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");
292 extern int start_simple_client(char* msg);
293 extern char* maru_convert_path(char* msg, const char *path);
296 static bool check_throttle_config(ThrottleConfig *cfg, Error **errp)
298 if (throttle_conflicting(cfg)) {
299 error_setg(errp, "bps/iops/max total values and read/write values"
300 " cannot be used at the same time");
304 if (!throttle_is_valid(cfg)) {
305 error_setg(errp, "bps/iops/maxs values must be 0 or greater");
312 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
314 /* Takes the ownership of bs_opts */
315 static DriveInfo *blockdev_init(const char *file, QDict *bs_opts,
322 int on_read_error, on_write_error;
331 bool has_driver_specific_opts;
332 BlockDriver *drv = NULL;
334 /* Check common options by copying from bs_opts to opts, all other options
335 * stay in bs_opts for processing by bdrv_open(). */
336 id = qdict_get_try_str(bs_opts, "id");
337 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
339 error_propagate(errp, error);
343 qemu_opts_absorb_qdict(opts, bs_opts, &error);
345 error_propagate(errp, error);
350 qdict_del(bs_opts, "id");
353 has_driver_specific_opts = !!qdict_size(bs_opts);
355 /* extract parameters */
356 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
357 ro = qemu_opt_get_bool(opts, "read-only", 0);
358 copy_on_read = qemu_opt_get_bool(opts, "copy-on-read", false);
360 serial = qemu_opt_get(opts, "serial");
362 if ((buf = qemu_opt_get(opts, "discard")) != NULL) {
363 if (bdrv_parse_discard_flags(buf, &bdrv_flags) != 0) {
364 error_setg(errp, "invalid discard option");
369 if (qemu_opt_get_bool(opts, "cache.writeback", true)) {
370 bdrv_flags |= BDRV_O_CACHE_WB;
372 if (qemu_opt_get_bool(opts, "cache.direct", false)) {
373 bdrv_flags |= BDRV_O_NOCACHE;
375 if (qemu_opt_get_bool(opts, "cache.no-flush", false)) {
376 bdrv_flags |= BDRV_O_NO_FLUSH;
379 #ifdef CONFIG_LINUX_AIO
380 if ((buf = qemu_opt_get(opts, "aio")) != NULL) {
381 if (!strcmp(buf, "native")) {
382 bdrv_flags |= BDRV_O_NATIVE_AIO;
383 } else if (!strcmp(buf, "threads")) {
384 /* this is the default */
386 error_setg(errp, "invalid aio option");
392 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
393 if (is_help_option(buf)) {
394 error_printf("Supported formats:");
395 bdrv_iterate_format(bdrv_format_print, NULL);
400 drv = bdrv_find_format(buf);
402 error_setg(errp, "'%s' invalid format", buf);
407 /* disk I/O throttling */
408 memset(&cfg, 0, sizeof(cfg));
409 cfg.buckets[THROTTLE_BPS_TOTAL].avg =
410 qemu_opt_get_number(opts, "throttling.bps-total", 0);
411 cfg.buckets[THROTTLE_BPS_READ].avg =
412 qemu_opt_get_number(opts, "throttling.bps-read", 0);
413 cfg.buckets[THROTTLE_BPS_WRITE].avg =
414 qemu_opt_get_number(opts, "throttling.bps-write", 0);
415 cfg.buckets[THROTTLE_OPS_TOTAL].avg =
416 qemu_opt_get_number(opts, "throttling.iops-total", 0);
417 cfg.buckets[THROTTLE_OPS_READ].avg =
418 qemu_opt_get_number(opts, "throttling.iops-read", 0);
419 cfg.buckets[THROTTLE_OPS_WRITE].avg =
420 qemu_opt_get_number(opts, "throttling.iops-write", 0);
422 cfg.buckets[THROTTLE_BPS_TOTAL].max =
423 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
424 cfg.buckets[THROTTLE_BPS_READ].max =
425 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
426 cfg.buckets[THROTTLE_BPS_WRITE].max =
427 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
428 cfg.buckets[THROTTLE_OPS_TOTAL].max =
429 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
430 cfg.buckets[THROTTLE_OPS_READ].max =
431 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
432 cfg.buckets[THROTTLE_OPS_WRITE].max =
433 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
435 cfg.op_size = qemu_opt_get_number(opts, "throttling.iops-size", 0);
437 if (!check_throttle_config(&cfg, &error)) {
438 error_propagate(errp, error);
442 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
443 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
444 on_write_error = parse_block_error_action(buf, 0, &error);
446 error_propagate(errp, error);
451 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
452 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
453 on_read_error = parse_block_error_action(buf, 1, &error);
455 error_propagate(errp, error);
460 if (bdrv_find_node(qemu_opts_id(opts))) {
461 error_setg(errp, "device id=%s is conflicting with a node-name",
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;
473 if (serial != NULL) {
474 dinfo->serial = g_strdup(serial);
476 QTAILQ_INSERT_TAIL(&drives, dinfo, next);
478 bdrv_set_on_error(dinfo->bdrv, on_read_error, on_write_error);
480 /* disk I/O throttling */
481 if (throttle_enabled(&cfg)) {
482 bdrv_io_limits_enable(dinfo->bdrv);
483 bdrv_set_io_limits(dinfo->bdrv, &cfg);
486 if (!file || !*file) {
487 if (has_driver_specific_opts) {
496 /* always use cache=unsafe with snapshot */
497 bdrv_flags &= ~BDRV_O_CACHE_MASK;
498 bdrv_flags |= (BDRV_O_SNAPSHOT|BDRV_O_CACHE_WB|BDRV_O_NO_FLUSH);
502 bdrv_flags |= BDRV_O_COPY_ON_READ;
505 if (runstate_check(RUN_STATE_INMIGRATE)) {
506 bdrv_flags |= BDRV_O_INCOMING;
509 bdrv_flags |= ro ? 0 : BDRV_O_RDWR;
512 ret = bdrv_open(&dinfo->bdrv, file, NULL, bs_opts, bdrv_flags, drv, &error);
516 const char _msg[] = "Failed to load disk file from the following path. Check if the file is corrupted or missing.\n\n";
517 char* err_msg = NULL;
519 err_msg = maru_convert_path((char*)_msg, file);
521 start_simple_client(err_msg);
526 error_setg(errp, "could not open disk image %s: %s",
527 file ?: dinfo->id, error_get_pretty(error));
532 if (bdrv_key_required(dinfo->bdrv))
541 bdrv_unref(dinfo->bdrv);
543 QTAILQ_REMOVE(&drives, dinfo, next);
551 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to)
555 value = qemu_opt_get(opts, from);
557 qemu_opt_set(opts, to, value);
558 qemu_opt_unset(opts, from);
562 QemuOptsList qemu_legacy_drive_opts = {
564 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
