meson.build: Make keyutils independent from keyring
[qemu/kevin.git] / blockdev.c
blob325b7a3bef655ff4a87abea4879a157fc3dedf0a
1 /*
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
10 * permission notice:
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
30 * THE SOFTWARE.
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)
73 GLOBAL_STATE_CODE();
74 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
77 static const char *const if_name[IF_COUNT] = {
78 [IF_NONE] = "none",
79 [IF_IDE] = "ide",
80 [IF_SCSI] = "scsi",
81 [IF_FLOPPY] = "floppy",
82 [IF_PFLASH] = "pflash",
83 [IF_MTD] = "mtd",
84 [IF_SD] = "sd",
85 [IF_VIRTIO] = "virtio",
86 [IF_XEN] = "xen",
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
97 * values.
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.
104 [IF_IDE] = 2,
105 [IF_SCSI] = 7,
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)
114 BlockBackend *blk;
115 DriveInfo *dinfo;
117 GLOBAL_STATE_CODE();
119 if (max_devs <= 0) {
120 return;
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);
146 BlockJob *job;
148 GLOBAL_STATE_CODE();
150 if (!dinfo) {
151 return;
154 JOB_LOCK_GUARD();
156 do {
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);
164 if (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);
171 } while (job);
173 dinfo->auto_del = 1;
176 void blockdev_auto_del(BlockBackend *blk)
178 DriveInfo *dinfo = blk_legacy_dinfo(blk);
179 GLOBAL_STATE_CODE();
181 if (dinfo && dinfo->auto_del) {
182 monitor_remove_blk(blk);
183 blk_unref(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,
200 const char *optstr)
202 QemuOpts *opts;
204 GLOBAL_STATE_CODE();
206 opts = qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
207 if (!opts) {
208 return NULL;
210 if (type != IF_DEFAULT) {
211 qemu_opt_set(opts, "if", if_name[type], &error_abort);
213 if (index >= 0) {
214 qemu_opt_set_number(opts, "index", index, &error_abort);
216 if (file)
217 qemu_opt_set(opts, "file", file, &error_abort);
218 return opts;
221 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
223 BlockBackend *blk;
224 DriveInfo *dinfo;
226 GLOBAL_STATE_CODE();
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) {
232 return dinfo;
236 return NULL;
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)
245 BlockBackend *blk;
246 DriveInfo *dinfo;
247 Location loc;
248 bool orphans = false;
250 GLOBAL_STATE_CODE();
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
257 * users fault.
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) {
265 continue;
267 if (!blk_get_attached_dev(blk)) {
268 loc_push_none(&loc);
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);
273 loc_pop(&loc);
274 orphans = true;
278 if (orphans) {
279 exit(1);
283 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
285 GLOBAL_STATE_CODE();
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)
293 int max_bus;
294 BlockBackend *blk;
295 DriveInfo *dinfo;
297 GLOBAL_STATE_CODE();
299 max_bus = -1;
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;
306 return max_bus;
309 static void bdrv_format_print(void *opaque, const char *name)
311 qemu_printf(" %s", name);
314 typedef struct {
315 QEMUBH *bh;
316 BlockDriverState *bs;
317 } BDRVPutRefBH;
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;
329 } else {
330 error_setg(errp, "'%s' invalid %s error action",
331 buf, is_read ? "read" : "write");
332 return -1;
336 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
337 Error **errp)
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 uint64_t length;
345 const char *str = qstring_get_str(qobject_to(QString,
346 entry->value));
347 if (parse_uint_full(str, 10, &length) == 0 &&
348 length > 0 && length <= UINT_MAX) {
349 block_acct_add_interval(stats, (unsigned) length);
350 } else {
351 error_setg(errp, "Invalid interval length: %s", str);
352 return false;
354 break;
357 case QTYPE_QNUM: {
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);
362 } else {
363 error_setg(errp, "Invalid interval length: %" PRId64, length);
364 return false;
366 break;
369 default:
370 error_setg(errp, "The specification of stats-intervals is invalid");
371 return false;
374 return true;
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;
385 const char *aio;
387 if (bdrv_flags) {
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");
395 return;
400 /* disk I/O throttling */
401 if (throttling_group) {
402 *throttling_group = qemu_opt_get(opts, "throttling.group");
405 if (throttle_cfg) {
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)) {
450 return;
454 if (detect_zeroes) {
455 *detect_zeroes =
456 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
457 qemu_opt_get(opts, "detect-zeroes"),
458 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
459 &local_error);
460 if (local_error) {
461 error_propagate(errp, local_error);
462 return;
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,
480 Error **errp)
482 const char *buf;
483 int bdrv_flags = 0;
484 int on_read_error, on_write_error;
485 OnOffAuto account_invalid, account_failed;
486 bool writethrough, read_only;
487 BlockBackend *blk;
488 BlockDriverState *bs;
489 ThrottleConfig cfg;
490 int snapshot = 0;
491 Error *error = NULL;
492 QemuOpts *opts;
493 QDict *interval_dict = NULL;
494 QList *interval_list = NULL;
495 const char *id;
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);
504 if (!opts) {
505 goto err_no_opts;
508 if (!qemu_opts_absorb_qdict(opts, bs_opts, errp)) {
509 goto early_err;
512 if (id) {
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);
532 goto early_err;
535 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
536 &detect_zeroes, &error);
537 if (error) {
538 error_propagate(errp, error);
539 goto early_err;
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);
548 qemu_printf("\n");
549 goto early_err;
552 if (qdict_haskey(bs_opts, "driver")) {
553 error_setg(errp, "Cannot specify both 'driver' and 'format'");
554 goto early_err;
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);
562 if (error) {
563 error_propagate(errp, error);
564 goto early_err;
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);
571 if (error) {
572 error_propagate(errp, error);
573 goto early_err;
577 if (snapshot) {
578 bdrv_flags |= BDRV_O_SNAPSHOT;
581 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
583 /* init */
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);
593 } else {
594 if (file && !*file) {
595 file = NULL;
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);
613 if (!blk) {
614 goto err_no_bs_opts;
616 bs = blk_bs(blk);
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)) {
623 blk_unref(blk);
624 blk = NULL;
625 goto err_no_bs_opts;
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)) {
642 blk_unref(blk);
643 blk = NULL;
644 goto err_no_bs_opts;
647 err_no_bs_opts:
648 qemu_opts_del(opts);
649 qobject_unref(interval_dict);
650 qobject_unref(interval_list);
651 return blk;
653 early_err:
654 qemu_opts_del(opts);
655 qobject_unref(interval_dict);
656 qobject_unref(interval_list);
657 err_no_opts:
658 qobject_unref(bs_opts);
659 return NULL;
662 /* Takes the ownership of bs_opts */
663 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
665 BlockDriverState *bs;
666 int bdrv_flags = 0;
668 GLOBAL_STATE_CODE();
669 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
670 * with other callers) rather than what we want as the real defaults.
671 * Apply the defaults here instead. */
672 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
673 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
674 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
676 if (runstate_check(RUN_STATE_INMIGRATE)) {
677 bdrv_flags |= BDRV_O_INACTIVE;
680 aio_context_acquire(qemu_get_aio_context());
681 bs = bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
682 aio_context_release(qemu_get_aio_context());
684 return bs;
687 void blockdev_close_all_bdrv_states(void)
689 BlockDriverState *bs, *next_bs;
691 GLOBAL_STATE_CODE();
692 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
693 AioContext *ctx = bdrv_get_aio_context(bs);
695 aio_context_acquire(ctx);
696 bdrv_unref(bs);
697 aio_context_release(ctx);
701 /* Iterates over the list of monitor-owned BlockDriverStates */
702 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
704 GLOBAL_STATE_CODE();
705 return bs ? QTAILQ_NEXT(bs, monitor_list)
706 : QTAILQ_FIRST(&monitor_bdrv_states);
709 static bool qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
710 Error **errp)
712 const char *value;
714 value = qemu_opt_get(opts, from);
715 if (value) {
716 if (qemu_opt_find(opts, to)) {
717 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
718 "same time", to, from);
719 return false;
723 /* rename all items in opts */
724 while ((value = qemu_opt_get(opts, from))) {
725 qemu_opt_set(opts, to, value, &error_abort);
726 qemu_opt_unset(opts, from);
728 return true;
731 QemuOptsList qemu_legacy_drive_opts = {
732 .name = "drive",
733 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
734 .desc = {
736 .name = "bus",
737 .type = QEMU_OPT_NUMBER,
738 .help = "bus number",
740 .name = "unit",
741 .type = QEMU_OPT_NUMBER,
742 .help = "unit number (i.e. lun for scsi)",
744 .name = "index",
745 .type = QEMU_OPT_NUMBER,
746 .help = "index number",
748 .name = "media",
749 .type = QEMU_OPT_STRING,
750 .help = "media type (disk, cdrom)",
752 .name = "if",
753 .type = QEMU_OPT_STRING,
754 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
756 .name = "file",
757 .type = QEMU_OPT_STRING,
758 .help = "file name",
761 /* Options that are passed on, but have special semantics with -drive */
763 .name = BDRV_OPT_READ_ONLY,
764 .type = QEMU_OPT_BOOL,
765 .help = "open drive file as read-only",
767 .name = "rerror",
768 .type = QEMU_OPT_STRING,
769 .help = "read error action",
771 .name = "werror",
772 .type = QEMU_OPT_STRING,
773 .help = "write error action",
775 .name = "copy-on-read",
776 .type = QEMU_OPT_BOOL,
777 .help = "copy read data from backing file into image file",
780 { /* end of list */ }
784 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
785 Error **errp)
787 const char *value;
788 BlockBackend *blk;
789 DriveInfo *dinfo = NULL;
790 QDict *bs_opts;
791 QemuOpts *legacy_opts;
792 DriveMediaType media = MEDIA_DISK;
793 BlockInterfaceType type;
794 int max_devs, bus_id, unit_id, index;
795 const char *werror, *rerror;
796 bool read_only = false;
797 bool copy_on_read;
798 const char *filename;
799 int i;
801 GLOBAL_STATE_CODE();
803 /* Change legacy command line options into QMP ones */
804 static const struct {
805 const char *from;
806 const char *to;
807 } opt_renames[] = {
808 { "iops", "throttling.iops-total" },
809 { "iops_rd", "throttling.iops-read" },
810 { "iops_wr", "throttling.iops-write" },
812 { "bps", "throttling.bps-total" },
813 { "bps_rd", "throttling.bps-read" },
814 { "bps_wr", "throttling.bps-write" },
816 { "iops_max", "throttling.iops-total-max" },
817 { "iops_rd_max", "throttling.iops-read-max" },
818 { "iops_wr_max", "throttling.iops-write-max" },
820 { "bps_max", "throttling.bps-total-max" },
821 { "bps_rd_max", "throttling.bps-read-max" },
822 { "bps_wr_max", "throttling.bps-write-max" },
824 { "iops_size", "throttling.iops-size" },
826 { "group", "throttling.group" },
828 { "readonly", BDRV_OPT_READ_ONLY },
831 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
832 if (!qemu_opt_rename(all_opts, opt_renames[i].from,
833 opt_renames[i].to, errp)) {
834 return NULL;
838 value = qemu_opt_get(all_opts, "cache");
839 if (value) {
840 int flags = 0;
841 bool writethrough;
843 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
844 error_setg(errp, "invalid cache option");
845 return NULL;
848 /* Specific options take precedence */
849 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
850 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
851 !writethrough, &error_abort);
853 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
854 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
855 !!(flags & BDRV_O_NOCACHE), &error_abort);
857 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
858 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
859 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
861 qemu_opt_unset(all_opts, "cache");
864 /* Get a QDict for processing the options */
865 bs_opts = qdict_new();
866 qemu_opts_to_qdict(all_opts, bs_opts);
868 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
869 &error_abort);
870 if (!qemu_opts_absorb_qdict(legacy_opts, bs_opts, errp)) {
871 goto fail;
874 /* Media type */
875 value = qemu_opt_get(legacy_opts, "media");
876 if (value) {
877 if (!strcmp(value, "disk")) {
878 media = MEDIA_DISK;
879 } else if (!strcmp(value, "cdrom")) {
880 media = MEDIA_CDROM;
881 read_only = true;
882 } else {
883 error_setg(errp, "'%s' invalid media", value);
884 goto fail;
888 /* copy-on-read is disabled with a warning for read-only devices */
889 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
890 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
892 if (read_only && copy_on_read) {
893 warn_report("disabling copy-on-read on read-only drive");
894 copy_on_read = false;
897 qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
898 qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
900 /* Controller type */
901 value = qemu_opt_get(legacy_opts, "if");
902 if (value) {
903 for (type = 0;
904 type < IF_COUNT && strcmp(value, if_name[type]);
905 type++) {
907 if (type == IF_COUNT) {
908 error_setg(errp, "unsupported bus type '%s'", value);
909 goto fail;
911 } else {
912 type = block_default_type;
915 /* Device address specified by bus/unit or index.
