iotests: 194: wait for migration completion on target too
[qemu/kevin.git] / blockdev.c
blob72df193ca73b30d1722a130fbddf12823ad3556c
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/qdict.h"
39 #include "block/throttle-groups.h"
40 #include "monitor/monitor.h"
41 #include "qemu/error-report.h"
42 #include "qemu/option.h"
43 #include "qemu/qemu-print.h"
44 #include "qemu/config-file.h"
45 #include "qapi/qapi-commands-block.h"
46 #include "qapi/qapi-commands-transaction.h"
47 #include "qapi/qapi-visit-block-core.h"
48 #include "qapi/qmp/qdict.h"
49 #include "qapi/qmp/qnum.h"
50 #include "qapi/qmp/qstring.h"
51 #include "qapi/error.h"
52 #include "qapi/qmp/qerror.h"
53 #include "qapi/qmp/qlist.h"
54 #include "qapi/qobject-output-visitor.h"
55 #include "sysemu/sysemu.h"
56 #include "sysemu/iothread.h"
57 #include "block/block_int.h"
58 #include "block/trace.h"
59 #include "sysemu/arch_init.h"
60 #include "sysemu/qtest.h"
61 #include "sysemu/runstate.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 QTAILQ_HEAD(, BlockDriverState) monitor_bdrv_states =
68 QTAILQ_HEAD_INITIALIZER(monitor_bdrv_states);
70 void bdrv_set_monitor_owned(BlockDriverState *bs)
72 QTAILQ_INSERT_TAIL(&monitor_bdrv_states, bs, monitor_list);
75 static const char *const if_name[IF_COUNT] = {
76 [IF_NONE] = "none",
77 [IF_IDE] = "ide",
78 [IF_SCSI] = "scsi",
79 [IF_FLOPPY] = "floppy",
80 [IF_PFLASH] = "pflash",
81 [IF_MTD] = "mtd",
82 [IF_SD] = "sd",
83 [IF_VIRTIO] = "virtio",
84 [IF_XEN] = "xen",
87 static int if_max_devs[IF_COUNT] = {
89 * Do not change these numbers! They govern how drive option
90 * index maps to unit and bus. That mapping is ABI.
92 * All controllers used to implement if=T drives need to support
93 * if_max_devs[T] units, for any T with if_max_devs[T] != 0.
94 * Otherwise, some index values map to "impossible" bus, unit
95 * values.
97 * For instance, if you change [IF_SCSI] to 255, -drive
98 * if=scsi,index=12 no longer means bus=1,unit=5, but
99 * bus=0,unit=12. With an lsi53c895a controller (7 units max),
100 * the drive can't be set up. Regression.
102 [IF_IDE] = 2,
103 [IF_SCSI] = 7,
107 * Boards may call this to offer board-by-board overrides
108 * of the default, global values.
110 void override_max_devs(BlockInterfaceType type, int max_devs)
112 BlockBackend *blk;
113 DriveInfo *dinfo;
115 if (max_devs <= 0) {
116 return;
119 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
120 dinfo = blk_legacy_dinfo(blk);
121 if (dinfo->type == type) {
122 fprintf(stderr, "Cannot override units-per-bus property of"
123 " the %s interface, because a drive of that type has"
124 " already been added.\n", if_name[type]);
125 g_assert_not_reached();
129 if_max_devs[type] = max_devs;
133 * We automatically delete the drive when a device using it gets
134 * unplugged. Questionable feature, but we can't just drop it.
135 * Device models call blockdev_mark_auto_del() to schedule the
136 * automatic deletion, and generic qdev code calls blockdev_auto_del()
137 * when deletion is actually safe.
139 void blockdev_mark_auto_del(BlockBackend *blk)
141 DriveInfo *dinfo = blk_legacy_dinfo(blk);
142 BlockJob *job;
144 if (!dinfo) {
145 return;
148 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
149 if (block_job_has_bdrv(job, blk_bs(blk))) {
150 AioContext *aio_context = job->job.aio_context;
151 aio_context_acquire(aio_context);
153 job_cancel(&job->job, false);
155 aio_context_release(aio_context);
159 dinfo->auto_del = 1;
162 void blockdev_auto_del(BlockBackend *blk)
164 DriveInfo *dinfo = blk_legacy_dinfo(blk);
166 if (dinfo && dinfo->auto_del) {
167 monitor_remove_blk(blk);
168 blk_unref(blk);
173 * Returns the current mapping of how many units per bus
174 * a particular interface can support.
176 * A positive integer indicates n units per bus.
177 * 0 implies the mapping has not been established.
178 * -1 indicates an invalid BlockInterfaceType was given.
180 int drive_get_max_devs(BlockInterfaceType type)
182 if (type >= IF_IDE && type < IF_COUNT) {
183 return if_max_devs[type];
186 return -1;
189 static int drive_index_to_bus_id(BlockInterfaceType type, int index)
191 int max_devs = if_max_devs[type];
192 return max_devs ? index / max_devs : 0;
195 static int drive_index_to_unit_id(BlockInterfaceType type, int index)
197 int max_devs = if_max_devs[type];
198 return max_devs ? index % max_devs : index;
201 QemuOpts *drive_def(const char *optstr)
203 return qemu_opts_parse_noisily(qemu_find_opts("drive"), optstr, false);
206 QemuOpts *drive_add(BlockInterfaceType type, int index, const char *file,
207 const char *optstr)
209 QemuOpts *opts;
211 opts = drive_def(optstr);
212 if (!opts) {
213 return NULL;
215 if (type != IF_DEFAULT) {
216 qemu_opt_set(opts, "if", if_name[type], &error_abort);
218 if (index >= 0) {
219 qemu_opt_set_number(opts, "index", index, &error_abort);
221 if (file)
222 qemu_opt_set(opts, "file", file, &error_abort);
223 return opts;
226 DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
228 BlockBackend *blk;
229 DriveInfo *dinfo;
231 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
232 dinfo = blk_legacy_dinfo(blk);
233 if (dinfo && dinfo->type == type
234 && dinfo->bus == bus && dinfo->unit == unit) {
235 return dinfo;
239 return NULL;
242 void drive_check_orphaned(void)
244 BlockBackend *blk;
245 DriveInfo *dinfo;
246 Location loc;
247 bool orphans = false;
249 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
250 dinfo = blk_legacy_dinfo(blk);
251 if (!blk_get_attached_dev(blk) && !dinfo->is_default &&
252 dinfo->type != IF_NONE) {
253 loc_push_none(&loc);
254 qemu_opts_loc_restore(dinfo->opts);
255 error_report("machine type does not support"
256 " if=%s,bus=%d,unit=%d",
257 if_name[dinfo->type], dinfo->bus, dinfo->unit);
258 loc_pop(&loc);
259 orphans = true;
263 if (orphans) {
264 exit(1);
268 DriveInfo *drive_get_by_index(BlockInterfaceType type, int index)
270 return drive_get(type,
271 drive_index_to_bus_id(type, index),
272 drive_index_to_unit_id(type, index));
275 int drive_get_max_bus(BlockInterfaceType type)
277 int max_bus;
278 BlockBackend *blk;
279 DriveInfo *dinfo;
281 max_bus = -1;
282 for (blk = blk_next(NULL); blk; blk = blk_next(blk)) {
283 dinfo = blk_legacy_dinfo(blk);
284 if (dinfo && dinfo->type == type && dinfo->bus > max_bus) {
285 max_bus = dinfo->bus;
288 return max_bus;
291 /* Get a block device. This should only be used for single-drive devices
292 (e.g. SD/Floppy/MTD). Multi-disk devices (scsi/ide) should use the
293 appropriate bus. */
294 DriveInfo *drive_get_next(BlockInterfaceType type)
296 static int next_block_unit[IF_COUNT];
298 return drive_get(type, 0, next_block_unit[type]++);
301 static void bdrv_format_print(void *opaque, const char *name)
303 qemu_printf(" %s", name);
306 typedef struct {
307 QEMUBH *bh;
308 BlockDriverState *bs;
309 } BDRVPutRefBH;
311 static int parse_block_error_action(const char *buf, bool is_read, Error **errp)
313 if (!strcmp(buf, "ignore")) {
314 return BLOCKDEV_ON_ERROR_IGNORE;
315 } else if (!is_read && !strcmp(buf, "enospc")) {
316 return BLOCKDEV_ON_ERROR_ENOSPC;
317 } else if (!strcmp(buf, "stop")) {
318 return BLOCKDEV_ON_ERROR_STOP;
319 } else if (!strcmp(buf, "report")) {
320 return BLOCKDEV_ON_ERROR_REPORT;
321 } else {
322 error_setg(errp, "'%s' invalid %s error action",
323 buf, is_read ? "read" : "write");
324 return -1;
328 static bool parse_stats_intervals(BlockAcctStats *stats, QList *intervals,
329 Error **errp)
331 const QListEntry *entry;
332 for (entry = qlist_first(intervals); entry; entry = qlist_next(entry)) {
333 switch (qobject_type(entry->value)) {
335 case QTYPE_QSTRING: {
336 unsigned long long length;
337 const char *str = qstring_get_str(qobject_to(QString,
338 entry->value));
339 if (parse_uint_full(str, &length, 10) == 0 &&
340 length > 0 && length <= UINT_MAX) {
341 block_acct_add_interval(stats, (unsigned) length);
342 } else {
343 error_setg(errp, "Invalid interval length: %s", str);
344 return false;
346 break;
349 case QTYPE_QNUM: {
350 int64_t length = qnum_get_int(qobject_to(QNum, entry->value));
352 if (length > 0 && length <= UINT_MAX) {
353 block_acct_add_interval(stats, (unsigned) length);
354 } else {
355 error_setg(errp, "Invalid interval length: %" PRId64, length);
356 return false;
358 break;
361 default:
362 error_setg(errp, "The specification of stats-intervals is invalid");
363 return false;
366 return true;
369 typedef enum { MEDIA_DISK, MEDIA_CDROM } DriveMediaType;
371 /* All parameters but @opts are optional and may be set to NULL. */
372 static void extract_common_blockdev_options(QemuOpts *opts, int *bdrv_flags,
373 const char **throttling_group, ThrottleConfig *throttle_cfg,
374 BlockdevDetectZeroesOptions *detect_zeroes, Error **errp)
376 Error *local_error = NULL;
377 const char *aio;
379 if (bdrv_flags) {
380 if (qemu_opt_get_bool(opts, "copy-on-read", false)) {
381 *bdrv_flags |= BDRV_O_COPY_ON_READ;
384 if ((aio = qemu_opt_get(opts, "aio")) != NULL) {
385 if (bdrv_parse_aio(aio, bdrv_flags) < 0) {
386 error_setg(errp, "invalid aio option");
387 return;
392 /* disk I/O throttling */
393 if (throttling_group) {
394 *throttling_group = qemu_opt_get(opts, "throttling.group");
397 if (throttle_cfg) {
398 throttle_config_init(throttle_cfg);
399 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].avg =
400 qemu_opt_get_number(opts, "throttling.bps-total", 0);
401 throttle_cfg->buckets[THROTTLE_BPS_READ].avg =
402 qemu_opt_get_number(opts, "throttling.bps-read", 0);
403 throttle_cfg->buckets[THROTTLE_BPS_WRITE].avg =
404 qemu_opt_get_number(opts, "throttling.bps-write", 0);
405 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].avg =
406 qemu_opt_get_number(opts, "throttling.iops-total", 0);
407 throttle_cfg->buckets[THROTTLE_OPS_READ].avg =
408 qemu_opt_get_number(opts, "throttling.iops-read", 0);
409 throttle_cfg->buckets[THROTTLE_OPS_WRITE].avg =
410 qemu_opt_get_number(opts, "throttling.iops-write", 0);
412 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].max =
413 qemu_opt_get_number(opts, "throttling.bps-total-max", 0);
414 throttle_cfg->buckets[THROTTLE_BPS_READ].max =
415 qemu_opt_get_number(opts, "throttling.bps-read-max", 0);
416 throttle_cfg->buckets[THROTTLE_BPS_WRITE].max =
417 qemu_opt_get_number(opts, "throttling.bps-write-max", 0);
418 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].max =
419 qemu_opt_get_number(opts, "throttling.iops-total-max", 0);
420 throttle_cfg->buckets[THROTTLE_OPS_READ].max =
421 qemu_opt_get_number(opts, "throttling.iops-read-max", 0);
422 throttle_cfg->buckets[THROTTLE_OPS_WRITE].max =
423 qemu_opt_get_number(opts, "throttling.iops-write-max", 0);
425 throttle_cfg->buckets[THROTTLE_BPS_TOTAL].burst_length =
426 qemu_opt_get_number(opts, "throttling.bps-total-max-length", 1);
427 throttle_cfg->buckets[THROTTLE_BPS_READ].burst_length =
428 qemu_opt_get_number(opts, "throttling.bps-read-max-length", 1);
429 throttle_cfg->buckets[THROTTLE_BPS_WRITE].burst_length =
430 qemu_opt_get_number(opts, "throttling.bps-write-max-length", 1);
431 throttle_cfg->buckets[THROTTLE_OPS_TOTAL].burst_length =
432 qemu_opt_get_number(opts, "throttling.iops-total-max-length", 1);
433 throttle_cfg->buckets[THROTTLE_OPS_READ].burst_length =
434 qemu_opt_get_number(opts, "throttling.iops-read-max-length", 1);
435 throttle_cfg->buckets[THROTTLE_OPS_WRITE].burst_length =
436 qemu_opt_get_number(opts, "throttling.iops-write-max-length", 1);
438 throttle_cfg->op_size =
439 qemu_opt_get_number(opts, "throttling.iops-size", 0);
441 if (!throttle_is_valid(throttle_cfg, errp)) {
442 return;
446 if (detect_zeroes) {
447 *detect_zeroes =
448 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup,
449 qemu_opt_get(opts, "detect-zeroes"),
450 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF,
451 &local_error);
452 if (local_error) {
453 error_propagate(errp, local_error);