568 .type = QEMU_OPT_NUMBER,
569 .help = "bus number",
572 .type = QEMU_OPT_NUMBER,
573 .help = "unit number (i.e. lun for scsi)",
576 .type = QEMU_OPT_NUMBER,
577 .help = "index number",
580 .type = QEMU_OPT_STRING,
581 .help = "media type (disk, cdrom)",
584 .type = QEMU_OPT_STRING,
585 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
588 .type = QEMU_OPT_NUMBER,
589 .help = "number of cylinders (ide disk geometry)",
592 .type = QEMU_OPT_NUMBER,
593 .help = "number of heads (ide disk geometry)",
596 .type = QEMU_OPT_NUMBER,
597 .help = "number of sectors (ide disk geometry)",
600 .type = QEMU_OPT_STRING,
601 .help = "chs translation (auto, lba, none)",
604 .type = QEMU_OPT_BOOL,
605 .help = "(deprecated, ignored)",
608 .type = QEMU_OPT_STRING,
609 .help = "pci address (virtio only)",
612 .type = QEMU_OPT_STRING,
616 /* Options that are passed on, but have special semantics with -drive */
619 .type = QEMU_OPT_BOOL,
620 .help = "open drive file as read-only",
623 .type = QEMU_OPT_STRING,
624 .help = "read error action",
627 .type = QEMU_OPT_STRING,
628 .help = "write error action",
630 .name = "copy-on-read",
631 .type = QEMU_OPT_BOOL,
632 .help = "copy read data from backing file into image file",
635 { /* end of list */ }
639 DriveInfo *drive_init(QemuOpts *all_opts, BlockInterfaceType block_default_type)
642 DriveInfo *dinfo = NULL;
644 QemuOpts *legacy_opts;
645 DriveMediaType media = MEDIA_DISK;
646 BlockInterfaceType type;
647 int cyls, heads, secs, translation;
648 int max_devs, bus_id, unit_id, index;
650 const char *werror, *rerror;
651 bool read_only = false;
653 const char *filename;
654 Error *local_err = NULL;
656 /* Change legacy command line options into QMP ones */
657 qemu_opt_rename(all_opts, "iops", "throttling.iops-total");
658 qemu_opt_rename(all_opts, "iops_rd", "throttling.iops-read");
659 qemu_opt_rename(all_opts, "iops_wr", "throttling.iops-write");
661 qemu_opt_rename(all_opts, "bps", "throttling.bps-total");
662 qemu_opt_rename(all_opts, "bps_rd", "throttling.bps-read");
663 qemu_opt_rename(all_opts, "bps_wr", "throttling.bps-write");
665 qemu_opt_rename(all_opts, "iops_max", "throttling.iops-total-max");
666 qemu_opt_rename(all_opts, "iops_rd_max", "throttling.iops-read-max");
667 qemu_opt_rename(all_opts, "iops_wr_max", "throttling.iops-write-max");
669 qemu_opt_rename(all_opts, "bps_max", "throttling.bps-total-max");
670 qemu_opt_rename(all_opts, "bps_rd_max", "throttling.bps-read-max");
671 qemu_opt_rename(all_opts, "bps_wr_max", "throttling.bps-write-max");
673 qemu_opt_rename(all_opts,
674 "iops_size", "throttling.iops-size");
676 qemu_opt_rename(all_opts, "readonly", "read-only");
678 value = qemu_opt_get(all_opts, "cache");
682 if (bdrv_parse_cache_flags(value, &flags) != 0) {
683 error_report("invalid cache option");
687 /* Specific options take precedence */
688 if (!qemu_opt_get(all_opts, "cache.writeback")) {
689 qemu_opt_set_bool(all_opts, "cache.writeback",
690 !!(flags & BDRV_O_CACHE_WB));
692 if (!qemu_opt_get(all_opts, "cache.direct")) {
693 qemu_opt_set_bool(all_opts, "cache.direct",
694 !!(flags & BDRV_O_NOCACHE));
696 if (!qemu_opt_get(all_opts, "cache.no-flush")) {
697 qemu_opt_set_bool(all_opts, "cache.no-flush",
698 !!(flags & BDRV_O_NO_FLUSH));
700 qemu_opt_unset(all_opts, "cache");
703 /* Get a QDict for processing the options */
704 bs_opts = qdict_new();
705 qemu_opts_to_qdict(all_opts, bs_opts);
707 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
709 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
711 qerror_report_err(local_err);
712 error_free(local_err);
716 /* Deprecated option boot=[on|off] */
717 if (qemu_opt_get(legacy_opts, "boot") != NULL) {
718 fprintf(stderr, "qemu-kvm: boot=on|off is deprecated and will be "
719 "ignored. Future versions will reject this parameter. Please "
720 "update your scripts.\n");
724 value = qemu_opt_get(legacy_opts, "media");
726 if (!strcmp(value, "disk")) {
728 } else if (!strcmp(value, "cdrom")) {
732 error_report("'%s' invalid media", value);
737 /* copy-on-read is disabled with a warning for read-only devices */
738 read_only |= qemu_opt_get_bool(legacy_opts, "read-only", false);
739 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
741 if (read_only && copy_on_read) {
742 error_report("warning: disabling copy-on-read on read-only drive");
743 copy_on_read = false;
746 qdict_put(bs_opts, "read-only",
747 qstring_from_str(read_only ? "on" : "off"));
748 qdict_put(bs_opts, "copy-on-read",
749 qstring_from_str(copy_on_read ? "on" :"off"));
751 /* Controller type */
752 value = qemu_opt_get(legacy_opts, "if");
755 type < IF_COUNT && strcmp(value, if_name[type]);
758 if (type == IF_COUNT) {
759 error_report("unsupported bus type '%s'", value);
763 type = block_default_type;
767 cyls = qemu_opt_get_number(legacy_opts, "cyls", 0);
768 heads = qemu_opt_get_number(legacy_opts, "heads", 0);
769 secs = qemu_opt_get_number(legacy_opts, "secs", 0);
771 if (cyls || heads || secs) {
773 error_report("invalid physical cyls number");
777 error_report("invalid physical heads number");
781 error_report("invalid physical secs number");
786 translation = BIOS_ATA_TRANSLATION_AUTO;
787 value = qemu_opt_get(legacy_opts, "trans");
790 error_report("'%s' trans must be used with cyls, heads and secs",
794 if (!strcmp(value, "none")) {
795 translation = BIOS_ATA_TRANSLATION_NONE;
796 } else if (!strcmp(value, "lba")) {
797 translation = BIOS_ATA_TRANSLATION_LBA;
798 } else if (!strcmp(value, "large")) {
799 translation = BIOS_ATA_TRANSLATION_LARGE;
800 } else if (!strcmp(value, "rechs")) {
801 translation = BIOS_ATA_TRANSLATION_RECHS;
802 } else if (!strcmp(value, "auto")) {
803 translation = BIOS_ATA_TRANSLATION_AUTO;
805 error_report("'%s' invalid translation type", value);
810 if (media == MEDIA_CDROM) {
811 if (cyls || secs || heads) {
812 error_report("CHS can't be set with media=cdrom");
817 /* Device address specified by bus/unit or index.