916 * If none was specified, try to find the first free one. */
917 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
918 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
919 index = qemu_opt_get_number(legacy_opts, "index", -1);
921 max_devs = if_max_devs[type];
923 if (index != -1) {
924 if (bus_id != 0 || unit_id != -1) {
925 error_setg(errp, "index cannot be used with bus and unit");
926 goto fail;
928 bus_id = drive_index_to_bus_id(type, index);
929 unit_id = drive_index_to_unit_id(type, index);
932 if (unit_id == -1) {
933 unit_id = 0;
934 while (drive_get(type, bus_id, unit_id) != NULL) {
935 unit_id++;
936 if (max_devs && unit_id >= max_devs) {
937 unit_id -= max_devs;
938 bus_id++;
943 if (max_devs && unit_id >= max_devs) {
944 error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
945 goto fail;
948 if (drive_get(type, bus_id, unit_id) != NULL) {
949 error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
950 bus_id, unit_id, index);
951 goto fail;
954 /* no id supplied -> create one */
955 if (qemu_opts_id(all_opts) == NULL) {
956 char *new_id;
957 const char *mediastr = "";
958 if (type == IF_IDE || type == IF_SCSI) {
959 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
961 if (max_devs) {
962 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
963 mediastr, unit_id);
964 } else {
965 new_id = g_strdup_printf("%s%s%i", if_name[type],
966 mediastr, unit_id);
968 qdict_put_str(bs_opts, "id", new_id);
969 g_free(new_id);
972 /* Add virtio block device */
973 if (type == IF_VIRTIO) {
974 QemuOpts *devopts;
975 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
976 &error_abort);
977 qemu_opt_set(devopts, "driver", "virtio-blk", &error_abort);
978 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
979 &error_abort);
982 filename = qemu_opt_get(legacy_opts, "file");
984 /* Check werror/rerror compatibility with if=... */
985 werror = qemu_opt_get(legacy_opts, "werror");
986 if (werror != NULL) {
987 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
988 type != IF_NONE) {
989 error_setg(errp, "werror is not supported by this bus type");
990 goto fail;
992 qdict_put_str(bs_opts, "werror", werror);
995 rerror = qemu_opt_get(legacy_opts, "rerror");
996 if (rerror != NULL) {
997 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
998 type != IF_NONE) {
999 error_setg(errp, "rerror is not supported by this bus type");
1000 goto fail;
1002 qdict_put_str(bs_opts, "rerror", rerror);
1005 /* Actual block device init: Functionality shared with blockdev-add */
1006 blk = blockdev_init(filename, bs_opts, errp);
1007 bs_opts = NULL;
1008 if (!blk) {
1009 goto fail;
1012 /* Create legacy DriveInfo */
1013 dinfo = g_malloc0(sizeof(*dinfo));
1014 dinfo->opts = all_opts;
1016 dinfo->type = type;
1017 dinfo->bus = bus_id;
1018 dinfo->unit = unit_id;
1020 blk_set_legacy_dinfo(blk, dinfo);
1022 switch(type) {
1023 case IF_IDE:
1024 case IF_SCSI:
1025 case IF_XEN:
1026 case IF_NONE:
1027 dinfo->media_cd = media == MEDIA_CDROM;
1028 break;
1029 default:
1030 break;
1033 fail:
1034 qemu_opts_del(legacy_opts);
1035 qobject_unref(bs_opts);
1036 return dinfo;
1039 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1041 BlockDriverState *bs;
1042 AioContext *aio_context;
1044 bs = bdrv_lookup_bs(name, name, errp);
1045 if (bs == NULL) {
1046 return NULL;
1049 if (!bdrv_is_root_node(bs)) {
1050 error_setg(errp, "Need a root block node");
1051 return NULL;
1054 aio_context = bdrv_get_aio_context(bs);
1055 aio_context_acquire(aio_context);
1057 if (!bdrv_is_inserted(bs)) {
1058 error_setg(errp, "Device has no medium");
1059 bs = NULL;
1062 aio_context_release(aio_context);
1064 return bs;
1067 static void blockdev_do_action(TransactionAction *action, Error **errp)
1069 TransactionActionList list;
1071 list.value = action;
1072 list.next = NULL;
1073 qmp_transaction(&list, NULL, errp);
1076 void qmp_blockdev_snapshot_sync(const char *device, const char *node_name,
1077 const char *snapshot_file,
1078 const char *snapshot_node_name,
1079 const char *format,
1080 bool has_mode, NewImageMode mode, Error **errp)
1082 BlockdevSnapshotSync snapshot = {
1083 .device = (char *) device,
1084 .node_name = (char *) node_name,
1085 .snapshot_file = (char *) snapshot_file,
1086 .snapshot_node_name = (char *) snapshot_node_name,
1087 .format = (char *) format,
1088 .has_mode = has_mode,
1089 .mode = mode,
1091 TransactionAction action = {
1092 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1093 .u.blockdev_snapshot_sync.data = &snapshot,
1095 blockdev_do_action(&action, errp);
1098 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1099 Error **errp)
1101 BlockdevSnapshot snapshot_data = {
1102 .node = (char *) node,
1103 .overlay = (char *) overlay
1105 TransactionAction action = {
1106 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1107 .u.blockdev_snapshot.data = &snapshot_data,
1109 blockdev_do_action(&action, errp);
1112 void qmp_blockdev_snapshot_internal_sync(const char *device,
1113 const char *name,
1114 Error **errp)
1116 BlockdevSnapshotInternal snapshot = {
1117 .device = (char *) device,
1118 .name = (char *) name
1120 TransactionAction action = {
1121 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1122 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1124 blockdev_do_action(&action, errp);
1127 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1128 const char *id,
1129 const char *name,
1130 Error **errp)
1132 BlockDriverState *bs;
1133 AioContext *aio_context;
1134 QEMUSnapshotInfo sn;
1135 Error *local_err = NULL;
1136 SnapshotInfo *info = NULL;
1137 int ret;
1139 bs = qmp_get_root_bs(device, errp);
1140 if (!bs) {
1141 return NULL;
1143 aio_context = bdrv_get_aio_context(bs);
1144 aio_context_acquire(aio_context);
1146 if (!id && !name) {
1147 error_setg(errp, "Name or id must be provided");
1148 goto out_aio_context;
1151 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1152 goto out_aio_context;
1155 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1156 if (local_err) {
1157 error_propagate(errp, local_err);
1158 goto out_aio_context;
1160 if (!ret) {
1161 error_setg(errp,
1162 "Snapshot with id '%s' and name '%s' does not exist on "
1163 "device '%s'",
1164 STR_OR_NULL(id), STR_OR_NULL(name), device);
1165 goto out_aio_context;
1168 bdrv_snapshot_delete(bs, id, name, &local_err);
1169 if (local_err) {
1170 error_propagate(errp, local_err);
1171 goto out_aio_context;
1174 aio_context_release(aio_context);
1176 info = g_new0(SnapshotInfo, 1);
1177 info->id = g_strdup(sn.id_str);
1178 info->name = g_strdup(sn.name);
1179 info->date_nsec = sn.date_nsec;
1180 info->date_sec = sn.date_sec;
1181 info->vm_state_size = sn.vm_state_size;
1182 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1183 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1184 if (sn.icount != -1ULL) {
1185 info->icount = sn.icount;
1186 info->has_icount = true;
1189 return info;
1191 out_aio_context:
1192 aio_context_release(aio_context);
1193 return NULL;
1196 /* internal snapshot private data */
1197 typedef struct InternalSnapshotState {
1198 BlockDriverState *bs;
1199 QEMUSnapshotInfo sn;
1200 bool created;
1201 } InternalSnapshotState;
1203 static void internal_snapshot_abort(void *opaque);
1204 static void internal_snapshot_clean(void *opaque);
1205 TransactionActionDrv internal_snapshot_drv = {
1206 .abort = internal_snapshot_abort,
1207 .clean = internal_snapshot_clean,
1210 static void internal_snapshot_action(BlockdevSnapshotInternal *internal,
1211 Transaction *tran, Error **errp)
1213 Error *local_err = NULL;
1214 const char *device;
1215 const char *name;
1216 BlockDriverState *bs;
1217 QEMUSnapshotInfo old_sn, *sn;
1218 bool ret;
1219 int64_t rt;
1220 InternalSnapshotState *state = g_new0(InternalSnapshotState, 1);
1221 AioContext *aio_context;
1222 int ret1;
1224 tran_add(tran, &internal_snapshot_drv, state);
1226 device = internal->device;
1227 name = internal->name;
1229 bs = qmp_get_root_bs(device, errp);
1230 if (!bs) {
1231 return;
1234 aio_context = bdrv_get_aio_context(bs);
1235 aio_context_acquire(aio_context);
1237 state->bs = bs;
1239 /* Paired with .clean() */
1240 bdrv_drained_begin(bs);
1242 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1243 goto out;
1246 if (bdrv_is_read_only(bs)) {
1247 error_setg(errp, "Device '%s' is read only", device);
1248 goto out;
1251 if (!bdrv_can_snapshot(bs)) {
1252 error_setg(errp, "Block format '%s' used by device '%s' "
1253 "does not support internal snapshots",
1254 bs->drv->format_name, device);
1255 goto out;
1258 if (!strlen(name)) {
1259 error_setg(errp, "Name is empty");
1260 goto out;
1263 /* check whether a snapshot with name exist */
1264 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1265 &local_err);
1266 if (local_err) {
1267 error_propagate(errp, local_err);
1268 goto out;
1269 } else if (ret) {
1270 error_setg(errp,
1271 "Snapshot with name '%s' already exists on device '%s'",
1272 name, device);
1273 goto out;
1276 /* 3. take the snapshot */
1277 sn = &state->sn;
1278 pstrcpy(sn->name, sizeof(sn->name), name);
1279 rt = g_get_real_time();
1280 sn->date_sec = rt / G_USEC_PER_SEC;
1281 sn->date_nsec = (rt % G_USEC_PER_SEC) * 1000;
1282 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1283 if (replay_mode != REPLAY_MODE_NONE) {
1284 sn->icount = replay_get_current_icount();
1285 } else {
1286 sn->icount = -1ULL;
1289 ret1 = bdrv_snapshot_create(bs, sn);
1290 if (ret1 < 0) {
1291 error_setg_errno(errp, -ret1,
1292 "Failed to create snapshot '%s' on device '%s'",
1293 name, device);
1294 goto out;
1297 /* 4. succeed, mark a snapshot is created */
1298 state->created = true;
1300 out:
1301 aio_context_release(aio_context);
1304 static void internal_snapshot_abort(void *opaque)
1306 InternalSnapshotState *state = opaque;
1307 BlockDriverState *bs = state->bs;
1308 QEMUSnapshotInfo *sn = &state->sn;
1309 AioContext *aio_context;
1310 Error *local_error = NULL;
1312 if (!state->created) {
1313 return;
1316 aio_context = bdrv_get_aio_context(state->bs);
1317 aio_context_acquire(aio_context);
1319 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1320 error_reportf_err(local_error,
