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 "qemu/osdep.h"
34 #include "sysemu/block-backend.h"
35 #include "sysemu/blockdev.h"
36 #include "hw/block/block.h"
37 #include "block/blockjob.h"
38 #include "block/dirty-bitmap.h"
39 #include "block/qdict.h"
40 #include "block/throttle-groups.h"
41 #include "monitor/monitor.h"
42 #include "qemu/error-report.h"
43 #include "qemu/option.h"
44 #include "qemu/qemu-print.h"
45 #include "qemu/config-file.h"
46 #include "qapi/qapi-commands-block.h"
47 #include "qapi/qapi-commands-transaction.h"
48 #include "qapi/qapi-visit-block-core.h"
49 #include "qapi/qmp/qdict.h"
50 #include "qapi/qmp/qnum.h"
51 #include "qapi/qmp/qstring.h"
52 #include "qapi/error.h"
53 #include "qapi/qmp/qerror.h"
54 #include "qapi/qmp/qlist.h"
55 #include "qapi/qobject-output-visitor.h"
56 #include "sysemu/sysemu.h"
57 #include "sysemu/iothread.h"
58 #include "block/block_int.h"
59 #include "block/trace.h"
60 #include "sysemu/runstate.h"
61 #include "sysemu/replay.h"
62 #include "qemu/cutils.h"
63 #include "qemu/help_option.h"
64 #include "qemu/main-loop.h"
65 #include "qemu/throttle-options.h"
67 /* Protected by BQL */
68 QTAILQ_HEAD(, BlockDriverState) monitor_bdrv_states =
69 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states);
71 void bdrv_set_monitor_owned(BlockDriverState *bs)
74 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
77 static const char *const if_name[IF_COUNT] = {
81 [IF_FLOPPY] = "floppy",
82 [IF_PFLASH] = "pflash",
85 [IF_VIRTIO] = "virtio",
89 static int if_max_devs[IF_COUNT] = {
91 * Do not change these numbers! They govern how drive option
92 * index maps to unit and bus. That mapping is ABI.
94 * All controllers used to implement if=T drives need to support
95 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
96 * Otherwise, some index values map to "impossible" bus, unit
99 * For instance, if you change [IF_SCSI] to 255, -drive
100 * if=scsi,index=12 no longer means bus=1,unit=5, but
101 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
102 * the drive can't be set up. Regression.
109 * Boards may call this to offer board-by-board overrides
110 * of the default, global values.
112 void override_max_devs(BlockInterfaceType type, int max_devs)
123 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
124 dinfo = blk_legacy_dinfo(blk);
125 if (dinfo->type == type) {
126 fprintf(stderr, "Cannot override units-per-bus property of"
127 " the %s interface, because a drive of that type has"
128 " already been added.\n", if_name[type]);
129 g_assert_not_reached();
133 if_max_devs[type] = max_devs;
137 * We automatically delete the drive when a device using it gets
138 * unplugged. Questionable feature, but we can't just drop it.
139 * Device models call blockdev_mark_auto_del() to schedule the
140 * automatic deletion, and generic qdev code calls blockdev_auto_del()
141 * when deletion is actually safe.
143 void blockdev_mark_auto_del(BlockBackend *blk)
145 DriveInfo *dinfo = blk_legacy_dinfo(blk);
157 job = block_job_next_locked(NULL);
158 while (job && (job->job.cancelled ||
159 job->job.deferred_to_main_loop ||
160 !block_job_has_bdrv(job, blk_bs(blk))))
162 job = block_job_next_locked(job);
166 * This drops the job lock temporarily and polls, so we need to
167 * restart processing the list from the start after this.
169 job_cancel_locked(&job->job, false);
176 void blockdev_auto_del(BlockBackend *blk)
178 DriveInfo *dinfo = blk_legacy_dinfo(blk);
181 if (dinfo && dinfo->auto_del) {
182 monitor_remove_blk(blk);
187 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
189 int max_devs = if_max_devs[type];
190 return max_devs ? index / max_devs : 0;
193 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
195 int max_devs = if_max_devs[type];
196 return max_devs ? index % max_devs : index;
199 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
206 opts = qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
210 if (type != IF_DEFAULT) {
211 qemu_opt_set(opts, "if", if_name[type], &error_abort);
214 qemu_opt_set_number(opts, "index", index, &error_abort);
217 qemu_opt_set(opts, "file", file, &error_abort);
221 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
228 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
229 dinfo = blk_legacy_dinfo(blk);
230 if (dinfo && dinfo->type == type
231 && dinfo->bus == bus && dinfo->unit == unit) {
240 * Check board claimed all -drive that are meant to be claimed.
241 * Fatal error if any remain unclaimed.
243 void drive_check_orphaned(void)
248 bool orphans = false;
252 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
253 dinfo = blk_legacy_dinfo(blk);
255 * Ignore default drives, because we create certain default
256 * drives unconditionally, then leave them unclaimed. Not the
258 * Ignore IF_VIRTIO, because it gets desugared into -device,
259 * so we can leave failing to -device.
260 * Ignore IF_NONE, because leaving unclaimed IF_NONE remains
261 * available for device_add is a feature.
263 if (dinfo->is_default || dinfo->type == IF_VIRTIO
264 || dinfo->type == IF_NONE) {
267 if (!blk_get_attached_dev(blk)) {
269 qemu_opts_loc_restore(dinfo->opts);
270 error_report("machine type does not support"
271 " if=%s,bus=%d,unit=%d",
272 if_name[dinfo->type], dinfo->bus, dinfo->unit);
283 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
286 return drive_get(type,
287 drive_index_to_bus_id(type, index),
288 drive_index_to_unit_id(type, index));
291 int drive_get_max_bus(BlockInterfaceType type)
300 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
301 dinfo = blk_legacy_dinfo(blk);
302 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
303 max_bus = dinfo->bus;
309 static void bdrv_format_print(void *opaque, const char *name)
311 qemu_printf(" %s", name);
316 BlockDriverState *bs;
319 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
321 if (!strcmp(buf, "ignore")) {
322 return BLOCKDEV_ON_ERROR_IGNORE;
323 } else if (!is_read && !strcmp(buf, "enospc")) {
324 return BLOCKDEV_ON_ERROR_ENOSPC;
325 } else if (!strcmp(buf, "stop")) {
326 return BLOCKDEV_ON_ERROR_STOP;
327 } else if (!strcmp(buf, "report")) {
328 return BLOCKDEV_ON_ERROR_REPORT;
330 error_setg(errp, "'%s' invalid %s error action",
331 buf, is_read ? "read" : "write");
336 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
339 const QListEntry *entry;
340 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
341 switch (qobject_type(entry->value)) {
343 case QTYPE_QSTRING: {
344 unsigned long long length;
345 const char *str = qstring_get_str(qobject_to(QString,
347 if (parse_uint_full(str, &length, 10) == 0 &&
348 length > 0 && length <= UINT_MAX) {
349 block_acct_add_interval(stats, (unsigned) length);
351 error_setg(errp, "Invalid interval length: %s", str);
358 int64_t length = qnum_get_int(qobject_to(QNum, entry->value));
360 if (length > 0 && length <= UINT_MAX) {
361 block_acct_add_interval(stats, (unsigned) length);
363 error_setg(errp, "Invalid interval length: %" PRId64, length);
370 error_setg(errp, "The specification of stats-intervals is invalid");
377 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
379 /* All parameters but @opts are optional and may be set to NULL. */
380 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
381 const char **throttling_group, ThrottleConfig *throttle_cfg,
382 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
384 Error *local_error = NULL;
388 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
389 *bdrv_flags |= BDRV_O_COPY_ON_READ;
392 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
393 if (bdrv_parse_aio(aio, bdrv_flags) < 0) {
394 error_setg(errp, "invalid aio option");
400 /* disk I/O throttling */
401 if (throttling_group) {
402 *throttling_group = qemu_opt_get(opts, "throttling.group");
406 throttle_config_init(throttle_cfg);
407 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
408 qemu_opt_get_number(opts, "throttling.bps-total", 0);
409 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
410 qemu_opt_get_number(opts, "throttling.bps-read", 0);
411 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
412 qemu_opt_get_number(opts, "throttling.bps-write", 0);
413 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
414 qemu_opt_get_number(opts, "throttling.iops-total", 0);
415 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
416 qemu_opt_get_number(opts, "throttling.iops-read", 0);
417 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
418 qemu_opt_get_number(opts, "throttling.iops-write", 0);
420 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
421 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
422 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
423 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
424 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
425 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
426 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
427 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
428 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
429 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
430 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
431 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
433 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
434 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
435 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
436 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
437 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
438 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
439 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
440 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
441 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
442 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
443 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
444 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
446 throttle_cfg->op_size =
447 qemu_opt_get_number(opts, "throttling.iops-size", 0);
449 if (!throttle_is_valid(throttle_cfg, errp)) {
456 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
457 qemu_opt_get(opts, "detect-zeroes"),
458 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
461 error_propagate(errp, local_error);
467 static OnOffAuto account_get_opt(QemuOpts *opts, const char *name)
469 if (!qemu_opt_find(opts, name)) {
470 return ON_OFF_AUTO_AUTO;
472 if (qemu_opt_get_bool(opts, name, true)) {
473 return ON_OFF_AUTO_ON;
475 return ON_OFF_AUTO_OFF;
478 /* Takes the ownership of bs_opts */
479 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
484 int on_read_error, on_write_error;
485 OnOffAuto account_invalid, account_failed;
486 bool writethrough, read_only;
488 BlockDriverState *bs;
493 QDict *interval_dict = NULL;
494 QList *interval_list = NULL;
496 BlockdevDetectZeroesOptions detect_zeroes =
497 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
498 const char *throttling_group = NULL;
500 /* Check common options by copying from bs_opts to opts, all other options
501 * stay in bs_opts for processing by bdrv_open(). */
502 id = qdict_get_try_str(bs_opts, "id");
503 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, errp);
508 if (!qemu_opts_absorb_qdict(opts, bs_opts, errp)) {
513 qdict_del(bs_opts, "id");
516 /* extract parameters */
517 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
519 account_invalid = account_get_opt(opts, "stats-account-invalid");
520 account_failed = account_get_opt(opts, "stats-account-failed");
522 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
524 id = qemu_opts_id(opts);
526 qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
527 qdict_array_split(interval_dict, &interval_list);
529 if (qdict_size(interval_dict) != 0) {
530 error_setg(errp, "Invalid option stats-intervals.%s",
531 qdict_first(interval_dict)->key);
535 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
536 &detect_zeroes, &error);
538 error_propagate(errp, error);
542 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
543 if (is_help_option(buf)) {
544 qemu_printf("Supported formats:");
545 bdrv_iterate_format(bdrv_format_print, NULL, false);
546 qemu_printf("\nSupported formats (read-only):");
547 bdrv_iterate_format(bdrv_format_print, NULL, true);
552 if (qdict_haskey(bs_opts, "driver")) {
553 error_setg(errp, "Cannot specify both 'driver' and 'format'");
556 qdict_put_str(bs_opts, "driver", buf);
559 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
560 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
561 on_write_error = parse_block_error_action(buf, 0, &error);
563 error_propagate(errp, error);
568 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
569 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
570 on_read_error = parse_block_error_action(buf, 1, &error);
572 error_propagate(errp, error);
578 bdrv_flags |= BDRV_O_SNAPSHOT;
581 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
584 if ((!file || !*file) && !qdict_size(bs_opts)) {
585 BlockBackendRootState *blk_rs;
587 blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
588 blk_rs = blk_get_root_state(blk);
589 blk_rs->open_flags = bdrv_flags | (read_only ? 0 : BDRV_O_RDWR);
590 blk_rs->detect_zeroes = detect_zeroes;
592 qobject_unref(bs_opts);
594 if (file && !*file) {
598 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
599 * with other callers) rather than what we want as the real defaults.