818 * If none was specified, try to find the first free one. */
819 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
820 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
821 index = qemu_opt_get_number(legacy_opts, "index", -1);
823 max_devs = if_max_devs[type];
826 if (bus_id != 0 || unit_id != -1) {
827 error_report("index cannot be used with bus and unit");
830 bus_id = drive_index_to_bus_id(type, index);
831 unit_id = drive_index_to_unit_id(type, index);
836 while (drive_get(type, bus_id, unit_id) != NULL) {
838 if (max_devs && unit_id >= max_devs) {
845 if (max_devs && unit_id >= max_devs) {
846 error_report("unit %d too big (max is %d)", unit_id, max_devs - 1);
850 if (drive_get(type, bus_id, unit_id) != NULL) {
851 error_report("drive with bus=%d, unit=%d (index=%d) exists",
852 bus_id, unit_id, index);
856 /* no id supplied -> create one */
857 if (qemu_opts_id(all_opts) == NULL) {
859 const char *mediastr = "";
860 if (type == IF_IDE || type == IF_SCSI) {
861 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
864 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
867 new_id = g_strdup_printf("%s%s%i", if_name[type],
870 qdict_put(bs_opts, "id", qstring_from_str(new_id));
874 /* Add virtio block device */
875 devaddr = qemu_opt_get(legacy_opts, "addr");
876 if (devaddr && type != IF_VIRTIO) {
877 error_report("addr is not supported by this bus type");
881 if (type == IF_VIRTIO) {
883 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
885 if (arch_type == QEMU_ARCH_S390X) {
886 qemu_opt_set(devopts, "driver", "virtio-blk-s390");
888 qemu_opt_set(devopts, "driver", "virtio-blk-pci");
890 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"));
892 qemu_opt_set(devopts, "addr", devaddr);
896 filename = qemu_opt_get(legacy_opts, "file");
898 /* Check werror/rerror compatibility with if=... */
899 werror = qemu_opt_get(legacy_opts, "werror");
900 if (werror != NULL) {
901 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
903 error_report("werror is not supported by this bus type");
906 qdict_put(bs_opts, "werror", qstring_from_str(werror));
909 rerror = qemu_opt_get(legacy_opts, "rerror");
910 if (rerror != NULL) {
911 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
913 error_report("rerror is not supported by this bus type");
916 qdict_put(bs_opts, "rerror", qstring_from_str(rerror));
919 /* Actual block device init: Functionality shared with blockdev-add */
920 dinfo = blockdev_init(filename, bs_opts, &local_err);
923 qerror_report_err(local_err);
924 error_free(local_err);
931 /* Set legacy DriveInfo fields */
932 dinfo->enable_auto_del = true;
933 dinfo->opts = all_opts;
936 dinfo->heads = heads;
938 dinfo->trans = translation;
942 dinfo->unit = unit_id;
943 dinfo->devaddr = devaddr;
950 dinfo->media_cd = media == MEDIA_CDROM;
957 qemu_opts_del(legacy_opts);
961 void do_commit(Monitor *mon, const QDict *qdict)
963 const char *device = qdict_get_str(qdict, "device");
964 BlockDriverState *bs;
967 if (!strcmp(device, "all")) {
968 ret = bdrv_commit_all();
970 bs = bdrv_find(device);
972 monitor_printf(mon, "Device '%s' not found\n", device);
975 ret = bdrv_commit(bs);
978 monitor_printf(mon, "'commit' error for '%s': %s\n", device,
983 static void blockdev_do_action(int kind, void *data, Error **errp)
985 TransactionAction action;
986 TransactionActionList list;
990 list.value = &action;
992 qmp_transaction(&list, errp);
995 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
996 bool has_node_name, const char *node_name,
997 const char *snapshot_file,
998 bool has_snapshot_node_name,
999 const char *snapshot_node_name,
1000 bool has_format, const char *format,
1001 bool has_mode, NewImageMode mode, Error **errp)
1003 BlockdevSnapshot snapshot = {
1004 .has_device = has_device,
1005 .device = (char *) device,
1006 .has_node_name = has_node_name,
1007 .node_name = (char *) node_name,
1008 .snapshot_file = (char *) snapshot_file,
1009 .has_snapshot_node_name = has_snapshot_node_name,
1010 .snapshot_node_name = (char *) snapshot_node_name,
1011 .has_format = has_format,
1012 .format = (char *) format,
1013 .has_mode = has_mode,
1016 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1020 void qmp_blockdev_snapshot_internal_sync(const char *device,
1024 BlockdevSnapshotInternal snapshot = {
1025 .device = (char *) device,
1026 .name = (char *) name
1029 blockdev_do_action(TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1033 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1040 BlockDriverState *bs = bdrv_find(device);
1041 QEMUSnapshotInfo sn;
1042 Error *local_err = NULL;
1043 SnapshotInfo *info = NULL;
1047 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1060 error_setg(errp, "Name or id must be provided");
1064 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1066 error_propagate(errp, local_err);
1071 "Snapshot with id '%s' and name '%s' does not exist on "
1073 STR_OR_NULL(id), STR_OR_NULL(name), device);
1077 bdrv_snapshot_delete(bs, id, name, &local_err);
1079 error_propagate(errp, local_err);
1083 info = g_malloc0(sizeof(SnapshotInfo));
1084 info->id = g_strdup(sn.id_str);
1085 info->name = g_strdup(sn.name);
1086 info->date_nsec = sn.date_nsec;
1087 info->date_sec = sn.date_sec;
1088 info->vm_state_size = sn.vm_state_size;
1089 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1090 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1095 /* New and old BlockDriverState structs for group snapshots */
1097 typedef struct BlkTransactionState BlkTransactionState;
1099 /* Only prepare() may fail. In a single transaction, only one of commit() or
1100 abort() will be called, clean() will always be called if it present. */
1101 typedef struct BdrvActionOps {
1102 /* Size of state struct, in bytes. */
1103 size_t instance_size;
1104 /* Prepare the work, must NOT be NULL. */
1105 void (*prepare)(BlkTransactionState *common, Error **errp);
1106 /* Commit the changes, can be NULL. */
1107 void (*commit)(BlkTransactionState *common);
1108 /* Abort the changes on fail, can be NULL. */
1109 void (*abort)(BlkTransactionState *common);
1110 /* Clean up resource in the end, can be NULL. */
1111 void (*clean)(BlkTransactionState *common);
1115 * This structure must be arranged as first member in child type, assuming
1116 * that compiler will also arrange it to the same address with parent instance.
1117 * Later it will be used in free().