1321 "Failed to delete snapshot with id '%s' and "
1322 "name '%s' on device '%s' in abort: ",
1323 sn->id_str, sn->name,
1324 bdrv_get_device_name(bs));
1327 aio_context_release(aio_context);
1330 static void internal_snapshot_clean(void *opaque)
1332 g_autofree InternalSnapshotState *state = opaque;
1333 AioContext *aio_context;
1335 if (!state->bs) {
1336 return;
1339 aio_context = bdrv_get_aio_context(state->bs);
1340 aio_context_acquire(aio_context);
1342 bdrv_drained_end(state->bs);
1344 aio_context_release(aio_context);
1347 /* external snapshot private data */
1348 typedef struct ExternalSnapshotState {
1349 BlockDriverState *old_bs;
1350 BlockDriverState *new_bs;
1351 bool overlay_appended;
1352 } ExternalSnapshotState;
1354 static void external_snapshot_commit(void *opaque);
1355 static void external_snapshot_abort(void *opaque);
1356 static void external_snapshot_clean(void *opaque);
1357 TransactionActionDrv external_snapshot_drv = {
1358 .commit = external_snapshot_commit,
1359 .abort = external_snapshot_abort,
1360 .clean = external_snapshot_clean,
1363 static void external_snapshot_action(TransactionAction *action,
1364 Transaction *tran, Error **errp)
1366 int ret;
1367 int flags = 0;
1368 QDict *options = NULL;
1369 Error *local_err = NULL;
1370 /* Device and node name of the image to generate the snapshot from */
1371 const char *device;
1372 const char *node_name;
1373 /* Reference to the new image (for 'blockdev-snapshot') */
1374 const char *snapshot_ref;
1375 /* File name of the new image (for 'blockdev-snapshot-sync') */
1376 const char *new_image_file;
1377 ExternalSnapshotState *state = g_new0(ExternalSnapshotState, 1);
1378 AioContext *aio_context;
1379 uint64_t perm, shared;
1381 /* TODO We'll eventually have to take a writer lock in this function */
1382 GRAPH_RDLOCK_GUARD_MAINLOOP();
1384 tran_add(tran, &external_snapshot_drv, state);
1386 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1387 * purpose but a different set of parameters */
1388 switch (action->type) {
1389 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1391 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1392 device = s->node;
1393 node_name = s->node;
1394 new_image_file = NULL;
1395 snapshot_ref = s->overlay;
1397 break;
1398 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1400 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1401 device = s->device;
1402 node_name = s->node_name;
1403 new_image_file = s->snapshot_file;
1404 snapshot_ref = NULL;
1406 break;
1407 default:
1408 g_assert_not_reached();
1411 /* start processing */
1413 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1414 if (!state->old_bs) {
1415 return;
1418 aio_context = bdrv_get_aio_context(state->old_bs);
1419 aio_context_acquire(aio_context);
1421 /* Paired with .clean() */
1422 bdrv_drained_begin(state->old_bs);
1424 if (!bdrv_is_inserted(state->old_bs)) {
1425 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1426 goto out;
1429 if (bdrv_op_is_blocked(state->old_bs,
1430 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1431 goto out;
1434 if (!bdrv_is_read_only(state->old_bs)) {
1435 if (bdrv_flush(state->old_bs)) {
1436 error_setg(errp, QERR_IO_ERROR);
1437 goto out;
1441 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1442 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1443 const char *format = s->format ?: "qcow2";
1444 enum NewImageMode mode;
1445 const char *snapshot_node_name = s->snapshot_node_name;
1447 if (node_name && !snapshot_node_name) {
1448 error_setg(errp, "New overlay node-name missing");
1449 goto out;
1452 if (snapshot_node_name &&
1453 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1454 error_setg(errp, "New overlay node-name already in use");
1455 goto out;
1458 flags = state->old_bs->open_flags;
1459 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1460 flags |= BDRV_O_NO_BACKING;
1462 /* create new image w/backing file */
1463 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1464 if (mode != NEW_IMAGE_MODE_EXISTING) {
1465 int64_t size = bdrv_getlength(state->old_bs);
1466 if (size < 0) {
1467 error_setg_errno(errp, -size, "bdrv_getlength failed");
1468 goto out;
1470 bdrv_refresh_filename(state->old_bs);
1472 aio_context_release(aio_context);
1473 bdrv_img_create(new_image_file, format,
1474 state->old_bs->filename,
1475 state->old_bs->drv->format_name,
1476 NULL, size, flags, false, &local_err);
1477 aio_context_acquire(aio_context);
1479 if (local_err) {
1480 error_propagate(errp, local_err);
1481 goto out;
1485 options = qdict_new();
1486 if (snapshot_node_name) {
1487 qdict_put_str(options, "node-name", snapshot_node_name);
1489 qdict_put_str(options, "driver", format);
1491 aio_context_release(aio_context);
1493 aio_context_acquire(qemu_get_aio_context());
1494 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1495 errp);
1496 aio_context_release(qemu_get_aio_context());
1498 /* We will manually add the backing_hd field to the bs later */
1499 if (!state->new_bs) {
1500 return;
1503 aio_context_acquire(aio_context);
1506 * Allow attaching a backing file to an overlay that's already in use only
1507 * if the parents don't assume that they are already seeing a valid image.
1508 * (Specifically, allow it as a mirror target, which is write-only access.)
1510 bdrv_get_cumulative_perm(state->new_bs, &perm, &shared);
1511 if (perm & BLK_PERM_CONSISTENT_READ) {
1512 error_setg(errp, "The overlay is already in use");
1513 goto out;
1516 if (state->new_bs->drv->is_filter) {
1517 error_setg(errp, "Filters cannot be used as overlays");
1518 goto out;
1521 if (bdrv_cow_child(state->new_bs)) {
1522 error_setg(errp, "The overlay already has a backing image");
1523 goto out;
1526 if (!state->new_bs->drv->supports_backing) {
1527 error_setg(errp, "The overlay does not support backing images");
1528 goto out;
1531 ret = bdrv_append(state->new_bs, state->old_bs, errp);
1532 if (ret < 0) {
1533 goto out;
1535 state->overlay_appended = true;
1537 out:
1538 aio_context_release(aio_context);
1541 static void external_snapshot_commit(void *opaque)
1543 ExternalSnapshotState *state = opaque;
1544 AioContext *aio_context;
1546 aio_context = bdrv_get_aio_context(state->old_bs);
1547 aio_context_acquire(aio_context);
1549 /* We don't need (or want) to use the transactional
1550 * bdrv_reopen_multiple() across all the entries at once, because we
1551 * don't want to abort all of them if one of them fails the reopen */
1552 if (!qatomic_read(&state->old_bs->copy_on_read)) {
1553 bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1556 aio_context_release(aio_context);
1559 static void external_snapshot_abort(void *opaque)
1561 ExternalSnapshotState *state = opaque;
1562 if (state->new_bs) {
1563 if (state->overlay_appended) {
1564 AioContext *aio_context;
1565 AioContext *tmp_context;
1566 int ret;
1568 aio_context = bdrv_get_aio_context(state->old_bs);
1569 aio_context_acquire(aio_context);
1571 bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd()
1572 close state->old_bs; we need it */
1573 bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1576 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1577 * the main AioContext. As we're still going to be using it, return
1578 * it to the AioContext it was before.
1580 tmp_context = bdrv_get_aio_context(state->old_bs);
1581 if (aio_context != tmp_context) {
1582 aio_context_release(aio_context);
1583 aio_context_acquire(tmp_context);
1585 ret = bdrv_try_change_aio_context(state->old_bs,
1586 aio_context, NULL, NULL);
1587 assert(ret == 0);
1589 aio_context_release(tmp_context);
1590 aio_context_acquire(aio_context);
1593 bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1594 bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1596 aio_context_release(aio_context);
1601 static void external_snapshot_clean(void *opaque)
1603 g_autofree ExternalSnapshotState *state = opaque;
1604 AioContext *aio_context;
1606 if (!state->old_bs) {
1607 return;
1610 aio_context = bdrv_get_aio_context(state->old_bs);
1611 aio_context_acquire(aio_context);
1613 bdrv_drained_end(state->old_bs);
1614 bdrv_unref(state->new_bs);
1616 aio_context_release(aio_context);
1619 typedef struct DriveBackupState {
1620 BlockDriverState *bs;
1621 BlockJob *job;
1622 } DriveBackupState;
1624 static BlockJob *do_backup_common(BackupCommon *backup,
1625 BlockDriverState *bs,
1626 BlockDriverState *target_bs,
1627 AioContext *aio_context,
1628 JobTxn *txn, Error **errp);
1630 static void drive_backup_commit(void *opaque);
1631 static void drive_backup_abort(void *opaque);
1632 static void drive_backup_clean(void *opaque);
1633 TransactionActionDrv drive_backup_drv = {
1634 .commit = drive_backup_commit,
1635 .abort = drive_backup_abort,
1636 .clean = drive_backup_clean,
1639 static void drive_backup_action(DriveBackup *backup,
1640 JobTxn *block_job_txn,
1641 Transaction *tran, Error **errp)
1643 DriveBackupState *state = g_new0(DriveBackupState, 1);
1644 BlockDriverState *bs;
1645 BlockDriverState *target_bs;
1646 BlockDriverState *source = NULL;
1647 AioContext *aio_context;
1648 AioContext *old_context;
1649 const char *format;
1650 QDict *options;
1651 Error *local_err = NULL;
1652 int flags;
1653 int64_t size;
1654 bool set_backing_hd = false;
1655 int ret;
1657 tran_add(tran, &drive_backup_drv, state);
1659 if (!backup->has_mode) {
1660 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1663 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1664 if (!bs) {
1665 return;
1668 if (!bs->drv) {
1669 error_setg(errp, "Device has no medium");
1670 return;
1673 aio_context = bdrv_get_aio_context(bs);
1674 aio_context_acquire(aio_context);
1676 state->bs = bs;
1677 /* Paired with .clean() */
1678 bdrv_drained_begin(bs);
1680 format = backup->format;
1681 if (!format && backup->mode != NEW_IMAGE_MODE_EXISTING) {
1682 format = bs->drv->format_name;
1685 /* Early check to avoid creating target */
1686 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
1687 goto out;
1690 flags = bs->open_flags | BDRV_O_RDWR;
1693 * See if we have a backing HD we can use to create our new image
1694 * on top of.
