684 if (has_bps_wr_max) {
1685 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1688 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1690 if (has_iops_rd_max) {
1691 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1693 if (has_iops_wr_max) {
1694 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1697 if (has_iops_size) {
1698 cfg.op_size = iops_size;
1701 if (!check_throttle_config(&cfg, errp)) {
1705 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1706 bdrv_io_limits_enable(bs);
1707 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1708 bdrv_io_limits_disable(bs);
1711 if (bs->io_limits_enabled) {
1712 bdrv_set_io_limits(bs, &cfg);
1716 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1718 const char *id = qdict_get_str(qdict, "id");
1719 BlockDriverState *bs;
1723 qerror_report(QERR_DEVICE_NOT_FOUND, id);
1726 if (bdrv_in_use(bs)) {
1727 qerror_report(QERR_DEVICE_IN_USE, id);
1731 /* quiesce block driver; prevent further io */
1736 /* if we have a device attached to this BlockDriverState
1737 * then we need to make the drive anonymous until the device
1738 * can be removed. If this is a drive with no device backing
1739 * then we can just get rid of the block driver state right here.
1741 if (bdrv_get_attached_dev(bs)) {
1744 /* Further I/O must not pause the guest */
1745 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1746 BLOCKDEV_ON_ERROR_REPORT);
1748 drive_uninit(drive_get_by_blockdev(bs));
1754 void qmp_block_resize(bool has_device, const char *device,
1755 bool has_node_name, const char *node_name,
1756 int64_t size, Error **errp)
1758 Error *local_err = NULL;
1759 BlockDriverState *bs;
1762 bs = bdrv_lookup_bs(has_device ? device : NULL,
1763 has_node_name ? node_name : NULL,
1766 error_propagate(errp, local_err);
1770 if (!bdrv_is_first_non_filter(bs)) {
1771 error_set(errp, QERR_FEATURE_DISABLED, "resize");
1776 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1780 /* complete all in-flight operations before resizing the device */
1783 ret = bdrv_truncate(bs, size);
1788 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1791 error_set(errp, QERR_UNSUPPORTED);
1794 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1797 error_set(errp, QERR_DEVICE_IN_USE, device);
1800 error_setg_errno(errp, -ret, "Could not resize");
1805 static void block_job_cb(void *opaque, int ret)
1807 BlockDriverState *bs = opaque;
1810 trace_block_job_cb(bs, bs->job, ret);
1813 obj = qobject_from_block_job(bs->job);
1815 QDict *dict = qobject_to_qdict(obj);
1816 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1819 if (block_job_is_cancelled(bs->job)) {
1820 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1822 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1824 qobject_decref(obj);
1826 bdrv_put_ref_bh_schedule(bs);
1829 void qmp_block_stream(const char *device, bool has_base,
1830 const char *base, bool has_speed, int64_t speed,
1831 bool has_on_error, BlockdevOnError on_error,
1834 BlockDriverState *bs;
1835 BlockDriverState *base_bs = NULL;
1836 Error *local_err = NULL;
1838 if (!has_on_error) {
1839 on_error = BLOCKDEV_ON_ERROR_REPORT;
1842 bs = bdrv_find(device);
1844 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1849 base_bs = bdrv_find_backing_image(bs, base);
1850 if (base_bs == NULL) {
1851 error_set(errp, QERR_BASE_NOT_FOUND, base);
1856 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1857 on_error, block_job_cb, bs, &local_err);
1859 error_propagate(errp, local_err);
1863 trace_qmp_block_stream(bs, bs->job);
1866 void qmp_block_commit(const char *device,
1867 bool has_base, const char *base, const char *top,
1868 bool has_speed, int64_t speed,
1871 BlockDriverState *bs;
1872 BlockDriverState *base_bs, *top_bs;
1873 Error *local_err = NULL;
1874 /* This will be part of the QMP command, if/when the
1875 * BlockdevOnError change for blkmirror makes it in