1119 struct BlkTransactionState {
1120 TransactionAction *action;
1121 const BdrvActionOps *ops;
1122 QSIMPLEQ_ENTRY(BlkTransactionState) entry;
1125 /* internal snapshot private data */
1126 typedef struct InternalSnapshotState {
1127 BlkTransactionState common;
1128 BlockDriverState *bs;
1129 QEMUSnapshotInfo sn;
1130 } InternalSnapshotState;
1132 static void internal_snapshot_prepare(BlkTransactionState *common,
1137 BlockDriverState *bs;
1138 QEMUSnapshotInfo old_sn, *sn;
1141 BlockdevSnapshotInternal *internal;
1142 InternalSnapshotState *state;
1145 g_assert(common->action->kind ==
1146 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1147 internal = common->action->blockdev_snapshot_internal_sync;
1148 state = DO_UPCAST(InternalSnapshotState, common, common);
1150 /* 1. parse input */
1151 device = internal->device;
1152 name = internal->name;
1154 /* 2. check for validation */
1155 bs = bdrv_find(device);
1157 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1161 if (!bdrv_is_inserted(bs)) {
1162 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1166 if (bdrv_is_read_only(bs)) {
1167 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1171 if (!bdrv_can_snapshot(bs)) {
1172 error_set(errp, QERR_BLOCK_FORMAT_FEATURE_NOT_SUPPORTED,
1173 bs->drv->format_name, device, "internal snapshot");
1177 if (!strlen(name)) {
1178 error_setg(errp, "Name is empty");
1182 /* check whether a snapshot with name exist */
1183 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn, errp);
1184 if (error_is_set(errp)) {
1188 "Snapshot with name '%s' already exists on device '%s'",
1193 /* 3. take the snapshot */
1195 pstrcpy(sn->name, sizeof(sn->name), name);
1196 qemu_gettimeofday(&tv);
1197 sn->date_sec = tv.tv_sec;
1198 sn->date_nsec = tv.tv_usec * 1000;
1199 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1201 ret1 = bdrv_snapshot_create(bs, sn);
1203 error_setg_errno(errp, -ret1,
1204 "Failed to create snapshot '%s' on device '%s'",
1209 /* 4. succeed, mark a snapshot is created */
1213 static void internal_snapshot_abort(BlkTransactionState *common)
1215 InternalSnapshotState *state =
1216 DO_UPCAST(InternalSnapshotState, common, common);
1217 BlockDriverState *bs = state->bs;
1218 QEMUSnapshotInfo *sn = &state->sn;
1219 Error *local_error = NULL;
1225 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1226 error_report("Failed to delete snapshot with id '%s' and name '%s' on "
1227 "device '%s' in abort: %s",
1230 bdrv_get_device_name(bs),
1231 error_get_pretty(local_error));
1232 error_free(local_error);
1236 /* external snapshot private data */
1237 typedef struct ExternalSnapshotState {
1238 BlkTransactionState common;
1239 BlockDriverState *old_bs;
1240 BlockDriverState *new_bs;
1241 } ExternalSnapshotState;
1243 static void external_snapshot_prepare(BlkTransactionState *common,
1248 QDict *options = NULL;
1249 Error *local_err = NULL;
1250 bool has_device = false;
1252 bool has_node_name = false;
1253 const char *node_name;
1254 bool has_snapshot_node_name = false;
1255 const char *snapshot_node_name;
1256 const char *new_image_file;
1257 const char *format = "qcow2";
1258 enum NewImageMode mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1259 ExternalSnapshotState *state =
1260 DO_UPCAST(ExternalSnapshotState, common, common);
1261 TransactionAction *action = common->action;
1263 /* get parameters */
1264 g_assert(action->kind == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC);
1266 has_device = action->blockdev_snapshot_sync->has_device;
1267 device = action->blockdev_snapshot_sync->device;
1268 has_node_name = action->blockdev_snapshot_sync->has_node_name;
1269 node_name = action->blockdev_snapshot_sync->node_name;
1270 has_snapshot_node_name =
1271 action->blockdev_snapshot_sync->has_snapshot_node_name;
1272 snapshot_node_name = action->blockdev_snapshot_sync->snapshot_node_name;
1274 new_image_file = action->blockdev_snapshot_sync->snapshot_file;
1275 if (action->blockdev_snapshot_sync->has_format) {
1276 format = action->blockdev_snapshot_sync->format;
1278 if (action->blockdev_snapshot_sync->has_mode) {
1279 mode = action->blockdev_snapshot_sync->mode;
1282 /* start processing */
1283 drv = bdrv_find_format(format);
1285 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1289 state->old_bs = bdrv_lookup_bs(has_device ? device : NULL,
1290 has_node_name ? node_name : NULL,
1293 error_propagate(errp, local_err);
1297 if (has_node_name && !has_snapshot_node_name) {
1298 error_setg(errp, "New snapshot node name missing");
1302 if (has_snapshot_node_name && bdrv_find_node(snapshot_node_name)) {
1303 error_setg(errp, "New snapshot node name already existing");
1307 if (!bdrv_is_inserted(state->old_bs)) {
1308 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1312 if (bdrv_in_use(state->old_bs)) {
1313 error_set(errp, QERR_DEVICE_IN_USE, device);
1317 if (!bdrv_is_read_only(state->old_bs)) {
1318 if (bdrv_flush(state->old_bs)) {
1319 error_set(errp, QERR_IO_ERROR);
1324 if (!bdrv_is_first_non_filter(state->old_bs)) {
1325 error_set(errp, QERR_FEATURE_DISABLED, "snapshot");
1329 flags = state->old_bs->open_flags;
1331 /* create new image w/backing file */
1332 if (mode != NEW_IMAGE_MODE_EXISTING) {
1333 bdrv_img_create(new_image_file, format,
1334 state->old_bs->filename,
1335 state->old_bs->drv->format_name,
1336 NULL, -1, flags, &local_err, false);
1338 error_propagate(errp, local_err);
1343 if (has_snapshot_node_name) {
1344 options = qdict_new();
1345 qdict_put(options, "node-name",
1346 qstring_from_str(snapshot_node_name));
1349 /* TODO Inherit bs->options or only take explicit options with an
1350 * extended QMP command? */
1351 assert(state->new_bs == NULL);
1352 ret = bdrv_open(&state->new_bs, new_image_file, NULL, options,
1353 flags | BDRV_O_NO_BACKING, drv, &local_err);
1354 /* We will manually add the backing_hd field to the bs later */
1356 error_propagate(errp, local_err);
1360 static void external_snapshot_commit(BlkTransactionState *common)
1362 ExternalSnapshotState *state =
1363 DO_UPCAST(ExternalSnapshotState, common, common);
1365 /* This removes our old bs and adds the new bs */
1366 bdrv_append(state->new_bs, state->old_bs);
1367 /* We don't need (or want) to use the transactional
1368 * bdrv_reopen_multiple() across all the entries at once, because we
1369 * don't want to abort all of them if one of them fails the reopen */