1696 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
1698 * Backup will not replace the source by the target, so none
1699 * of the filters skipped here will be removed (in contrast to
1700 * mirror). Therefore, we can skip all of them when looking
1701 * for the first COW relationship.
1703 source = bdrv_cow_bs(bdrv_skip_filters(bs));
1704 if (!source) {
1705 backup->sync = MIRROR_SYNC_MODE_FULL;
1708 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
1709 source = bs;
1710 flags |= BDRV_O_NO_BACKING;
1711 set_backing_hd = true;
1714 size = bdrv_getlength(bs);
1715 if (size < 0) {
1716 error_setg_errno(errp, -size, "bdrv_getlength failed");
1717 goto out;
1720 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
1721 assert(format);
1722 if (source) {
1723 /* Implicit filters should not appear in the filename */
1724 BlockDriverState *explicit_backing =
1725 bdrv_skip_implicit_filters(source);
1727 bdrv_refresh_filename(explicit_backing);
1728 bdrv_img_create(backup->target, format,
1729 explicit_backing->filename,
1730 explicit_backing->drv->format_name, NULL,
1731 size, flags, false, &local_err);
1732 } else {
1733 bdrv_img_create(backup->target, format, NULL, NULL, NULL,
1734 size, flags, false, &local_err);
1738 if (local_err) {
1739 error_propagate(errp, local_err);
1740 goto out;
1743 options = qdict_new();
1744 qdict_put_str(options, "discard", "unmap");
1745 qdict_put_str(options, "detect-zeroes", "unmap");
1746 if (format) {
1747 qdict_put_str(options, "driver", format);
1749 aio_context_release(aio_context);
1751 aio_context_acquire(qemu_get_aio_context());
1752 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
1753 aio_context_release(qemu_get_aio_context());
1755 if (!target_bs) {
1756 return;
1759 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1760 old_context = bdrv_get_aio_context(target_bs);
1761 aio_context_acquire(old_context);
1763 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
1764 if (ret < 0) {
1765 bdrv_unref(target_bs);
1766 aio_context_release(old_context);
1767 return;
1770 aio_context_release(old_context);
1771 aio_context_acquire(aio_context);
1773 if (set_backing_hd) {
1774 if (bdrv_set_backing_hd(target_bs, source, errp) < 0) {
1775 goto unref;
1779 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1780 bs, target_bs, aio_context,
1781 block_job_txn, errp);
1783 unref:
1784 bdrv_unref(target_bs);
1785 out:
1786 aio_context_release(aio_context);
1789 static void drive_backup_commit(void *opaque)
1791 DriveBackupState *state = opaque;
1792 AioContext *aio_context;
1794 aio_context = bdrv_get_aio_context(state->bs);
1795 aio_context_acquire(aio_context);
1797 assert(state->job);
1798 job_start(&state->job->job);
1800 aio_context_release(aio_context);
1803 static void drive_backup_abort(void *opaque)
1805 DriveBackupState *state = opaque;
1807 if (state->job) {
1808 job_cancel_sync(&state->job->job, true);
1812 static void drive_backup_clean(void *opaque)
1814 g_autofree DriveBackupState *state = opaque;
1815 AioContext *aio_context;
1817 if (!state->bs) {
1818 return;
1821 aio_context = bdrv_get_aio_context(state->bs);
1822 aio_context_acquire(aio_context);
1824 bdrv_drained_end(state->bs);
1826 aio_context_release(aio_context);
1829 typedef struct BlockdevBackupState {
1830 BlockDriverState *bs;
1831 BlockJob *job;
1832 } BlockdevBackupState;
1834 static void blockdev_backup_commit(void *opaque);
1835 static void blockdev_backup_abort(void *opaque);
1836 static void blockdev_backup_clean(void *opaque);
1837 TransactionActionDrv blockdev_backup_drv = {
1838 .commit = blockdev_backup_commit,
1839 .abort = blockdev_backup_abort,
1840 .clean = blockdev_backup_clean,
1843 static void blockdev_backup_action(BlockdevBackup *backup,
1844 JobTxn *block_job_txn,
1845 Transaction *tran, Error **errp)
1847 BlockdevBackupState *state = g_new0(BlockdevBackupState, 1);
1848 BlockDriverState *bs;
1849 BlockDriverState *target_bs;
1850 AioContext *aio_context;
1851 AioContext *old_context;
1852 int ret;
1854 tran_add(tran, &blockdev_backup_drv, state);
1856 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1857 if (!bs) {
1858 return;
1861 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1862 if (!target_bs) {
1863 return;
1866 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
1867 aio_context = bdrv_get_aio_context(bs);
1868 old_context = bdrv_get_aio_context(target_bs);
1869 aio_context_acquire(old_context);
1871 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
1872 if (ret < 0) {
1873 aio_context_release(old_context);
1874 return;
1877 aio_context_release(old_context);
1878 aio_context_acquire(aio_context);
1879 state->bs = bs;
1881 /* Paired with .clean() */
1882 bdrv_drained_begin(state->bs);
1884 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1885 bs, target_bs, aio_context,
1886 block_job_txn, errp);
1888 aio_context_release(aio_context);
1891 static void blockdev_backup_commit(void *opaque)
1893 BlockdevBackupState *state = opaque;
1894 AioContext *aio_context;
1896 aio_context = bdrv_get_aio_context(state->bs);
1897 aio_context_acquire(aio_context);
1899 assert(state->job);
1900 job_start(&state->job->job);
1902 aio_context_release(aio_context);
1905 static void blockdev_backup_abort(void *opaque)
1907 BlockdevBackupState *state = opaque;
1909 if (state->job) {
1910 job_cancel_sync(&state->job->job, true);
1914 static void blockdev_backup_clean(void *opaque)
1916 g_autofree BlockdevBackupState *state = opaque;
1917 AioContext *aio_context;
1919 if (!state->bs) {
1920 return;
1923 aio_context = bdrv_get_aio_context(state->bs);
1924 aio_context_acquire(aio_context);
1926 bdrv_drained_end(state->bs);
1928 aio_context_release(aio_context);
1931 typedef struct BlockDirtyBitmapState {
1932 BdrvDirtyBitmap *bitmap;
1933 BlockDriverState *bs;
1934 HBitmap *backup;
1935 bool was_enabled;
1936 } BlockDirtyBitmapState;
1938 static void block_dirty_bitmap_add_abort(void *opaque);
1939 TransactionActionDrv block_dirty_bitmap_add_drv = {
1940 .abort = block_dirty_bitmap_add_abort,
1941 .clean = g_free,
1944 static void block_dirty_bitmap_add_action(BlockDirtyBitmapAdd *action,
1945 Transaction *tran, Error **errp)
1947 Error *local_err = NULL;
1948 BlockDirtyBitmapState *state = g_new0(BlockDirtyBitmapState, 1);
1950 tran_add(tran, &block_dirty_bitmap_add_drv, state);
1952 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1953 qmp_block_dirty_bitmap_add(action->node, action->name,
1954 action->has_granularity, action->granularity,
1955 action->has_persistent, action->persistent,
1956 action->has_disabled, action->disabled,
1957 &local_err);
1959 if (!local_err) {
1960 state->bitmap = block_dirty_bitmap_lookup(action->node, action->name,
1961 NULL, &error_abort);
1962 } else {
1963 error_propagate(errp, local_err);
1967 static void block_dirty_bitmap_add_abort(void *opaque)
1969 BlockDirtyBitmapState *state = opaque;
1971 if (state->bitmap) {
1972 bdrv_release_dirty_bitmap(state->bitmap);
1976 static void block_dirty_bitmap_restore(void *opaque);
1977 static void block_dirty_bitmap_free_backup(void *opaque);
1978 TransactionActionDrv block_dirty_bitmap_clear_drv = {
1979 .abort = block_dirty_bitmap_restore,
1980 .commit = block_dirty_bitmap_free_backup,
1981 .clean = g_free,
1984 static void block_dirty_bitmap_clear_action(BlockDirtyBitmap *action,
1985 Transaction *tran, Error **errp)
1987 BlockDirtyBitmapState *state = g_new0(BlockDirtyBitmapState, 1);
1989 tran_add(tran, &block_dirty_bitmap_clear_drv, state);
1991 state->bitmap = block_dirty_bitmap_lookup(action->node,
1992 action->name,
1993 &state->bs,
1994 errp);
1995 if (!state->bitmap) {
1996 return;
1999 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2000 return;
2003 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2006 static void block_dirty_bitmap_restore(void *opaque)
2008 BlockDirtyBitmapState *state = opaque;
2010 if (state->backup) {
2011 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2015 static void block_dirty_bitmap_free_backup(void *opaque)
2017 BlockDirtyBitmapState *state = opaque;
2019 hbitmap_free(state->backup);
2022 static void block_dirty_bitmap_enable_abort(void *opaque);
2023 TransactionActionDrv block_dirty_bitmap_enable_drv = {
2024 .abort = block_dirty_bitmap_enable_abort,
2025 .clean = g_free,
2028 static void block_dirty_bitmap_enable_action(BlockDirtyBitmap *action,
2029 Transaction *tran, Error **errp)
2031 BlockDirtyBitmapState *state = g_new0(BlockDirtyBitmapState, 1);
2033 tran_add(tran, &block_dirty_bitmap_enable_drv, state);
2035 state->bitmap = block_dirty_bitmap_lookup(action->node,
2036 action->name,
2037 NULL,
2038 errp);
2039 if (!state->bitmap) {
2040 return;
2043 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2044 return;
2047 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2048 bdrv_enable_dirty_bitmap(state->bitmap);
2051 static void block_dirty_bitmap_enable_abort(void *opaque)
2053 BlockDirtyBitmapState *state = opaque;
2055 if (!state->was_enabled) {
2056 bdrv_disable_dirty_bitmap(state->bitmap);
2060 static void block_dirty_bitmap_disable_abort(void *opaque);
2061 TransactionActionDrv block_dirty_bitmap_disable_drv = {
2062 .abort = block_dirty_bitmap_disable_abort,
2063 .clean = g_free,
2066 static void block_dirty_bitmap_disable_action(BlockDirtyBitmap *action,
2067 Transaction *tran, Error **errp)
2069 BlockDirtyBitmapState *state = g_new0(BlockDirtyBitmapState, 1);
2071 tran_add(tran, &block_dirty_bitmap_disable_drv, state);
2073 state->bitmap = block_dirty_bitmap_lookup(action->node,
2074 action->name,
2075 NULL,
2076 errp);
2077 if (!state->bitmap) {
2078 return;
2081 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2082 return;
2085 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2086 bdrv_disable_dirty_bitmap(state->bitmap);
2089 static void block_dirty_bitmap_disable_abort(void *opaque)
2091 BlockDirtyBitmapState *state = opaque;
2093 if (state->was_enabled) {
2094 bdrv_enable_dirty_bitmap(state->bitmap);
2098 TransactionActionDrv block_dirty_bitmap_merge_drv = {
2099 .commit = block_dirty_bitmap_free_backup,
2100 .abort = block_dirty_bitmap_restore,
2101 .clean = g_free,
2104 static void block_dirty_bitmap_merge_action(BlockDirtyBitmapMerge *action,
2105 Transaction *tran, Error **errp)
2107 BlockDirtyBitmapState *state = g_new0(BlockDirtyBitmapState, 1);
2109 tran_add(tran, &block_dirty_bitmap_merge_drv, state);