600 * Apply the defaults here instead. */
601 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
602 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
603 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
604 read_only ? "on" : "off");
605 qdict_set_default_str(bs_opts, BDRV_OPT_AUTO_READ_ONLY, "on");
606 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
608 if (runstate_check(RUN_STATE_INMIGRATE)) {
609 bdrv_flags |= BDRV_O_INACTIVE;
612 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
618 bs->detect_zeroes = detect_zeroes;
620 block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
622 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
629 /* disk I/O throttling */
630 if (throttle_enabled(&cfg)) {
631 if (!throttling_group) {
632 throttling_group = id;
634 blk_io_limits_enable(blk, throttling_group);
635 blk_set_io_limits(blk, &cfg);
638 blk_set_enable_write_cache(blk, !writethrough);
639 blk_set_on_error(blk, on_read_error, on_write_error);
641 if (!monitor_add_blk(blk, id, errp)) {
649 qobject_unref(interval_dict);
650 qobject_unref(interval_list);
655 qobject_unref(interval_dict);
656 qobject_unref(interval_list);
658 qobject_unref(bs_opts);
662 /* Takes the ownership of bs_opts */
663 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
668 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
669 * with other callers) rather than what we want as the real defaults.
670 * Apply the defaults here instead. */
671 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
672 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
673 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
675 if (runstate_check(RUN_STATE_INMIGRATE)) {
676 bdrv_flags |= BDRV_O_INACTIVE;
679 return bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
682 void blockdev_close_all_bdrv_states(void)
684 BlockDriverState *bs, *next_bs;
687 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
688 AioContext *ctx = bdrv_get_aio_context(bs);
690 aio_context_acquire(ctx);
692 aio_context_release(ctx);
696 /* Iterates over the list of monitor-owned BlockDriverStates */
697 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
700 return bs ? QTAILQ_NEXT(bs, monitor_list)
701 : QTAILQ_FIRST(&monitor_bdrv_states);
704 static bool qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
709 value = qemu_opt_get(opts, from);
711 if (qemu_opt_find(opts, to)) {
712 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
713 "same time", to, from);
718 /* rename all items in opts */
719 while ((value = qemu_opt_get(opts, from))) {
720 qemu_opt_set(opts, to, value, &error_abort);
721 qemu_opt_unset(opts, from);
726 QemuOptsList qemu_legacy_drive_opts = {
728 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
732 .type = QEMU_OPT_NUMBER,
733 .help = "bus number",
736 .type = QEMU_OPT_NUMBER,
737 .help = "unit number (i.e. lun for scsi)",
740 .type = QEMU_OPT_NUMBER,
741 .help = "index number",
744 .type = QEMU_OPT_STRING,
745 .help = "media type (disk, cdrom)",
748 .type = QEMU_OPT_STRING,
749 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
752 .type = QEMU_OPT_STRING,
756 /* Options that are passed on, but have special semantics with -drive */
758 .name = BDRV_OPT_READ_ONLY,
759 .type = QEMU_OPT_BOOL,
760 .help = "open drive file as read-only",
763 .type = QEMU_OPT_STRING,
764 .help = "read error action",
767 .type = QEMU_OPT_STRING,
768 .help = "write error action",
770 .name = "copy-on-read",
771 .type = QEMU_OPT_BOOL,
772 .help = "copy read data from backing file into image file",
775 { /* end of list */ }
779 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
784 DriveInfo *dinfo = NULL;
786 QemuOpts *legacy_opts;
787 DriveMediaType media = MEDIA_DISK;
788 BlockInterfaceType type;
789 int max_devs, bus_id, unit_id, index;
790 const char *werror, *rerror;
791 bool read_only = false;
793 const char *filename;
798 /* Change legacy command line options into QMP ones */
799 static const struct {
803 { "iops", "throttling.iops-total" },
804 { "iops_rd", "throttling.iops-read" },
805 { "iops_wr", "throttling.iops-write" },
807 { "bps", "throttling.bps-total" },
808 { "bps_rd", "throttling.bps-read" },
809 { "bps_wr", "throttling.bps-write" },
811 { "iops_max", "throttling.iops-total-max" },
812 { "iops_rd_max", "throttling.iops-read-max" },
813 { "iops_wr_max", "throttling.iops-write-max" },
815 { "bps_max", "throttling.bps-total-max" },
816 { "bps_rd_max", "throttling.bps-read-max" },
817 { "bps_wr_max", "throttling.bps-write-max" },
819 { "iops_size", "throttling.iops-size" },
821 { "group", "throttling.group" },
823 { "readonly", BDRV_OPT_READ_ONLY },
826 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
827 if (!qemu_opt_rename(all_opts, opt_renames[i].from,
828 opt_renames[i].to, errp)) {
833 value = qemu_opt_get(all_opts, "cache");
838 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
839 error_setg(errp, "invalid cache option");
843 /* Specific options take precedence */
844 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
845 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
846 !writethrough, &error_abort);
848 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
849 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
850 !!(flags & BDRV_O_NOCACHE), &error_abort);
852 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
853 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
854 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
856 qemu_opt_unset(all_opts, "cache");
859 /* Get a QDict for processing the options */
860 bs_opts = qdict_new();
861 qemu_opts_to_qdict(all_opts, bs_opts);
863 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
865 if (!qemu_opts_absorb_qdict(legacy_opts, bs_opts, errp)) {
870 value = qemu_opt_get(legacy_opts, "media");
872 if (!strcmp(value, "disk")) {
874 } else if (!strcmp(value, "cdrom")) {
878 error_setg(errp, "'%s' invalid media", value);
883 /* copy-on-read is disabled with a warning for read-only devices */
884 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
885 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
887 if (read_only && copy_on_read) {
888 warn_report("disabling copy-on-read on read-only drive");
889 copy_on_read = false;
892 qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
893 qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
895 /* Controller type */
896 value = qemu_opt_get(legacy_opts, "if");
899 type < IF_COUNT && strcmp(value, if_name[type]);
902 if (type == IF_COUNT) {
903 error_setg(errp, "unsupported bus type '%s'", value);
907 type = block_default_type;
910 /* Device address specified by bus/unit or index.
911 * If none was specified, try to find the first free one. */
912 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
913 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
914 index = qemu_opt_get_number(legacy_opts, "index", -1);
916 max_devs = if_max_devs[type];
919 if (bus_id != 0 || unit_id != -1) {
920 error_setg(errp, "index cannot be used with bus and unit");
923 bus_id = drive_index_to_bus_id(type, index);
924 unit_id = drive_index_to_unit_id(type, index);
929 while (drive_get(type, bus_id, unit_id) != NULL) {
931 if (max_devs && unit_id >= max_devs) {
938 if (max_devs && unit_id >= max_devs) {
939 error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
943 if (drive_get(type, bus_id, unit_id) != NULL) {
944 error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
945 bus_id, unit_id, index);
949 /* no id supplied -> create one */
950 if (qemu_opts_id(all_opts) == NULL) {
952 const char *mediastr = "";
953 if (type == IF_IDE || type == IF_SCSI) {
954 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
957 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
960 new_id = g_strdup_printf("%s%s%i", if_name[type],
963 qdict_put_str(bs_opts, "id", new_id);
967 /* Add virtio block device */
968 if (type == IF_VIRTIO) {
970 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
972 qemu_opt_set(devopts, "driver", "virtio-blk", &error_abort);
973 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
977 filename = qemu_opt_get(legacy_opts, "file");
979 /* Check werror/rerror compatibility with if=... */
980 werror = qemu_opt_get(legacy_opts, "werror");
981 if (werror != NULL) {
982 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
984 error_setg(errp, "werror is not supported by this bus type");
987 qdict_put_str(bs_opts, "werror", werror);
990 rerror = qemu_opt_get(legacy_opts, "rerror");
991 if (rerror != NULL) {
992 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
994 error_setg(errp, "rerror is not supported by this bus type");
997 qdict_put_str(bs_opts, "rerror", rerror);
1000 /* Actual block device init: Functionality shared with blockdev-add */
1001 blk = blockdev_init(filename, bs_opts, errp);
1007 /* Create legacy DriveInfo */
1008 dinfo = g_malloc0(sizeof(*dinfo));
1009 dinfo->opts = all_opts;
1012 dinfo->bus = bus_id;
1013 dinfo->unit = unit_id;
1015 blk_set_legacy_dinfo(blk, dinfo);
1022 dinfo->media_cd = media == MEDIA_CDROM;
1029 qemu_opts_del(legacy_opts);
1030 qobject_unref(bs_opts);
1034 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1036 BlockDriverState *bs;
1037 AioContext *aio_context;
1039 bs = bdrv_lookup_bs(name, name, errp);
1044 if (!bdrv_is_root_node(bs)) {
1045 error_setg(errp, "Need a root block node");
1049 aio_context = bdrv_get_aio_context(bs);
1050 aio_context_acquire(aio_context);
1052 if (!bdrv_is_inserted(bs)) {
1053 error_setg(errp, "Device has no medium");
1057 aio_context_release(aio_context);
1062 static void blockdev_do_action(TransactionAction *action, Error **errp)
1064 TransactionActionList list;
1066 list.value = action;
1068 qmp_transaction(&list, NULL, errp);
1071 void qmp_blockdev_snapshot_sync(const char *device, const char *node_name,
1072 const char *snapshot_file,
1073 const char *snapshot_node_name,
1075 bool has_mode, NewImageMode mode, Error **errp)
1077 BlockdevSnapshotSync snapshot = {
1078 .device = (char *) device,
1079 .node_name = (char *) node_name,
1080 .snapshot_file = (char *) snapshot_file,
1081 .snapshot_node_name = (char *) snapshot_node_name,
1082 .format = (char *) format,
1083 .has_mode = has_mode,
1086 TransactionAction action = {
1087 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1088 .u.blockdev_snapshot_sync.data = &snapshot,
1090 blockdev_do_action(&action, errp);
1093 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1096 BlockdevSnapshot snapshot_data = {
1097 .node = (char *) node,
1098 .overlay = (char *) overlay
1100 TransactionAction action = {
1101 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1102 .u.blockdev_snapshot.data = &snapshot_data,
1104 blockdev_do_action(&action, errp);
1107 void qmp_blockdev_snapshot_internal_sync(const char *device,
1111 BlockdevSnapshotInternal snapshot = {
1112 .device = (char *) device,
1113 .name = (char *) name
1115 TransactionAction action = {
1116 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1117 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1119 blockdev_do_action(&action, errp);
1122 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1127 BlockDriverState *bs;
1128 AioContext *aio_context;
1129 QEMUSnapshotInfo sn;
1130 Error *local_err = NULL;
1131 SnapshotInfo *info = NULL;
1134 bs = qmp_get_root_bs(device, errp);
1138 aio_context = bdrv_get_aio_context(bs);
1139 aio_context_acquire(aio_context);
1142 error_setg(errp, "Name or id must be provided");
1143 goto out_aio_context;
1146 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1147 goto out_aio_context;
1150 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1152 error_propagate(errp, local_err);
1153 goto out_aio_context;
1157 "Snapshot with id '%s' and name '%s' does not exist on "
1159 STR_OR_NULL(id), STR_OR_NULL(name), device);
1160 goto out_aio_context;
1163 bdrv_snapshot_delete(bs, id, name, &local_err);
1165 error_propagate(errp, local_err);
1166 goto out_aio_context;
1169 aio_context_release(aio_context);
1171 info = g_new0(SnapshotInfo, 1);
1172 info->id = g_strdup(sn.id_str);
1173 info->name = g_strdup(sn.name);
1174 info->date_nsec = sn.date_nsec;
1175 info->date_sec = sn.date_sec;
1176 info->vm_state_size = sn.vm_state_size;
1177 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1178 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1179 if (sn.icount != -1ULL) {
1180 info->icount = sn.icount;
1181 info->has_icount = true;
1187 aio_context_release(aio_context);
1191 /* New and old BlockDriverState structs for atomic group operations */
1193 typedef struct BlkActionState BlkActionState;
1197 * Table of operations that define an Action.