454 return;
459 /* Takes the ownership of bs_opts */
460 static BlockBackend *blockdev_init(const char *file, QDict *bs_opts,
461 Error **errp)
463 const char *buf;
464 int bdrv_flags = 0;
465 int on_read_error, on_write_error;
466 bool account_invalid, account_failed;
467 bool writethrough, read_only;
468 BlockBackend *blk;
469 BlockDriverState *bs;
470 ThrottleConfig cfg;
471 int snapshot = 0;
472 Error *error = NULL;
473 QemuOpts *opts;
474 QDict *interval_dict = NULL;
475 QList *interval_list = NULL;
476 const char *id;
477 BlockdevDetectZeroesOptions detect_zeroes =
478 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF;
479 const char *throttling_group = NULL;
481 /* Check common options by copying from bs_opts to opts, all other options
482 * stay in bs_opts for processing by bdrv_open(). */
483 id = qdict_get_try_str(bs_opts, "id");
484 opts = qemu_opts_create(&qemu_common_drive_opts, id, 1, &error);
485 if (error) {
486 error_propagate(errp, error);
487 goto err_no_opts;
490 qemu_opts_absorb_qdict(opts, bs_opts, &error);
491 if (error) {
492 error_propagate(errp, error);
493 goto early_err;
496 if (id) {
497 qdict_del(bs_opts, "id");
500 /* extract parameters */
501 snapshot = qemu_opt_get_bool(opts, "snapshot", 0);
503 account_invalid = qemu_opt_get_bool(opts, "stats-account-invalid", true);
504 account_failed = qemu_opt_get_bool(opts, "stats-account-failed", true);
506 writethrough = !qemu_opt_get_bool(opts, BDRV_OPT_CACHE_WB, true);
508 id = qemu_opts_id(opts);
510 qdict_extract_subqdict(bs_opts, &interval_dict, "stats-intervals.");
511 qdict_array_split(interval_dict, &interval_list);
513 if (qdict_size(interval_dict) != 0) {
514 error_setg(errp, "Invalid option stats-intervals.%s",
515 qdict_first(interval_dict)->key);
516 goto early_err;
519 extract_common_blockdev_options(opts, &bdrv_flags, &throttling_group, &cfg,
520 &detect_zeroes, &error);
521 if (error) {
522 error_propagate(errp, error);
523 goto early_err;
526 if ((buf = qemu_opt_get(opts, "format")) != NULL) {
527 if (is_help_option(buf)) {
528 qemu_printf("Supported formats:");
529 bdrv_iterate_format(bdrv_format_print, NULL, false);
530 qemu_printf("\nSupported formats (read-only):");
531 bdrv_iterate_format(bdrv_format_print, NULL, true);
532 qemu_printf("\n");
533 goto early_err;
536 if (qdict_haskey(bs_opts, "driver")) {
537 error_setg(errp, "Cannot specify both 'driver' and 'format'");
538 goto early_err;
540 qdict_put_str(bs_opts, "driver", buf);
543 on_write_error = BLOCKDEV_ON_ERROR_ENOSPC;
544 if ((buf = qemu_opt_get(opts, "werror")) != NULL) {
545 on_write_error = parse_block_error_action(buf, 0, &error);
546 if (error) {
547 error_propagate(errp, error);
548 goto early_err;
552 on_read_error = BLOCKDEV_ON_ERROR_REPORT;
553 if ((buf = qemu_opt_get(opts, "rerror")) != NULL) {
554 on_read_error = parse_block_error_action(buf, 1, &error);
555 if (error) {
556 error_propagate(errp, error);
557 goto early_err;
561 if (snapshot) {
562 bdrv_flags |= BDRV_O_SNAPSHOT;
565 read_only = qemu_opt_get_bool(opts, BDRV_OPT_READ_ONLY, false);
567 /* init */
568 if ((!file || !*file) && !qdict_size(bs_opts)) {
569 BlockBackendRootState *blk_rs;
571 blk = blk_new(qemu_get_aio_context(), 0, BLK_PERM_ALL);
572 blk_rs = blk_get_root_state(blk);
573 blk_rs->open_flags = bdrv_flags;
574 blk_rs->read_only = read_only;
575 blk_rs->detect_zeroes = detect_zeroes;
577 qobject_unref(bs_opts);
578 } else {
579 if (file && !*file) {
580 file = NULL;
583 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
584 * with other callers) rather than what we want as the real defaults.
585 * Apply the defaults here instead. */
586 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
587 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
588 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY,
589 read_only ? "on" : "off");
590 qdict_set_default_str(bs_opts, BDRV_OPT_AUTO_READ_ONLY, "on");
591 assert((bdrv_flags & BDRV_O_CACHE_MASK) == 0);
593 if (runstate_check(RUN_STATE_INMIGRATE)) {
594 bdrv_flags |= BDRV_O_INACTIVE;
597 blk = blk_new_open(file, NULL, bs_opts, bdrv_flags, errp);
598 if (!blk) {
599 goto err_no_bs_opts;
601 bs = blk_bs(blk);
603 bs->detect_zeroes = detect_zeroes;
605 block_acct_setup(blk_get_stats(blk), account_invalid, account_failed);
607 if (!parse_stats_intervals(blk_get_stats(blk), interval_list, errp)) {
608 blk_unref(blk);
609 blk = NULL;
610 goto err_no_bs_opts;
614 /* disk I/O throttling */
615 if (throttle_enabled(&cfg)) {
616 if (!throttling_group) {
617 throttling_group = id;
619 blk_io_limits_enable(blk, throttling_group);
620 blk_set_io_limits(blk, &cfg);
623 blk_set_enable_write_cache(blk, !writethrough);
624 blk_set_on_error(blk, on_read_error, on_write_error);
626 if (!monitor_add_blk(blk, id, errp)) {
627 blk_unref(blk);
628 blk = NULL;
629 goto err_no_bs_opts;
632 err_no_bs_opts:
633 qemu_opts_del(opts);
634 qobject_unref(interval_dict);
635 qobject_unref(interval_list);
636 return blk;
638 early_err:
639 qemu_opts_del(opts);
640 qobject_unref(interval_dict);
641 qobject_unref(interval_list);
642 err_no_opts:
643 qobject_unref(bs_opts);
644 return NULL;
647 /* Takes the ownership of bs_opts */
648 BlockDriverState *bds_tree_init(QDict *bs_opts, Error **errp)
650 int bdrv_flags = 0;
652 /* bdrv_open() defaults to the values in bdrv_flags (for compatibility
653 * with other callers) rather than what we want as the real defaults.
654 * Apply the defaults here instead. */
655 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_DIRECT, "off");
656 qdict_set_default_str(bs_opts, BDRV_OPT_CACHE_NO_FLUSH, "off");
657 qdict_set_default_str(bs_opts, BDRV_OPT_READ_ONLY, "off");
659 if (runstate_check(RUN_STATE_INMIGRATE)) {
660 bdrv_flags |= BDRV_O_INACTIVE;
663 return bdrv_open(NULL, NULL, bs_opts, bdrv_flags, errp);
666 void blockdev_close_all_bdrv_states(void)
668 BlockDriverState *bs, *next_bs;
670 QTAILQ_FOREACH_SAFE(bs, &monitor_bdrv_states, monitor_list, next_bs) {
671 AioContext *ctx = bdrv_get_aio_context(bs);
673 aio_context_acquire(ctx);
674 bdrv_unref(bs);
675 aio_context_release(ctx);
679 /* Iterates over the list of monitor-owned BlockDriverStates */
680 BlockDriverState *bdrv_next_monitor_owned(BlockDriverState *bs)
682 return bs ? QTAILQ_NEXT(bs, monitor_list)
683 : QTAILQ_FIRST(&monitor_bdrv_states);
686 static void qemu_opt_rename(QemuOpts *opts, const char *from, const char *to,
687 Error **errp)
689 const char *value;
691 value = qemu_opt_get(opts, from);
692 if (value) {
693 if (qemu_opt_find(opts, to)) {
694 error_setg(errp, "'%s' and its alias '%s' can't be used at the "
695 "same time", to, from);
696 return;
700 /* rename all items in opts */
701 while ((value = qemu_opt_get(opts, from))) {
702 qemu_opt_set(opts, to, value, &error_abort);
703 qemu_opt_unset(opts, from);
707 QemuOptsList qemu_legacy_drive_opts = {
708 .name = "drive",
709 .head = QTAILQ_HEAD_INITIALIZER(qemu_legacy_drive_opts.head),
710 .desc = {
712 .name = "bus",
713 .type = QEMU_OPT_NUMBER,
714 .help = "bus number",
716 .name = "unit",
717 .type = QEMU_OPT_NUMBER,
718 .help = "unit number (i.e. lun for scsi)",
720 .name = "index",
721 .type = QEMU_OPT_NUMBER,
722 .help = "index number",
724 .name = "media",
725 .type = QEMU_OPT_STRING,
726 .help = "media type (disk, cdrom)",
728 .name = "if",
729 .type = QEMU_OPT_STRING,
730 .help = "interface (ide, scsi, sd, mtd, floppy, pflash, virtio)",
732 .name = "file",
733 .type = QEMU_OPT_STRING,
734 .help = "file name",
737 /* Options that are passed on, but have special semantics with -drive */
739 .name = BDRV_OPT_READ_ONLY,
740 .type = QEMU_OPT_BOOL,
741 .help = "open drive file as read-only",
743 .name = "rerror",
744 .type = QEMU_OPT_STRING,
745 .help = "read error action",
747 .name = "werror",
748 .type = QEMU_OPT_STRING,
749 .help = "write error action",
751 .name = "copy-on-read",
752 .type = QEMU_OPT_BOOL,
753 .help = "copy read data from backing file into image file",
756 { /* end of list */ }
760 DriveInfo *drive_new(QemuOpts *all_opts, BlockInterfaceType block_default_type,
761 Error **errp)
763 const char *value;
764 BlockBackend *blk;
765 DriveInfo *dinfo = NULL;
766 QDict *bs_opts;
767 QemuOpts *legacy_opts;
768 DriveMediaType media = MEDIA_DISK;
769 BlockInterfaceType type;
770 int max_devs, bus_id, unit_id, index;
771 const char *werror, *rerror;
772 bool read_only = false;
773 bool copy_on_read;
774 const char *filename;
775 Error *local_err = NULL;
776 int i;
778 /* Change legacy command line options into QMP ones */
779 static const struct {
780 const char *from;
781 const char *to;
782 } opt_renames[] = {
783 { "iops", "throttling.iops-total" },
784 { "iops_rd", "throttling.iops-read" },
785 { "iops_wr", "throttling.iops-write" },
787 { "bps", "throttling.bps-total" },
788 { "bps_rd", "throttling.bps-read" },
789 { "bps_wr", "throttling.bps-write" },
791 { "iops_max", "throttling.iops-total-max" },
792 { "iops_rd_max", "throttling.iops-read-max" },
793 { "iops_wr_max", "throttling.iops-write-max" },
795 { "bps_max", "throttling.bps-total-max" },
796 { "bps_rd_max", "throttling.bps-read-max" },
797 { "bps_wr_max", "throttling.bps-write-max" },
799 { "iops_size", "throttling.iops-size" },
801 { "group", "throttling.group" },
803 { "readonly", BDRV_OPT_READ_ONLY },
806 for (i = 0; i < ARRAY_SIZE(opt_renames); i++) {
807 qemu_opt_rename(all_opts, opt_renames[i].from, opt_renames[i].to,
808 &local_err);
809 if (local_err) {
810 error_propagate(errp, local_err);
811 return NULL;
815 value = qemu_opt_get(all_opts, "cache");
816 if (value) {
817 int flags = 0;
818 bool writethrough;
820 if (bdrv_parse_cache_mode(value, &flags, &writethrough) != 0) {
821 error_setg(errp, "invalid cache option");
822 return NULL;
825 /* Specific options take precedence */
826 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_WB)) {
827 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_WB,
828 !writethrough, &error_abort);
830 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_DIRECT)) {
831 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_DIRECT,
832 !!(flags & BDRV_O_NOCACHE), &error_abort);
834 if (!qemu_opt_get(all_opts, BDRV_OPT_CACHE_NO_FLUSH)) {
835 qemu_opt_set_bool(all_opts, BDRV_OPT_CACHE_NO_FLUSH,
836 !!(flags & BDRV_O_NO_FLUSH), &error_abort);
838 qemu_opt_unset(all_opts, "cache");
841 /* Get a QDict for processing the options */
842 bs_opts = qdict_new();
843 qemu_opts_to_qdict(all_opts, bs_opts);
845 legacy_opts = qemu_opts_create(&qemu_legacy_drive_opts, NULL, 0,
846 &error_abort);
847 qemu_opts_absorb_qdict(legacy_opts, bs_opts, &local_err);
848 if (local_err) {
849 error_propagate(errp, local_err);
850 goto fail;
853 /* Media type */
854 value = qemu_opt_get(legacy_opts, "media");
855 if (value) {
856 if (!strcmp(value, "disk")) {
857 media = MEDIA_DISK;
858 } else if (!strcmp(value, "cdrom")) {
859 media = MEDIA_CDROM;
860 read_only = true;
861 } else {
862 error_setg(errp, "'%s' invalid media", value);
863 goto fail;
867 /* copy-on-read is disabled with a warning for read-only devices */
868 read_only |= qemu_opt_get_bool(legacy_opts, BDRV_OPT_READ_ONLY, false);
869 copy_on_read = qemu_opt_get_bool(legacy_opts, "copy-on-read", false);
871 if (read_only && copy_on_read) {
872 warn_report("disabling copy-on-read on read-only drive");
873 copy_on_read = false;
876 qdict_put_str(bs_opts, BDRV_OPT_READ_ONLY, read_only ? "on" : "off");
877 qdict_put_str(bs_opts, "copy-on-read", copy_on_read ? "on" : "off");
879 /* Controller type */
880 value = qemu_opt_get(legacy_opts, "if");
881 if (value) {
882 for (type = 0;
883 type < IF_COUNT && strcmp(value, if_name[type]);
884 type++) {
886 if (type == IF_COUNT) {
887 error_setg(errp, "unsupported bus type '%s'", value);
888 goto fail;
890 } else {
891 type = block_default_type;
894 /* Device address specified by bus/unit or index.