1370 bdrv_reopen(state->new_bs, state->new_bs->open_flags & ~BDRV_O_RDWR,
1374 static void external_snapshot_abort(BlkTransactionState *common)
1376 ExternalSnapshotState *state =
1377 DO_UPCAST(ExternalSnapshotState, common, common);
1378 if (state->new_bs) {
1379 bdrv_unref(state->new_bs);
1383 typedef struct DriveBackupState {
1384 BlkTransactionState common;
1385 BlockDriverState *bs;
1389 static void drive_backup_prepare(BlkTransactionState *common, Error **errp)
1391 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1392 DriveBackup *backup;
1393 Error *local_err = NULL;
1395 assert(common->action->kind == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1396 backup = common->action->drive_backup;
1398 qmp_drive_backup(backup->device, backup->target,
1399 backup->has_format, backup->format,
1401 backup->has_mode, backup->mode,
1402 backup->has_speed, backup->speed,
1403 backup->has_on_source_error, backup->on_source_error,
1404 backup->has_on_target_error, backup->on_target_error,
1407 error_propagate(errp, local_err);
1413 state->bs = bdrv_find(backup->device);
1414 state->job = state->bs->job;
1417 static void drive_backup_abort(BlkTransactionState *common)
1419 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1420 BlockDriverState *bs = state->bs;
1422 /* Only cancel if it's the job we started */
1423 if (bs && bs->job && bs->job == state->job) {
1424 block_job_cancel_sync(bs->job);
1428 static void abort_prepare(BlkTransactionState *common, Error **errp)
1430 error_setg(errp, "Transaction aborted using Abort action");
1433 static void abort_commit(BlkTransactionState *common)
1435 g_assert_not_reached(); /* this action never succeeds */
1438 static const BdrvActionOps actions[] = {
1439 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
1440 .instance_size = sizeof(ExternalSnapshotState),
1441 .prepare = external_snapshot_prepare,
1442 .commit = external_snapshot_commit,
1443 .abort = external_snapshot_abort,
1445 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
1446 .instance_size = sizeof(DriveBackupState),
1447 .prepare = drive_backup_prepare,
1448 .abort = drive_backup_abort,
1450 [TRANSACTION_ACTION_KIND_ABORT] = {
1451 .instance_size = sizeof(BlkTransactionState),
1452 .prepare = abort_prepare,
1453 .commit = abort_commit,
1455 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
1456 .instance_size = sizeof(InternalSnapshotState),
1457 .prepare = internal_snapshot_prepare,
1458 .abort = internal_snapshot_abort,
1463 * 'Atomic' group snapshots. The snapshots are taken as a set, and if any fail
1464 * then we do not pivot any of the devices in the group, and abandon the
1467 void qmp_transaction(TransactionActionList *dev_list, Error **errp)
1469 TransactionActionList *dev_entry = dev_list;
1470 BlkTransactionState *state, *next;
1471 Error *local_err = NULL;
1473 QSIMPLEQ_HEAD(snap_bdrv_states, BlkTransactionState) snap_bdrv_states;
1474 QSIMPLEQ_INIT(&snap_bdrv_states);
1476 /* drain all i/o before any snapshots */
1479 /* We don't do anything in this loop that commits us to the snapshot */
1480 while (NULL != dev_entry) {
1481 TransactionAction *dev_info = NULL;
1482 const BdrvActionOps *ops;
1484 dev_info = dev_entry->value;
1485 dev_entry = dev_entry->next;
1487 assert(dev_info->kind < ARRAY_SIZE(actions));
1489 ops = &actions[dev_info->kind];
1490 assert(ops->instance_size > 0);
1492 state = g_malloc0(ops->instance_size);
1494 state->action = dev_info;
1495 QSIMPLEQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
1497 state->ops->prepare(state, &local_err);
1499 error_propagate(errp, local_err);
1500 goto delete_and_fail;
1504 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1505 if (state->ops->commit) {
1506 state->ops->commit(state);
1515 * failure, and it is all-or-none; abandon each new bs, and keep using
1516 * the original bs for all images
1518 QSIMPLEQ_FOREACH(state, &snap_bdrv_states, entry) {
1519 if (state->ops->abort) {
1520 state->ops->abort(state);
1524 QSIMPLEQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
1525 if (state->ops->clean) {
1526 state->ops->clean(state);
1533 static void eject_device(BlockDriverState *bs, int force, Error **errp)
1535 if (bdrv_in_use(bs)) {
1536 error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
1539 if (!bdrv_dev_has_removable_media(bs)) {
1540 error_set(errp, QERR_DEVICE_NOT_REMOVABLE, bdrv_get_device_name(bs));
1544 if (bdrv_dev_is_medium_locked(bs) && !bdrv_dev_is_tray_open(bs)) {
1545 bdrv_dev_eject_request(bs, force);
1547 error_set(errp, QERR_DEVICE_LOCKED, bdrv_get_device_name(bs));
1555 void qmp_eject(const char *device, bool has_force, bool force, Error **errp)
1557 BlockDriverState *bs;
1559 bs = bdrv_find(device);
1561 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1565 eject_device(bs, force, errp);
1568 void qmp_block_passwd(bool has_device, const char *device,
1569 bool has_node_name, const char *node_name,
1570 const char *password, Error **errp)
1572 Error *local_err = NULL;
1573 BlockDriverState *bs;
1576 bs = bdrv_lookup_bs(has_device ? device : NULL,
1577 has_node_name ? node_name : NULL,
1580 error_propagate(errp, local_err);
1584 err = bdrv_set_key(bs, password);
1585 if (err == -EINVAL) {
1586 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1588 } else if (err < 0) {
1589 error_set(errp, QERR_INVALID_PASSWORD);
1594 static void qmp_bdrv_open_encrypted(BlockDriverState *bs, const char *filename,
1595 int bdrv_flags, BlockDriver *drv,
1596 const char *password, Error **errp)
1598 Error *local_err = NULL;
1601 ret = bdrv_open(&bs, filename, NULL, NULL, bdrv_flags, drv, &local_err);
1603 error_propagate(errp, local_err);
1607 if (bdrv_key_required(bs)) {
1609 if (bdrv_set_key(bs, password) < 0) {
1610 error_set(errp, QERR_INVALID_PASSWORD);
1613 error_set(errp, QERR_DEVICE_ENCRYPTED, bdrv_get_device_name(bs),
1614 bdrv_get_encrypted_filename(bs));
1616 } else if (password) {
1617 error_set(errp, QERR_DEVICE_NOT_ENCRYPTED, bdrv_get_device_name(bs));
1621 void qmp_change_blockdev(const char *device, const char *filename,
1622 const char *format, Error **errp)
1624 BlockDriverState *bs;
1625 BlockDriver *drv = NULL;
1629 bs = bdrv_find(device);
1631 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1636 drv = bdrv_find_whitelisted_format(format, bs->read_only);
1638 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1643 eject_device(bs, 0, &err);
1645 error_propagate(errp, err);