1199 * @instance_size: Size of state struct, in bytes.
1200 * @prepare: Prepare the work, must NOT be NULL.
1201 * @commit: Commit the changes, can be NULL.
1202 * @abort: Abort the changes on fail, can be NULL.
1203 * @clean: Clean up resources after all transaction actions have called
1204 * commit() or abort(). Can be NULL.
1206 * Only prepare() may fail. In a single transaction, only one of commit() or
1207 * abort() will be called. clean() will always be called if it is present.
1209 * Always run under BQL.
1211 typedef struct BlkActionOps {
1212 size_t instance_size;
1213 void (*action)(BlkActionState *common, Transaction *tran, Error **errp);
1218 * Describes one Action's state within a Transaction.
1220 * @action: QAPI-defined enum identifying which Action to perform.
1221 * @ops: Table of ActionOps this Action can perform.
1222 * @block_job_txn: Transaction which this action belongs to.
1223 * @entry: List membership for all Actions in this Transaction.
1225 * This structure must be arranged as first member in a subclassed type,
1226 * assuming that the compiler will also arrange it to the same offsets as the
1229 struct BlkActionState {
1230 TransactionAction *action;
1231 const BlkActionOps *ops;
1232 JobTxn *block_job_txn;
1233 TransactionProperties *txn_props;
1234 QTAILQ_ENTRY(BlkActionState) entry;
1237 /* internal snapshot private data */
1238 typedef struct InternalSnapshotState {
1239 BlkActionState common;
1240 BlockDriverState *bs;
1241 QEMUSnapshotInfo sn;
1243 } InternalSnapshotState;
1245 static void internal_snapshot_abort(void *opaque);
1246 static void internal_snapshot_clean(void *opaque);
1247 TransactionActionDrv internal_snapshot_drv = {
1248 .abort = internal_snapshot_abort,
1249 .clean = internal_snapshot_clean,
1252 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1254 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1256 "Action '%s' does not support Transaction property "
1257 "completion-mode = %s",
1258 TransactionActionKind_str(s->action->type),
1259 ActionCompletionMode_str(s->txn_props->completion_mode));
1265 static void internal_snapshot_action(BlkActionState *common,
1266 Transaction *tran, Error **errp)
1268 Error *local_err = NULL;
1271 BlockDriverState *bs;
1272 QEMUSnapshotInfo old_sn, *sn;
1275 BlockdevSnapshotInternal *internal;
1276 InternalSnapshotState *state;
1277 AioContext *aio_context;
1280 g_assert(common->action->type ==
1281 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1282 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1283 state = DO_UPCAST(InternalSnapshotState, common, common);
1285 tran_add(tran, &internal_snapshot_drv, state);
1287 /* 1. parse input */
1288 device = internal->device;
1289 name = internal->name;
1291 /* 2. check for validation */
1292 if (action_check_completion_mode(common, errp) < 0) {
1296 bs = qmp_get_root_bs(device, errp);
1301 aio_context = bdrv_get_aio_context(bs);
1302 aio_context_acquire(aio_context);
1306 /* Paired with .clean() */
1307 bdrv_drained_begin(bs);
1309 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1313 if (bdrv_is_read_only(bs)) {
1314 error_setg(errp, "Device '%s' is read only", device);
1318 if (!bdrv_can_snapshot(bs)) {
1319 error_setg(errp, "Block format '%s' used by device '%s' "
1320 "does not support internal snapshots",
1321 bs->drv->format_name, device);
1325 if (!strlen(name)) {
1326 error_setg(errp, "Name is empty");
1330 /* check whether a snapshot with name exist */
1331 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1334 error_propagate(errp, local_err);
1338 "Snapshot with name '%s' already exists on device '%s'",
1343 /* 3. take the snapshot */
1345 pstrcpy(sn->name, sizeof(sn->name), name);
1346 rt = g_get_real_time();
1347 sn->date_sec = rt / G_USEC_PER_SEC;
1348 sn->date_nsec = (rt % G_USEC_PER_SEC) * 1000;
1349 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1350 if (replay_mode != REPLAY_MODE_NONE) {
1351 sn->icount = replay_get_current_icount();
1356 ret1 = bdrv_snapshot_create(bs, sn);
1358 error_setg_errno(errp, -ret1,
1359 "Failed to create snapshot '%s' on device '%s'",
1364 /* 4. succeed, mark a snapshot is created */
1365 state->created = true;
1368 aio_context_release(aio_context);
1371 static void internal_snapshot_abort(void *opaque)
1373 InternalSnapshotState *state = opaque;
1374 BlockDriverState *bs = state->bs;
1375 QEMUSnapshotInfo *sn = &state->sn;
1376 AioContext *aio_context;
1377 Error *local_error = NULL;
1379 if (!state->created) {
1383 aio_context = bdrv_get_aio_context(state->bs);
1384 aio_context_acquire(aio_context);
1386 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1387 error_reportf_err(local_error,
1388 "Failed to delete snapshot with id '%s' and "
1389 "name '%s' on device '%s' in abort: ",
1390 sn->id_str, sn->name,
1391 bdrv_get_device_name(bs));
1394 aio_context_release(aio_context);
1397 static void internal_snapshot_clean(void *opaque)
1399 g_autofree InternalSnapshotState *state = opaque;
1400 AioContext *aio_context;
1406 aio_context = bdrv_get_aio_context(state->bs);
1407 aio_context_acquire(aio_context);
1409 bdrv_drained_end(state->bs);
1411 aio_context_release(aio_context);
1414 /* external snapshot private data */
1415 typedef struct ExternalSnapshotState {
1416 BlkActionState common;
1417 BlockDriverState *old_bs;
1418 BlockDriverState *new_bs;
1419 bool overlay_appended;
1420 } ExternalSnapshotState;
1422 static void external_snapshot_commit(void *opaque);
1423 static void external_snapshot_abort(void *opaque);
1424 static void external_snapshot_clean(void *opaque);
1425 TransactionActionDrv external_snapshot_drv = {
1426 .commit = external_snapshot_commit,
1427 .abort = external_snapshot_abort,
1428 .clean = external_snapshot_clean,
1431 static void external_snapshot_action(BlkActionState *common, Transaction *tran,
1436 QDict *options = NULL;
1437 Error *local_err = NULL;
1438 /* Device and node name of the image to generate the snapshot from */
1440 const char *node_name;
1441 /* Reference to the new image (for 'blockdev-snapshot') */
1442 const char *snapshot_ref;
1443 /* File name of the new image (for 'blockdev-snapshot-sync') */
1444 const char *new_image_file;
1445 ExternalSnapshotState *state =
1446 DO_UPCAST(ExternalSnapshotState, common, common);
1447 TransactionAction *action = common->action;
1448 AioContext *aio_context;
1449 uint64_t perm, shared;
1451 tran_add(tran, &external_snapshot_drv, state);
1453 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1454 * purpose but a different set of parameters */
1455 switch (action->type) {
1456 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1458 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1460 node_name = s->node;
1461 new_image_file = NULL;
1462 snapshot_ref = s->overlay;
1465 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1467 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1469 node_name = s->node_name;
1470 new_image_file = s->snapshot_file;
1471 snapshot_ref = NULL;
1475 g_assert_not_reached();
1478 /* start processing */
1479 if (action_check_completion_mode(common, errp) < 0) {
1483 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1484 if (!state->old_bs) {
1488 aio_context = bdrv_get_aio_context(state->old_bs);
1489 aio_context_acquire(aio_context);
1491 /* Paired with .clean() */
1492 bdrv_drained_begin(state->old_bs);
1494 if (!bdrv_is_inserted(state->old_bs)) {
1495 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1499 if (bdrv_op_is_blocked(state->old_bs,
1500 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1504 if (!bdrv_is_read_only(state->old_bs)) {
1505 if (bdrv_flush(state->old_bs)) {
1506 error_setg(errp, QERR_IO_ERROR);
1511 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1512 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1513 const char *format = s->format ?: "qcow2";
1514 enum NewImageMode mode;
1515 const char *snapshot_node_name = s->snapshot_node_name;
1517 if (node_name && !snapshot_node_name) {
1518 error_setg(errp, "New overlay node-name missing");
1522 if (snapshot_node_name &&
1523 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1524 error_setg(errp, "New overlay node-name already in use");
1528 flags = state->old_bs->open_flags;
1529 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1530 flags |= BDRV_O_NO_BACKING;
1532 /* create new image w/backing file */
1533 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1534 if (mode != NEW_IMAGE_MODE_EXISTING) {
1535 int64_t size = bdrv_getlength(state->old_bs);
1537 error_setg_errno(errp, -size, "bdrv_getlength failed");
1540 bdrv_refresh_filename(state->old_bs);
1542 aio_context_release(aio_context);
1543 bdrv_img_create(new_image_file, format,
1544 state->old_bs->filename,
1545 state->old_bs->drv->format_name,
1546 NULL, size, flags, false, &local_err);
1547 aio_context_acquire(aio_context);
1550 error_propagate(errp, local_err);
1555 options = qdict_new();
1556 if (snapshot_node_name) {
1557 qdict_put_str(options, "node-name", snapshot_node_name);
1559 qdict_put_str(options, "driver", format);
1562 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1564 /* We will manually add the backing_hd field to the bs later */
1565 if (!state->new_bs) {
1570 * Allow attaching a backing file to an overlay that's already in use only
1571 * if the parents don't assume that they are already seeing a valid image.