895 * If none was specified, try to find the first free one. */
896 bus_id = qemu_opt_get_number(legacy_opts, "bus", 0);
897 unit_id = qemu_opt_get_number(legacy_opts, "unit", -1);
898 index = qemu_opt_get_number(legacy_opts, "index", -1);
900 max_devs = if_max_devs[type];
902 if (index != -1) {
903 if (bus_id != 0 || unit_id != -1) {
904 error_setg(errp, "index cannot be used with bus and unit");
905 goto fail;
907 bus_id = drive_index_to_bus_id(type, index);
908 unit_id = drive_index_to_unit_id(type, index);
911 if (unit_id == -1) {
912 unit_id = 0;
913 while (drive_get(type, bus_id, unit_id) != NULL) {
914 unit_id++;
915 if (max_devs && unit_id >= max_devs) {
916 unit_id -= max_devs;
917 bus_id++;
922 if (max_devs && unit_id >= max_devs) {
923 error_setg(errp, "unit %d too big (max is %d)", unit_id, max_devs - 1);
924 goto fail;
927 if (drive_get(type, bus_id, unit_id) != NULL) {
928 error_setg(errp, "drive with bus=%d, unit=%d (index=%d) exists",
929 bus_id, unit_id, index);
930 goto fail;
933 /* no id supplied -> create one */
934 if (qemu_opts_id(all_opts) == NULL) {
935 char *new_id;
936 const char *mediastr = "";
937 if (type == IF_IDE || type == IF_SCSI) {
938 mediastr = (media == MEDIA_CDROM) ? "-cd" : "-hd";
940 if (max_devs) {
941 new_id = g_strdup_printf("%s%i%s%i", if_name[type], bus_id,
942 mediastr, unit_id);
943 } else {
944 new_id = g_strdup_printf("%s%s%i", if_name[type],
945 mediastr, unit_id);
947 qdict_put_str(bs_opts, "id", new_id);
948 g_free(new_id);
951 /* Add virtio block device */
952 if (type == IF_VIRTIO) {
953 QemuOpts *devopts;
954 devopts = qemu_opts_create(qemu_find_opts("device"), NULL, 0,
955 &error_abort);
956 if (arch_type == QEMU_ARCH_S390X) {
957 qemu_opt_set(devopts, "driver", "virtio-blk-ccw", &error_abort);
958 } else {
959 qemu_opt_set(devopts, "driver", "virtio-blk-pci", &error_abort);
961 qemu_opt_set(devopts, "drive", qdict_get_str(bs_opts, "id"),
962 &error_abort);
965 filename = qemu_opt_get(legacy_opts, "file");
967 /* Check werror/rerror compatibility with if=... */
968 werror = qemu_opt_get(legacy_opts, "werror");
969 if (werror != NULL) {
970 if (type != IF_IDE && type != IF_SCSI && type != IF_VIRTIO &&
971 type != IF_NONE) {
972 error_setg(errp, "werror is not supported by this bus type");
973 goto fail;
975 qdict_put_str(bs_opts, "werror", werror);
978 rerror = qemu_opt_get(legacy_opts, "rerror");
979 if (rerror != NULL) {
980 if (type != IF_IDE && type != IF_VIRTIO && type != IF_SCSI &&
981 type != IF_NONE) {
982 error_setg(errp, "rerror is not supported by this bus type");
983 goto fail;
985 qdict_put_str(bs_opts, "rerror", rerror);
988 /* Actual block device init: Functionality shared with blockdev-add */
989 blk = blockdev_init(filename, bs_opts, &local_err);
990 bs_opts = NULL;
991 if (!blk) {
992 error_propagate(errp, local_err);
993 goto fail;
994 } else {
995 assert(!local_err);
998 /* Create legacy DriveInfo */
999 dinfo = g_malloc0(sizeof(*dinfo));
1000 dinfo->opts = all_opts;
1002 dinfo->type = type;
1003 dinfo->bus = bus_id;
1004 dinfo->unit = unit_id;
1006 blk_set_legacy_dinfo(blk, dinfo);
1008 switch(type) {
1009 case IF_IDE:
1010 case IF_SCSI:
1011 case IF_XEN:
1012 case IF_NONE:
1013 dinfo->media_cd = media == MEDIA_CDROM;
1014 break;
1015 default:
1016 break;
1019 fail:
1020 qemu_opts_del(legacy_opts);
1021 qobject_unref(bs_opts);
1022 return dinfo;
1025 static BlockDriverState *qmp_get_root_bs(const char *name, Error **errp)
1027 BlockDriverState *bs;
1029 bs = bdrv_lookup_bs(name, name, errp);
1030 if (bs == NULL) {
1031 return NULL;
1034 if (!bdrv_is_root_node(bs)) {
1035 error_setg(errp, "Need a root block node");
1036 return NULL;
1039 if (!bdrv_is_inserted(bs)) {
1040 error_setg(errp, "Device has no medium");
1041 return NULL;
1044 return bs;
1047 static void blockdev_do_action(TransactionAction *action, Error **errp)
1049 TransactionActionList list;
1051 list.value = action;
1052 list.next = NULL;
1053 qmp_transaction(&list, false, NULL, errp);
1056 void qmp_blockdev_snapshot_sync(bool has_device, const char *device,
1057 bool has_node_name, const char *node_name,
1058 const char *snapshot_file,
1059 bool has_snapshot_node_name,
1060 const char *snapshot_node_name,
1061 bool has_format, const char *format,
1062 bool has_mode, NewImageMode mode, Error **errp)
1064 BlockdevSnapshotSync snapshot = {
1065 .has_device = has_device,
1066 .device = (char *) device,
1067 .has_node_name = has_node_name,
1068 .node_name = (char *) node_name,
1069 .snapshot_file = (char *) snapshot_file,
1070 .has_snapshot_node_name = has_snapshot_node_name,
1071 .snapshot_node_name = (char *) snapshot_node_name,
1072 .has_format = has_format,
1073 .format = (char *) format,
1074 .has_mode = has_mode,
1075 .mode = mode,
1077 TransactionAction action = {
1078 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC,
1079 .u.blockdev_snapshot_sync.data = &snapshot,
1081 blockdev_do_action(&action, errp);
1084 void qmp_blockdev_snapshot(const char *node, const char *overlay,
1085 Error **errp)
1087 BlockdevSnapshot snapshot_data = {
1088 .node = (char *) node,
1089 .overlay = (char *) overlay
1091 TransactionAction action = {
1092 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT,
1093 .u.blockdev_snapshot.data = &snapshot_data,
1095 blockdev_do_action(&action, errp);
1098 void qmp_blockdev_snapshot_internal_sync(const char *device,
1099 const char *name,
1100 Error **errp)
1102 BlockdevSnapshotInternal snapshot = {
1103 .device = (char *) device,
1104 .name = (char *) name
1106 TransactionAction action = {
1107 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC,
1108 .u.blockdev_snapshot_internal_sync.data = &snapshot,
1110 blockdev_do_action(&action, errp);
1113 SnapshotInfo *qmp_blockdev_snapshot_delete_internal_sync(const char *device,
1114 bool has_id,
1115 const char *id,
1116 bool has_name,
1117 const char *name,
1118 Error **errp)
1120 BlockDriverState *bs;
1121 AioContext *aio_context;
1122 QEMUSnapshotInfo sn;
1123 Error *local_err = NULL;
1124 SnapshotInfo *info = NULL;
1125 int ret;
1127 bs = qmp_get_root_bs(device, errp);
1128 if (!bs) {
1129 return NULL;
1131 aio_context = bdrv_get_aio_context(bs);
1132 aio_context_acquire(aio_context);
1134 if (!has_id) {
1135 id = NULL;
1138 if (!has_name) {
1139 name = NULL;
1142 if (!id && !name) {
1143 error_setg(errp, "Name or id must be provided");
1144 goto out_aio_context;
1147 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT_DELETE, errp)) {
1148 goto out_aio_context;
1151 ret = bdrv_snapshot_find_by_id_and_name(bs, id, name, &sn, &local_err);
1152 if (local_err) {
1153 error_propagate(errp, local_err);
1154 goto out_aio_context;
1156 if (!ret) {
1157 error_setg(errp,
1158 "Snapshot with id '%s' and name '%s' does not exist on "
1159 "device '%s'",
1160 STR_OR_NULL(id), STR_OR_NULL(name), device);
1161 goto out_aio_context;
1164 bdrv_snapshot_delete(bs, id, name, &local_err);
1165 if (local_err) {
1166 error_propagate(errp, local_err);
1167 goto out_aio_context;
1170 aio_context_release(aio_context);
1172 info = g_new0(SnapshotInfo, 1);
1173 info->id = g_strdup(sn.id_str);
1174 info->name = g_strdup(sn.name);
1175 info->date_nsec = sn.date_nsec;
1176 info->date_sec = sn.date_sec;
1177 info->vm_state_size = sn.vm_state_size;
1178 info->vm_clock_nsec = sn.vm_clock_nsec % 1000000000;
1179 info->vm_clock_sec = sn.vm_clock_nsec / 1000000000;
1181 return info;
1183 out_aio_context:
1184 aio_context_release(aio_context);
1185 return NULL;
1188 /* New and old BlockDriverState structs for atomic group operations */
1190 typedef struct BlkActionState BlkActionState;
1193 * BlkActionOps:
1194 * Table of operations that define an Action.
1196 * @instance_size: Size of state struct, in bytes.
1197 * @prepare: Prepare the work, must NOT be NULL.
1198 * @commit: Commit the changes, can be NULL.
1199 * @abort: Abort the changes on fail, can be NULL.
1200 * @clean: Clean up resources after all transaction actions have called
1201 * commit() or abort(). Can be NULL.
1203 * Only prepare() may fail. In a single transaction, only one of commit() or
1204 * abort() will be called. clean() will always be called if it is present.
1206 typedef struct BlkActionOps {
1207 size_t instance_size;
1208 void (*prepare)(BlkActionState *common, Error **errp);
1209 void (*commit)(BlkActionState *common);
1210 void (*abort)(BlkActionState *common);
1211 void (*clean)(BlkActionState *common);
1212 } BlkActionOps;
1215 * BlkActionState:
1216 * Describes one Action's state within a Transaction.
1218 * @action: QAPI-defined enum identifying which Action to perform.
1219 * @ops: Table of ActionOps this Action can perform.
1220 * @block_job_txn: Transaction which this action belongs to.
1221 * @entry: List membership for all Actions in this Transaction.
1223 * This structure must be arranged as first member in a subclassed type,
1224 * assuming that the compiler will also arrange it to the same offsets as the
1225 * base class.