1649 bdrv_flags = bdrv_is_read_only(bs) ? 0 : BDRV_O_RDWR;
1650 bdrv_flags |= bdrv_is_snapshot(bs) ? BDRV_O_SNAPSHOT : 0;
1652 qmp_bdrv_open_encrypted(bs, filename, bdrv_flags, drv, NULL, errp);
1655 /* throttling disk I/O limits */
1656 void qmp_block_set_io_throttle(const char *device, int64_t bps, int64_t bps_rd,
1663 bool has_bps_rd_max,
1665 bool has_bps_wr_max,
1669 bool has_iops_rd_max,
1670 int64_t iops_rd_max,
1671 bool has_iops_wr_max,
1672 int64_t iops_wr_max,
1674 int64_t iops_size, Error **errp)
1677 BlockDriverState *bs;
1679 bs = bdrv_find(device);
1681 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1685 memset(&cfg, 0, sizeof(cfg));
1686 cfg.buckets[THROTTLE_BPS_TOTAL].avg = bps;
1687 cfg.buckets[THROTTLE_BPS_READ].avg = bps_rd;
1688 cfg.buckets[THROTTLE_BPS_WRITE].avg = bps_wr;
1690 cfg.buckets[THROTTLE_OPS_TOTAL].avg = iops;
1691 cfg.buckets[THROTTLE_OPS_READ].avg = iops_rd;
1692 cfg.buckets[THROTTLE_OPS_WRITE].avg = iops_wr;
1695 cfg.buckets[THROTTLE_BPS_TOTAL].max = bps_max;
1697 if (has_bps_rd_max) {
1698 cfg.buckets[THROTTLE_BPS_READ].max = bps_rd_max;
1700 if (has_bps_wr_max) {
1701 cfg.buckets[THROTTLE_BPS_WRITE].max = bps_wr_max;
1704 cfg.buckets[THROTTLE_OPS_TOTAL].max = iops_max;
1706 if (has_iops_rd_max) {
1707 cfg.buckets[THROTTLE_OPS_READ].max = iops_rd_max;
1709 if (has_iops_wr_max) {
1710 cfg.buckets[THROTTLE_OPS_WRITE].max = iops_wr_max;
1713 if (has_iops_size) {
1714 cfg.op_size = iops_size;
1717 if (!check_throttle_config(&cfg, errp)) {
1721 if (!bs->io_limits_enabled && throttle_enabled(&cfg)) {
1722 bdrv_io_limits_enable(bs);
1723 } else if (bs->io_limits_enabled && !throttle_enabled(&cfg)) {
1724 bdrv_io_limits_disable(bs);
1727 if (bs->io_limits_enabled) {
1728 bdrv_set_io_limits(bs, &cfg);
1732 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data)
1734 const char *id = qdict_get_str(qdict, "id");
1735 BlockDriverState *bs;
1739 qerror_report(QERR_DEVICE_NOT_FOUND, id);
1742 if (bdrv_in_use(bs)) {
1743 qerror_report(QERR_DEVICE_IN_USE, id);
1747 /* quiesce block driver; prevent further io */
1752 /* if we have a device attached to this BlockDriverState
1753 * then we need to make the drive anonymous until the device
1754 * can be removed. If this is a drive with no device backing
1755 * then we can just get rid of the block driver state right here.
1757 if (bdrv_get_attached_dev(bs)) {
1760 /* Further I/O must not pause the guest */
1761 bdrv_set_on_error(bs, BLOCKDEV_ON_ERROR_REPORT,
1762 BLOCKDEV_ON_ERROR_REPORT);
1764 drive_uninit(drive_get_by_blockdev(bs));
1770 void qmp_block_resize(bool has_device, const char *device,
1771 bool has_node_name, const char *node_name,
1772 int64_t size, Error **errp)
1774 Error *local_err = NULL;
1775 BlockDriverState *bs;
1778 bs = bdrv_lookup_bs(has_device ? device : NULL,
1779 has_node_name ? node_name : NULL,
1782 error_propagate(errp, local_err);
1786 if (!bdrv_is_first_non_filter(bs)) {
1787 error_set(errp, QERR_FEATURE_DISABLED, "resize");
1792 error_set(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
1796 /* complete all in-flight operations before resizing the device */
1799 ret = bdrv_truncate(bs, size);
1804 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1807 error_set(errp, QERR_UNSUPPORTED);
1810 error_set(errp, QERR_DEVICE_IS_READ_ONLY, device);
1813 error_set(errp, QERR_DEVICE_IN_USE, device);
1816 error_setg_errno(errp, -ret, "Could not resize");
1821 static void block_job_cb(void *opaque, int ret)
1823 BlockDriverState *bs = opaque;
1826 trace_block_job_cb(bs, bs->job, ret);
1829 obj = qobject_from_block_job(bs->job);
1831 QDict *dict = qobject_to_qdict(obj);
1832 qdict_put(dict, "error", qstring_from_str(strerror(-ret)));
1835 if (block_job_is_cancelled(bs->job)) {
1836 monitor_protocol_event(QEVENT_BLOCK_JOB_CANCELLED, obj);
1838 monitor_protocol_event(QEVENT_BLOCK_JOB_COMPLETED, obj);
1840 qobject_decref(obj);
1842 bdrv_put_ref_bh_schedule(bs);
1845 void qmp_block_stream(const char *device, bool has_base,
1846 const char *base, bool has_speed, int64_t speed,
1847 bool has_on_error, BlockdevOnError on_error,
1850 BlockDriverState *bs;
1851 BlockDriverState *base_bs = NULL;
1852 Error *local_err = NULL;
1854 if (!has_on_error) {
1855 on_error = BLOCKDEV_ON_ERROR_REPORT;
1858 bs = bdrv_find(device);
1860 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1865 base_bs = bdrv_find_backing_image(bs, base);
1866 if (base_bs == NULL) {
1867 error_set(errp, QERR_BASE_NOT_FOUND, base);
1872 stream_start(bs, base_bs, base, has_speed ? speed : 0,
1873 on_error, block_job_cb, bs, &local_err);
1875 error_propagate(errp, local_err);
1879 trace_qmp_block_stream(bs, bs->job);
1882 void qmp_block_commit(const char *device,
1883 bool has_base, const char *base, const char *top,
1884 bool has_speed, int64_t speed,
1887 BlockDriverState *bs;
1888 BlockDriverState *base_bs, *top_bs;
1889 Error *local_err = NULL;
1890 /* This will be part of the QMP command, if/when the
1891 * BlockdevOnError change for blkmirror makes it in
1893 BlockdevOnError on_error = BLOCKDEV_ON_ERROR_REPORT;
1899 /* drain all i/o before commits */
1902 bs = bdrv_find(device);
1904 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1908 /* default top_bs is the active layer */
1912 if (strcmp(bs->filename, top) != 0) {
1913 top_bs = bdrv_find_backing_image(bs, top);
1917 if (top_bs == NULL) {
1918 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
1922 if (has_base && base) {
1923 base_bs = bdrv_find_backing_image(top_bs, base);
1925 base_bs = bdrv_find_base(top_bs);
1928 if (base_bs == NULL) {
1929 error_set(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
1934 commit_active_start(bs, base_bs, speed, on_error, block_job_cb,
1937 commit_start(bs, base_bs, top_bs, speed, on_error, block_job_cb, bs,
1940 if (local_err != NULL) {
1941 error_propagate(errp, local_err);
1946 void qmp_drive_backup(const char *device, const char *target,
1947 bool has_format, const char *format,
1948 enum MirrorSyncMode sync,
1949 bool has_mode, enum NewImageMode mode,
1950 bool has_speed, int64_t speed,
1951 bool has_on_source_error, BlockdevOnError on_source_error,
1952 bool has_on_target_error, BlockdevOnError on_target_error,
1955 BlockDriverState *bs;
1956 BlockDriverState *target_bs;
1957 BlockDriverState *source = NULL;
1958 BlockDriver *drv = NULL;
1959 Error *local_err = NULL;
1967 if (!has_on_source_error) {
1968 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