1572 * (Specifically, allow it as a mirror target, which is write-only access.)
1574 bdrv_get_cumulative_perm(state->new_bs, &perm, &shared);
1575 if (perm & BLK_PERM_CONSISTENT_READ) {
1576 error_setg(errp, "The overlay is already in use");
1580 if (state->new_bs->drv->is_filter) {
1581 error_setg(errp, "Filters cannot be used as overlays");
1585 if (bdrv_cow_child(state->new_bs)) {
1586 error_setg(errp, "The overlay already has a backing image");
1590 if (!state->new_bs->drv->supports_backing) {
1591 error_setg(errp, "The overlay does not support backing images");
1595 ret = bdrv_append(state->new_bs, state->old_bs, errp);
1599 state->overlay_appended = true;
1602 aio_context_release(aio_context);
1605 static void external_snapshot_commit(void *opaque)
1607 ExternalSnapshotState *state = opaque;
1608 AioContext *aio_context;
1610 aio_context = bdrv_get_aio_context(state->old_bs);
1611 aio_context_acquire(aio_context);
1613 /* We don't need (or want) to use the transactional
1614 * bdrv_reopen_multiple() across all the entries at once, because we
1615 * don't want to abort all of them if one of them fails the reopen */
1616 if (!qatomic_read(&state->old_bs->copy_on_read)) {
1617 bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1620 aio_context_release(aio_context);
1623 static void external_snapshot_abort(void *opaque)
1625 ExternalSnapshotState *state = opaque;
1626 if (state->new_bs) {
1627 if (state->overlay_appended) {
1628 AioContext *aio_context;
1629 AioContext *tmp_context;
1632 aio_context = bdrv_get_aio_context(state->old_bs);
1633 aio_context_acquire(aio_context);
1635 bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd()
1636 close state->old_bs; we need it */
1637 bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1640 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1641 * the main AioContext. As we're still going to be using it, return
1642 * it to the AioContext it was before.
1644 tmp_context = bdrv_get_aio_context(state->old_bs);
1645 if (aio_context != tmp_context) {
1646 aio_context_release(aio_context);
1647 aio_context_acquire(tmp_context);
1649 ret = bdrv_try_change_aio_context(state->old_bs,
1650 aio_context, NULL, NULL);
1653 aio_context_release(tmp_context);
1654 aio_context_acquire(aio_context);
1657 bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1658 bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1660 aio_context_release(aio_context);
1665 static void external_snapshot_clean(void *opaque)
1667 g_autofree ExternalSnapshotState *state = opaque;
1668 AioContext *aio_context;
1670 if (!state->old_bs) {
1674 aio_context = bdrv_get_aio_context(state->old_bs);
1675 aio_context_acquire(aio_context);
1677 bdrv_drained_end(state->old_bs);
1678 bdrv_unref(state->new_bs);
1680 aio_context_release(aio_context);
1683 typedef struct DriveBackupState {
1684 BlkActionState common;
1685 BlockDriverState *bs;
1689 static BlockJob *do_backup_common(BackupCommon *backup,
1690 BlockDriverState *bs,
1691 BlockDriverState *target_bs,
1692 AioContext *aio_context,
1693 JobTxn *txn, Error **errp);
1695 static void drive_backup_commit(void *opaque);
1696 static void drive_backup_abort(void *opaque);
1697 static void drive_backup_clean(void *opaque);
1698 TransactionActionDrv drive_backup_drv = {
1699 .commit = drive_backup_commit,
1700 .abort = drive_backup_abort,
1701 .clean = drive_backup_clean,
1704 static void drive_backup_action(BlkActionState *common, Transaction *tran,
1707 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1708 DriveBackup *backup;
1709 BlockDriverState *bs;
1710 BlockDriverState *target_bs;
1711 BlockDriverState *source = NULL;
1712 AioContext *aio_context;
1713 AioContext *old_context;
1716 Error *local_err = NULL;
1719 bool set_backing_hd = false;
1722 tran_add(tran, &drive_backup_drv, state);
1724 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1725 backup = common->action->u.drive_backup.data;
1727 if (!backup->has_mode) {
1728 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1731 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1737 error_setg(errp, "Device has no medium");
1741 aio_context = bdrv_get_aio_context(bs);
1742 aio_context_acquire(aio_context);
1745 /* Paired with .clean() */
1746 bdrv_drained_begin(bs);
1748 format = backup->format;
1749 if (!format && backup->mode != NEW_IMAGE_MODE_EXISTING) {
1750 format = bs->drv->format_name;
1753 /* Early check to avoid creating target */
1754 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
1758 flags = bs->open_flags | BDRV_O_RDWR;
1761 * See if we have a backing HD we can use to create our new image
1764 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
1766 * Backup will not replace the source by the target, so none
1767 * of the filters skipped here will be removed (in contrast to
1768 * mirror). Therefore, we can skip all of them when looking
1769 * for the first COW relationship.
1771 source = bdrv_cow_bs(bdrv_skip_filters(bs));
1773 backup->sync = MIRROR_SYNC_MODE_FULL;
1776 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
1778 flags |= BDRV_O_NO_BACKING;
1779 set_backing_hd = true;
1782 size = bdrv_getlength(bs);
1784 error_setg_errno(errp, -size, "bdrv_getlength failed");
1788 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
1791 /* Implicit filters should not appear in the filename */
1792 BlockDriverState *explicit_backing =
1793 bdrv_skip_implicit_filters(source);
1795 bdrv_refresh_filename(explicit_backing);
1796 bdrv_img_create(backup->target, format,
1797 explicit_backing->filename,
1798 explicit_backing->drv->format_name, NULL,
1799 size, flags, false, &local_err);
1801 bdrv_img_create(backup->target, format, NULL, NULL, NULL,
1802 size, flags, false, &local_err);
1807 error_propagate(errp, local_err);
1811 options = qdict_new();
1812 qdict_put_str(options, "discard", "unmap");
1813 qdict_put_str(options, "detect-zeroes", "unmap");
1815 qdict_put_str(options, "driver", format);
1818 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
1823 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1824 old_context = bdrv_get_aio_context(target_bs);
1825 aio_context_release(aio_context);
1826 aio_context_acquire(old_context);
1828 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
1830 bdrv_unref(target_bs);
1831 aio_context_release(old_context);
1835 aio_context_release(old_context);
1836 aio_context_acquire(aio_context);
1838 if (set_backing_hd) {
1839 if (bdrv_set_backing_hd(target_bs, source, errp) < 0) {
1844 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1845 bs, target_bs, aio_context,
1846 common->block_job_txn, errp);
1849 bdrv_unref(target_bs);
1851 aio_context_release(aio_context);
1854 static void drive_backup_commit(void *opaque)
1856 DriveBackupState *state = opaque;
1857 AioContext *aio_context;
1859 aio_context = bdrv_get_aio_context(state->bs);
1860 aio_context_acquire(aio_context);
1863 job_start(&state->job->job);
1865 aio_context_release(aio_context);
1868 static void drive_backup_abort(void *opaque)
1870 DriveBackupState *state = opaque;
1873 job_cancel_sync(&state->job->job, true);
1877 static void drive_backup_clean(void *opaque)
1879 g_autofree DriveBackupState *state = opaque;
1880 AioContext *aio_context;
1886 aio_context = bdrv_get_aio_context(state->bs);
1887 aio_context_acquire(aio_context);
1889 bdrv_drained_end(state->bs);
1891 aio_context_release(aio_context);
1894 typedef struct BlockdevBackupState {
1895 BlkActionState common;
1896 BlockDriverState *bs;
1898 } BlockdevBackupState;
1900 static void blockdev_backup_commit(void *opaque);
1901 static void blockdev_backup_abort(void *opaque);
1902 static void blockdev_backup_clean(void *opaque);
1903 TransactionActionDrv blockdev_backup_drv = {
1904 .commit = blockdev_backup_commit,
1905 .abort = blockdev_backup_abort,
1906 .clean = blockdev_backup_clean,
1909 static void blockdev_backup_action(BlkActionState *common, Transaction *tran,
1912 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1913 BlockdevBackup *backup;
1914 BlockDriverState *bs;
1915 BlockDriverState *target_bs;
1916 AioContext *aio_context;
1917 AioContext *old_context;
1920 tran_add(tran, &blockdev_backup_drv, state);
1922 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1923 backup = common->action->u.blockdev_backup.data;
1925 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1930 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1935 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1936 aio_context = bdrv_get_aio_context(bs);
1937 old_context = bdrv_get_aio_context(target_bs);
1938 aio_context_acquire(old_context);
1940 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
1942 aio_context_release(old_context);
1946 aio_context_release(old_context);
1947 aio_context_acquire(aio_context);
1950 /* Paired with .clean() */
1951 bdrv_drained_begin(state->bs);
1953 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1954 bs, target_bs, aio_context,
1955 common->block_job_txn, errp);
1957 aio_context_release(aio_context);
1960 static void blockdev_backup_commit(void *opaque)
1962 BlockdevBackupState *state = opaque;
1963 AioContext *aio_context;
1965 aio_context = bdrv_get_aio_context(state->bs);
1966 aio_context_acquire(aio_context);
1969 job_start(&state->job->job);
1971 aio_context_release(aio_context);
1974 static void blockdev_backup_abort(void *opaque)
1976 BlockdevBackupState *state = opaque;
1979 job_cancel_sync(&state->job->job, true);
1983 static void blockdev_backup_clean(void *opaque)
1985 g_autofree BlockdevBackupState *state = opaque;
1986 AioContext *aio_context;
1992 aio_context = bdrv_get_aio_context(state->bs);
1993 aio_context_acquire(aio_context);
1995 bdrv_drained_end(state->bs);
1997 aio_context_release(aio_context);
2000 typedef struct BlockDirtyBitmapState {
2001 BlkActionState common;
2002 BdrvDirtyBitmap *bitmap;
2003 BlockDriverState *bs;
2007 } BlockDirtyBitmapState;
2009 static void block_dirty_bitmap_add_abort(void *opaque);
2010 TransactionActionDrv block_dirty_bitmap_add_drv = {
2011 .abort = block_dirty_bitmap_add_abort,
2015 static void block_dirty_bitmap_add_action(BlkActionState *common,
2016 Transaction *tran, Error **errp)
2018 Error *local_err = NULL;
2019 BlockDirtyBitmapAdd *action;
2020 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2023 tran_add(tran, &block_dirty_bitmap_add_drv, state);
2025 if (action_check_completion_mode(common, errp) < 0) {
2029 action = common->action->u.block_dirty_bitmap_add.data;
2030 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
2031 qmp_block_dirty_bitmap_add(action->node, action->name,
2032 action->has_granularity, action->granularity,
2033 action->has_persistent, action->persistent,
2034 action->has_disabled, action->disabled,
2038 state->prepared = true;
2040 error_propagate(errp, local_err);
2044 static void block_dirty_bitmap_add_abort(void *opaque)
2046 BlockDirtyBitmapAdd *action;
2047 BlockDirtyBitmapState *state = opaque;
2049 action = state->common.action->u.block_dirty_bitmap_add.data;
2050 /* Should not be able to fail: IF the bitmap was added via .prepare(),
2051 * then the node reference and bitmap name must have been valid.