1227 struct BlkActionState {
1228 TransactionAction *action;
1229 const BlkActionOps *ops;
1230 JobTxn *block_job_txn;
1231 TransactionProperties *txn_props;
1232 QTAILQ_ENTRY(BlkActionState) entry;
1235 /* internal snapshot private data */
1236 typedef struct InternalSnapshotState {
1237 BlkActionState common;
1238 BlockDriverState *bs;
1239 QEMUSnapshotInfo sn;
1240 bool created;
1241 } InternalSnapshotState;
1244 static int action_check_completion_mode(BlkActionState *s, Error **errp)
1246 if (s->txn_props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
1247 error_setg(errp,
1248 "Action '%s' does not support Transaction property "
1249 "completion-mode = %s",
1250 TransactionActionKind_str(s->action->type),
1251 ActionCompletionMode_str(s->txn_props->completion_mode));
1252 return -1;
1254 return 0;
1257 static void internal_snapshot_prepare(BlkActionState *common,
1258 Error **errp)
1260 Error *local_err = NULL;
1261 const char *device;
1262 const char *name;
1263 BlockDriverState *bs;
1264 QEMUSnapshotInfo old_sn, *sn;
1265 bool ret;
1266 qemu_timeval tv;
1267 BlockdevSnapshotInternal *internal;
1268 InternalSnapshotState *state;
1269 AioContext *aio_context;
1270 int ret1;
1272 g_assert(common->action->type ==
1273 TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC);
1274 internal = common->action->u.blockdev_snapshot_internal_sync.data;
1275 state = DO_UPCAST(InternalSnapshotState, common, common);
1277 /* 1. parse input */
1278 device = internal->device;
1279 name = internal->name;
1281 /* 2. check for validation */
1282 if (action_check_completion_mode(common, errp) < 0) {
1283 return;
1286 bs = qmp_get_root_bs(device, errp);
1287 if (!bs) {
1288 return;
1291 aio_context = bdrv_get_aio_context(bs);
1292 aio_context_acquire(aio_context);
1294 state->bs = bs;
1296 /* Paired with .clean() */
1297 bdrv_drained_begin(bs);
1299 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) {
1300 goto out;
1303 if (bdrv_is_read_only(bs)) {
1304 error_setg(errp, "Device '%s' is read only", device);
1305 goto out;
1308 if (!bdrv_can_snapshot(bs)) {
1309 error_setg(errp, "Block format '%s' used by device '%s' "
1310 "does not support internal snapshots",
1311 bs->drv->format_name, device);
1312 goto out;
1315 if (!strlen(name)) {
1316 error_setg(errp, "Name is empty");
1317 goto out;
1320 /* check whether a snapshot with name exist */
1321 ret = bdrv_snapshot_find_by_id_and_name(bs, NULL, name, &old_sn,
1322 &local_err);
1323 if (local_err) {
1324 error_propagate(errp, local_err);
1325 goto out;
1326 } else if (ret) {
1327 error_setg(errp,
1328 "Snapshot with name '%s' already exists on device '%s'",
1329 name, device);
1330 goto out;
1333 /* 3. take the snapshot */
1334 sn = &state->sn;
1335 pstrcpy(sn->name, sizeof(sn->name), name);
1336 qemu_gettimeofday(&tv);
1337 sn->date_sec = tv.tv_sec;
1338 sn->date_nsec = tv.tv_usec * 1000;
1339 sn->vm_clock_nsec = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
1341 ret1 = bdrv_snapshot_create(bs, sn);
1342 if (ret1 < 0) {
1343 error_setg_errno(errp, -ret1,
1344 "Failed to create snapshot '%s' on device '%s'",
1345 name, device);
1346 goto out;
1349 /* 4. succeed, mark a snapshot is created */
1350 state->created = true;
1352 out:
1353 aio_context_release(aio_context);
1356 static void internal_snapshot_abort(BlkActionState *common)
1358 InternalSnapshotState *state =
1359 DO_UPCAST(InternalSnapshotState, common, common);
1360 BlockDriverState *bs = state->bs;
1361 QEMUSnapshotInfo *sn = &state->sn;
1362 AioContext *aio_context;
1363 Error *local_error = NULL;
1365 if (!state->created) {
1366 return;
1369 aio_context = bdrv_get_aio_context(state->bs);
1370 aio_context_acquire(aio_context);
1372 if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) {
1373 error_reportf_err(local_error,
1374 "Failed to delete snapshot with id '%s' and "
1375 "name '%s' on device '%s' in abort: ",
1376 sn->id_str, sn->name,
1377 bdrv_get_device_name(bs));
1380 aio_context_release(aio_context);
1383 static void internal_snapshot_clean(BlkActionState *common)
1385 InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState,
1386 common, common);
1387 AioContext *aio_context;
1389 if (!state->bs) {
1390 return;
1393 aio_context = bdrv_get_aio_context(state->bs);
1394 aio_context_acquire(aio_context);
1396 bdrv_drained_end(state->bs);
1398 aio_context_release(aio_context);
1401 /* external snapshot private data */
1402 typedef struct ExternalSnapshotState {
1403 BlkActionState common;
1404 BlockDriverState *old_bs;
1405 BlockDriverState *new_bs;
1406 bool overlay_appended;
1407 } ExternalSnapshotState;
1409 static void external_snapshot_prepare(BlkActionState *common,
1410 Error **errp)
1412 int flags = 0;
1413 QDict *options = NULL;
1414 Error *local_err = NULL;
1415 /* Device and node name of the image to generate the snapshot from */
1416 const char *device;
1417 const char *node_name;
1418 /* Reference to the new image (for 'blockdev-snapshot') */
1419 const char *snapshot_ref;
1420 /* File name of the new image (for 'blockdev-snapshot-sync') */
1421 const char *new_image_file;
1422 ExternalSnapshotState *state =
1423 DO_UPCAST(ExternalSnapshotState, common, common);
1424 TransactionAction *action = common->action;
1425 AioContext *aio_context;
1426 uint64_t perm, shared;
1428 /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar
1429 * purpose but a different set of parameters */
1430 switch (action->type) {
1431 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT:
1433 BlockdevSnapshot *s = action->u.blockdev_snapshot.data;
1434 device = s->node;
1435 node_name = s->node;
1436 new_image_file = NULL;
1437 snapshot_ref = s->overlay;
1439 break;
1440 case TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC:
1442 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1443 device = s->has_device ? s->device : NULL;
1444 node_name = s->has_node_name ? s->node_name : NULL;
1445 new_image_file = s->snapshot_file;
1446 snapshot_ref = NULL;
1448 break;
1449 default:
1450 g_assert_not_reached();
1453 /* start processing */
1454 if (action_check_completion_mode(common, errp) < 0) {
1455 return;
1458 state->old_bs = bdrv_lookup_bs(device, node_name, errp);
1459 if (!state->old_bs) {
1460 return;
1463 aio_context = bdrv_get_aio_context(state->old_bs);
1464 aio_context_acquire(aio_context);
1466 /* Paired with .clean() */
1467 bdrv_drained_begin(state->old_bs);
1469 if (!bdrv_is_inserted(state->old_bs)) {
1470 error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device);
1471 goto out;
1474 if (bdrv_op_is_blocked(state->old_bs,
1475 BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) {
1476 goto out;
1479 if (!bdrv_is_read_only(state->old_bs)) {
1480 if (bdrv_flush(state->old_bs)) {
1481 error_setg(errp, QERR_IO_ERROR);
1482 goto out;
1486 if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) {
1487 BlockdevSnapshotSync *s = action->u.blockdev_snapshot_sync.data;
1488 const char *format = s->has_format ? s->format : "qcow2";
1489 enum NewImageMode mode;
1490 const char *snapshot_node_name =
1491 s->has_snapshot_node_name ? s->snapshot_node_name : NULL;
1493 if (node_name && !snapshot_node_name) {
1494 error_setg(errp, "New overlay node name missing");
1495 goto out;
1498 if (snapshot_node_name &&
1499 bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) {
1500 error_setg(errp, "New overlay node name already in use");
1501 goto out;
1504 flags = state->old_bs->open_flags;
1505 flags &= ~(BDRV_O_SNAPSHOT | BDRV_O_COPY_ON_READ);
1506 flags |= BDRV_O_NO_BACKING;
1508 /* create new image w/backing file */
1509 mode = s->has_mode ? s->mode : NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1510 if (mode != NEW_IMAGE_MODE_EXISTING) {
1511 int64_t size = bdrv_getlength(state->old_bs);
1512 if (size < 0) {
1513 error_setg_errno(errp, -size, "bdrv_getlength failed");
1514 goto out;
1516 bdrv_refresh_filename(state->old_bs);
1517 bdrv_img_create(new_image_file, format,
1518 state->old_bs->filename,
1519 state->old_bs->drv->format_name,
1520 NULL, size, flags, false, &local_err);
1521 if (local_err) {
1522 error_propagate(errp, local_err);
1523 goto out;
1527 options = qdict_new();
1528 if (snapshot_node_name) {
1529 qdict_put_str(options, "node-name", snapshot_node_name);
1531 qdict_put_str(options, "driver", format);
1534 state->new_bs = bdrv_open(new_image_file, snapshot_ref, options, flags,
1535 errp);
1536 /* We will manually add the backing_hd field to the bs later */
1537 if (!state->new_bs) {
1538 goto out;
1542 * Allow attaching a backing file to an overlay that's already in use only
1543 * if the parents don't assume that they are already seeing a valid image.
1544 * (Specifically, allow it as a mirror target, which is write-only access.)
1546 bdrv_get_cumulative_perm(state->new_bs, &perm, &shared);
1547 if (perm & BLK_PERM_CONSISTENT_READ) {
1548 error_setg(errp, "The overlay is already in use");
1549 goto out;
1552 if (state->new_bs->backing != NULL) {
1553 error_setg(errp, "The overlay already has a backing image");
1554 goto out;
1557 if (!state->new_bs->drv->supports_backing) {
1558 error_setg(errp, "The overlay does not support backing images");
1559 goto out;
1562 /* This removes our old bs and adds the new bs. This is an operation that
1563 * can fail, so we need to do it in .prepare; undoing it for abort is
1564 * always possible. */
1565 bdrv_ref(state->new_bs);
1566 bdrv_append(state->new_bs, state->old_bs, &local_err);
1567 if (local_err) {
1568 error_propagate(errp, local_err);
1569 goto out;
1571 state->overlay_appended = true;
1573 out:
1574 aio_context_release(aio_context);
1577 static void external_snapshot_commit(BlkActionState *common)
1579 ExternalSnapshotState *state =
1580 DO_UPCAST(ExternalSnapshotState, common, common);
1581 AioContext *aio_context;
1583 aio_context = bdrv_get_aio_context(state->old_bs);
1584 aio_context_acquire(aio_context);
1586 /* We don't need (or want) to use the transactional
1587 * bdrv_reopen_multiple() across all the entries at once, because we
1588 * don't want to abort all of them if one of them fails the reopen */
1589 if (!atomic_read(&state->old_bs->copy_on_read)) {
1590 bdrv_reopen_set_read_only(state->old_bs, true, NULL);
1593 aio_context_release(aio_context);
1596 static void external_snapshot_abort(BlkActionState *common)
1598 ExternalSnapshotState *state =
1599 DO_UPCAST(ExternalSnapshotState, common, common);
1600 if (state->new_bs) {
1601 if (state->overlay_appended) {
1602 AioContext *aio_context;
1603 AioContext *tmp_context;
1604 int ret;
1606 aio_context = bdrv_get_aio_context(state->old_bs);
1607 aio_context_acquire(aio_context);
1609 bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd()
1610 close state->old_bs; we need it */
1611 bdrv_set_backing_hd(state->new_bs, NULL, &error_abort);
1614 * The call to bdrv_set_backing_hd() above returns state->old_bs to
1615 * the main AioContext. As we're still going to be using it, return
1616 * it to the AioContext it was before.
1618 tmp_context = bdrv_get_aio_context(state->old_bs);
1619 if (aio_context != tmp_context) {
1620 aio_context_release(aio_context);
1621 aio_context_acquire(tmp_context);
1623 ret = bdrv_try_set_aio_context(state->old_bs,
1624 aio_context, NULL);
1625 assert(ret == 0);
1627 aio_context_release(tmp_context);
1628 aio_context_acquire(aio_context);
1631 bdrv_replace_node(state->new_bs, state->old_bs, &error_abort);
1632 bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */
1634 aio_context_release(aio_context);
1639 static void external_snapshot_clean(BlkActionState *common)
1641 ExternalSnapshotState *state =
1642 DO_UPCAST(ExternalSnapshotState, common, common);
1643 AioContext *aio_context;
1645 if (!state->old_bs) {
1646 return;
1649 aio_context = bdrv_get_aio_context(state->old_bs);
1650 aio_context_acquire(aio_context);
1652 bdrv_drained_end(state->old_bs);
1653 bdrv_unref(state->new_bs);
1655 aio_context_release(aio_context);
1658 typedef struct DriveBackupState {
1659 BlkActionState common;
1660 BlockDriverState *bs;
1661 BlockJob *job;
1662 } DriveBackupState;
1664 static BlockJob *do_backup_common(BackupCommon *backup,
1665 BlockDriverState *bs,
1666 BlockDriverState *target_bs,
1667 AioContext *aio_context,
1668 JobTxn *txn, Error **errp);
1670 static void drive_backup_prepare(BlkActionState *common, Error **errp)
1672 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1673 DriveBackup *backup;
1674 BlockDriverState *bs;
1675 BlockDriverState *target_bs;
1676 BlockDriverState *source = NULL;
1677 AioContext *aio_context;
1678 AioContext *old_context;
1679 QDict *options;
1680 Error *local_err = NULL;
1681 int flags;
1682 int64_t size;
1683 bool set_backing_hd = false;
1684 int ret;
1686 assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP);
1687 backup = common->action->u.drive_backup.data;
1689 if (!backup->has_mode) {
1690 backup->mode = NEW_IMAGE_MODE_ABSOLUTE_PATHS;
1693 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1694 if (!bs) {
1695 return;
1698 if (!bs->drv) {
1699 error_setg(errp, "Device has no medium");
1700 return;
1703 aio_context = bdrv_get_aio_context(bs);
1704 aio_context_acquire(aio_context);
1706 /* Paired with .clean() */
1707 bdrv_drained_begin(bs);
1709 if (!backup->has_format) {
1710 backup->format = backup->mode == NEW_IMAGE_MODE_EXISTING ?
1711 NULL : (char *) bs->drv->format_name;
1714 /* Early check to avoid creating target */
1715 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_BACKUP_SOURCE, errp)) {
1716 goto out;
1719 flags = bs->open_flags | BDRV_O_RDWR;
1722 * See if we have a backing HD we can use to create our new image
1723 * on top of.