1970 if (!has_on_target_error) {
1971 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
1974 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1977 bs = bdrv_find(device);
1979 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
1983 if (!bdrv_is_inserted(bs)) {
1984 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1989 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
1992 drv = bdrv_find_format(format);
1994 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
1999 if (bdrv_in_use(bs)) {
2000 error_set(errp, QERR_DEVICE_IN_USE, device);
2004 flags = bs->open_flags | BDRV_O_RDWR;
2006 /* See if we have a backing HD we can use to create our new image
2008 if (sync == MIRROR_SYNC_MODE_TOP) {
2009 source = bs->backing_hd;
2011 sync = MIRROR_SYNC_MODE_FULL;
2014 if (sync == MIRROR_SYNC_MODE_NONE) {
2018 size = bdrv_getlength(bs);
2020 error_setg_errno(errp, -size, "bdrv_getlength failed");
2024 if (mode != NEW_IMAGE_MODE_EXISTING) {
2025 assert(format && drv);
2027 bdrv_img_create(target, format, source->filename,
2028 source->drv->format_name, NULL,
2029 size, flags, &local_err, false);
2031 bdrv_img_create(target, format, NULL, NULL, NULL,
2032 size, flags, &local_err, false);
2037 error_propagate(errp, local_err);
2042 ret = bdrv_open(&target_bs, target, NULL, NULL, flags, drv, &local_err);
2044 error_propagate(errp, local_err);
2048 backup_start(bs, target_bs, speed, sync, on_source_error, on_target_error,
2049 block_job_cb, bs, &local_err);
2050 if (local_err != NULL) {
2051 bdrv_unref(target_bs);
2052 error_propagate(errp, local_err);
2057 BlockDeviceInfoList *qmp_query_named_block_nodes(Error **errp)
2059 return bdrv_named_nodes_list();
2062 #define DEFAULT_MIRROR_BUF_SIZE (10 << 20)
2064 void qmp_drive_mirror(const char *device, const char *target,
2065 bool has_format, const char *format,
2066 enum MirrorSyncMode sync,
2067 bool has_mode, enum NewImageMode mode,
2068 bool has_speed, int64_t speed,
2069 bool has_granularity, uint32_t granularity,
2070 bool has_buf_size, int64_t buf_size,
2071 bool has_on_source_error, BlockdevOnError on_source_error,
2072 bool has_on_target_error, BlockdevOnError on_target_error,
2075 BlockDriverState *bs;
2076 BlockDriverState *source, *target_bs;
2077 BlockDriver *drv = NULL;
2078 Error *local_err = NULL;
2086 if (!has_on_source_error) {
2087 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2089 if (!has_on_target_error) {
2090 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2093 mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
2095 if (!has_granularity) {
2098 if (!has_buf_size) {
2099 buf_size = DEFAULT_MIRROR_BUF_SIZE;
2102 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2103 error_set(errp, QERR_INVALID_PARAMETER, device);
2106 if (granularity & (granularity - 1)) {
2107 error_set(errp, QERR_INVALID_PARAMETER, device);
2111 bs = bdrv_find(device);
2113 error_set(errp, QERR_DEVICE_NOT_FOUND, device);
2117 if (!bdrv_is_inserted(bs)) {
2118 error_set(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
2123 format = mode == NEW_IMAGE_MODE_EXISTING ? NULL : bs->drv->format_name;
2126 drv = bdrv_find_format(format);
2128 error_set(errp, QERR_INVALID_BLOCK_FORMAT, format);
2133 if (bdrv_in_use(bs)) {
2134 error_set(errp, QERR_DEVICE_IN_USE, device);
2138 flags = bs->open_flags | BDRV_O_RDWR;
2139 source = bs->backing_hd;
2140 if (!source && sync == MIRROR_SYNC_MODE_TOP) {
2141 sync = MIRROR_SYNC_MODE_FULL;
2143 if (sync == MIRROR_SYNC_MODE_NONE) {
2147 size = bdrv_getlength(bs);
2149 error_setg_errno(errp, -size, "bdrv_getlength failed");
2153 if ((sync == MIRROR_SYNC_MODE_FULL || !source)
2154 && mode != NEW_IMAGE_MODE_EXISTING)
2156 /* create new image w/o backing file */
2157 assert(format && drv);
2158 bdrv_img_create(target, format,
2159 NULL, NULL, NULL, size, flags, &local_err, false);
2162 case NEW_IMAGE_MODE_EXISTING:
2164 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
2165 /* create new image with backing file */
2166 bdrv_img_create(target, format,
2168 source->drv->format_name,
2169 NULL, size, flags, &local_err, false);
2177 error_propagate(errp, local_err);
2181 /* Mirroring takes care of copy-on-write using the source's backing
2185 ret = bdrv_open(&target_bs, target, NULL, NULL, flags | BDRV_O_NO_BACKING,
2188 error_propagate(errp, local_err);
2192 mirror_start(bs, target_bs, speed, granularity, buf_size, sync,
2193 on_source_error, on_target_error,
2194 block_job_cb, bs, &local_err);
2195 if (local_err != NULL) {
2196 bdrv_unref(target_bs);
2197 error_propagate(errp, local_err);
2202 static BlockJob *find_block_job(const char *device)
2204 BlockDriverState *bs;
2206 bs = bdrv_find(device);
2207 if (!bs || !bs->job) {
2213 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
2215 BlockJob *job = find_block_job(device);
2218 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2222 block_job_set_speed(job, speed, errp);
2225 void qmp_block_job_cancel(const char *device,
2226 bool has_force, bool force, Error **errp)
2228 BlockJob *job = find_block_job(device);
2235 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2238 if (job->paused && !force) {
2239 error_set(errp, QERR_BLOCK_JOB_PAUSED, device);
2243 trace_qmp_block_job_cancel(job);
2244 block_job_cancel(job);
2247 void qmp_block_job_pause(const char *device, Error **errp)
2249 BlockJob *job = find_block_job(device);
2252 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2256 trace_qmp_block_job_pause(job);
2257 block_job_pause(job);
2260 void qmp_block_job_resume(const char *device, Error **errp)
2262 BlockJob *job = find_block_job(device);
2265 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2269 trace_qmp_block_job_resume(job);
2270 block_job_resume(job);
2273 void qmp_block_job_complete(const char *device, Error **errp)
2275 BlockJob *job = find_block_job(device);
2278 error_set(errp, QERR_BLOCK_JOB_NOT_ACTIVE, device);
2282 trace_qmp_block_job_complete(job);
2283 block_job_complete(job, errp);
2286 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
2288 QmpOutputVisitor *ov = qmp_output_visitor_new();
2292 Error *local_err = NULL;
2294 /* Require an ID in the top level */
2295 if (!options->has_id) {
2296 error_setg(errp, "Block device needs an ID");
2300 /* TODO Sort it out in raw-posix and drive_init: Reject aio=native with
2301 * cache.direct=false instead of silently switching to aio=threads, except
2302 * if called from drive_init.