1877 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1883 /* drain all i/o before commits */
1886 bs = bdrv_find(device);
1888 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1892 /* default top_bs is the active layer */
1896 if (strcmp(bs->filename, top) != 0) {
1897 top_bs = bdrv_find_backing_image(bs, top);
1901 if (top_bs == NULL) {
1902 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1906 if (has_base && base) {
1907 base_bs = bdrv_find_backing_image(top_bs, base);
1909 base_bs = bdrv_find_base(top_bs);
1912 if (base_bs == NULL) {
1913 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1918 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1921 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1924 if (local_err != NULL) {
1925 error_propagate(errp, local_err);
1930 void qmp_drive_backup(const char *device, const char *target,
1931 bool has_format, const char *format,
1932 enum MirrorSyncMode sync,
1933 bool has_mode, enum NewImageMode mode,
1934 bool has_speed, int64_t speed,
1935 bool has_on_source_error, BlockdevOnError on_source_error,
1936 bool has_on_target_error, BlockdevOnError on_target_error,
1939 BlockDriverState *bs;
1940 BlockDriverState *target_bs;
1941 BlockDriverState *source = NULL;
1942 BlockDriver *drv = NULL;
1943 Error *local_err = NULL;
1951 if (!has_on_source_error) {
1952 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1954 if (!has_on_target_error) {
1955 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1958 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1961 bs = bdrv_find(device);
1963 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1967 if (!bdrv_is_inserted(bs)) {
1968 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1973 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1976 drv = bdrv_find_format(format);
1978 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1983 if (bdrv_in_use(bs)) {
1984 error_set(errp, QERR_DEVICE_IN_USE, device);
1988 flags = bs->open_flags | BDRV_O_RDWR;
1990 /* See if we have a backing HD we can use to create our new image
1992 if (sync == MIRROR_SYNC_MODE_TOP) {
1993 source = bs->backing_hd;
1995 sync = MIRROR_SYNC_MODE_FULL;
1998 if (sync == MIRROR_SYNC_MODE_NONE) {
2002 size = bdrv_getlength(bs);
2004 error_setg_errno(errp, -size, "bdrv_getlength failed");
2008 if (mode != NEW_IMAGE_MODE_EXISTING) {
2009 assert(format && drv);
2011 bdrv_img_create(target, format, source->filename,
2012 source->drv->format_name, NULL,
2013 size, flags, &local_err, false);
2015 bdrv_img_create(target, format, NULL, NULL, NULL,
2016 size, flags, &local_err, false);
2021 error_propagate(errp, local_err);
2026 ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2028 error_propagate(errp, local_err);
2032 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2033 block_job_cb, bs, &local_err);
2034 if (local_err != NULL) {
2035 bdrv_unref(target_bs);
2036 error_propagate(errp, local_err);
2041 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2043 return bdrv_named_nodes_list();
2046 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2048 void qmp_drive_mirror(const char *device, const char *target,
2049 bool has_format, const char *format,
2050 enum MirrorSyncMode sync,
2051 bool has_mode, enum NewImageMode mode,
2052 bool has_speed, int64_t speed,
2053 bool has_granularity, uint32_t granularity,
2054 bool has_buf_size, int64_t buf_size,
2055 bool has_on_source_error, BlockdevOnError on_source_error,
2056 bool has_on_target_error, BlockdevOnError on_target_error,
2059 BlockDriverState *bs;
2060 BlockDriverState *source, *target_bs;
2061 BlockDriver *drv = NULL;
2062 Error *local_err = NULL;
2070 if (!has_on_source_error) {
2071 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2073 if (!has_on_target_error) {
2074 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2077 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2079 if (!has_granularity) {