1725 if (backup->sync == MIRROR_SYNC_MODE_TOP) {
1726 source = backing_bs(bs);
1727 if (!source) {
1728 backup->sync = MIRROR_SYNC_MODE_FULL;
1731 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
1732 source = bs;
1733 flags |= BDRV_O_NO_BACKING;
1734 set_backing_hd = true;
1737 size = bdrv_getlength(bs);
1738 if (size < 0) {
1739 error_setg_errno(errp, -size, "bdrv_getlength failed");
1740 goto out;
1743 if (backup->mode != NEW_IMAGE_MODE_EXISTING) {
1744 assert(backup->format);
1745 if (source) {
1746 bdrv_refresh_filename(source);
1747 bdrv_img_create(backup->target, backup->format, source->filename,
1748 source->drv->format_name, NULL,
1749 size, flags, false, &local_err);
1750 } else {
1751 bdrv_img_create(backup->target, backup->format, NULL, NULL, NULL,
1752 size, flags, false, &local_err);
1756 if (local_err) {
1757 error_propagate(errp, local_err);
1758 goto out;
1761 options = qdict_new();
1762 qdict_put_str(options, "discard", "unmap");
1763 qdict_put_str(options, "detect-zeroes", "unmap");
1764 if (backup->format) {
1765 qdict_put_str(options, "driver", backup->format);
1768 target_bs = bdrv_open(backup->target, NULL, options, flags, errp);
1769 if (!target_bs) {
1770 goto out;
1773 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1774 old_context = bdrv_get_aio_context(target_bs);
1775 aio_context_release(aio_context);
1776 aio_context_acquire(old_context);
1778 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1779 if (ret < 0) {
1780 bdrv_unref(target_bs);
1781 aio_context_release(old_context);
1782 return;
1785 aio_context_release(old_context);
1786 aio_context_acquire(aio_context);
1788 if (set_backing_hd) {
1789 bdrv_set_backing_hd(target_bs, source, &local_err);
1790 if (local_err) {
1791 goto unref;
1795 state->bs = bs;
1797 state->job = do_backup_common(qapi_DriveBackup_base(backup),
1798 bs, target_bs, aio_context,
1799 common->block_job_txn, errp);
1801 unref:
1802 bdrv_unref(target_bs);
1803 out:
1804 aio_context_release(aio_context);
1807 static void drive_backup_commit(BlkActionState *common)
1809 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1810 AioContext *aio_context;
1812 aio_context = bdrv_get_aio_context(state->bs);
1813 aio_context_acquire(aio_context);
1815 assert(state->job);
1816 job_start(&state->job->job);
1818 aio_context_release(aio_context);
1821 static void drive_backup_abort(BlkActionState *common)
1823 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1825 if (state->job) {
1826 AioContext *aio_context;
1828 aio_context = bdrv_get_aio_context(state->bs);
1829 aio_context_acquire(aio_context);
1831 job_cancel_sync(&state->job->job);
1833 aio_context_release(aio_context);
1837 static void drive_backup_clean(BlkActionState *common)
1839 DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common);
1840 AioContext *aio_context;
1842 if (!state->bs) {
1843 return;
1846 aio_context = bdrv_get_aio_context(state->bs);
1847 aio_context_acquire(aio_context);
1849 bdrv_drained_end(state->bs);
1851 aio_context_release(aio_context);
1854 typedef struct BlockdevBackupState {
1855 BlkActionState common;
1856 BlockDriverState *bs;
1857 BlockJob *job;
1858 } BlockdevBackupState;
1860 static void blockdev_backup_prepare(BlkActionState *common, Error **errp)
1862 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1863 BlockdevBackup *backup;
1864 BlockDriverState *bs;
1865 BlockDriverState *target_bs;
1866 AioContext *aio_context;
1867 AioContext *old_context;
1868 int ret;
1870 assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP);
1871 backup = common->action->u.blockdev_backup.data;
1873 bs = bdrv_lookup_bs(backup->device, backup->device, errp);
1874 if (!bs) {
1875 return;
1878 target_bs = bdrv_lookup_bs(backup->target, backup->target, errp);
1879 if (!target_bs) {
1880 return;
1883 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
1884 aio_context = bdrv_get_aio_context(bs);
1885 old_context = bdrv_get_aio_context(target_bs);
1886 aio_context_acquire(old_context);
1888 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
1889 if (ret < 0) {
1890 aio_context_release(old_context);
1891 return;
1894 aio_context_release(old_context);
1895 aio_context_acquire(aio_context);
1896 state->bs = bs;
1898 /* Paired with .clean() */
1899 bdrv_drained_begin(state->bs);
1901 state->job = do_backup_common(qapi_BlockdevBackup_base(backup),
1902 bs, target_bs, aio_context,
1903 common->block_job_txn, errp);
1905 aio_context_release(aio_context);
1908 static void blockdev_backup_commit(BlkActionState *common)
1910 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1911 AioContext *aio_context;
1913 aio_context = bdrv_get_aio_context(state->bs);
1914 aio_context_acquire(aio_context);
1916 assert(state->job);
1917 job_start(&state->job->job);
1919 aio_context_release(aio_context);
1922 static void blockdev_backup_abort(BlkActionState *common)
1924 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1926 if (state->job) {
1927 AioContext *aio_context;
1929 aio_context = bdrv_get_aio_context(state->bs);
1930 aio_context_acquire(aio_context);
1932 job_cancel_sync(&state->job->job);
1934 aio_context_release(aio_context);
1938 static void blockdev_backup_clean(BlkActionState *common)
1940 BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common);
1941 AioContext *aio_context;
1943 if (!state->bs) {
1944 return;
1947 aio_context = bdrv_get_aio_context(state->bs);
1948 aio_context_acquire(aio_context);
1950 bdrv_drained_end(state->bs);
1952 aio_context_release(aio_context);
1955 typedef struct BlockDirtyBitmapState {
1956 BlkActionState common;
1957 BdrvDirtyBitmap *bitmap;
1958 BlockDriverState *bs;
1959 HBitmap *backup;
1960 bool prepared;
1961 bool was_enabled;
1962 } BlockDirtyBitmapState;
1964 static void block_dirty_bitmap_add_prepare(BlkActionState *common,
1965 Error **errp)
1967 Error *local_err = NULL;
1968 BlockDirtyBitmapAdd *action;
1969 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1970 common, common);
1972 if (action_check_completion_mode(common, errp) < 0) {
1973 return;
1976 action = common->action->u.block_dirty_bitmap_add.data;
1977 /* AIO context taken and released within qmp_block_dirty_bitmap_add */
1978 qmp_block_dirty_bitmap_add(action->node, action->name,
1979 action->has_granularity, action->granularity,
1980 action->has_persistent, action->persistent,
1981 action->has_disabled, action->disabled,
1982 &local_err);
1984 if (!local_err) {
1985 state->prepared = true;
1986 } else {
1987 error_propagate(errp, local_err);
1991 static void block_dirty_bitmap_add_abort(BlkActionState *common)
1993 BlockDirtyBitmapAdd *action;
1994 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
1995 common, common);
1997 action = common->action->u.block_dirty_bitmap_add.data;
1998 /* Should not be able to fail: IF the bitmap was added via .prepare(),
1999 * then the node reference and bitmap name must have been valid.
2001 if (state->prepared) {
2002 qmp_block_dirty_bitmap_remove(action->node, action->name, &error_abort);
2006 static void block_dirty_bitmap_clear_prepare(BlkActionState *common,
2007 Error **errp)
2009 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2010 common, common);
2011 BlockDirtyBitmap *action;
2013 if (action_check_completion_mode(common, errp) < 0) {
2014 return;
2017 action = common->action->u.block_dirty_bitmap_clear.data;
2018 state->bitmap = block_dirty_bitmap_lookup(action->node,
2019 action->name,
2020 &state->bs,
2021 errp);
2022 if (!state->bitmap) {
2023 return;
2026 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_DEFAULT, errp)) {
2027 return;
2030 bdrv_clear_dirty_bitmap(state->bitmap, &state->backup);
2033 static void block_dirty_bitmap_restore(BlkActionState *common)
2035 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2036 common, common);
2038 if (state->backup) {
2039 bdrv_restore_dirty_bitmap(state->bitmap, state->backup);
2043 static void block_dirty_bitmap_free_backup(BlkActionState *common)
2045 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2046 common, common);
2048 hbitmap_free(state->backup);
2051 static void block_dirty_bitmap_enable_prepare(BlkActionState *common,
2052 Error **errp)
2054 BlockDirtyBitmap *action;
2055 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2056 common, common);
2058 if (action_check_completion_mode(common, errp) < 0) {
2059 return;
2062 action = common->action->u.block_dirty_bitmap_enable.data;
2063 state->bitmap = block_dirty_bitmap_lookup(action->node,
2064 action->name,
2065 NULL,
2066 errp);
2067 if (!state->bitmap) {
2068 return;
2071 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2072 return;
2075 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2076 bdrv_enable_dirty_bitmap(state->bitmap);
2079 static void block_dirty_bitmap_enable_abort(BlkActionState *common)
2081 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2082 common, common);
2084 if (!state->was_enabled) {
2085 bdrv_disable_dirty_bitmap(state->bitmap);
2089 static void block_dirty_bitmap_disable_prepare(BlkActionState *common,
2090 Error **errp)
2092 BlockDirtyBitmap *action;
2093 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2094 common, common);
2096 if (action_check_completion_mode(common, errp) < 0) {
2097 return;
2100 action = common->action->u.block_dirty_bitmap_disable.data;
2101 state->bitmap = block_dirty_bitmap_lookup(action->node,
2102 action->name,
2103 NULL,
2104 errp);
2105 if (!state->bitmap) {
2106 return;
2109 if (bdrv_dirty_bitmap_check(state->bitmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2110 return;
2113 state->was_enabled = bdrv_dirty_bitmap_enabled(state->bitmap);
2114 bdrv_disable_dirty_bitmap(state->bitmap);
2117 static void block_dirty_bitmap_disable_abort(BlkActionState *common)
2119 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2120 common, common);
2122 if (state->was_enabled) {
2123 bdrv_enable_dirty_bitmap(state->bitmap);
2127 static void block_dirty_bitmap_merge_prepare(BlkActionState *common,
2128 Error **errp)
2130 BlockDirtyBitmapMerge *action;
2131 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2132 common, common);
2134 if (action_check_completion_mode(common, errp) < 0) {
2135 return;
2138 action = common->action->u.block_dirty_bitmap_merge.data;
2140 state->bitmap = block_dirty_bitmap_merge(action->node, action->target,
2141 action->bitmaps, &state->backup,
2142 errp);
2145 static void block_dirty_bitmap_remove_prepare(BlkActionState *common,
2146 Error **errp)
2148 BlockDirtyBitmap *action;
2149 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2150 common, common);
2152 if (action_check_completion_mode(common, errp) < 0) {
2153 return;
2156 action = common->action->u.block_dirty_bitmap_remove.data;
2158 state->bitmap = block_dirty_bitmap_remove(action->node, action->name,
2159 false, &state->bs, errp);
2160 if (state->bitmap) {
2161 bdrv_dirty_bitmap_skip_store(state->bitmap, true);
2162 bdrv_dirty_bitmap_set_busy(state->bitmap, true);
2166 static void block_dirty_bitmap_remove_abort(BlkActionState *common)
2168 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2169 common, common);
2171 if (state->bitmap) {
2172 bdrv_dirty_bitmap_skip_store(state->bitmap, false);
2173 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2177 static void block_dirty_bitmap_remove_commit(BlkActionState *common)
2179 BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState,
2180 common, common);
2182 bdrv_dirty_bitmap_set_busy(state->bitmap, false);
2183 bdrv_release_dirty_bitmap(state->bitmap);
2186 static void abort_prepare(BlkActionState *common, Error **errp)
2188 error_setg(errp, "Transaction aborted using Abort action");
2191 static void abort_commit(BlkActionState *common)
2193 g_assert_not_reached(); /* this action never succeeds */
2196 static const BlkActionOps actions[] = {
2197 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT] = {
2198 .instance_size = sizeof(ExternalSnapshotState),
2199 .prepare = external_snapshot_prepare,
2200 .commit = external_snapshot_commit,
2201 .abort = external_snapshot_abort,
2202 .clean = external_snapshot_clean,
2204 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC] = {
2205 .instance_size = sizeof(ExternalSnapshotState),
2206 .prepare = external_snapshot_prepare,
2207 .commit = external_snapshot_commit,
2208 .abort = external_snapshot_abort,
2209 .clean = external_snapshot_clean,
2211 [TRANSACTION_ACTION_KIND_DRIVE_BACKUP] = {
2212 .instance_size = sizeof(DriveBackupState),
2213 .prepare = drive_backup_prepare,
2214 .commit = drive_backup_commit,
2215 .abort = drive_backup_abort,
2216 .clean = drive_backup_clean,
2218 [TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP] = {
2219 .instance_size = sizeof(BlockdevBackupState),
2220 .prepare = blockdev_backup_prepare,
2221 .commit = blockdev_backup_commit,
2222 .abort = blockdev_backup_abort,
2223 .clean = blockdev_backup_clean,
2225 [TRANSACTION_ACTION_KIND_ABORT] = {
2226 .instance_size = sizeof(BlkActionState),
2227 .prepare = abort_prepare,
2228 .commit = abort_commit,
2230 [TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_INTERNAL_SYNC] = {
2231 .instance_size = sizeof(InternalSnapshotState),
2232 .prepare = internal_snapshot_prepare,
2233 .abort = internal_snapshot_abort,
2234 .clean = internal_snapshot_clean,
2236 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ADD] = {
2237 .instance_size = sizeof(BlockDirtyBitmapState),
2238 .prepare = block_dirty_bitmap_add_prepare,
2239 .abort = block_dirty_bitmap_add_abort,
2241 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_CLEAR] = {
2242 .instance_size = sizeof(BlockDirtyBitmapState),
2243 .prepare = block_dirty_bitmap_clear_prepare,
2244 .commit = block_dirty_bitmap_free_backup,
2245 .abort = block_dirty_bitmap_restore,
2247 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_ENABLE] = {
2248 .instance_size = sizeof(BlockDirtyBitmapState),
2249 .prepare = block_dirty_bitmap_enable_prepare,
2250 .abort = block_dirty_bitmap_enable_abort,
2252 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_DISABLE] = {
2253 .instance_size = sizeof(BlockDirtyBitmapState),
2254 .prepare = block_dirty_bitmap_disable_prepare,
2255 .abort = block_dirty_bitmap_disable_abort,
2257 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_MERGE] = {
2258 .instance_size = sizeof(BlockDirtyBitmapState),
2259 .prepare = block_dirty_bitmap_merge_prepare,
2260 .commit = block_dirty_bitmap_free_backup,
2261 .abort = block_dirty_bitmap_restore,
2263 [TRANSACTION_ACTION_KIND_BLOCK_DIRTY_BITMAP_REMOVE] = {
2264 .instance_size = sizeof(BlockDirtyBitmapState),
2265 .prepare = block_dirty_bitmap_remove_prepare,
2266 .commit = block_dirty_bitmap_remove_commit,
2267 .abort = block_dirty_bitmap_remove_abort,
2269 /* Where are transactions for MIRROR, COMMIT and STREAM?
2270 * Although these blockjobs use transaction callbacks like the backup job,
2271 * these jobs do not necessarily adhere to transaction semantics.
2272 * These jobs may not fully undo all of their actions on abort, nor do they
2273 * necessarily work in transactions with more than one job in them.
2278 * Allocate a TransactionProperties structure if necessary, and fill
2279 * that structure with desired defaults if they are unset.
2281 static TransactionProperties *get_transaction_properties(
2282 TransactionProperties *props)
2284 if (!props) {
2285 props = g_new0(TransactionProperties, 1);
2288 if (!props->has_completion_mode) {
2289 props->has_completion_mode = true;
2290 props->completion_mode = ACTION_COMPLETION_MODE_INDIVIDUAL;
2293 return props;
2297 * 'Atomic' group operations. The operations are performed as a set, and if
2298 * any fail then we roll back all operations in the group.