2053 if (state->prepared) {
2054 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2058 static void block_dirty_bitmap_restore(void *opaque);
2059 static void block_dirty_bitmap_free_backup(void *opaque);
2060 TransactionActionDrv block_dirty_bitmap_clear_drv = {
2061 .abort = block_dirty_bitmap_restore,
2062 .commit = block_dirty_bitmap_free_backup,
2066 static void block_dirty_bitmap_clear_action(BlkActionState *common,
2067 Transaction *tran, Error **errp)
2069 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2071 BlockDirtyBitmap *action;
2073 tran_add(tran, &block_dirty_bitmap_clear_drv, state);
2075 if (action_check_completion_mode(common, errp) < 0) {
2079 action = common->action->u.block_dirty_bitmap_clear.data;
2080 state->bitmap = block_dirty_bitmap_lookup(action->node,
2084 if (!state->bitmap) {
2088 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2092 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2095 static void block_dirty_bitmap_restore(void *opaque)
2097 BlockDirtyBitmapState *state = opaque;
2099 if (state->backup) {
2100 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2104 static void block_dirty_bitmap_free_backup(void *opaque)
2106 BlockDirtyBitmapState *state = opaque;
2108 hbitmap_free(state->backup);
2111 static void block_dirty_bitmap_enable_abort(void *opaque);
2112 TransactionActionDrv block_dirty_bitmap_enable_drv = {
2113 .abort = block_dirty_bitmap_enable_abort,
2117 static void block_dirty_bitmap_enable_action(BlkActionState *common,
2118 Transaction *tran, Error **errp)
2120 BlockDirtyBitmap *action;
2121 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2124 tran_add(tran, &block_dirty_bitmap_enable_drv, state);
2126 if (action_check_completion_mode(common, errp) < 0) {
2130 action = common->action->u.block_dirty_bitmap_enable.data;
2131 state->bitmap = block_dirty_bitmap_lookup(action->node,
2135 if (!state->bitmap) {
2139 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2143 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2144 bdrv_enable_dirty_bitmap(state->bitmap);
2147 static void block_dirty_bitmap_enable_abort(void *opaque)
2149 BlockDirtyBitmapState *state = opaque;
2151 if (!state->was_enabled) {
2152 bdrv_disable_dirty_bitmap(state->bitmap);
2156 static void block_dirty_bitmap_disable_abort(void *opaque);
2157 TransactionActionDrv block_dirty_bitmap_disable_drv = {
2158 .abort = block_dirty_bitmap_disable_abort,
2162 static void block_dirty_bitmap_disable_action(BlkActionState *common,
2163 Transaction *tran, Error **errp)
2165 BlockDirtyBitmap *action;
2166 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2169 tran_add(tran, &block_dirty_bitmap_disable_drv, state);
2171 if (action_check_completion_mode(common, errp) < 0) {
2175 action = common->action->u.block_dirty_bitmap_disable.data;
2176 state->bitmap = block_dirty_bitmap_lookup(action->node,
2180 if (!state->bitmap) {
2184 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2188 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2189 bdrv_disable_dirty_bitmap(state->bitmap);
2192 static void block_dirty_bitmap_disable_abort(void *opaque)
2194 BlockDirtyBitmapState *state = opaque;
2196 if (state->was_enabled) {
2197 bdrv_enable_dirty_bitmap(state->bitmap);
2201 TransactionActionDrv block_dirty_bitmap_merge_drv = {
2202 .commit = block_dirty_bitmap_free_backup,
2203 .abort = block_dirty_bitmap_restore,
2207 static void block_dirty_bitmap_merge_action(BlkActionState *common,
2208 Transaction *tran, Error **errp)
2210 BlockDirtyBitmapMerge *action;
2211 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2214 tran_add(tran, &block_dirty_bitmap_merge_drv, state);
2216 if (action_check_completion_mode(common, errp) < 0) {
2220 action = common->action->u.block_dirty_bitmap_merge.data;
2222 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2223 action->bitmaps, &state->backup,
2227 static void block_dirty_bitmap_remove_commit(void *opaque);
2228 static void block_dirty_bitmap_remove_abort(void *opaque);
2229 TransactionActionDrv block_dirty_bitmap_remove_drv = {
2230 .commit = block_dirty_bitmap_remove_commit,
2231 .abort = block_dirty_bitmap_remove_abort,
2235 static void block_dirty_bitmap_remove_action(BlkActionState *common,
2236 Transaction *tran, Error **errp)
2238 BlockDirtyBitmap *action;
2239 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2242 tran_add(tran, &block_dirty_bitmap_remove_drv, state);
2244 if (action_check_completion_mode(common, errp) < 0) {
2248 action = common->action->u.block_dirty_bitmap_remove.data;
2250 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2251 false, &state->bs, errp);
2252 if (state->bitmap) {
2253 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2254 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2258 static void block_dirty_bitmap_remove_abort(void *opaque)
2260 BlockDirtyBitmapState *state = opaque;
2262 if (state->bitmap) {
2263 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2264 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2268 static void block_dirty_bitmap_remove_commit(void *opaque)
2270 BlockDirtyBitmapState *state = opaque;
2272 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2273 bdrv_release_dirty_bitmap(state->bitmap);
2276 static void abort_commit(void *opaque);
2277 TransactionActionDrv abort_drv = {
2278 .commit = abort_commit,
2282 static void abort_action(BlkActionState *common, Transaction *tran,
2285 tran_add(tran, &abort_drv, common);
2286 error_setg(errp, "Transaction aborted using Abort action");
2289 static void abort_commit(void *opaque)
2291 g_assert_not_reached(); /* this action never succeeds */
2294 static const BlkActionOps actions[] = {
2295 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2296 .instance_size = sizeof(ExternalSnapshotState),
2297 .action = external_snapshot_action,
2299 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2300 .instance_size = sizeof(ExternalSnapshotState),
2301 .action = external_snapshot_action,
2303 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2304 .instance_size = sizeof(DriveBackupState),
2305 .action = drive_backup_action,
2307 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2308 .instance_size = sizeof(BlockdevBackupState),
2309 .action = blockdev_backup_action,
2311 [TRANSACTION_ACTION_KIND_ABORT] = {
2312 .instance_size = sizeof(BlkActionState),
2313 .action = abort_action,
2315 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2316 .instance_size = sizeof(InternalSnapshotState),
2317 .action = internal_snapshot_action,
2319 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2320 .instance_size = sizeof(BlockDirtyBitmapState),
2321 .action = block_dirty_bitmap_add_action,
2323 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2324 .instance_size = sizeof(BlockDirtyBitmapState),
2325 .action = block_dirty_bitmap_clear_action,
2327 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2328 .instance_size = sizeof(BlockDirtyBitmapState),
2329 .action = block_dirty_bitmap_enable_action,
2331 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2332 .instance_size = sizeof(BlockDirtyBitmapState),
2333 .action = block_dirty_bitmap_disable_action,
2335 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2336 .instance_size = sizeof(BlockDirtyBitmapState),
2337 .action = block_dirty_bitmap_merge_action,
2339 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE] = {
2340 .instance_size = sizeof(BlockDirtyBitmapState),
2341 .action = block_dirty_bitmap_remove_action,
2343 /* Where are transactions for MIRROR, COMMIT and STREAM?
2344 * Although these blockjobs use transaction callbacks like the backup job,
2345 * these jobs do not necessarily adhere to transaction semantics.
2346 * These jobs may not fully undo all of their actions on abort, nor do they
2347 * necessarily work in transactions with more than one job in them.
2352 * Allocate a TransactionProperties structure if necessary, and fill
2353 * that structure with desired defaults if they are unset.
2355 static TransactionProperties *get_transaction_properties(
2356 TransactionProperties *props)
2359 props = g_new0(TransactionProperties, 1);
2362 if (!props->has_completion_mode) {
2363 props->has_completion_mode = true;
2364 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2371 * 'Atomic' group operations. The operations are performed as a set, and if
2372 * any fail then we roll back all operations in the group.
2374 * Always run under BQL.
2376 void qmp_transaction(TransactionActionList *dev_list,
2377 struct TransactionProperties *props,
2380 TransactionActionList *dev_entry = dev_list;
2381 bool has_props = !!props;
2382 JobTxn *block_job_txn = NULL;
2383 Error *local_err = NULL;
2384 Transaction *tran = tran_new();
2386 GLOBAL_STATE_CODE();
2388 /* Does this transaction get canceled as a group on failure?
2389 * If not, we don't really need to make a JobTxn.