2082 if (!has_buf_size) {
2083 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2086 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2087 error_set(errp, QERR_INVALID_PARAMETER, device);
2090 if (granularity & (granularity - 1)) {
2091 error_set(errp, QERR_INVALID_PARAMETER, device);
2095 bs = bdrv_find(device);
2097 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2101 if (!bdrv_is_inserted(bs)) {
2102 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2107 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2110 drv = bdrv_find_format(format);
2112 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2117 if (bdrv_in_use(bs)) {
2118 error_set(errp, QERR_DEVICE_IN_USE, device);
2122 flags = bs->open_flags | BDRV_O_RDWR;
2123 source = bs->backing_hd;
2124 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2125 sync = MIRROR_SYNC_MODE_FULL;
2127 if (sync == MIRROR_SYNC_MODE_NONE) {
2131 size = bdrv_getlength(bs);
2133 error_setg_errno(errp, -size, "bdrv_getlength failed");
2137 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2138 && mode != NEW_IMAGE_MODE_EXISTING)
2140 /* create new image w/o backing file */
2141 assert(format && drv);
2142 bdrv_img_create(target, format,
2143 NULL, NULL, NULL, size, flags, &local_err, false);
2146 case NEW_IMAGE_MODE_EXISTING:
2148 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2149 /* create new image with backing file */
2150 bdrv_img_create(target, format,
2152 source->drv->format_name,
2153 NULL, size, flags, &local_err, false);
2161 error_propagate(errp, local_err);
2165 /* Mirroring takes care of copy-on-write using the source's backing
2169 ret = bdrv_open(&target_bs, target, NULL, NULL, flags | BDRV_O_NO_BACKING,
2172 error_propagate(errp, local_err);
2176 mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2177 on_source_error, on_target_error,
2178 block_job_cb, bs, &local_err);
2179 if (local_err != NULL) {
2180 bdrv_unref(target_bs);
2181 error_propagate(errp, local_err);
2186 static BlockJob *find_block_job(const char *device)
2188 BlockDriverState *bs;
2190 bs = bdrv_find(device);
2191 if (!bs || !bs->job) {
2197 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2199 BlockJob *job = find_block_job(device);
2202 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2206 block_job_set_speed(job, speed, errp);
2209 void qmp_block_job_cancel(const char *device,
2210 bool has_force, bool force, Error **errp)
2212 BlockJob *job = find_block_job(device);
2219 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2222 if (job->paused && !force) {
2223 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
2227 trace_qmp_block_job_cancel(job);
2228 block_job_cancel(job);
2231 void qmp_block_job_pause(const char *device, Error **errp)
2233 BlockJob *job = find_block_job(device);
2236 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2240 trace_qmp_block_job_pause(job);
2241 block_job_pause(job);
2244 void qmp_block_job_resume(const char *device, Error **errp)
2246 BlockJob *job = find_block_job(device);
2249 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2253 trace_qmp_block_job_resume(job);
2254 block_job_resume(job);
2257 void qmp_block_job_complete(const char *device, Error **errp)
2259 BlockJob *job = find_block_job(device);
2262 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2266 trace_qmp_block_job_complete(job);
2267 block_job_complete(job, errp);
2270 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2272 QmpOutputVisitor *ov = qmp_output_visitor_new();
2276 Error *local_err = NULL;
2278 /* Require an ID in the top level */
2279 if (!options->has_id) {
2280 error_setg(errp, "Block device needs an ID");
2284 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2285 * cache.direct=false instead of silently switching to aio=threads, except
2286 * if called from drive_init.