2300 void qmp_transaction(TransactionActionList *dev_list,
2301 bool has_props,
2302 struct TransactionProperties *props,
2303 Error **errp)
2305 TransactionActionList *dev_entry = dev_list;
2306 JobTxn *block_job_txn = NULL;
2307 BlkActionState *state, *next;
2308 Error *local_err = NULL;
2310 QTAILQ_HEAD(, BlkActionState) snap_bdrv_states;
2311 QTAILQ_INIT(&snap_bdrv_states);
2313 /* Does this transaction get canceled as a group on failure?
2314 * If not, we don't really need to make a JobTxn.
2316 props = get_transaction_properties(props);
2317 if (props->completion_mode != ACTION_COMPLETION_MODE_INDIVIDUAL) {
2318 block_job_txn = job_txn_new();
2321 /* drain all i/o before any operations */
2322 bdrv_drain_all();
2324 /* We don't do anything in this loop that commits us to the operations */
2325 while (NULL != dev_entry) {
2326 TransactionAction *dev_info = NULL;
2327 const BlkActionOps *ops;
2329 dev_info = dev_entry->value;
2330 dev_entry = dev_entry->next;
2332 assert(dev_info->type < ARRAY_SIZE(actions));
2334 ops = &actions[dev_info->type];
2335 assert(ops->instance_size > 0);
2337 state = g_malloc0(ops->instance_size);
2338 state->ops = ops;
2339 state->action = dev_info;
2340 state->block_job_txn = block_job_txn;
2341 state->txn_props = props;
2342 QTAILQ_INSERT_TAIL(&snap_bdrv_states, state, entry);
2344 state->ops->prepare(state, &local_err);
2345 if (local_err) {
2346 error_propagate(errp, local_err);
2347 goto delete_and_fail;
2351 QTAILQ_FOREACH(state, &snap_bdrv_states, entry) {
2352 if (state->ops->commit) {
2353 state->ops->commit(state);
2357 /* success */
2358 goto exit;
2360 delete_and_fail:
2361 /* failure, and it is all-or-none; roll back all operations */
2362 QTAILQ_FOREACH_REVERSE(state, &snap_bdrv_states, entry) {
2363 if (state->ops->abort) {
2364 state->ops->abort(state);
2367 exit:
2368 QTAILQ_FOREACH_SAFE(state, &snap_bdrv_states, entry, next) {
2369 if (state->ops->clean) {
2370 state->ops->clean(state);
2372 g_free(state);
2374 if (!has_props) {
2375 qapi_free_TransactionProperties(props);
2377 job_txn_unref(block_job_txn);
2380 void qmp_block_passwd(bool has_device, const char *device,
2381 bool has_node_name, const char *node_name,
2382 const char *password, Error **errp)
2384 error_setg(errp,
2385 "Setting block passwords directly is no longer supported");
2388 BlockDirtyBitmapSha256 *qmp_x_debug_block_dirty_bitmap_sha256(const char *node,
2389 const char *name,
2390 Error **errp)
2392 BdrvDirtyBitmap *bitmap;
2393 BlockDriverState *bs;
2394 BlockDirtyBitmapSha256 *ret = NULL;
2395 char *sha256;
2397 bitmap = block_dirty_bitmap_lookup(node, name, &bs, errp);
2398 if (!bitmap || !bs) {
2399 return NULL;
2402 sha256 = bdrv_dirty_bitmap_sha256(bitmap, errp);
2403 if (sha256 == NULL) {
2404 return NULL;
2407 ret = g_new(BlockDirtyBitmapSha256, 1);
2408 ret->sha256 = sha256;
2410 return ret;
2413 void qmp_block_resize(bool has_device, const char *device,
2414 bool has_node_name, const char *node_name,
2415 int64_t size, Error **errp)
2417 Error *local_err = NULL;
2418 BlockBackend *blk = NULL;
2419 BlockDriverState *bs;
2420 AioContext *aio_context;
2422 bs = bdrv_lookup_bs(has_device ? device : NULL,
2423 has_node_name ? node_name : NULL,
2424 &local_err);
2425 if (local_err) {
2426 error_propagate(errp, local_err);
2427 return;
2430 aio_context = bdrv_get_aio_context(bs);
2431 aio_context_acquire(aio_context);
2433 if (size < 0) {
2434 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "size", "a >0 size");
2435 goto out;
2438 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_RESIZE, NULL)) {
2439 error_setg(errp, QERR_DEVICE_IN_USE, device);
2440 goto out;
2443 blk = blk_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp);
2444 if (!blk) {
2445 goto out;
2448 bdrv_drained_begin(bs);
2449 blk_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp);
2450 bdrv_drained_end(bs);
2452 out:
2453 blk_unref(blk);
2454 aio_context_release(aio_context);
2457 void qmp_block_stream(bool has_job_id, const char *job_id, const char *device,
2458 bool has_base, const char *base,
2459 bool has_base_node, const char *base_node,
2460 bool has_backing_file, const char *backing_file,
2461 bool has_speed, int64_t speed,
2462 bool has_on_error, BlockdevOnError on_error,
2463 bool has_auto_finalize, bool auto_finalize,
2464 bool has_auto_dismiss, bool auto_dismiss,
2465 Error **errp)
2467 BlockDriverState *bs, *iter;
2468 BlockDriverState *base_bs = NULL;
2469 AioContext *aio_context;
2470 Error *local_err = NULL;
2471 const char *base_name = NULL;
2472 int job_flags = JOB_DEFAULT;
2474 if (!has_on_error) {
2475 on_error = BLOCKDEV_ON_ERROR_REPORT;
2478 bs = bdrv_lookup_bs(device, device, errp);
2479 if (!bs) {
2480 return;
2483 aio_context = bdrv_get_aio_context(bs);
2484 aio_context_acquire(aio_context);
2486 if (has_base && has_base_node) {
2487 error_setg(errp, "'base' and 'base-node' cannot be specified "
2488 "at the same time");
2489 goto out;
2492 if (has_base) {
2493 base_bs = bdrv_find_backing_image(bs, base);
2494 if (base_bs == NULL) {
2495 error_setg(errp, QERR_BASE_NOT_FOUND, base);
2496 goto out;
2498 assert(bdrv_get_aio_context(base_bs) == aio_context);
2499 base_name = base;
2502 if (has_base_node) {
2503 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2504 if (!base_bs) {
2505 goto out;
2507 if (bs == base_bs || !bdrv_chain_contains(bs, base_bs)) {
2508 error_setg(errp, "Node '%s' is not a backing image of '%s'",
2509 base_node, device);
2510 goto out;
2512 assert(bdrv_get_aio_context(base_bs) == aio_context);
2513 bdrv_refresh_filename(base_bs);
2514 base_name = base_bs->filename;
2517 /* Check for op blockers in the whole chain between bs and base */
2518 for (iter = bs; iter && iter != base_bs; iter = backing_bs(iter)) {
2519 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_STREAM, errp)) {
2520 goto out;
2524 /* if we are streaming the entire chain, the result will have no backing
2525 * file, and specifying one is therefore an error */
2526 if (base_bs == NULL && has_backing_file) {
2527 error_setg(errp, "backing file specified, but streaming the "
2528 "entire chain");
2529 goto out;
2532 /* backing_file string overrides base bs filename */
2533 base_name = has_backing_file ? backing_file : base_name;
2535 if (has_auto_finalize && !auto_finalize) {
2536 job_flags |= JOB_MANUAL_FINALIZE;
2538 if (has_auto_dismiss && !auto_dismiss) {
2539 job_flags |= JOB_MANUAL_DISMISS;
2542 stream_start(has_job_id ? job_id : NULL, bs, base_bs, base_name,
2543 job_flags, has_speed ? speed : 0, on_error, &local_err);
2544 if (local_err) {
2545 error_propagate(errp, local_err);
2546 goto out;
2549 trace_qmp_block_stream(bs);
2551 out:
2552 aio_context_release(aio_context);
2555 void qmp_block_commit(bool has_job_id, const char *job_id, const char *device,
2556 bool has_base_node, const char *base_node,
2557 bool has_base, const char *base,
2558 bool has_top_node, const char *top_node,
2559 bool has_top, const char *top,
2560 bool has_backing_file, const char *backing_file,
2561 bool has_speed, int64_t speed,
2562 bool has_on_error, BlockdevOnError on_error,
2563 bool has_filter_node_name, const char *filter_node_name,
2564 bool has_auto_finalize, bool auto_finalize,
2565 bool has_auto_dismiss, bool auto_dismiss,
2566 Error **errp)
2568 BlockDriverState *bs;
2569 BlockDriverState *iter;
2570 BlockDriverState *base_bs, *top_bs;
2571 AioContext *aio_context;
2572 Error *local_err = NULL;
2573 int job_flags = JOB_DEFAULT;
2575 if (!has_speed) {
2576 speed = 0;
2578 if (!has_on_error) {
2579 on_error = BLOCKDEV_ON_ERROR_REPORT;
2581 if (!has_filter_node_name) {
2582 filter_node_name = NULL;
2584 if (has_auto_finalize && !auto_finalize) {
2585 job_flags |= JOB_MANUAL_FINALIZE;
2587 if (has_auto_dismiss && !auto_dismiss) {
2588 job_flags |= JOB_MANUAL_DISMISS;
2591 /* Important Note:
2592 * libvirt relies on the DeviceNotFound error class in order to probe for
2593 * live commit feature versions; for this to work, we must make sure to
2594 * perform the device lookup before any generic errors that may occur in a
2595 * scenario in which all optional arguments are omitted. */
2596 bs = qmp_get_root_bs(device, &local_err);
2597 if (!bs) {
2598 bs = bdrv_lookup_bs(device, device, NULL);
2599 if (!bs) {
2600 error_free(local_err);
2601 error_set(errp, ERROR_CLASS_DEVICE_NOT_FOUND,
2602 "Device '%s' not found", device);
2603 } else {
2604 error_propagate(errp, local_err);
2606 return;
2609 aio_context = bdrv_get_aio_context(bs);
2610 aio_context_acquire(aio_context);
2612 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_COMMIT_SOURCE, errp)) {
2613 goto out;
2616 /* default top_bs is the active layer */
2617 top_bs = bs;
2619 if (has_top_node && has_top) {
2620 error_setg(errp, "'top-node' and 'top' are mutually exclusive");
2621 goto out;
2622 } else if (has_top_node) {
2623 top_bs = bdrv_lookup_bs(NULL, top_node, errp);
2624 if (top_bs == NULL) {
2625 goto out;
2627 if (!bdrv_chain_contains(bs, top_bs)) {
2628 error_setg(errp, "'%s' is not in this backing file chain",
2629 top_node);
2630 goto out;
2632 } else if (has_top && top) {
2633 /* This strcmp() is just a shortcut, there is no need to
2634 * refresh @bs's filename. If it mismatches,
2635 * bdrv_find_backing_image() will do the refresh and may still
2636 * return @bs. */
2637 if (strcmp(bs->filename, top) != 0) {
2638 top_bs = bdrv_find_backing_image(bs, top);
2642 if (top_bs == NULL) {
2643 error_setg(errp, "Top image file %s not found", top ? top : "NULL");
2644 goto out;
2647 assert(bdrv_get_aio_context(top_bs) == aio_context);
2649 if (has_base_node && has_base) {
2650 error_setg(errp, "'base-node' and 'base' are mutually exclusive");
2651 goto out;
2652 } else if (has_base_node) {
2653 base_bs = bdrv_lookup_bs(NULL, base_node, errp);
2654 if (base_bs == NULL) {
2655 goto out;
2657 if (!bdrv_chain_contains(top_bs, base_bs)) {
2658 error_setg(errp, "'%s' is not in this backing file chain",
2659 base_node);
2660 goto out;
2662 } else if (has_base && base) {
2663 base_bs = bdrv_find_backing_image(top_bs, base);
2664 } else {
2665 base_bs = bdrv_find_base(top_bs);
2668 if (base_bs == NULL) {
2669 error_setg(errp, QERR_BASE_NOT_FOUND, base ? base : "NULL");
2670 goto out;
2673 assert(bdrv_get_aio_context(base_bs) == aio_context);
2675 for (iter = top_bs; iter != backing_bs(base_bs); iter = backing_bs(iter)) {
2676 if (bdrv_op_is_blocked(iter, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2677 goto out;
2681 /* Do not allow attempts to commit an image into itself */
2682 if (top_bs == base_bs) {
2683 error_setg(errp, "cannot commit an image into itself");
2684 goto out;
2687 if (top_bs == bs) {
2688 if (has_backing_file) {
2689 error_setg(errp, "'backing-file' specified,"
2690 " but 'top' is the active layer");
2691 goto out;
2693 commit_active_start(has_job_id ? job_id : NULL, bs, base_bs,
2694 job_flags, speed, on_error,
2695 filter_node_name, NULL, NULL, false, &local_err);
2696 } else {
2697 BlockDriverState *overlay_bs = bdrv_find_overlay(bs, top_bs);
2698 if (bdrv_op_is_blocked(overlay_bs, BLOCK_OP_TYPE_COMMIT_TARGET, errp)) {
2699 goto out;
2701 commit_start(has_job_id ? job_id : NULL, bs, base_bs, top_bs, job_flags,
2702 speed, on_error, has_backing_file ? backing_file : NULL,
2703 filter_node_name, &local_err);
2705 if (local_err != NULL) {
2706 error_propagate(errp, local_err);
2707 goto out;
2710 out:
2711 aio_context_release(aio_context);
2714 /* Common QMP interface for drive-backup and blockdev-backup */
2715 static BlockJob *do_backup_common(BackupCommon *backup,
2716 BlockDriverState *bs,
2717 BlockDriverState *target_bs,
2718 AioContext *aio_context,
2719 JobTxn *txn, Error **errp)
2721 BlockJob *job = NULL;
2722 BdrvDirtyBitmap *bmap = NULL;
2723 int job_flags = JOB_DEFAULT;
2725 if (!backup->has_speed) {
2726 backup->speed = 0;
2728 if (!backup->has_on_source_error) {
2729 backup->on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2731 if (!backup->has_on_target_error) {
2732 backup->on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2734 if (!backup->has_job_id) {
2735 backup->job_id = NULL;
2737 if (!backup->has_auto_finalize) {
2738 backup->auto_finalize = true;
2740 if (!backup->has_auto_dismiss) {
2741 backup->auto_dismiss = true;
2743 if (!backup->has_compress) {
2744 backup->compress = false;
2747 if ((backup->sync == MIRROR_SYNC_MODE_BITMAP) ||
2748 (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL)) {
2749 /* done before desugaring 'incremental' to print the right message */
2750 if (!backup->has_bitmap) {
2751 error_setg(errp, "must provide a valid bitmap name for "
2752 "'%s' sync mode", MirrorSyncMode_str(backup->sync));
2753 return NULL;
2757 if (backup->sync == MIRROR_SYNC_MODE_INCREMENTAL) {
2758 if (backup->has_bitmap_mode &&