2111 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2112 action->bitmaps, &state->backup,
2113 errp);
2116 static void block_dirty_bitmap_remove_commit(void *opaque);
2117 static void block_dirty_bitmap_remove_abort(void *opaque);
2118 TransactionActionDrv block_dirty_bitmap_remove_drv = {
2119 .commit = block_dirty_bitmap_remove_commit,
2120 .abort = block_dirty_bitmap_remove_abort,
2121 .clean = g_free,
2124 static void block_dirty_bitmap_remove_action(BlockDirtyBitmap *action,
2125 Transaction *tran, Error **errp)
2127 BlockDirtyBitmapState *state = g_new0(BlockDirtyBitmapState, 1);
2129 tran_add(tran, &block_dirty_bitmap_remove_drv, state);
2132 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2133 false, &state->bs, errp);
2134 if (state->bitmap) {
2135 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2136 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2140 static void block_dirty_bitmap_remove_abort(void *opaque)
2142 BlockDirtyBitmapState *state = opaque;
2144 if (state->bitmap) {
2145 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2146 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2150 static void block_dirty_bitmap_remove_commit(void *opaque)
2152 BlockDirtyBitmapState *state = opaque;
2154 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2155 bdrv_release_dirty_bitmap(state->bitmap);
2158 static void abort_commit(void *opaque);
2159 TransactionActionDrv abort_drv = {
2160 .commit = abort_commit,
2163 static void abort_action(Transaction *tran, Error **errp)
2165 tran_add(tran, &abort_drv, NULL);
2166 error_setg(errp, "Transaction aborted using Abort action");
2169 static void abort_commit(void *opaque)
2171 g_assert_not_reached(); /* this action never succeeds */
2174 static void transaction_action(TransactionAction *act, JobTxn *block_job_txn,
2175 Transaction *tran, Error **errp)
2177 switch (act->type) {
2178 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
2179 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
2180 external_snapshot_action(act, tran, errp);
2181 return;
2182 case TRANSACTION_ACTION_KIND_DRIVE_BACKUP:
2183 drive_backup_action(act->u.drive_backup.data,
2184 block_job_txn, tran, errp);
2185 return;
2186 case TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP:
2187 blockdev_backup_action(act->u.blockdev_backup.data,
2188 block_job_txn, tran, errp);
2189 return;
2190 case TRANSACTION_ACTION_KIND_ABORT:
2191 abort_action(tran, errp);
2192 return;
2193 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC:
2194 internal_snapshot_action(act->u.blockdev_snapshot_internal_sync.data,
2195 tran, errp);
2196 return;
2197 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD:
2198 block_dirty_bitmap_add_action(act->u.block_dirty_bitmap_add.data,
2199 tran, errp);
2200 return;
2201 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR:
2202 block_dirty_bitmap_clear_action(act->u.block_dirty_bitmap_clear.data,
2203 tran, errp);
2204 return;
2205 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE:
2206 block_dirty_bitmap_enable_action(act->u.block_dirty_bitmap_enable.data,
2207 tran, errp);
2208 return;
2209 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE:
2210 block_dirty_bitmap_disable_action(
2211 act->u.block_dirty_bitmap_disable.data, tran, errp);
2212 return;
2213 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE:
2214 block_dirty_bitmap_merge_action(act->u.block_dirty_bitmap_merge.data,
2215 tran, errp);
2216 return;
2217 case TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE:
2218 block_dirty_bitmap_remove_action(act->u.block_dirty_bitmap_remove.data,
2219 tran, errp);
2220 return;
2222 * Where are transactions for MIRROR, COMMIT and STREAM?
2223 * Although these blockjobs use transaction callbacks like the backup job,
2224 * these jobs do not necessarily adhere to transaction semantics.
2225 * These jobs may not fully undo all of their actions on abort, nor do they
2226 * necessarily work in transactions with more than one job in them.
2228 case TRANSACTION_ACTION_KIND__MAX:
2229 default:
2230 g_assert_not_reached();
2236 * 'Atomic' group operations. The operations are performed as a set, and if
2237 * any fail then we roll back all operations in the group.
2239 * Always run under BQL.
2241 void qmp_transaction(TransactionActionList *actions,
2242 struct TransactionProperties *properties,
2243 Error **errp)
2245 TransactionActionList *act;
2246 JobTxn *block_job_txn = NULL;
2247 Error *local_err = NULL;
2248 Transaction *tran;
2249 ActionCompletionMode comp_mode =
2250 properties ? properties->completion_mode :
2251 ACTION_COMPLETION_MODE_INDIVIDUAL;
2253 GLOBAL_STATE_CODE();
2255 /* Does this transaction get canceled as a group on failure?
2256 * If not, we don't really need to make a JobTxn.
2258 if (comp_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2259 for (act = actions; act; act = act->next) {
2260 TransactionActionKind type = act->value->type;
2262 if (type != TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP &&
2263 type != TRANSACTION_ACTION_KIND_DRIVE_BACKUP)
2265 error_setg(errp,
2266 "Action '%s' does not support transaction property "
2267 "completion-mode = %s",
2268 TransactionActionKind_str(type),
2269 ActionCompletionMode_str(comp_mode));
2270 return;
2274 block_job_txn = job_txn_new();
2277 /* drain all i/o before any operations */
2278 bdrv_drain_all();
2280 tran = tran_new();
2282 /* We don't do anything in this loop that commits us to the operations */
2283 for (act = actions; act; act = act->next) {
2284 transaction_action(act->value, block_job_txn, tran, &local_err);
2285 if (local_err) {
2286 error_propagate(errp, local_err);
2287 goto delete_and_fail;
2291 tran_commit(tran);
2293 /* success */
2294 goto exit;
2296 delete_and_fail:
2297 /* failure, and it is all-or-none; roll back all operations */
2298 tran_abort(tran);
2299 exit:
2300 job_txn_unref(block_job_txn);
2303 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2304 const char *name,
2305 Error **errp)
2307 BdrvDirtyBitmap *bitmap;
2308 BlockDriverState *bs;
2309 BlockDirtyBitmapSha256 *ret = NULL;
2310 char *sha256;
2312 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2313 if (!bitmap || !bs) {
2314 return NULL;
2317 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2318 if (sha256 == NULL) {
2319 return NULL;
2322 ret = g_new(BlockDirtyBitmapSha256, 1);
2323 ret->sha256 = sha256;
2325 return ret;
2328 void coroutine_fn qmp_block_resize(const char *device, const char *node_name,
2329 int64_t size, Error **errp)
2331 Error *local_err = NULL;
2332 BlockBackend *blk;
2333 BlockDriverState *bs;
2334 AioContext *old_ctx;
2336 bs = bdrv_lookup_bs(device, node_name, &local_err);
2337 if (local_err) {
2338 error_propagate(errp, local_err);
2339 return;
2342 if (size < 0) {
2343 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2344 return;
2347 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2348 error_setg(errp, QERR_DEVICE_IN_USE, device);
2349 return;
2352 blk = blk_co_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2353 if (!blk) {
2354 return;
2357 bdrv_co_lock(bs);
2358 bdrv_drained_begin(bs);
2359 bdrv_co_unlock(bs);
2361 old_ctx = bdrv_co_enter(bs);
2362 blk_co_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2363 bdrv_co_leave(bs, old_ctx);
2365 bdrv_co_lock(bs);
2366 bdrv_drained_end(bs);
2367 blk_co_unref(blk);
2368 bdrv_co_unlock(bs);
2371 void qmp_block_stream(const char *job_id, const char *device,
2372 const char *base,
2373 const char *base_node,
2374 const char *backing_file,
2375 const char *bottom,
2376 bool has_speed, int64_t speed,
2377 bool has_on_error, BlockdevOnError on_error,
2378 const char *filter_node_name,
2379 bool has_auto_finalize, bool auto_finalize,
2380 bool has_auto_dismiss, bool auto_dismiss,
2381 Error **errp)
2383 BlockDriverState *bs, *iter, *iter_end;
2384 BlockDriverState *base_bs = NULL;
2385 BlockDriverState *bottom_bs = NULL;
2386 AioContext *aio_context;
2387 Error *local_err = NULL;
2388 int job_flags = JOB_DEFAULT;
2390 if (base && base_node) {
2391 error_setg(errp, "'base' and 'base-node' cannot be specified "
2392 "at the same time");
2393 return;
2396 if (base && bottom) {
2397 error_setg(errp, "'base' and 'bottom' cannot be specified "
2398 "at the same time");
2399 return;
2402 if (bottom && base_node) {
2403 error_setg(errp, "'bottom' and 'base-node' cannot be specified "
2404 "at the same time");
2405 return;
2408 if (!has_on_error) {
2409 on_error = BLOCKDEV_ON_ERROR_REPORT;
2412 bs = bdrv_lookup_bs(device, device, errp);
2413 if (!bs) {
2414 return;
2417 aio_context = bdrv_get_aio_context(bs);
2418 aio_context_acquire(aio_context);
2420 if (base) {
2421 base_bs = bdrv_find_backing_image(bs, base);
2422 if (base_bs == NULL) {
2423 error_setg(errp, "Can't find '%s' in the backing chain", base);
2424 goto out;
2426 assert(bdrv_get_aio_context(base_bs) == aio_context);
2429 if (base_node) {
2430 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2431 if (!base_bs) {
2432 goto out;
2434 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2435 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2436 base_node, device);
2437 goto out;
2439 assert(bdrv_get_aio_context(base_bs) == aio_context);
2440 bdrv_refresh_filename(base_bs);
2443 if (bottom) {
2444 bottom_bs = bdrv_lookup_bs(NULL, bottom, errp);
2445 if (!bottom_bs) {
2446 goto out;
2448 if (!bottom_bs->drv) {
2449 error_setg(errp, "Node '%s' is not open", bottom);
2450 goto out;
2452 if (bottom_bs->drv->is_filter) {
2453 error_setg(errp, "Node '%s' is a filter, use a non-filter node "
2454 "as 'bottom'", bottom);
2455 goto out;
2457 if (!bdrv_chain_contains(bs, bottom_bs)) {
2458 error_setg(errp, "Node '%s' is not in a chain starting from '%s'",
2459 bottom, device);
2460 goto out;
2462 assert(bdrv_get_aio_context(bottom_bs) == aio_context);
2466 * Check for op blockers in the whole chain between bs and base (or bottom)
2468 iter_end = bottom ? bdrv_filter_or_cow_bs(bottom_bs) : base_bs;
2469 for (iter = bs; iter && iter != iter_end;
2470 iter = bdrv_filter_or_cow_bs(iter))
2472 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2473 goto out;
2477 /* if we are streaming the entire chain, the result will have no backing
2478 * file, and specifying one is therefore an error */
2479 if (!base_bs && backing_file) {
2480 error_setg(errp, "backing file specified, but streaming the "
2481 "entire chain");
2482 goto out;
2485 if (has_auto_finalize && !auto_finalize) {