2288 * For now, simply forbidding the combination for all drivers will do. */
2289 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2290 bool direct = options->has_cache &&
2291 options->cache->has_direct &&
2292 options->cache->direct;
2294 error_setg(errp, "aio=native requires cache.direct=true");
2299 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2300 &options, NULL, &local_err);
2302 error_propagate(errp, local_err);
2306 obj = qmp_output_get_qobject(ov);
2307 qdict = qobject_to_qdict(obj);
2309 qdict_flatten(qdict);
2311 dinfo = blockdev_init(NULL, qdict, &local_err);
2313 error_propagate(errp, local_err);
2317 if (bdrv_key_required(dinfo->bdrv)) {
2318 drive_uninit(dinfo);
2319 error_setg(errp, "blockdev-add doesn't support encrypted devices");
2324 qmp_output_visitor_cleanup(ov);
2327 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2329 BlockJobInfoList **prev = opaque;
2330 BlockJob *job = bs->job;
2333 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2334 elem->value = block_job_query(bs->job);
2335 (*prev)->next = elem;
2340 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2342 /* Dummy is a fake list element for holding the head pointer */
2343 BlockJobInfoList dummy = {};
2344 BlockJobInfoList *prev = &dummy;
2345 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2349 QemuOptsList qemu_common_drive_opts = {
2351 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2355 .type = QEMU_OPT_BOOL,
2356 .help = "enable/disable snapshot mode",
2359 .type = QEMU_OPT_STRING,
2360 .help = "discard operation (ignore/off, unmap/on)",
2362 .name = "cache.writeback",
2363 .type = QEMU_OPT_BOOL,
2364 .help = "enables writeback mode for any caches",
2366 .name = "cache.direct",
2367 .type = QEMU_OPT_BOOL,
2368 .help = "enables use of O_DIRECT (bypass the host page cache)",
2370 .name = "cache.no-flush",
2371 .type = QEMU_OPT_BOOL,
2372 .help = "ignore any flush requests for the device",
2375 .type = QEMU_OPT_STRING,
2376 .help = "host AIO implementation (threads, native)",
2379 .type = QEMU_OPT_STRING,
2380 .help = "disk format (raw, qcow2, ...)",
2383 .type = QEMU_OPT_STRING,
2384 .help = "disk serial number",
2387 .type = QEMU_OPT_STRING,
2388 .help = "read error action",
2391 .type = QEMU_OPT_STRING,
2392 .help = "write error action",
2394 .name = "read-only",
2395 .type = QEMU_OPT_BOOL,
2396 .help = "open drive file as read-only",
2398 .name = "throttling.iops-total",
2399 .type = QEMU_OPT_NUMBER,
2400 .help = "limit total I/O operations per second",
2402 .name = "throttling.iops-read",
2403 .type = QEMU_OPT_NUMBER,
2404 .help = "limit read operations per second",
2406 .name = "throttling.iops-write",
2407 .type = QEMU_OPT_NUMBER,
2408 .help = "limit write operations per second",
2410 .name = "throttling.bps-total",
2411 .type = QEMU_OPT_NUMBER,
2412 .help = "limit total bytes per second",
2414 .name = "throttling.bps-read",
2415 .type = QEMU_OPT_NUMBER,
2416 .help = "limit read bytes per second",
2418 .name = "throttling.bps-write",
2419 .type = QEMU_OPT_NUMBER,
2420 .help = "limit write bytes per second",
2422 .name = "throttling.iops-total-max",
2423 .type = QEMU_OPT_NUMBER,
2424 .help = "I/O operations burst",
2426 .name = "throttling.iops-read-max",
2427 .type = QEMU_OPT_NUMBER,
2428 .help = "I/O operations read burst",
2430 .name = "throttling.iops-write-max",
2431 .type = QEMU_OPT_NUMBER,
2432 .help = "I/O operations write burst",
2434 .name = "throttling.bps-total-max",
2435 .type = QEMU_OPT_NUMBER,
2436 .help = "total bytes burst",
2438 .name = "throttling.bps-read-max",
2439 .type = QEMU_OPT_NUMBER,
2440 .help = "total bytes read burst",
2442 .name = "throttling.bps-write-max",
2443 .type = QEMU_OPT_NUMBER,
2444 .help = "total bytes write burst",
2446 .name = "throttling.iops-size",
2447 .type = QEMU_OPT_NUMBER,
2448 .help = "when limiting by iops max size of an I/O in bytes",
2450 .name = "copy-on-read",
2451 .type = QEMU_OPT_BOOL,
2452 .help = "copy read data from backing file into image file",
2454 { /* end of list */ }
2458 QemuOptsList qemu_drive_opts = {
2460 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2463 * no elements => accept any params
2464 * validation will happen later
2466 { /* end of list */ }