2759 backup->bitmap_mode != BITMAP_SYNC_MODE_ON_SUCCESS) {
2760 error_setg(errp, "Bitmap sync mode must be '%s' "
2761 "when using sync mode '%s'",
2762 BitmapSyncMode_str(BITMAP_SYNC_MODE_ON_SUCCESS),
2763 MirrorSyncMode_str(backup->sync));
2764 return NULL;
2766 backup->has_bitmap_mode = true;
2767 backup->sync = MIRROR_SYNC_MODE_BITMAP;
2768 backup->bitmap_mode = BITMAP_SYNC_MODE_ON_SUCCESS;
2771 if (backup->has_bitmap) {
2772 bmap = bdrv_find_dirty_bitmap(bs, backup->bitmap);
2773 if (!bmap) {
2774 error_setg(errp, "Bitmap '%s' could not be found", backup->bitmap);
2775 return NULL;
2777 if (!backup->has_bitmap_mode) {
2778 error_setg(errp, "Bitmap sync mode must be given "
2779 "when providing a bitmap");
2780 return NULL;
2782 if (bdrv_dirty_bitmap_check(bmap, BDRV_BITMAP_ALLOW_RO, errp)) {
2783 return NULL;
2786 /* This does not produce a useful bitmap artifact: */
2787 if (backup->sync == MIRROR_SYNC_MODE_NONE) {
2788 error_setg(errp, "sync mode '%s' does not produce meaningful bitmap"
2789 " outputs", MirrorSyncMode_str(backup->sync));
2790 return NULL;
2793 /* If the bitmap isn't used for input or output, this is useless: */
2794 if (backup->bitmap_mode == BITMAP_SYNC_MODE_NEVER &&
2795 backup->sync != MIRROR_SYNC_MODE_BITMAP) {
2796 error_setg(errp, "Bitmap sync mode '%s' has no meaningful effect"
2797 " when combined with sync mode '%s'",
2798 BitmapSyncMode_str(backup->bitmap_mode),
2799 MirrorSyncMode_str(backup->sync));
2800 return NULL;
2804 if (!backup->has_bitmap && backup->has_bitmap_mode) {
2805 error_setg(errp, "Cannot specify bitmap sync mode without a bitmap");
2806 return NULL;
2809 if (!backup->auto_finalize) {
2810 job_flags |= JOB_MANUAL_FINALIZE;
2812 if (!backup->auto_dismiss) {
2813 job_flags |= JOB_MANUAL_DISMISS;
2816 job = backup_job_create(backup->job_id, bs, target_bs, backup->speed,
2817 backup->sync, bmap, backup->bitmap_mode,
2818 backup->compress,
2819 backup->filter_node_name,
2820 backup->on_source_error,
2821 backup->on_target_error,
2822 job_flags, NULL, NULL, txn, errp);
2823 return job;
2826 void qmp_drive_backup(DriveBackup *backup, Error **errp)
2828 TransactionAction action = {
2829 .type = TRANSACTION_ACTION_KIND_DRIVE_BACKUP,
2830 .u.drive_backup.data = backup,
2832 blockdev_do_action(&action, errp);
2835 BlockDeviceInfoList *qmp_query_named_block_nodes(bool has_flat,
2836 bool flat,
2837 Error **errp)
2839 bool return_flat = has_flat && flat;
2841 return bdrv_named_nodes_list(return_flat, errp);
2844 XDbgBlockGraph *qmp_x_debug_query_block_graph(Error **errp)
2846 return bdrv_get_xdbg_block_graph(errp);
2849 void qmp_blockdev_backup(BlockdevBackup *backup, Error **errp)
2851 TransactionAction action = {
2852 .type = TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP,
2853 .u.blockdev_backup.data = backup,
2855 blockdev_do_action(&action, errp);
2858 /* Parameter check and block job starting for drive mirroring.
2859 * Caller should hold @device and @target's aio context (must be the same).
2861 static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs,
2862 BlockDriverState *target,
2863 bool has_replaces, const char *replaces,
2864 enum MirrorSyncMode sync,
2865 BlockMirrorBackingMode backing_mode,
2866 bool zero_target,
2867 bool has_speed, int64_t speed,
2868 bool has_granularity, uint32_t granularity,
2869 bool has_buf_size, int64_t buf_size,
2870 bool has_on_source_error,
2871 BlockdevOnError on_source_error,
2872 bool has_on_target_error,
2873 BlockdevOnError on_target_error,
2874 bool has_unmap, bool unmap,
2875 bool has_filter_node_name,
2876 const char *filter_node_name,
2877 bool has_copy_mode, MirrorCopyMode copy_mode,
2878 bool has_auto_finalize, bool auto_finalize,
2879 bool has_auto_dismiss, bool auto_dismiss,
2880 Error **errp)
2882 int job_flags = JOB_DEFAULT;
2884 if (!has_speed) {
2885 speed = 0;
2887 if (!has_on_source_error) {
2888 on_source_error = BLOCKDEV_ON_ERROR_REPORT;
2890 if (!has_on_target_error) {
2891 on_target_error = BLOCKDEV_ON_ERROR_REPORT;
2893 if (!has_granularity) {
2894 granularity = 0;
2896 if (!has_buf_size) {
2897 buf_size = 0;
2899 if (!has_unmap) {
2900 unmap = true;
2902 if (!has_filter_node_name) {
2903 filter_node_name = NULL;
2905 if (!has_copy_mode) {
2906 copy_mode = MIRROR_COPY_MODE_BACKGROUND;
2908 if (has_auto_finalize && !auto_finalize) {
2909 job_flags |= JOB_MANUAL_FINALIZE;
2911 if (has_auto_dismiss && !auto_dismiss) {
2912 job_flags |= JOB_MANUAL_DISMISS;
2915 if (granularity != 0 && (granularity < 512 || granularity > 1048576 * 64)) {
2916 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2917 "a value in range [512B, 64MB]");
2918 return;
2920 if (granularity & (granularity - 1)) {
2921 error_setg(errp, QERR_INVALID_PARAMETER_VALUE, "granularity",
2922 "power of 2");
2923 return;
2926 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_MIRROR_SOURCE, errp)) {
2927 return;
2929 if (bdrv_op_is_blocked(target, BLOCK_OP_TYPE_MIRROR_TARGET, errp)) {
2930 return;
2933 if (!bs->backing && sync == MIRROR_SYNC_MODE_TOP) {
2934 sync = MIRROR_SYNC_MODE_FULL;
2937 if (has_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 has_replaces ? replaces : NULL, 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 *source, *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->has_format) {
3013 format = (arg->mode == NEW_IMAGE_MODE_EXISTING
3014 ? NULL : bs->drv->format_name);
3017 flags = bs->open_flags | BDRV_O_RDWR;
3018 source = backing_bs(bs);
3019 if (!source && arg->sync == MIRROR_SYNC_MODE_TOP) {
3020 arg->sync = MIRROR_SYNC_MODE_FULL;
3022 if (arg->sync == MIRROR_SYNC_MODE_NONE) {
3023 source = 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->has_replaces) {
3033 if (!arg->has_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 || !source)
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 switch (arg->mode) {
3058 case NEW_IMAGE_MODE_EXISTING:
3059 break;
3060 case NEW_IMAGE_MODE_ABSOLUTE_PATHS:
3061 /* create new image with backing file */
3062 bdrv_refresh_filename(source);
3063 bdrv_img_create(arg->target, format,
3064 source->filename,
3065 source->drv->format_name,
3066 NULL, size, flags, false, &local_err);
3067 break;
3068 default:
3069 abort();
3073 if (local_err) {
3074 error_propagate(errp, local_err);
3075 goto out;
3078 options = qdict_new();
3079 if (arg->has_node_name) {
3080 qdict_put_str(options, "node-name", arg->node_name);
3082 if (format) {
3083 qdict_put_str(options, "driver", format);
3086 /* Mirroring takes care of copy-on-write using the source's backing
3087 * file.
3089 target_bs = bdrv_open(arg->target, NULL, options, flags, errp);
3090 if (!target_bs) {
3091 goto out;
3094 zero_target = (arg->sync == MIRROR_SYNC_MODE_FULL &&
3095 (arg->mode == NEW_IMAGE_MODE_EXISTING ||
3096 !bdrv_has_zero_init(target_bs)));
3099 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3100 old_context = bdrv_get_aio_context(target_bs);
3101 aio_context_release(aio_context);
3102 aio_context_acquire(old_context);
3104 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3105 if (ret < 0) {
3106 bdrv_unref(target_bs);
3107 aio_context_release(old_context);
3108 return;
3111 aio_context_release(old_context);
3112 aio_context_acquire(aio_context);
3114 blockdev_mirror_common(arg->has_job_id ? arg->job_id : NULL, bs, target_bs,
3115 arg->has_replaces, arg->replaces, arg->sync,
3116 backing_mode, zero_target,
3117 arg->has_speed, arg->speed,
3118 arg->has_granularity, arg->granularity,
3119 arg->has_buf_size, arg->buf_size,
3120 arg->has_on_source_error, arg->on_source_error,
3121 arg->has_on_target_error, arg->on_target_error,
3122 arg->has_unmap, arg->unmap,
3123 false, NULL,
3124 arg->has_copy_mode, arg->copy_mode,
3125 arg->has_auto_finalize, arg->auto_finalize,
3126 arg->has_auto_dismiss, arg->auto_dismiss,
3127 &local_err);
3128 bdrv_unref(target_bs);
3129 error_propagate(errp, local_err);
3130 out:
3131 aio_context_release(aio_context);
3134 void qmp_blockdev_mirror(bool has_job_id, const char *job_id,
3135 const char *device, const char *target,
3136 bool has_replaces, const char *replaces,
3137 MirrorSyncMode sync,
3138 bool has_speed, int64_t speed,
3139 bool has_granularity, uint32_t granularity,
3140 bool has_buf_size, int64_t buf_size,
3141 bool has_on_source_error,
3142 BlockdevOnError on_source_error,
3143 bool has_on_target_error,
3144 BlockdevOnError on_target_error,
3145 bool has_filter_node_name,
3146 const char *filter_node_name,
3147 bool has_copy_mode, MirrorCopyMode copy_mode,
3148 bool has_auto_finalize, bool auto_finalize,
3149 bool has_auto_dismiss, bool auto_dismiss,
3150 Error **errp)
3152 BlockDriverState *bs;
3153 BlockDriverState *target_bs;
3154 AioContext *aio_context;
3155 AioContext *old_context;
3156 BlockMirrorBackingMode backing_mode = MIRROR_LEAVE_BACKING_CHAIN;
3157 Error *local_err = NULL;
3158 bool zero_target;
3159 int ret;
3161 bs = qmp_get_root_bs(device, errp);
3162 if (!bs) {
3163 return;
3166 target_bs = bdrv_lookup_bs(target, target, errp);
3167 if (!target_bs) {
3168 return;
3171 zero_target = (sync == MIRROR_SYNC_MODE_FULL);
3173 /* Honor bdrv_try_set_aio_context() context acquisition requirements. */
3174 old_context = bdrv_get_aio_context(target_bs);
3175 aio_context = bdrv_get_aio_context(bs);
3176 aio_context_acquire(old_context);
3178 ret = bdrv_try_set_aio_context(target_bs, aio_context, errp);
3180 aio_context_release(old_context);
3181 aio_context_acquire(aio_context);
3183 if (ret < 0) {
3184 goto out;
3187 blockdev_mirror_common(has_job_id ? job_id : NULL, bs, target_bs,
3188 has_replaces, replaces, sync, backing_mode,
3189 zero_target, has_speed, speed,
3190 has_granularity, granularity,
3191 has_buf_size, buf_size,
3192 has_on_source_error, on_source_error,
3193 has_on_target_error, on_target_error,
3194 true, true,
3195 has_filter_node_name, filter_node_name,
3196 has_copy_mode, copy_mode,
3197 has_auto_finalize, auto_finalize,
3198 has_auto_dismiss, auto_dismiss,
3199 &local_err);
3200 error_propagate(errp, local_err);
3201 out:
3202 aio_context_release(aio_context);
3205 /* Get a block job using its ID and acquire its AioContext */
3206 static BlockJob *find_block_job(const char *id, AioContext **aio_context,
3207 Error **errp)
3209 BlockJob *job;
3211 assert(id != NULL);
3213 *aio_context = NULL;
3215 job = block_job_get(id);
3217 if (!job) {
3218 error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
3219 "Block job '%s' not found", id);
3220 return NULL;
3223 *aio_context = blk_get_aio_context(job->blk);
3224 aio_context_acquire(*aio_context);
3226 return job;
3229 void qmp_block_job_set_speed(const char *device, int64_t speed, Error **errp)
3231 AioContext *aio_context;
3232 BlockJob *job = find_block_job(device, &aio_context, errp);
3234 if (!job) {
3235 return;
3238 block_job_set_speed(job, speed, errp);
3239 aio_context_release(aio_context);
3242 void qmp_block_job_cancel(const char *device,
3243 bool has_force, bool force, Error **errp)
3245 AioContext *aio_context;
3246 BlockJob *job = find_block_job(device, &aio_context, errp);
3248 if (!job) {
3249 return;
3252 if (!has_force) {
3253 force = false;
3256 if (job_user_paused(&job->job) && !force) {
3257 error_setg(errp, "The block job for device '%s' is currently paused",
3258 device);
3259 goto out;
3262 trace_qmp_block_job_cancel(job);
3263 job_user_cancel(&job->job, force, errp);
3264 out:
3265 aio_context_release(aio_context);
3268 void qmp_block_job_pause(const char *device, Error **errp)
3270 AioContext *aio_context;
3271 BlockJob *job = find_block_job(device, &aio_context, errp);
3273 if (!job) {
3274 return;
3277 trace_qmp_block_job_pause(job);
3278 job_user_pause(&job->job, errp);
3279 aio_context_release(aio_context);
3282 void qmp_block_job_resume(const char *device, Error **errp)
3284 AioContext *aio_context;
3285 BlockJob *job = find_block_job(device, &aio_context, errp);
3287 if (!job) {
3288 return;
3291 trace_qmp_block_job_resume(job);
3292 job_user_resume(&job->job, errp);
3293 aio_context_release(aio_context);
3296 void qmp_block_job_complete(const char *device, Error **errp)
3298 AioContext *aio_context;
3299 BlockJob *job = find_block_job(device, &aio_context, errp);
3301 if (!job) {
3302 return;
3305 trace_qmp_block_job_complete(job);
3306 job_complete(&job->job, errp);
3307 aio_context_release(aio_context);
3310 void qmp_block_job_finalize(const char *id, Error **errp)
3312 AioContext *aio_context;
3313 BlockJob *job = find_block_job(id, &aio_context, errp);
3315 if (!job) {
3316 return;
3319 trace_qmp_block_job_finalize(job);
3320 job_ref(&job->job);
3321 job_finalize(&job->job, errp);
3324 * Job's context might have changed via job_finalize (and job_txn_apply
3325 * automatically acquires the new one), so make sure we release the correct
3326 * one.