2486 job_flags |= JOB_MANUAL_FINALIZE;
2488 if (has_auto_dismiss && !auto_dismiss) {
2489 job_flags |= JOB_MANUAL_DISMISS;
2492 stream_start(job_id, bs, base_bs, backing_file,
2493 bottom_bs, job_flags, has_speed ? speed : 0, on_error,
2494 filter_node_name, &local_err);
2495 if (local_err) {
2496 error_propagate(errp, local_err);
2497 goto out;
2500 trace_qmp_block_stream(bs);
2502 out:
2503 aio_context_release(aio_context);
2506 void qmp_block_commit(const char *job_id, const char *device,
2507 const char *base_node,
2508 const char *base,
2509 const char *top_node,
2510 const char *top,
2511 const char *backing_file,
2512 bool has_speed, int64_t speed,
2513 bool has_on_error, BlockdevOnError on_error,
2514 const char *filter_node_name,
2515 bool has_auto_finalize, bool auto_finalize,
2516 bool has_auto_dismiss, bool auto_dismiss,
2517 Error **errp)
2519 BlockDriverState *bs;
2520 BlockDriverState *iter;
2521 BlockDriverState *base_bs, *top_bs;
2522 AioContext *aio_context;
2523 Error *local_err = NULL;
2524 int job_flags = JOB_DEFAULT;
2525 uint64_t top_perm, top_shared;
2527 /* TODO We'll eventually have to take a writer lock in this function */
2528 GRAPH_RDLOCK_GUARD_MAINLOOP();
2530 if (!has_speed) {
2531 speed = 0;
2533 if (!has_on_error) {
2534 on_error = BLOCKDEV_ON_ERROR_REPORT;
2536 if (has_auto_finalize && !auto_finalize) {
2537 job_flags |= JOB_MANUAL_FINALIZE;
2539 if (has_auto_dismiss && !auto_dismiss) {
2540 job_flags |= JOB_MANUAL_DISMISS;
2543 /* Important Note:
2544 * libvirt relies on the DeviceNotFound error class in order to probe for
2545 * live commit feature versions; for this to work, we must make sure to
2546 * perform the device lookup before any generic errors that may occur in a
2547 * scenario in which all optional arguments are omitted. */
2548 bs = qmp_get_root_bs(device, &local_err);
2549 if (!bs) {
2550 bs = bdrv_lookup_bs(device, device, NULL);
2551 if (!bs) {
2552 error_free(local_err);
2553 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2554 "Device '%s' not found", device);
2555 } else {
2556 error_propagate(errp, local_err);
2558 return;
2561 aio_context = bdrv_get_aio_context(bs);
2562 aio_context_acquire(aio_context);
2564 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2565 goto out;
2568 /* default top_bs is the active layer */
2569 top_bs = bs;
2571 if (top_node && top) {
2572 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2573 goto out;
2574 } else if (top_node) {
2575 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2576 if (top_bs == NULL) {
2577 goto out;
2579 if (!bdrv_chain_contains(bs, top_bs)) {
2580 error_setg(errp, "'%s' is not in this backing file chain",
2581 top_node);
2582 goto out;
2584 } else if (top) {
2585 /* This strcmp() is just a shortcut, there is no need to
2586 * refresh @bs's filename. If it mismatches,
2587 * bdrv_find_backing_image() will do the refresh and may still
2588 * return @bs. */
2589 if (strcmp(bs->filename, top) != 0) {
2590 top_bs = bdrv_find_backing_image(bs, top);
2594 if (top_bs == NULL) {
2595 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2596 goto out;
2599 assert(bdrv_get_aio_context(top_bs) == aio_context);
2601 if (base_node && base) {
2602 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2603 goto out;
2604 } else if (base_node) {
2605 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2606 if (base_bs == NULL) {
2607 goto out;
2609 if (!bdrv_chain_contains(top_bs, base_bs)) {
2610 error_setg(errp, "'%s' is not in this backing file chain",
2611 base_node);
2612 goto out;
2614 } else if (base) {
2615 base_bs = bdrv_find_backing_image(top_bs, base);
2616 if (base_bs == NULL) {
2617 error_setg(errp, "Can't find '%s' in the backing chain", base);
2618 goto out;
2620 } else {
2621 base_bs = bdrv_find_base(top_bs);
2622 if (base_bs == NULL) {
2623 error_setg(errp, "There is no backimg image");
2624 goto out;
2628 assert(bdrv_get_aio_context(base_bs) == aio_context);
2630 for (iter = top_bs; iter != bdrv_filter_or_cow_bs(base_bs);
2631 iter = bdrv_filter_or_cow_bs(iter))
2633 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2634 goto out;
2638 /* Do not allow attempts to commit an image into itself */
2639 if (top_bs == base_bs) {
2640 error_setg(errp, "cannot commit an image into itself");
2641 goto out;
2645 * Active commit is required if and only if someone has taken a
2646 * WRITE permission on the top node. Historically, we have always
2647 * used active commit for top nodes, so continue that practice
2648 * lest we possibly break clients that rely on this behavior, e.g.
2649 * to later attach this node to a writing parent.
2650 * (Active commit is never really wrong.)
2652 bdrv_get_cumulative_perm(top_bs, &top_perm, &top_shared);
2653 if (top_perm & BLK_PERM_WRITE ||
2654 bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs))
2656 if (backing_file) {
2657 if (bdrv_skip_filters(top_bs) == bdrv_skip_filters(bs)) {
2658 error_setg(errp, "'backing-file' specified,"
2659 " but 'top' is the active layer");
2660 } else {
2661 error_setg(errp, "'backing-file' specified, but 'top' has a "
2662 "writer on it");
2664 goto out;
2666 if (!job_id) {
2668 * Emulate here what block_job_create() does, because it
2669 * is possible that @bs != @top_bs (the block job should
2670 * be named after @bs, even if @top_bs is the actual
2671 * source)
2673 job_id = bdrv_get_device_name(bs);
2675 commit_active_start(job_id, top_bs, base_bs, job_flags, speed, on_error,
2676 filter_node_name, NULL, NULL, false, &local_err);
2677 } else {
2678 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2679 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2680 goto out;
2682 commit_start(job_id, bs, base_bs, top_bs, job_flags,
2683 speed, on_error, backing_file,
2684 filter_node_name, &local_err);
2686 if (local_err != NULL) {
2687 error_propagate(errp, local_err);
2688 goto out;
2691 out:
2692 aio_context_release(aio_context);
2695 /* Common QMP interface for drive-backup and blockdev-backup */
2696 static BlockJob *do_backup_common(BackupCommon *backup,
2697 BlockDriverState *bs,
2698 BlockDriverState *target_bs,
2699 AioContext *aio_context,
2700 JobTxn *txn, Error **errp)
2702 BlockJob *job = NULL;
2703 BdrvDirtyBitmap *bmap = NULL;
2704 BackupPerf perf = { .max_workers = 64 };
2705 int job_flags = JOB_DEFAULT;
2707 if (!backup->has_speed) {
2708 backup->speed = 0;
2710 if (!backup->has_on_source_error) {
2711 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2713 if (!backup->has_on_target_error) {
2714 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2716 if (!backup->has_auto_finalize) {
2717 backup->auto_finalize = true;
2719 if (!backup->has_auto_dismiss) {
2720 backup->auto_dismiss = true;
2722 if (!backup->has_compress) {
2723 backup->compress = false;
2726 if (backup->x_perf) {
2727 if (backup->x_perf->has_use_copy_range) {
2728 perf.use_copy_range = backup->x_perf->use_copy_range;
2730 if (backup->x_perf->has_max_workers) {
2731 perf.max_workers = backup->x_perf->max_workers;
2733 if (backup->x_perf->has_max_chunk) {
2734 perf.max_chunk = backup->x_perf->max_chunk;
2738 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2739 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2740 /* done before desugaring 'incremental' to print the right message */
2741 if (!backup->bitmap) {
2742 error_setg(errp, "must provide a valid bitmap name for "
2743 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2744 return NULL;
2748 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2749 if (backup->has_bitmap_mode &&
2750 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2751 error_setg(errp, "Bitmap sync mode must be '%s' "
2752 "when using sync mode '%s'",
2753 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2754 MirrorSyncMode_str(backup->sync));
2755 return NULL;
2757 backup->has_bitmap_mode = true;
2758 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2759 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2762 if (backup->bitmap) {
2763 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2764 if (!bmap) {
2765 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2766 return NULL;
2768 if (!backup->has_bitmap_mode) {
2769 error_setg(errp, "Bitmap sync mode must be given "
2770 "when providing a bitmap");
2771 return NULL;
2773 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2774 return NULL;
2777 /* This does not produce a useful bitmap artifact: */
2778 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2779 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2780 " outputs", MirrorSyncMode_str(backup->sync));
2781 return NULL;
2784 /* If the bitmap isn't used for input or output, this is useless: */
2785 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2786 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2787 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2788 " when combined with sync mode '%s'",
2789 BitmapSyncMode_str(backup->bitmap_mode),
2790 MirrorSyncMode_str(backup->sync));
2791 return NULL;
2795 if (!backup->bitmap && backup->has_bitmap_mode) {
2796 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2797 return NULL;
2800 if (!backup->auto_finalize) {
2801 job_flags |= JOB_MANUAL_FINALIZE;
2803 if (!backup->auto_dismiss) {
2804 job_flags |= JOB_MANUAL_DISMISS;
2807 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2808 backup->sync, bmap, backup->bitmap_mode,
2809 backup->compress,
2810 backup->filter_node_name,
2811 &perf,
2812 backup->on_source_error,
2813 backup->on_target_error,
2814 job_flags, NULL, NULL, txn, errp);
2815 return job;
2818 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2820 TransactionAction action = {
2821 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2822 .u.drive_backup.data = backup,
2824 blockdev_do_action(&action, errp);
2827 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2828 bool flat,
2829 Error **errp)
2831 bool return_flat = has_flat && flat;
2833 return bdrv_named_nodes_list(return_flat, errp);
2836 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2838 return bdrv_get_xdbg_block_graph(errp);
2841 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2843 TransactionAction action = {
2844 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2845 .u.blockdev_backup.data = backup,
2847 blockdev_do_action(&action, errp);
2850 /* Parameter check and block job starting for drive mirroring.
2851 * Caller should hold @device and @target's aio context (must be the same).