2304 * For now, simply forbidding the combination for all drivers will do. */
2305 if (options->has_aio && options->aio == BLOCKDEV_AIO_OPTIONS_NATIVE) {
2306 bool direct = options->has_cache &&
2307 options->cache->has_direct &&
2308 options->cache->direct;
2310 error_setg(errp, "aio=native requires cache.direct=true");
2315 visit_type_BlockdevOptions(qmp_output_get_visitor(ov),
2316 &options, NULL, &local_err);
2318 error_propagate(errp, local_err);
2322 obj = qmp_output_get_qobject(ov);
2323 qdict = qobject_to_qdict(obj);
2325 qdict_flatten(qdict);
2327 dinfo = blockdev_init(NULL, qdict, &local_err);
2329 error_propagate(errp, local_err);
2333 if (bdrv_key_required(dinfo->bdrv)) {
2334 drive_uninit(dinfo);
2335 error_setg(errp, "blockdev-add doesn't support encrypted devices");
2340 qmp_output_visitor_cleanup(ov);
2343 static void do_qmp_query_block_jobs_one(void *opaque, BlockDriverState *bs)
2345 BlockJobInfoList **prev = opaque;
2346 BlockJob *job = bs->job;
2349 BlockJobInfoList *elem = g_new0(BlockJobInfoList, 1);
2350 elem->value = block_job_query(bs->job);
2351 (*prev)->next = elem;
2356 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
2358 /* Dummy is a fake list element for holding the head pointer */
2359 BlockJobInfoList dummy = {};
2360 BlockJobInfoList *prev = &dummy;
2361 bdrv_iterate(do_qmp_query_block_jobs_one, &prev);
2365 QemuOptsList qemu_common_drive_opts = {
2367 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
2371 .type = QEMU_OPT_BOOL,
2372 .help = "enable/disable snapshot mode",
2375 .type = QEMU_OPT_STRING,
2376 .help = "discard operation (ignore/off, unmap/on)",
2378 .name = "cache.writeback",
2379 .type = QEMU_OPT_BOOL,
2380 .help = "enables writeback mode for any caches",
2382 .name = "cache.direct",
2383 .type = QEMU_OPT_BOOL,
2384 .help = "enables use of O_DIRECT (bypass the host page cache)",
2386 .name = "cache.no-flush",
2387 .type = QEMU_OPT_BOOL,
2388 .help = "ignore any flush requests for the device",
2391 .type = QEMU_OPT_STRING,
2392 .help = "host AIO implementation (threads, native)",
2395 .type = QEMU_OPT_STRING,
2396 .help = "disk format (raw, qcow2, ...)",
2399 .type = QEMU_OPT_STRING,
2400 .help = "disk serial number",
2403 .type = QEMU_OPT_STRING,
2404 .help = "read error action",
2407 .type = QEMU_OPT_STRING,
2408 .help = "write error action",
2410 .name = "read-only",
2411 .type = QEMU_OPT_BOOL,
2412 .help = "open drive file as read-only",
2414 .name = "throttling.iops-total",
2415 .type = QEMU_OPT_NUMBER,
2416 .help = "limit total I/O operations per second",
2418 .name = "throttling.iops-read",
2419 .type = QEMU_OPT_NUMBER,
2420 .help = "limit read operations per second",
2422 .name = "throttling.iops-write",
2423 .type = QEMU_OPT_NUMBER,
2424 .help = "limit write operations per second",
2426 .name = "throttling.bps-total",
2427 .type = QEMU_OPT_NUMBER,
2428 .help = "limit total bytes per second",
2430 .name = "throttling.bps-read",
2431 .type = QEMU_OPT_NUMBER,
2432 .help = "limit read bytes per second",
2434 .name = "throttling.bps-write",
2435 .type = QEMU_OPT_NUMBER,
2436 .help = "limit write bytes per second",
2438 .name = "throttling.iops-total-max",
2439 .type = QEMU_OPT_NUMBER,
2440 .help = "I/O operations burst",
2442 .name = "throttling.iops-read-max",
2443 .type = QEMU_OPT_NUMBER,
2444 .help = "I/O operations read burst",
2446 .name = "throttling.iops-write-max",
2447 .type = QEMU_OPT_NUMBER,
2448 .help = "I/O operations write burst",
2450 .name = "throttling.bps-total-max",
2451 .type = QEMU_OPT_NUMBER,
2452 .help = "total bytes burst",
2454 .name = "throttling.bps-read-max",
2455 .type = QEMU_OPT_NUMBER,
2456 .help = "total bytes read burst",
2458 .name = "throttling.bps-write-max",
2459 .type = QEMU_OPT_NUMBER,
2460 .help = "total bytes write burst",
2462 .name = "throttling.iops-size",
2463 .type = QEMU_OPT_NUMBER,
2464 .help = "when limiting by iops max size of an I/O in bytes",
2466 .name = "copy-on-read",
2467 .type = QEMU_OPT_BOOL,
2468 .help = "copy read data from backing file into image file",
2470 { /* end of list */ }
2474 QemuOptsList qemu_drive_opts = {
2476 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
2479 * no elements => accept any params
2480 * validation will happen later
2482 { /* end of list */ }