2391 props = get_transaction_properties(props);
2392 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2393 block_job_txn = job_txn_new();
2396 /* drain all i/o before any operations */
2399 /* We don't do anything in this loop that commits us to the operations */
2400 while (NULL != dev_entry) {
2401 TransactionAction *dev_info = NULL;
2402 const BlkActionOps *ops;
2403 BlkActionState *state;
2405 dev_info = dev_entry->value;
2406 dev_entry = dev_entry->next;
2408 assert(dev_info->type < ARRAY_SIZE(actions));
2410 ops = &actions[dev_info->type];
2411 assert(ops->instance_size > 0);
2413 state = g_malloc0(ops->instance_size);
2415 state->action = dev_info;
2416 state->block_job_txn = block_job_txn;
2417 state->txn_props = props;
2419 state->ops->action(state, tran, &local_err);
2421 error_propagate(errp, local_err);
2422 goto delete_and_fail;
2432 /* failure, and it is all-or-none; roll back all operations */
2436 qapi_free_TransactionProperties(props);
2438 job_txn_unref(block_job_txn);
2441 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2445 BdrvDirtyBitmap *bitmap;
2446 BlockDriverState *bs;
2447 BlockDirtyBitmapSha256 *ret = NULL;
2450 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2451 if (!bitmap || !bs) {
2455 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2456 if (sha256 == NULL) {
2460 ret = g_new(BlockDirtyBitmapSha256, 1);
2461 ret->sha256 = sha256;
2466 void coroutine_fn qmp_block_resize(const char *device, const char *node_name,
2467 int64_t size, Error **errp)
2469 Error *local_err = NULL;
2471 BlockDriverState *bs;
2472 AioContext *old_ctx;
2474 bs = bdrv_lookup_bs(device, node_name, &local_err);
2476 error_propagate(errp, local_err);
2481 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2485 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2486 error_setg(errp, QERR_DEVICE_IN_USE, device);
2490 blk = blk_co_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2496 bdrv_drained_begin(bs);
2499 old_ctx = bdrv_co_enter(bs);
2500 blk_co_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2501 bdrv_co_leave(bs, old_ctx);
2504 bdrv_drained_end(bs);
2509 void qmp_block_stream(const char *job_id, const char *device,
2511 const char *base_node,
2512 const char *backing_file,
2514 bool has_speed, int64_t speed,
2515 bool has_on_error, BlockdevOnError on_error,
2516 const char *filter_node_name,
2517 bool has_auto_finalize, bool auto_finalize,
2518 bool has_auto_dismiss, bool auto_dismiss,
2521 BlockDriverState *bs, *iter, *iter_end;
2522 BlockDriverState *base_bs = NULL;
2523 BlockDriverState *bottom_bs = NULL;
2524 AioContext *aio_context;
2525 Error *local_err = NULL;
2526 int job_flags = JOB_DEFAULT;
2528 if (base && base_node) {
2529 error_setg(errp, "'base' and 'base-node' cannot be specified "
2530 "at the same time");
2534 if (base && bottom) {
2535 error_setg(errp, "'base' and 'bottom' cannot be specified "
2536 "at the same time");
2540 if (bottom && base_node) {
2541 error_setg(errp, "'bottom' and 'base-node' cannot be specified "
2542 "at the same time");
2546 if (!has_on_error) {
2547 on_error = BLOCKDEV_ON_ERROR_REPORT;
2550 bs = bdrv_lookup_bs(device, device, errp);
2555 aio_context = bdrv_get_aio_context(bs);
2556 aio_context_acquire(aio_context);
2559 base_bs = bdrv_find_backing_image(bs, base);
2560 if (base_bs == NULL) {
2561 error_setg(errp, "Can't find '%s' in the backing chain", base);
2564 assert(bdrv_get_aio_context(base_bs) == aio_context);
2568 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2572 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2573 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2577 assert(bdrv_get_aio_context(base_bs) == aio_context);
2578 bdrv_refresh_filename(base_bs);
2582 bottom_bs = bdrv_lookup_bs(NULL, bottom, errp);
2586 if (!bottom_bs->drv) {
2587 error_setg(errp, "Node '%s' is not open", bottom);
2590 if (bottom_bs->drv->is_filter) {
2591 error_setg(errp, "Node '%s' is a filter, use a non-filter node "
2592 "as 'bottom'", bottom);
2595 if (!bdrv_chain_contains(bs, bottom_bs)) {
2596 error_setg(errp, "Node '%s' is not in a chain starting from '%s'",
2600 assert(bdrv_get_aio_context(bottom_bs) == aio_context);
2604 * Check for op blockers in the whole chain between bs and base (or bottom)
2606 iter_end = bottom ? bdrv_filter_or_cow_bs(bottom_bs) : base_bs;
2607 for (iter = bs; iter && iter != iter_end;
2608 iter = bdrv_filter_or_cow_bs(iter))
2610 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2615 /* if we are streaming the entire chain, the result will have no backing
2616 * file, and specifying one is therefore an error */
2617 if (!base_bs && backing_file) {
2618 error_setg(errp, "backing file specified, but streaming the "
2623 if (has_auto_finalize && !auto_finalize) {
2624 job_flags |= JOB_MANUAL_FINALIZE;
2626 if (has_auto_dismiss && !auto_dismiss) {
2627 job_flags |= JOB_MANUAL_DISMISS;
2630 stream_start(job_id, bs, base_bs, backing_file,
2631 bottom_bs, job_flags, has_speed ? speed : 0, on_error,
2632 filter_node_name, &local_err);
2634 error_propagate(errp, local_err);
2638 trace_qmp_block_stream(bs);
2641 aio_context_release(aio_context);
2644 void qmp_block_commit(const char *job_id, const char *device,
2645 const char *base_node,
2647 const char *top_node,
2649 const char *backing_file,
2650 bool has_speed, int64_t speed,
2651 bool has_on_error, BlockdevOnError on_error,
2652 const char *filter_node_name,
2653 bool has_auto_finalize, bool auto_finalize,
2654 bool has_auto_dismiss, bool auto_dismiss,
2657 BlockDriverState *bs;
2658 BlockDriverState *iter;
2659 BlockDriverState *base_bs, *top_bs;
2660 AioContext *aio_context;
2661 Error *local_err = NULL;
2662 int job_flags = JOB_DEFAULT;
2663 uint64_t top_perm, top_shared;
2668 if (!has_on_error) {
2669 on_error = BLOCKDEV_ON_ERROR_REPORT;
2671 if (has_auto_finalize && !auto_finalize) {
2672 job_flags |= JOB_MANUAL_FINALIZE;
2674 if (has_auto_dismiss && !auto_dismiss) {
2675 job_flags |= JOB_MANUAL_DISMISS;
2679 * libvirt relies on the DeviceNotFound error class in order to probe for
2680 * live commit feature versions; for this to work, we must make sure to
2681 * perform the device lookup before any generic errors that may occur in a
2682 * scenario in which all optional arguments are omitted. */
2683 bs = qmp_get_root_bs(device, &local_err);
2685 bs = bdrv_lookup_bs(device, device, NULL);
2687 error_free(local_err);
2688 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2689 "Device '%s' not found", device);
2691 error_propagate(errp, local_err);
2696 aio_context = bdrv_get_aio_context(bs);
2697 aio_context_acquire(aio_context);
2699 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2703 /* default top_bs is the active layer */
2706 if (top_node && top) {
2707 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2709 } else if (top_node) {
2710 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2711 if (top_bs == NULL) {
2714 if (!bdrv_chain_contains(bs, top_bs)) {
2715 error_setg(errp, "'%s' is not in this backing file chain",
2720 /* This strcmp() is just a shortcut, there is no need to
2721 * refresh @bs's filename. If it mismatches,
2722 * bdrv_find_backing_image() will do the refresh and may still
2724 if (strcmp(bs->filename, top) != 0) {
2725 top_bs = bdrv_find_backing_image(bs, top);
2729 if (top_bs == NULL) {
2730 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2734 assert(bdrv_get_aio_context(top_bs) == aio_context);
2736 if (base_node && base) {
2737 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2739 } else if (base_node) {
2740 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2741 if (base_bs == NULL) {
2744 if (!bdrv_chain_contains(top_bs, base_bs)) {
2745 error_setg(errp, "'%s' is not in this backing file chain",
2750 base_bs = bdrv_find_backing_image(top_bs, base);
2751 if (base_bs == NULL) {
2752 error_setg(errp, "Can't find '%s' in the backing chain", base);
2756 base_bs = bdrv_find_base(top_bs);
2757 if (base_bs == NULL) {
2758 error_setg(errp, "There is no backimg image");
2763 assert(bdrv_get_aio_context(base_bs) == aio_context);
2765 for (iter = top_bs; iter != bdrv_filter_or_cow_bs(base_bs);
2766 iter = bdrv_filter_or_cow_bs(iter))
2768 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2773 /* Do not allow attempts to commit an image into itself */
2774 if (top_bs == base_bs) {
2775 error_setg(errp, "cannot commit an image into itself");
2780 * Active commit is required if and only if someone has taken a
2781 * WRITE permission on the top node. Historically, we have always
2782 * used active commit for top nodes, so continue that practice
2783 * lest we possibly break clients that rely on this behavior, e.g.
2784 * to later attach this node to a writing parent.
2785 * (Active commit is never really wrong.)