2853 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2854 BlockDriverState *target,
2855 const char *replaces,
2856 enum MirrorSyncMode sync,
2857 BlockMirrorBackingMode backing_mode,
2858 bool zero_target,
2859 bool has_speed, int64_t speed,
2860 bool has_granularity, uint32_t granularity,
2861 bool has_buf_size, int64_t buf_size,
2862 bool has_on_source_error,
2863 BlockdevOnError on_source_error,
2864 bool has_on_target_error,
2865 BlockdevOnError on_target_error,
2866 bool has_unmap, bool unmap,
2867 const char *filter_node_name,
2868 bool has_copy_mode, MirrorCopyMode copy_mode,
2869 bool has_auto_finalize, bool auto_finalize,
2870 bool has_auto_dismiss, bool auto_dismiss,
2871 Error **errp)
2873 BlockDriverState *unfiltered_bs;
2874 int job_flags = JOB_DEFAULT;
2876 GLOBAL_STATE_CODE();
2877 GRAPH_RDLOCK_GUARD_MAINLOOP();
2879 if (!has_speed) {
2880 speed = 0;
2882 if (!has_on_source_error) {
2883 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2885 if (!has_on_target_error) {
2886 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2888 if (!has_granularity) {
2889 granularity = 0;
2891 if (!has_buf_size) {
2892 buf_size = 0;
2894 if (!has_unmap) {
2895 unmap = true;
2897 if (!has_copy_mode) {
2898 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
2900 if (has_auto_finalize && !auto_finalize) {
2901 job_flags |= JOB_MANUAL_FINALIZE;
2903 if (has_auto_dismiss && !auto_dismiss) {
2904 job_flags |= JOB_MANUAL_DISMISS;
2907 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2908 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2909 "a value in range [512B, 64MB]");
2910 return;
2912 if (granularity & (granularity - 1)) {
2913 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2914 "a power of 2");
2915 return;
2918 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
2919 return;
2921 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
2922 return;
2925 if (!bdrv_backing_chain_next(bs) && sync == MIRROR_SYNC_MODE_TOP) {
2926 sync = MIRROR_SYNC_MODE_FULL;
2929 if (!replaces) {
2930 /* We want to mirror from @bs, but keep implicit filters on top */
2931 unfiltered_bs = bdrv_skip_implicit_filters(bs);
2932 if (unfiltered_bs != bs) {
2933 replaces = unfiltered_bs->node_name;
2937 if (replaces) {
2938 BlockDriverState *to_replace_bs;
2939 AioContext *replace_aio_context;
2940 int64_t bs_size, replace_size;
2942 bs_size = bdrv_getlength(bs);
2943 if (bs_size < 0) {
2944 error_setg_errno(errp, -bs_size, "Failed to query device's size");
2945 return;
2948 to_replace_bs = check_to_replace_node(bs, replaces, errp);
2949 if (!to_replace_bs) {
2950 return;
2953 replace_aio_context = bdrv_get_aio_context(to_replace_bs);
2954 aio_context_acquire(replace_aio_context);
2955 replace_size = bdrv_getlength(to_replace_bs);
2956 aio_context_release(replace_aio_context);
2958 if (replace_size < 0) {
2959 error_setg_errno(errp, -replace_size,
2960 "Failed to query the replacement node's size");
2961 return;
2963 if (bs_size != replace_size) {
2964 error_setg(errp, "cannot replace image with a mirror image of "
2965 "different size");
2966 return;
2970 /* pass the node name to replace to mirror start since it's loose coupling
2971 * and will allow to check whether the node still exist at mirror completion
2973 mirror_start(job_id, bs, target,
2974 replaces, job_flags,
2975 speed, granularity, buf_size, sync, backing_mode, zero_target,
2976 on_source_error, on_target_error, unmap, filter_node_name,
2977 copy_mode, errp);
2980 void qmp_drive_mirror(DriveMirror *arg, Error **errp)
2982 BlockDriverState *bs;
2983 BlockDriverState *target_backing_bs, *target_bs;
2984 AioContext *aio_context;
2985 AioContext *old_context;
2986 BlockMirrorBackingMode backing_mode;
2987 Error *local_err = NULL;
2988 QDict *options = NULL;
2989 int flags;
2990 int64_t size;
2991 const char *format = arg->format;
2992 bool zero_target;
2993 int ret;
2995 bs = qmp_get_root_bs(arg->device, errp);
2996 if (!bs) {
2997 return;
3000 /* Early check to avoid creating target */
3001 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
3002 return;
3005 aio_context = bdrv_get_aio_context(bs);
3006 aio_context_acquire(aio_context);
3008 if (!arg->has_mode) {
3009 arg->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
3012 if (!arg->format) {
3013 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3014 ? NULL : bs->drv->format_name);
3017 flags = bs->open_flags | BDRV_O_RDWR;
3018 target_backing_bs = bdrv_cow_bs(bdrv_skip_filters(bs));
3019 if (!target_backing_bs && arg->sync == MIRROR_SYNC_MODE_TOP) {
3020 arg->sync = MIRROR_SYNC_MODE_FULL;
3022 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3023 target_backing_bs = bs;
3026 size = bdrv_getlength(bs);
3027 if (size < 0) {
3028 error_setg_errno(errp, -size, "bdrv_getlength failed");
3029 goto out;
3032 if (arg->replaces) {
3033 if (!arg->node_name) {
3034 error_setg(errp, "a node-name must be provided when replacing a"
3035 " named node of the graph");
3036 goto out;
3040 if (arg->mode == NEW_IMAGE_MODE_ABSOLUTE_PATHS) {
3041 backing_mode = MIRROR_SOURCE_BACKING_CHAIN;
3042 } else {
3043 backing_mode = MIRROR_OPEN_BACKING_CHAIN;
3046 /* Don't open backing image in create() */
3047 flags |= BDRV_O_NO_BACKING;
3049 if ((arg->sync == MIRROR_SYNC_MODE_FULL || !target_backing_bs)
3050 && arg->mode != NEW_IMAGE_MODE_EXISTING)
3052 /* create new image w/o backing file */
3053 assert(format);
3054 bdrv_img_create(arg->target, format,
3055 NULL, NULL, NULL, size, flags, false, &local_err);
3056 } else {
3057 /* Implicit filters should not appear in the filename */
3058 BlockDriverState *explicit_backing =
3059 bdrv_skip_implicit_filters(target_backing_bs);
3061 switch (arg->mode) {
3062 case NEW_IMAGE_MODE_EXISTING:
3063 break;
3064 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3065 /* create new image with backing file */
3066 bdrv_refresh_filename(explicit_backing);
3067 bdrv_img_create(arg->target, format,
3068 explicit_backing->filename,
3069 explicit_backing->drv->format_name,
3070 NULL, size, flags, false, &local_err);
3071 break;
3072 default:
3073 abort();
3077 if (local_err) {
3078 error_propagate(errp, local_err);
3079 goto out;
3082 options = qdict_new();
3083 if (arg->node_name) {
3084 qdict_put_str(options, "node-name", arg->node_name);
3086 if (format) {
3087 qdict_put_str(options, "driver", format);
3089 aio_context_release(aio_context);
3091 /* Mirroring takes care of copy-on-write using the source's backing
3092 * file.
3094 aio_context_acquire(qemu_get_aio_context());
3095 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3096 aio_context_release(qemu_get_aio_context());
3098 if (!target_bs) {
3099 return;
3102 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3103 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3104 !bdrv_has_zero_init(target_bs)));
3107 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3108 old_context = bdrv_get_aio_context(target_bs);
3109 aio_context_acquire(old_context);
3111 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
3112 if (ret < 0) {
3113 bdrv_unref(target_bs);
3114 aio_context_release(old_context);
3115 return;
3118 aio_context_release(old_context);
3119 aio_context_acquire(aio_context);
3121 blockdev_mirror_common(arg->job_id, bs, target_bs,
3122 arg->replaces, arg->sync,
3123 backing_mode, zero_target,
3124 arg->has_speed, arg->speed,
3125 arg->has_granularity, arg->granularity,
3126 arg->has_buf_size, arg->buf_size,
3127 arg->has_on_source_error, arg->on_source_error,
3128 arg->has_on_target_error, arg->on_target_error,
3129 arg->has_unmap, arg->unmap,
3130 NULL,
3131 arg->has_copy_mode, arg->copy_mode,
3132 arg->has_auto_finalize, arg->auto_finalize,
3133 arg->has_auto_dismiss, arg->auto_dismiss,
3134 errp);
3135 bdrv_unref(target_bs);
3136 out:
3137 aio_context_release(aio_context);
3140 void qmp_blockdev_mirror(const char *job_id,
3141 const char *device, const char *target,
3142 const char *replaces,
3143 MirrorSyncMode sync,
3144 bool has_speed, int64_t speed,
3145 bool has_granularity, uint32_t granularity,
3146 bool has_buf_size, int64_t buf_size,
3147 bool has_on_source_error,
3148 BlockdevOnError on_source_error,
3149 bool has_on_target_error,
3150 BlockdevOnError on_target_error,
3151 const char *filter_node_name,
3152 bool has_copy_mode, MirrorCopyMode copy_mode,
3153 bool has_auto_finalize, bool auto_finalize,
3154 bool has_auto_dismiss, bool auto_dismiss,
3155 Error **errp)
3157 BlockDriverState *bs;
3158 BlockDriverState *target_bs;
3159 AioContext *aio_context;
3160 AioContext *old_context;
3161 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3162 bool zero_target;
3163 int ret;
3165 bs = qmp_get_root_bs(device, errp);
3166 if (!bs) {
3167 return;
3170 target_bs = bdrv_lookup_bs(target, target, errp);
3171 if (!target_bs) {
3172 return;
3175 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3177 /* Honor bdrv_try_change_aio_context() context acquisition requirements. */
3178 old_context = bdrv_get_aio_context(target_bs);
3179 aio_context = bdrv_get_aio_context(bs);
3180 aio_context_acquire(old_context);
3182 ret = bdrv_try_change_aio_context(target_bs, aio_context, NULL, errp);
3184 aio_context_release(old_context);
3185 aio_context_acquire(aio_context);
3187 if (ret < 0) {
3188 goto out;
3191 blockdev_mirror_common(job_id, bs, target_bs,
3192 replaces, sync, backing_mode,
3193 zero_target, has_speed, speed,
3194 has_granularity, granularity,
3195 has_buf_size, buf_size,
3196 has_on_source_error, on_source_error,
3197 has_on_target_error, on_target_error,
3198 true, true, filter_node_name,
3199 has_copy_mode, copy_mode,
3200 has_auto_finalize, auto_finalize,
3201 has_auto_dismiss, auto_dismiss,
3202 errp);
3203 out:
3204 aio_context_release(aio_context);
3208 * Get a block job using its ID. Called with job_mutex held.