3328 aio_context = blk_get_aio_context(job->blk);
3329 job_unref(&job->job);
3330 aio_context_release(aio_context);
3333 void qmp_block_job_dismiss(const char *id, Error **errp)
3335 AioContext *aio_context;
3336 BlockJob *bjob = find_block_job(id, &aio_context, errp);
3337 Job *job;
3339 if (!bjob) {
3340 return;
3343 trace_qmp_block_job_dismiss(bjob);
3344 job = &bjob->job;
3345 job_dismiss(&job, errp);
3346 aio_context_release(aio_context);
3349 void qmp_change_backing_file(const char *device,
3350 const char *image_node_name,
3351 const char *backing_file,
3352 Error **errp)
3354 BlockDriverState *bs = NULL;
3355 AioContext *aio_context;
3356 BlockDriverState *image_bs = NULL;
3357 Error *local_err = NULL;
3358 bool ro;
3359 int ret;
3361 bs = qmp_get_root_bs(device, errp);
3362 if (!bs) {
3363 return;
3366 aio_context = bdrv_get_aio_context(bs);
3367 aio_context_acquire(aio_context);
3369 image_bs = bdrv_lookup_bs(NULL, image_node_name, &local_err);
3370 if (local_err) {
3371 error_propagate(errp, local_err);
3372 goto out;
3375 if (!image_bs) {
3376 error_setg(errp, "image file not found");
3377 goto out;
3380 if (bdrv_find_base(image_bs) == image_bs) {
3381 error_setg(errp, "not allowing backing file change on an image "
3382 "without a backing file");
3383 goto out;
3386 /* even though we are not necessarily operating on bs, we need it to
3387 * determine if block ops are currently prohibited on the chain */
3388 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_CHANGE, errp)) {
3389 goto out;
3392 /* final sanity check */
3393 if (!bdrv_chain_contains(bs, image_bs)) {
3394 error_setg(errp, "'%s' and image file are not in the same chain",
3395 device);
3396 goto out;
3399 /* if not r/w, reopen to make r/w */
3400 ro = bdrv_is_read_only(image_bs);
3402 if (ro) {
3403 if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) {
3404 goto out;
3408 ret = bdrv_change_backing_file(image_bs, backing_file,
3409 image_bs->drv ? image_bs->drv->format_name : "");
3411 if (ret < 0) {
3412 error_setg_errno(errp, -ret, "Could not change backing file to '%s'",
3413 backing_file);
3414 /* don't exit here, so we can try to restore open flags if
3415 * appropriate */
3418 if (ro) {
3419 bdrv_reopen_set_read_only(image_bs, true, &local_err);
3420 error_propagate(errp, local_err);
3423 out:
3424 aio_context_release(aio_context);
3427 void qmp_blockdev_add(BlockdevOptions *options, Error **errp)
3429 BlockDriverState *bs;
3430 QObject *obj;
3431 Visitor *v = qobject_output_visitor_new(&obj);
3432 QDict *qdict;
3434 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3435 visit_complete(v, &obj);
3436 qdict = qobject_to(QDict, obj);
3438 qdict_flatten(qdict);
3440 if (!qdict_get_try_str(qdict, "node-name")) {
3441 error_setg(errp, "'node-name' must be specified for the root node");
3442 goto fail;
3445 bs = bds_tree_init(qdict, errp);
3446 if (!bs) {
3447 goto fail;
3450 bdrv_set_monitor_owned(bs);
3452 fail:
3453 visit_free(v);
3456 void qmp_x_blockdev_reopen(BlockdevOptions *options, Error **errp)
3458 BlockDriverState *bs;
3459 AioContext *ctx;
3460 QObject *obj;
3461 Visitor *v = qobject_output_visitor_new(&obj);
3462 BlockReopenQueue *queue;
3463 QDict *qdict;
3465 /* Check for the selected node name */
3466 if (!options->has_node_name) {
3467 error_setg(errp, "Node name not specified");
3468 goto fail;
3471 bs = bdrv_find_node(options->node_name);
3472 if (!bs) {
3473 error_setg(errp, "Cannot find node named '%s'", options->node_name);
3474 goto fail;
3477 /* Put all options in a QDict and flatten it */
3478 visit_type_BlockdevOptions(v, NULL, &options, &error_abort);
3479 visit_complete(v, &obj);
3480 qdict = qobject_to(QDict, obj);
3482 qdict_flatten(qdict);
3484 /* Perform the reopen operation */
3485 ctx = bdrv_get_aio_context(bs);
3486 aio_context_acquire(ctx);
3487 bdrv_subtree_drained_begin(bs);
3488 queue = bdrv_reopen_queue(NULL, bs, qdict, false);
3489 bdrv_reopen_multiple(queue, errp);
3490 bdrv_subtree_drained_end(bs);
3491 aio_context_release(ctx);
3493 fail:
3494 visit_free(v);
3497 void qmp_blockdev_del(const char *node_name, Error **errp)
3499 AioContext *aio_context;
3500 BlockDriverState *bs;
3502 bs = bdrv_find_node(node_name);
3503 if (!bs) {
3504 error_setg(errp, "Cannot find node %s", node_name);
3505 return;
3507 if (bdrv_has_blk(bs)) {
3508 error_setg(errp, "Node %s is in use", node_name);
3509 return;
3511 aio_context = bdrv_get_aio_context(bs);
3512 aio_context_acquire(aio_context);
3514 if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_DRIVE_DEL, errp)) {
3515 goto out;
3518 if (!QTAILQ_IN_USE(bs, monitor_list)) {
3519 error_setg(errp, "Node %s is not owned by the monitor",
3520 bs->node_name);
3521 goto out;
3524 if (bs->refcnt > 1) {
3525 error_setg(errp, "Block device %s is in use",
3526 bdrv_get_device_or_node_name(bs));
3527 goto out;
3530 QTAILQ_REMOVE(&monitor_bdrv_states, bs, monitor_list);
3531 bdrv_unref(bs);
3533 out:
3534 aio_context_release(aio_context);
3537 static BdrvChild *bdrv_find_child(BlockDriverState *parent_bs,
3538 const char *child_name)
3540 BdrvChild *child;
3542 QLIST_FOREACH(child, &parent_bs->children, next) {
3543 if (strcmp(child->name, child_name) == 0) {
3544 return child;
3548 return NULL;
3551 void qmp_x_blockdev_change(const char *parent, bool has_child,
3552 const char *child, bool has_node,
3553 const char *node, Error **errp)
3555 BlockDriverState *parent_bs, *new_bs = NULL;
3556 BdrvChild *p_child;
3558 parent_bs = bdrv_lookup_bs(parent, parent, errp);
3559 if (!parent_bs) {
3560 return;
3563 if (has_child == has_node) {
3564 if (has_child) {
3565 error_setg(errp, "The parameters child and node are in conflict");
3566 } else {
3567 error_setg(errp, "Either child or node must be specified");
3569 return;
3572 if (has_child) {
3573 p_child = bdrv_find_child(parent_bs, child);
3574 if (!p_child) {
3575 error_setg(errp, "Node '%s' does not have child '%s'",
3576 parent, child);
3577 return;
3579 bdrv_del_child(parent_bs, p_child, errp);
3582 if (has_node) {
3583 new_bs = bdrv_find_node(node);
3584 if (!new_bs) {
3585 error_setg(errp, "Node '%s' not found", node);
3586 return;
3588 bdrv_add_child(parent_bs, new_bs, errp);
3592 BlockJobInfoList *qmp_query_block_jobs(Error **errp)
3594 BlockJobInfoList *head = NULL, **p_next = &head;
3595 BlockJob *job;
3597 for (job = block_job_next(NULL); job; job = block_job_next(job)) {
3598 BlockJobInfoList *elem;
3599 AioContext *aio_context;
3601 if (block_job_is_internal(job)) {
3602 continue;
3604 elem = g_new0(BlockJobInfoList, 1);
3605 aio_context = blk_get_aio_context(job->blk);
3606 aio_context_acquire(aio_context);
3607 elem->value = block_job_query(job, errp);
3608 aio_context_release(aio_context);
3609 if (!elem->value) {
3610 g_free(elem);
3611 qapi_free_BlockJobInfoList(head);
3612 return NULL;
3614 *p_next = elem;
3615 p_next = &elem->next;
3618 return head;
3621 void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread,
3622 bool has_force, bool force, Error **errp)
3624 AioContext *old_context;
3625 AioContext *new_context;
3626 BlockDriverState *bs;
3628 bs = bdrv_find_node(node_name);
3629 if (!bs) {
3630 error_setg(errp, "Cannot find node %s", node_name);
3631 return;
3634 /* Protects against accidents. */
3635 if (!(has_force && force) && bdrv_has_blk(bs)) {
3636 error_setg(errp, "Node %s is associated with a BlockBackend and could "
3637 "be in use (use force=true to override this check)",
3638 node_name);
3639 return;
3642 if (iothread->type == QTYPE_QSTRING) {
3643 IOThread *obj = iothread_by_id(iothread->u.s);
3644 if (!obj) {
3645 error_setg(errp, "Cannot find iothread %s", iothread->u.s);
3646 return;
3649 new_context = iothread_get_aio_context(obj);
3650 } else {
3651 new_context = qemu_get_aio_context();
3654 old_context = bdrv_get_aio_context(bs);
3655 aio_context_acquire(old_context);
3657 bdrv_try_set_aio_context(bs, new_context, errp);
3659 aio_context_release(old_context);
3662 QemuOptsList qemu_common_drive_opts = {
3663 .name = "drive",
3664 .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head),
3665 .desc = {
3667 .name = "snapshot",
3668 .type = QEMU_OPT_BOOL,
3669 .help = "enable/disable snapshot mode",
3671 .name = "aio",
3672 .type = QEMU_OPT_STRING,
3673 .help = "host AIO implementation (threads, native, io_uring)",
3675 .name = BDRV_OPT_CACHE_WB,
3676 .type = QEMU_OPT_BOOL,
3677 .help = "Enable writeback mode",
3679 .name = "format",
3680 .type = QEMU_OPT_STRING,
3681 .help = "disk format (raw, qcow2, ...)",
3683 .name = "rerror",
3684 .type = QEMU_OPT_STRING,
3685 .help = "read error action",
3687 .name = "werror",
3688 .type = QEMU_OPT_STRING,
3689 .help = "write error action",
3691 .name = BDRV_OPT_READ_ONLY,
3692 .type = QEMU_OPT_BOOL,
3693 .help = "open drive file as read-only",
3696 THROTTLE_OPTS,
3699 .name = "throttling.group",
3700 .type = QEMU_OPT_STRING,
3701 .help = "name of the block throttling group",
3703 .name = "copy-on-read",
3704 .type = QEMU_OPT_BOOL,
3705 .help = "copy read data from backing file into image file",
3707 .name = "detect-zeroes",
3708 .type = QEMU_OPT_STRING,
3709 .help = "try to optimize zero writes (off, on, unmap)",
3711 .name = "stats-account-invalid",
3712 .type = QEMU_OPT_BOOL,
3713 .help = "whether to account for invalid I/O operations "
3714 "in the statistics",
3716 .name = "stats-account-failed",
3717 .type = QEMU_OPT_BOOL,
3718 .help = "whether to account for failed I/O operations "
3719 "in the statistics",
3721 { /* end of list */ }
3725 QemuOptsList qemu_drive_opts = {
3726 .name = "drive",
3727 .head = QTAILQ_HEAD_INITIALIZER(qemu_drive_opts.head),
3728 .desc = {
3730 * no elements => accept any params
3731 * validation will happen later
3733 { /* end of list */ }