2787 bdrv_get_cumulative_perm(top_bs, &top_perm, &top_shared);
2788 if (top_perm & BLK_PERM_WRITE ||
2789 bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs))
2792 if (bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs)) {
2793 error_setg(errp, "'backing-file' specified,"
2794 " but 'top' is the active layer");
2796 error_setg(errp, "'backing-file' specified, but 'top' has a "
2803 * Emulate here what block_job_create() does, because it
2804 * is possible that @bs != @top_bs (the block job should
2805 * be named after @bs, even if @top_bs is the actual
2808 job_id = bdrv_get_device_name(bs);
2810 commit_active_start(job_id, top_bs, base_bs, job_flags, speed, on_error,
2811 filter_node_name, NULL, NULL, false, &local_err);
2813 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2814 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2817 commit_start(job_id, bs, base_bs, top_bs, job_flags,
2818 speed, on_error, backing_file,
2819 filter_node_name, &local_err);
2821 if (local_err != NULL) {
2822 error_propagate(errp, local_err);
2827 aio_context_release(aio_context);
2830 /* Common QMP interface for drive-backup and blockdev-backup */
2831 static BlockJob *do_backup_common(BackupCommon *backup,
2832 BlockDriverState *bs,
2833 BlockDriverState *target_bs,
2834 AioContext *aio_context,
2835 JobTxn *txn, Error **errp)
2837 BlockJob *job = NULL;
2838 BdrvDirtyBitmap *bmap = NULL;
2839 BackupPerf perf = { .max_workers = 64 };
2840 int job_flags = JOB_DEFAULT;
2842 if (!backup->has_speed) {
2845 if (!backup->has_on_source_error) {
2846 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2848 if (!backup->has_on_target_error) {
2849 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2851 if (!backup->has_auto_finalize) {
2852 backup->auto_finalize = true;
2854 if (!backup->has_auto_dismiss) {
2855 backup->auto_dismiss = true;
2857 if (!backup->has_compress) {
2858 backup->compress = false;
2861 if (backup->x_perf) {
2862 if (backup->x_perf->has_use_copy_range) {
2863 perf.use_copy_range = backup->x_perf->use_copy_range;
2865 if (backup->x_perf->has_max_workers) {
2866 perf.max_workers = backup->x_perf->max_workers;
2868 if (backup->x_perf->has_max_chunk) {
2869 perf.max_chunk = backup->x_perf->max_chunk;
2873 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2874 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2875 /* done before desugaring 'incremental' to print the right message */
2876 if (!backup->bitmap) {
2877 error_setg(errp, "must provide a valid bitmap name for "
2878 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2883 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2884 if (backup->has_bitmap_mode &&
2885 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2886 error_setg(errp, "Bitmap sync mode must be '%s' "
2887 "when using sync mode '%s'",
2888 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2889 MirrorSyncMode_str(backup->sync));
2892 backup->has_bitmap_mode = true;
2893 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2894 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2897 if (backup->bitmap) {
2898 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2900 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2903 if (!backup->has_bitmap_mode) {
2904 error_setg(errp, "Bitmap sync mode must be given "
2905 "when providing a bitmap");
2908 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2912 /* This does not produce a useful bitmap artifact: */
2913 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2914 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2915 " outputs", MirrorSyncMode_str(backup->sync));
2919 /* If the bitmap isn't used for input or output, this is useless: */
2920 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2921 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2922 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2923 " when combined with sync mode '%s'",
2924 BitmapSyncMode_str(backup->bitmap_mode),
2925 MirrorSyncMode_str(backup->sync));
2930 if (!backup->bitmap && backup->has_bitmap_mode) {
2931 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2935 if (!backup->auto_finalize) {
2936 job_flags |= JOB_MANUAL_FINALIZE;
2938 if (!backup->auto_dismiss) {
2939 job_flags |= JOB_MANUAL_DISMISS;
2942 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2943 backup->sync, bmap, backup->bitmap_mode,
2945 backup->filter_node_name,
2947 backup->on_source_error,
2948 backup->on_target_error,
2949 job_flags, NULL, NULL, txn, errp);
2953 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2955 TransactionAction action = {
2956 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2957 .u.drive_backup.data = backup,
2959 blockdev_do_action(&action, errp);
2962 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2966 bool return_flat = has_flat && flat;
2968 return bdrv_named_nodes_list(return_flat, errp);
2971 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2973 return bdrv_get_xdbg_block_graph(errp);
2976 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2978 TransactionAction action = {
2979 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2980 .u.blockdev_backup.data = backup,
2982 blockdev_do_action(&action, errp);
2985 /* Parameter check and block job starting for drive mirroring.
2986 * Caller should hold @device and @target's aio context (must be the same).
2988 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2989 BlockDriverState *target,
2990 const char *replaces,
2991 enum MirrorSyncMode sync,
2992 BlockMirrorBackingMode backing_mode,
2994 bool has_speed, int64_t speed,
2995 bool has_granularity, uint32_t granularity,
2996 bool has_buf_size, int64_t buf_size,
2997 bool has_on_source_error,
2998 BlockdevOnError on_source_error,
2999 bool has_on_target_error,
3000 BlockdevOnError on_target_error,
3001 bool has_unmap, bool unmap,
3002 const char *filter_node_name,
3003 bool has_copy_mode, MirrorCopyMode copy_mode,
3004 bool has_auto_finalize, bool auto_finalize,
3005 bool has_auto_dismiss, bool auto_dismiss,
3008 BlockDriverState *unfiltered_bs;
3009 int job_flags = JOB_DEFAULT;
3011 GLOBAL_STATE_CODE();
3012 GRAPH_RDLOCK_GUARD_MAINLOOP();
3017 if (!has_on_source_error) {
3018 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
3020 if (!has_on_target_error) {
3021 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
3023 if (!has_granularity) {
3026 if (!has_buf_size) {
3032 if (!has_copy_mode) {
3033 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
3035 if (has_auto_finalize && !auto_finalize) {
3036 job_flags |= JOB_MANUAL_FINALIZE;
3038 if (has_auto_dismiss && !auto_dismiss) {
3039 job_flags |= JOB_MANUAL_DISMISS;
3042 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
3043 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3044 "a value in range [512B, 64MB]");
3047 if (granularity & (granularity - 1)) {
3048 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
3053 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3056 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
3060 if (!bdrv_backing_chain_next(bs) && sync == MIRROR_SYNC_MODE_TOP) {
3061 sync = MIRROR_SYNC_MODE_FULL;
3065 /* We want to mirror from @bs, but keep implicit filters on top */
3066 unfiltered_bs = bdrv_skip_implicit_filters(bs);
3067 if (unfiltered_bs != bs) {
3068 replaces = unfiltered_bs->node_name;
3073 BlockDriverState *to_replace_bs;
3074 AioContext *replace_aio_context;
3075 int64_t bs_size, replace_size;
3077 bs_size = bdrv_getlength(bs);
3079 error_setg_errno(errp, -bs_size, "Failed to query device's size");
3083 to_replace_bs = check_to_replace_node(bs, replaces, errp);
3084 if (!to_replace_bs) {
3088 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
3089 aio_context_acquire(replace_aio_context);
3090 replace_size = bdrv_getlength(to_replace_bs);
3091 aio_context_release(replace_aio_context);
3093 if (replace_size < 0) {
3094 error_setg_errno(errp, -replace_size,
3095 "Failed to query the replacement node's size");
3098 if (bs_size != replace_size) {
3099 error_setg(errp, "cannot replace image with a mirror image of "
3105 /* pass the node name to replace to mirror start since it's loose coupling
3106 * and will allow to check whether the node still exist at mirror completion
3108 mirror_start(job_id, bs, target,
3109 replaces, job_flags,
3110 speed, granularity, buf_size, sync, backing_mode, zero_target,
3111 on_source_error, on_target_error, unmap, filter_node_name,
3115 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
3117 BlockDriverState *bs;
3118 BlockDriverState *target_backing_bs, *target_bs;
3119 AioContext *aio_context;
3120 AioContext *old_context;
3121 BlockMirrorBackingMode backing_mode;
3122 Error *local_err = NULL;
3123 QDict *options = NULL;
3126 const char *format = arg->format;
3130 bs = qmp_get_root_bs(arg->device, errp);
3135 /* Early check to avoid creating target */
3136 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3140 aio_context = bdrv_get_aio_context(bs);
3141 aio_context_acquire(aio_context);
3143 if (!arg->has_mode) {
3144 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3148 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3149 ? NULL : bs->drv->format_name);
3152 flags = bs->open_flags | BDRV_O_RDWR;
3153 target_backing_bs = bdrv_cow_bs(bdrv_skip_filters(bs));
3154 if (!target_backing_bs && arg->sync == MIRROR_SYNC_MODE_TOP) {
3155 arg->sync = MIRROR_SYNC_MODE_FULL;
3157 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3158 target_backing_bs = bs;
3161 size = bdrv_getlength(bs);
3163 error_setg_errno(errp, -size, "bdrv_getlength failed");
3167 if (arg->replaces) {
3168 if (!arg->node_name) {
3169 error_setg(errp, "a node-name must be provided when replacing a"
3170 " named node of the graph");
3175 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3176 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3178 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3181 /* Don't open backing image in create() */
3182 flags |= BDRV_O_NO_BACKING;
3184 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !target_backing_bs)
3185 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3187 /* create new image w/o backing file */
3189 bdrv_img_create(arg->target, format,
3190 NULL, NULL, NULL, size, flags, false, &local_err);
3192 /* Implicit filters should not appear in the filename */
3193 BlockDriverState *explicit_backing =
3194 bdrv_skip_implicit_filters(target_backing_bs);
3196 switch (arg->mode) {
3197 case NEW_IMAGE_MODE_EXISTING:
3199 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3200 /* create new image with backing file */
3201 bdrv_refresh_filename(explicit_backing);
3202 bdrv_img_create(arg->target, format,
3203 explicit_backing->filename,
3204 explicit_backing->drv->format_name,
3205 NULL, size, flags, false, &local_err);
3213 error_propagate(errp, local_err);
3217 options = qdict_new();
3218 if (arg->node_name) {
3219 qdict_put_str(options, "node-name", arg->node_name);
3222 qdict_put_str(options, "driver", format);
3225 /* Mirroring takes care of copy-on-write using the source's backing
3228 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3233 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3234 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3235 !bdrv_has_zero_init(target_bs)));
3238 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3239 old_context = bdrv_get_aio_context(target_bs);
3240 aio_context_release(aio_context);
3241 aio_context_acquire(old_context);
3243 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
3245 bdrv_unref(target_bs);
3246 aio_context_release(old_context);
3250 aio_context_release(old_context);
3251 aio_context_acquire(aio_context);
3253 blockdev_mirror_common(arg->job_id, bs, target_bs,
3254 arg->replaces, arg->sync,
3255 backing_mode, zero_target,
3256 arg->has_speed, arg->speed,
3257 arg->has_granularity, arg->granularity,
3258 arg->has_buf_size, arg->buf_size,
3259 arg->has_on_source_error, arg->on_source_error,
3260 arg->has_on_target_error, arg->on_target_error,
3261 arg->has_unmap, arg->unmap,
3263 arg->has_copy_mode, arg->copy_mode,
3264 arg->has_auto_finalize, arg->auto_finalize,
3265 arg->has_auto_dismiss, arg->auto_dismiss,
3267 bdrv_unref(target_bs);
3269 aio_context_release(aio_context);
3272 void qmp_blockdev_mirror(const char *job_id,
3273 const char *device, const char *target,
3274 const char *replaces,
3275 MirrorSyncMode sync,
3276 bool has_speed, int64_t speed,
3277 bool has_granularity, uint32_t granularity,
3278 bool has_buf_size, int64_t buf_size,
3279 bool has_on_source_error,
3280 BlockdevOnError on_source_error,
3281 bool has_on_target_error,
3282 BlockdevOnError on_target_error,
3283 const char *filter_node_name,
3284 bool has_copy_mode, MirrorCopyMode copy_mode,
3285 bool has_auto_finalize, bool auto_finalize,
3286 bool has_auto_dismiss, bool auto_dismiss,
3289 BlockDriverState *bs;
3290 BlockDriverState *target_bs;
3291 AioContext *aio_context;
3292 AioContext *old_context;
3293 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3297 bs = qmp_get_root_bs(device, errp);
3302 target_bs = bdrv_lookup_bs(target, target, errp);
3307 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3309 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3310 old_context = bdrv_get_aio_context(target_bs);
3311 aio_context = bdrv_get_aio_context(bs);
3312 aio_context_acquire(old_context);
3314 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
3316 aio_context_release(old_context);
3317 aio_context_acquire(aio_context);
3323 blockdev_mirror_common(job_id, bs, target_bs,
3324 replaces, sync, backing_mode,
3325 zero_target, has_speed, speed,
3326 has_granularity, granularity,
3327 has_buf_size, buf_size,
3328 has_on_source_error, on_source_error,
3329 has_on_target_error, on_target_error,
3330 true, true, filter_node_name,
3331 has_copy_mode, copy_mode,
3332 has_auto_finalize, auto_finalize,
3333 has_auto_dismiss, auto_dismiss,
3336 aio_context_release(aio_context);
3340 * Get a block job using its ID. Called with job_mutex held.