3210 static BlockJob *find_block_job_locked(const char *id, Error **errp)
3212 BlockJob *job;
3214 assert(id != NULL);
3216 job = block_job_get_locked(id);
3218 if (!job) {
3219 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3220 "Block job '%s' not found", id);
3221 return NULL;
3224 return job;
3227 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3229 BlockJob *job;
3231 JOB_LOCK_GUARD();
3232 job = find_block_job_locked(device, errp);
3234 if (!job) {
3235 return;
3238 block_job_set_speed_locked(job, speed, errp);
3241 void qmp_block_job_cancel(const char *device,
3242 bool has_force, bool force, Error **errp)
3244 BlockJob *job;
3246 JOB_LOCK_GUARD();
3247 job = find_block_job_locked(device, errp);
3249 if (!job) {
3250 return;
3253 if (!has_force) {
3254 force = false;
3257 if (job_user_paused_locked(&job->job) && !force) {
3258 error_setg(errp, "The block job for device '%s' is currently paused",
3259 device);
3260 return;
3263 trace_qmp_block_job_cancel(job);
3264 job_user_cancel_locked(&job->job, force, errp);
3267 void qmp_block_job_pause(const char *device, Error **errp)
3269 BlockJob *job;
3271 JOB_LOCK_GUARD();
3272 job = find_block_job_locked(device, errp);
3274 if (!job) {
3275 return;
3278 trace_qmp_block_job_pause(job);
3279 job_user_pause_locked(&job->job, errp);
3282 void qmp_block_job_resume(const char *device, Error **errp)
3284 BlockJob *job;
3286 JOB_LOCK_GUARD();
3287 job = find_block_job_locked(device, errp);
3289 if (!job) {
3290 return;
3293 trace_qmp_block_job_resume(job);
3294 job_user_resume_locked(&job->job, errp);
3297 void qmp_block_job_complete(const char *device, Error **errp)
3299 BlockJob *job;
3301 JOB_LOCK_GUARD();
3302 job = find_block_job_locked(device, errp);
3304 if (!job) {
3305 return;
3308 trace_qmp_block_job_complete(job);
3309 job_complete_locked(&job->job, errp);
3312 void qmp_block_job_finalize(const char *id, Error **errp)
3314 BlockJob *job;
3316 JOB_LOCK_GUARD();
3317 job = find_block_job_locked(id, errp);
3319 if (!job) {
3320 return;
3323 trace_qmp_block_job_finalize(job);
3324 job_ref_locked(&job->job);
3325 job_finalize_locked(&job->job, errp);
3327 job_unref_locked(&job->job);
3330 void qmp_block_job_dismiss(const char *id, Error **errp)
3332 BlockJob *bjob;
3333 Job *job;
3335 JOB_LOCK_GUARD();
3336 bjob = find_block_job_locked(id, errp);
3338 if (!bjob) {
3339 return;
3342 trace_qmp_block_job_dismiss(bjob);
3343 job = &bjob->job;
3344 job_dismiss_locked(&job, errp);
3347 void qmp_change_backing_file(const char *device,
3348 const char *image_node_name,
3349 const char *backing_file,
3350 Error **errp)
3352 BlockDriverState *bs = NULL;
3353 AioContext *aio_context;
3354 BlockDriverState *image_bs = NULL;
3355 Error *local_err = NULL;
3356 bool ro;
3357 int ret;
3359 bs = qmp_get_root_bs(device, errp);
3360 if (!bs) {
3361 return;
3364 aio_context = bdrv_get_aio_context(bs);
3365 aio_context_acquire(aio_context);
3367 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3368 if (local_err) {
3369 error_propagate(errp, local_err);
3370 goto out;
3373 if (!image_bs) {
3374 error_setg(errp, "image file not found");
3375 goto out;
3378 if (bdrv_find_base(image_bs) == image_bs) {
3379 error_setg(errp, "not allowing backing file change on an image "
3380 "without a backing file");
3381 goto out;
3384 /* even though we are not necessarily operating on bs, we need it to
3385 * determine if block ops are currently prohibited on the chain */
3386 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3387 goto out;
3390 /* final sanity check */
3391 if (!bdrv_chain_contains(bs, image_bs)) {
3392 error_setg(errp, "'%s' and image file are not in the same chain",
3393 device);
3394 goto out;
3397 /* if not r/w, reopen to make r/w */
3398 ro = bdrv_is_read_only(image_bs);
3400 if (ro) {
3401 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3402 goto out;
3406 ret = bdrv_change_backing_file(image_bs, backing_file,
3407 image_bs->drv ? image_bs->drv->format_name : "",
3408 false);
3410 if (ret < 0) {
3411 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3412 backing_file);
3413 /* don't exit here, so we can try to restore open flags if
3414 * appropriate */
3417 if (ro) {
3418 bdrv_reopen_set_read_only(image_bs, true, errp);
3421 out:
3422 aio_context_release(aio_context);
3425 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3427 BlockDriverState *bs;
3428 QObject *obj;
3429 Visitor *v = qobject_output_visitor_new(&obj);
3430 QDict *qdict;
3432 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3433 visit_complete(v, &obj);
3434 qdict = qobject_to(QDict, obj);
3436 qdict_flatten(qdict);
3438 if (!qdict_get_try_str(qdict, "node-name")) {
3439 error_setg(errp, "'node-name' must be specified for the root node");
3440 goto fail;
3443 bs = bds_tree_init(qdict, errp);
3444 if (!bs) {
3445 goto fail;
3448 bdrv_set_monitor_owned(bs);
3450 fail:
3451 visit_free(v);
3454 void qmp_blockdev_reopen(BlockdevOptionsList *reopen_list, Error **errp)
3456 BlockReopenQueue *queue = NULL;
3458 /* Add each one of the BDS that we want to reopen to the queue */
3459 for (; reopen_list != NULL; reopen_list = reopen_list->next) {
3460 BlockdevOptions *options = reopen_list->value;
3461 BlockDriverState *bs;
3462 AioContext *ctx;
3463 QObject *obj;
3464 Visitor *v;
3465 QDict *qdict;
3467 /* Check for the selected node name */
3468 if (!options->node_name) {
3469 error_setg(errp, "node-name not specified");
3470 goto fail;
3473 bs = bdrv_find_node(options->node_name);
3474 if (!bs) {
3475 error_setg(errp, "Failed to find node with node-name='%s'",
3476 options->node_name);
3477 goto fail;
3480 /* Put all options in a QDict and flatten it */
3481 v = qobject_output_visitor_new(&obj);
3482 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3483 visit_complete(v, &obj);
3484 visit_free(v);
3486 qdict = qobject_to(QDict, obj);
3488 qdict_flatten(qdict);
3490 ctx = bdrv_get_aio_context(bs);
3491 aio_context_acquire(ctx);
3493 queue = bdrv_reopen_queue(queue, bs, qdict, false);
3495 aio_context_release(ctx);
3498 /* Perform the reopen operation */
3499 bdrv_reopen_multiple(queue, errp);
3500 queue = NULL;
3502 fail:
3503 bdrv_reopen_queue_free(queue);
3506 void qmp_blockdev_del(const char *node_name, Error **errp)
3508 AioContext *aio_context;
3509 BlockDriverState *bs;
3511 GLOBAL_STATE_CODE();
3513 bs = bdrv_find_node(node_name);
3514 if (!bs) {
3515 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3516 return;
3518 if (bdrv_has_blk(bs)) {
3519 error_setg(errp, "Node %s is in use", node_name);
3520 return;
3522 aio_context = bdrv_get_aio_context(bs);
3523 aio_context_acquire(aio_context);
3525 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3526 goto out;
3529 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3530 error_setg(errp, "Node %s is not owned by the monitor",
3531 bs->node_name);
3532 goto out;
3535 if (bs->refcnt > 1) {
3536 error_setg(errp, "Block device %s is in use",
3537 bdrv_get_device_or_node_name(bs));
3538 goto out;
3541 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3542 bdrv_unref(bs);
3544 out:
3545 aio_context_release(aio_context);
3548 static BdrvChild * GRAPH_RDLOCK
3549 bdrv_find_child(BlockDriverState *parent_bs, const char *child_name)
3551 BdrvChild *child;
3553 QLIST_FOREACH(child, &parent_bs->children, next) {
3554 if (strcmp(child->name, child_name) == 0) {
3555 return child;
3559 return NULL;
3562 void qmp_x_blockdev_change(const char *parent, const char *child,
3563 const char *node, Error **errp)
3565 BlockDriverState *parent_bs, *new_bs = NULL;
3566 BdrvChild *p_child;
3568 bdrv_graph_wrlock(NULL);
3570 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3571 if (!parent_bs) {
3572 goto out;
3575 if (!child == !node) {
3576 if (child) {
3577 error_setg(errp, "The parameters child and node are in conflict");
3578 } else {
3579 error_setg(errp, "Either child or node must be specified");
3581 goto out;
3584 if (child) {
3585 p_child = bdrv_find_child(parent_bs, child);
3586 if (!p_child) {
3587 error_setg(errp, "Node '%s' does not have child '%s'",
3588 parent, child);
3589 goto out;
3591 bdrv_del_child(parent_bs, p_child, errp);
3594 if (node) {
3595 new_bs = bdrv_find_node(node);
3596 if (!new_bs) {
3597 error_setg(errp, "Node '%s' not found", node);
3598 goto out;
3600 bdrv_add_child(parent_bs, new_bs, errp);
3603 out:
3604 bdrv_graph_wrunlock();
3607 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3609 BlockJobInfoList *head = NULL, **tail = &head;
3610 BlockJob *job;
3612 JOB_LOCK_GUARD();
3614 for (job = block_job_next_locked(NULL); job;
3615 job = block_job_next_locked(job)) {
3616 BlockJobInfo *value;
3618 if (block_job_is_internal(job)) {
3619 continue;
3621 value = block_job_query_locked(job, errp);
3622 if (!value) {
3623 qapi_free_BlockJobInfoList(head);
3624 return NULL;
3626 QAPI_LIST_APPEND(tail, value);
3629 return head;
3632 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3633 bool has_force, bool force, Error **errp)
3635 AioContext *old_context;
3636 AioContext *new_context;
3637 BlockDriverState *bs;
3639 bs = bdrv_find_node(node_name);
3640 if (!bs) {
3641 error_setg(errp, "Failed to find node with node-name='%s'", node_name);
3642 return;
3645 /* Protects against accidents. */
3646 if (!(has_force && force) && bdrv_has_blk(bs)) {
3647 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3648 "be in use (use force=true to override this check)",
3649 node_name);
3650 return;
3653 if (iothread->type == QTYPE_QSTRING) {
3654 IOThread *obj = iothread_by_id(iothread->u.s);
3655 if (!obj) {
3656 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3657 return;
3660 new_context = iothread_get_aio_context(obj);
3661 } else {
3662 new_context = qemu_get_aio_context();
3665 old_context = bdrv_get_aio_context(bs);
3666 aio_context_acquire(old_context);
3668 bdrv_try_change_aio_context(bs, new_context, NULL, errp);
3670 aio_context_release(old_context);
3673 QemuOptsList qemu_common_drive_opts = {
3674 .name = "drive",
3675 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3676 .desc = {
3678 .name = "snapshot",
3679 .type = QEMU_OPT_BOOL,
3680 .help = "enable/disable snapshot mode",
3682 .name = "aio",
3683 .type = QEMU_OPT_STRING,
3684 .help = "host AIO implementation (threads, native, io_uring)",
3686 .name = BDRV_OPT_CACHE_WB,
3687 .type = QEMU_OPT_BOOL,
3688 .help = "Enable writeback mode",
3690 .name = "format",
3691 .type = QEMU_OPT_STRING,
3692 .help = "disk format (raw, qcow2, ...)",
3694 .name = "rerror",
3695 .type = QEMU_OPT_STRING,
3696 .help = "read error action",
3698 .name = "werror",
3699 .type = QEMU_OPT_STRING,
3700 .help = "write error action",
3702 .name = BDRV_OPT_READ_ONLY,
3703 .type = QEMU_OPT_BOOL,
3704 .help = "open drive file as read-only",
3707 THROTTLE_OPTS,
3710 .name = "throttling.group",
3711 .type = QEMU_OPT_STRING,
3712 .help = "name of the block throttling group",
3714 .name = "copy-on-read",
3715 .type = QEMU_OPT_BOOL,
3716 .help = "copy read data from backing file into image file",
3718 .name = "detect-zeroes",
3719 .type = QEMU_OPT_STRING,
3720 .help = "try to optimize zero writes (off, on, unmap)",
3722 .name = "stats-account-invalid",
3723 .type = QEMU_OPT_BOOL,
3724 .help = "whether to account for invalid I/O operations "
3725 "in the statistics",
3727 .name = "stats-account-failed",
3728 .type = QEMU_OPT_BOOL,
3729 .help = "whether to account for failed I/O operations "
3730 "in the statistics",
3732 { /* end of list */ }
3736 QemuOptsList qemu_drive_opts = {
3737 .name = "drive",
3738 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3739 .desc = {
3741 * no elements => accept any params
3742 * validation will happen later
3744 { /* end of list */ }