3342 static BlockJob *find_block_job_locked(const char *id, Error **errp)
3348 job = block_job_get_locked(id);
3351 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3352 "Block job '%s' not found", id);
3359 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3364 job = find_block_job_locked(device, errp);
3370 block_job_set_speed_locked(job, speed, errp);
3373 void qmp_block_job_cancel(const char *device,
3374 bool has_force, bool force, Error **errp)
3379 job = find_block_job_locked(device, errp);
3389 if (job_user_paused_locked(&job->job) && !force) {
3390 error_setg(errp, "The block job for device '%s' is currently paused",
3395 trace_qmp_block_job_cancel(job);
3396 job_user_cancel_locked(&job->job, force, errp);
3399 void qmp_block_job_pause(const char *device, Error **errp)
3404 job = find_block_job_locked(device, errp);
3410 trace_qmp_block_job_pause(job);
3411 job_user_pause_locked(&job->job, errp);
3414 void qmp_block_job_resume(const char *device, Error **errp)
3419 job = find_block_job_locked(device, errp);
3425 trace_qmp_block_job_resume(job);
3426 job_user_resume_locked(&job->job, errp);
3429 void qmp_block_job_complete(const char *device, Error **errp)
3434 job = find_block_job_locked(device, errp);
3440 trace_qmp_block_job_complete(job);
3441 job_complete_locked(&job->job, errp);
3444 void qmp_block_job_finalize(const char *id, Error **errp)
3449 job = find_block_job_locked(id, errp);
3455 trace_qmp_block_job_finalize(job);
3456 job_ref_locked(&job->job);
3457 job_finalize_locked(&job->job, errp);
3459 job_unref_locked(&job->job);
3462 void qmp_block_job_dismiss(const char *id, Error **errp)
3468 bjob = find_block_job_locked(id, errp);
3474 trace_qmp_block_job_dismiss(bjob);
3476 job_dismiss_locked(&job, errp);
3479 void qmp_change_backing_file(const char *device,
3480 const char *image_node_name,
3481 const char *backing_file,
3484 BlockDriverState *bs = NULL;
3485 AioContext *aio_context;
3486 BlockDriverState *image_bs = NULL;
3487 Error *local_err = NULL;
3491 bs = qmp_get_root_bs(device, errp);
3496 aio_context = bdrv_get_aio_context(bs);
3497 aio_context_acquire(aio_context);
3499 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3501 error_propagate(errp, local_err);
3506 error_setg(errp, "image file not found");
3510 if (bdrv_find_base(image_bs) == image_bs) {
3511 error_setg(errp, "not allowing backing file change on an image "
3512 "without a backing file");
3516 /* even though we are not necessarily operating on bs, we need it to
3517 * determine if block ops are currently prohibited on the chain */
3518 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3522 /* final sanity check */
3523 if (!bdrv_chain_contains(bs, image_bs)) {
3524 error_setg(errp, "'%s' and image file are not in the same chain",
3529 /* if not r/w, reopen to make r/w */
3530 ro = bdrv_is_read_only(image_bs);
3533 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3538 ret = bdrv_change_backing_file(image_bs, backing_file,
3539 image_bs->drv ? image_bs->drv->format_name : "",
3543 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3545 /* don't exit here, so we can try to restore open flags if
3550 bdrv_reopen_set_read_only(image_bs, true, errp);
3554 aio_context_release(aio_context);
3557 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3559 BlockDriverState *bs;
3561 Visitor *v = qobject_output_visitor_new(&obj);
3564 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3565 visit_complete(v, &obj);
3566 qdict = qobject_to(QDict, obj);
3568 qdict_flatten(qdict);
3570 if (!qdict_get_try_str(qdict, "node-name")) {
3571 error_setg(errp, "'node-name' must be specified for the root node");
3575 bs = bds_tree_init(qdict, errp);
3580 bdrv_set_monitor_owned(bs);
3586 void qmp_blockdev_reopen(BlockdevOptionsList *reopen_list, Error **errp)
3588 BlockReopenQueue *queue = NULL;
3590 /* Add each one of the BDS that we want to reopen to the queue */
3591 for (; reopen_list != NULL; reopen_list = reopen_list->next) {
3592 BlockdevOptions *options = reopen_list->value;
3593 BlockDriverState *bs;
3599 /* Check for the selected node name */
3600 if (!options->node_name) {
3601 error_setg(errp, "node-name not specified");
3605 bs = bdrv_find_node(options->node_name);
3607 error_setg(errp, "Failed to find node with node-name='%s'",
3608 options->node_name);
3612 /* Put all options in a QDict and flatten it */
3613 v = qobject_output_visitor_new(&obj);
3614 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3615 visit_complete(v, &obj);
3618 qdict = qobject_to(QDict, obj);
3620 qdict_flatten(qdict);
3622 ctx = bdrv_get_aio_context(bs);
3623 aio_context_acquire(ctx);
3625 queue = bdrv_reopen_queue(queue, bs, qdict, false);
3627 aio_context_release(ctx);
3630 /* Perform the reopen operation */
3631 bdrv_reopen_multiple(queue, errp);
3635 bdrv_reopen_queue_free(queue);
3638 void qmp_blockdev_del(const char *node_name, Error **errp)
3640 AioContext *aio_context;
3641 BlockDriverState *bs;
3643 GLOBAL_STATE_CODE();
3645 bs = bdrv_find_node(node_name);
3647 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3650 if (bdrv_has_blk(bs)) {
3651 error_setg(errp, "Node %s is in use", node_name);
3654 aio_context = bdrv_get_aio_context(bs);
3655 aio_context_acquire(aio_context);
3657 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3661 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3662 error_setg(errp, "Node %s is not owned by the monitor",
3667 if (bs->refcnt > 1) {
3668 error_setg(errp, "Block device %s is in use",
3669 bdrv_get_device_or_node_name(bs));
3673 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3677 aio_context_release(aio_context);
3680 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3681 const char *child_name)
3685 QLIST_FOREACH(child, &parent_bs->children, next) {
3686 if (strcmp(child->name, child_name) == 0) {
3694 void qmp_x_blockdev_change(const char *parent, const char *child,
3695 const char *node, Error **errp)
3697 BlockDriverState *parent_bs, *new_bs = NULL;
3700 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3705 if (!child == !node) {
3707 error_setg(errp, "The parameters child and node are in conflict");
3709 error_setg(errp, "Either child or node must be specified");
3715 p_child = bdrv_find_child(parent_bs, child);
3717 error_setg(errp, "Node '%s' does not have child '%s'",
3721 bdrv_del_child(parent_bs, p_child, errp);
3725 new_bs = bdrv_find_node(node);
3727 error_setg(errp, "Node '%s' not found", node);
3730 bdrv_add_child(parent_bs, new_bs, errp);
3734 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3736 BlockJobInfoList *head = NULL, **tail = &head;
3741 for (job = block_job_next_locked(NULL); job;
3742 job = block_job_next_locked(job)) {
3743 BlockJobInfo *value;
3745 if (block_job_is_internal(job)) {
3748 value = block_job_query_locked(job, errp);
3750 qapi_free_BlockJobInfoList(head);
3753 QAPI_LIST_APPEND(tail, value);
3759 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3760 bool has_force, bool force, Error **errp)
3762 AioContext *old_context;
3763 AioContext *new_context;
3764 BlockDriverState *bs;
3766 bs = bdrv_find_node(node_name);
3768 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3772 /* Protects against accidents. */
3773 if (!(has_force && force) && bdrv_has_blk(bs)) {
3774 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3775 "be in use (use force=true to override this check)",
3780 if (iothread->type == QTYPE_QSTRING) {
3781 IOThread *obj = iothread_by_id(iothread->u.s);
3783 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3787 new_context = iothread_get_aio_context(obj);
3789 new_context = qemu_get_aio_context();
3792 old_context = bdrv_get_aio_context(bs);
3793 aio_context_acquire(old_context);
3795 bdrv_try_change_aio_context(bs, new_context, NULL, errp);
3797 aio_context_release(old_context);
3800 QemuOptsList qemu_common_drive_opts = {
3802 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3806 .type = QEMU_OPT_BOOL,
3807 .help = "enable/disable snapshot mode",
3810 .type = QEMU_OPT_STRING,
3811 .help = "host AIO implementation (threads, native, io_uring)",
3813 .name = BDRV_OPT_CACHE_WB,
3814 .type = QEMU_OPT_BOOL,
3815 .help = "Enable writeback mode",
3818 .type = QEMU_OPT_STRING,
3819 .help = "disk format (raw, qcow2, ...)",
3822 .type = QEMU_OPT_STRING,
3823 .help = "read error action",
3826 .type = QEMU_OPT_STRING,
3827 .help = "write error action",
3829 .name = BDRV_OPT_READ_ONLY,
3830 .type = QEMU_OPT_BOOL,
3831 .help = "open drive file as read-only",
3837 .name = "throttling.group",
3838 .type = QEMU_OPT_STRING,
3839 .help = "name of the block throttling group",
3841 .name = "copy-on-read",
3842 .type = QEMU_OPT_BOOL,
3843 .help = "copy read data from backing file into image file",
3845 .name = "detect-zeroes",
3846 .type = QEMU_OPT_STRING,
3847 .help = "try to optimize zero writes (off, on, unmap)",
3849 .name = "stats-account-invalid",
3850 .type = QEMU_OPT_BOOL,
3851 .help = "whether to account for invalid I/O operations "
3852 "in the statistics",
3854 .name = "stats-account-failed",
3855 .type = QEMU_OPT_BOOL,
3856 .help = "whether to account for failed I/O operations "
3857 "in the statistics",
3859 { /* end of list */ }
3863 QemuOptsList qemu_drive_opts = {
3865 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3868 * no elements => accept any params
3869 * validation will happen later
3871 { /* end of list */ }